SMALL UNMANNED AIRCRAFT SYSTEMS AND THEIR PAYLOADS AS AERIAL DATA COLLECTION PLATFORMS FOR NATURAL RESOURCE BASED APPLICATIONS By MATTHEW ALEXANDER BURGESS A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY O F FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2017
2017 Matthew Alexander Burgess
To my loving wife, Brandy Leigh Burgess, and our Lab Baxter, Spencer, George, and Frank
4 ACKNOWLEDGMENTS This dissertation would not be possible without the help, guidance, and support of so many people. I am especially indebted to the members of my graduate committee: H. Fra nklin Percival, Raymond R. Carthy, Peter C. Frederick, Peter G. Ifju, and Benjamin E. Wilkinson. Each of them was instrumental in helping me navigate through the Ph.D. process, providing ways to make ends meet, and offering outstanding mentorship that has no doubt prepared me for the future. Other individuals contributed greatly toward the success of this dissertation who were not on my official graduate committee, but certainly could have been, includ ing Amr. H. Abd Elrahman, Mark I. Cook, Bon A. Dewitt, Jennifer S. Forbey, Susan Newman, Janet L. Rachlow, H. LeRoy Rodgers, Lisa A. Shipley, Scot E. Smith, and Christa L. Zweig. I am grateful to W. James Fleming, John F. Organ, John Thompson, Kevin G. Whalen, and B. Ken Williams, among others of the U.S. Ge ological Survey Cooperative Research Units Program for all that they do to prepare up and coming researchers and academics. Other U.S. Geological Survey employees (some current, some now retired) have helped me with my endeavors: Josip D. Adams, Gordon H. Anderson, Nicholas G. Aumen, Christopher A. Barnes, Mark A. Bauer, Cathy A. Beck, G. Ronnie Best, William D. Christiansen, Courtney J. Conway, Jill J. Cress, Donald E. Dennerline, Robert M. Dorazio, Thomas W. Doyle, Lisa M. Faust, Mark R. Feller, Leanne H anson, Kristen M. Hart, Vic Hines, Christopher L. Holmquist Johnson, Dane H. Huge, Michael E. Hutt, David A. Johncox, Fred A. Johnson, John W. Jones, Siddiq S. Kalaly, Suzette M. Kimball, Anne E. Kinsinger, Wiley M. Kitchens, Elizabeth Martin, Julien Marti n, James D. Nichols, Abby N. Powell, Bruce K. Quirk, Kenneth G. Rice, Timothy S. Saucier, Kate
5 Schoenecker, J. Michael Scott, Matthew Sexson, Jeff L. Sloan, E. Terrence Slonecker, Bradley M. Stith, Jonathan Stock, Lynn J. Torak, and Scott A. Wilson. Thanks to personnel at the Federal Aviation Administration: Byron Chew, Andrew Collins, Douglas Gould, Ashley Hale, Mark A. Jordan, Earl Lawrence, Keith Lusk, Joseph Maibach, Douglas R. Murphy, Lynda G. Otting, Thomas Rampulla, Dina Reyes Garcia, Ardyth M. Willi ams, James H. Williams, Randy Willis, and Michael K. Wilson, who are just a few of the many people with who m I have had the pleasure of interacting with Thanks to Harry J. Kieling, Mark L. Bathrick, Erin Horsburgh, Bradley S. Koeckeritz, Keith C. Raley, B laine Moriarty, Martha Watkins, Frank Crump III and Shari Moultrie and others at the US Department of the Interior, Office of Aviation Services, for their assistance A special thanks to Alexander J. Brostek, James Denigris, Robert A. Flathmann, Gary Han sen, W. Lee McBrien, AnnMarie Muscardin, Ivan Ortiz Sepulveda, David Reinert, and Joseph K. Wells at the South Florida Water Management District Flight Operations Unit for their patience, understanding, and willingness to try new ideas. T he following peopl e provid ed support, assistance, and/or opportunities along the way: Guy W. Adema, Brad Alcorn, David W. Allen, Robyn P. Angliss, Bill D. Arnsberg, Joel Arrieta, William Auer, Suzanne C. Baird, Steve Banks, Gail Banta, M. Lee Barber, Kim Bassos Hull, Susan Bates, Suzanne Beauchaine, Daniel Beck, Megan Benish, Courtney Bensey, Bruce M. Bicknell, Aitor Bidaburu, Ronald R. Bielefeld, David M. Bird, Mark Blanks, Matthew R. Bobo, Robin G. Boughton, Lance R. Brady, Laura A. Brandt, Gary Brennan, Danielle Broussard Todd A. Burton, Amanda Campbell, Joshua Castro,
6 Dominique M. Chabot, Ryan M. Chabot, John C. Coffey, Bradley B. Compton, Kendra Cope, Jeremy M. Crossland, Glenn Cullingford, Pam Darty, Hany Derias, Joe DeVivo, Victor R. Doig, Joe DuPont, Andrew Eastwick, Holly H. Edwards, Susan N. Ellis Felege, Matthew M. Fladeland, George L. Flynn, Hassan Foroosh, Cindy Fury, Margaret Gallagher, Michael S. Gallagher, G. Daniel Gann, Helena C. Gianniai, Phillip A. Groves, Susan Goplen, Patti Gorman, Andrew G. Gude, Philip G. Hall, Daren Harmon, David Harmon, Peter B. Heasley, Dennis W. Heinemann, Van T. Helker, Michael T. Hensch, Art W. Hinaman, Steven M. Hogan, Susan M. Hohner, Robbie E. Hood, Peter T. Hull, David A. Hunnicutt, Delia B. Ivanoff, Jim W. Jeffords, Robert L. Johnson, Joel Kerley, John K. Kilpatrick, Joyce M. Kleen, Kristen R. Kneifl, William R. Koski, Jeffrey Krueger, Peter W. Kubiak, Jon S. Lane, Anthony Lascano, Brad J. Laubach, Brent Lignell, Christopher Long, Marguerite M. Madden, Michael G. Manna, Deepak R. Mishra, Michael J. Moore, Lisa Moore LaRoe, Erin E. Moreland, Jon M. Morton, Colleen E. Moulton, Margarita Mulero Pzmny, Brian J. Mullin, Shawn Nagle, Tom Noble, Antonio Pernas, Wayne L. Perryman, Andrew E. Pontzer, Jane Powers, Kyle E. Rambo, Stephe n V. Rauch, Jed R. Redwine, Jonathan J. Rees, Glenn G. Rhett, Jennifer H. Richards, Orien M.W. Richmond, David J. Robar, David W. Roberts, Jon Rollens, Robert Roth, Shaun Sanchez, Colin J. Saunders, Raymond M. Sauvajot, Monette V. Schwoerer, Jennifer R. Se avey, Kristin L. Seitz, Roger Semler, Fred H. Sklar, Jacqueline C. Smith, Ray W. Snow, Lawrence J. Spencer, Thomas M. Spencer, Andrew Sterner, Michael H. Story, Kathryn L. Sweeney, Stuart S. Taft, Marshall F. Tappen, Adam N. Tarplee, Larry E. Taylor, James Traub, Karen K. Trevino, David A. Viker, Betsy Von Holle, Martha C. Wackenhut, Paul Wackenhut, Robert M. Wallace, Richard
7 W. Ward, Vicki Ward, John F. Weishampel, Nick Wiley, Victor L. Wilhelm, Larry A. Woodward, James S. Wortham, Peter E. Zager, and Will iam C. Zattau. This work would not have been possible without the assistance of Travis J. Whitley, Joseph G. DiRodio, Chad S. Tripp, H. Andrew Lassiter, Yun Ye, Tyler S. Ward, Abraham Balmori, Rodney M. Hunt, Jesse A. Durrance, Zoltan Szantoi, Spencer J. I ngley, Damon A. Wolfe, John H. Perry, Brandon S. Evers, Thomas J. Rambo, Thomas S. Reed, Kyuho Lee, Joshua A. Childs, Hector Y. Rodriguez Asilis, John C. Simon, Adam C. Watts, R. Scott Bowman, Michael J. Morton, Jamie A. Duberstein, George P. Jones, Ahmed H. Mohamed, Leonard G. Pearlstine, and Kenneth D. Meyer, among others as well. Thanks to Becky Abel, Jason Beck, Toby Boudreau, Steven M. Bradbury, Arnie F. Brimmer, Kenneth Bugler, Mark A. Chappell, Miranda M. Crowell, Donna M. Delparte, Joe DuPont, Wendy A. Estes Zumpf, Marcella R. Fremgen, Aaron P. Garcia, Kristina Gehlken, Gifford L. Gillette, Nancy F. Glenn, Jeff Gould, Jon K. Heggen, Pam R. Johnson, Meghan J. Leiper, Zachary B. Lockyer, Jeff Lonneker, Robert C. Lonsinger, Kris Millgate, Charlotte R. M illing, Jessica J. Mitchell, Frank Mullins, Jordan D. Nobler, Peter J. Olsoy, Amanda J. Price, Scott Putnam, Jon Rachael, Brecken C. Robb, Rex Sallabanks, Gael Sanchez, William C. Schrader, David M. Teuscher, Daniel H. Thornton, Amy C. Ulappa, Jamie L. Utz Natasha L. Wiggins, and James H. Witham, for assistance with various efforts conducted in and around Idaho. I am indebted to University of Florida administrators and faculty including: Michael S. Allen, Brandi K. Boniface, Lyn C. Branch, Jacqueline K. Bu rns, Stephen F. Coates, Irene M. Cooke, Terra A. DuBois, B. Dianne Farb, Dennis Fleetwood, William
8 M. Giuliano, Joseph Glover, David W. Hahn, Amy Meyers Hass, John P. Hayes, Eric C. Hellgren, Steven A. Johnson, Joseph C. Joyce, Jamie L. Keith, Jeanna M. Ma strodicasa, Kenneth W. McCain, Robert A. McCleery, Christina S. Moore, Christopher J. Moran, David P. Norton, Jack M. Payne, Elizabeth F. Pienaar, William E. Pine, Brian E. Prindle, William S. Properzio, Kathryn E. Sieving, Gary E. Vallad, and Tim L. White who have provided advice, education, and support through the course of exponential growth in small unmanned aircraft systems. Thanks to M. Gay Hale, Jennifer L. Miller, Hannah M. Taylor, Kathleen Van Riet, Alexis T. Martin, Janet L. Fay, Amanda S. Burnet t, Joan G. Hill, and the entire Florida Cooperative Fish and Wildlife Research Unit staff. I would also like to thank Tom J. Barnash, Heather l. Bradley, Clare A. Condon Grade, Kyle D. Cook, J. Elaine Culpepper, Kelley J. Cunningham, Fiona I. Hogan, Samue l A. Jones, Monica M. Lindberg, Caprice M. McRae, Kaleigh N. Riley, Claire C. Williams, and all the University of Florida Department of Wildlife Ecology and Conservation support staff over the years. Thanks also to Krystan A. Wilkinson, Travis M. Thomas, B rian M. Jeffery, Nichole D. Bishop, Brian J. Smith, Nicole J. Wood, Rachel M. Cassoff, Noah S. Burrell, Auriel M.V. Fournier, Brian E. Reichert, Chelsey R. Faller, Benjamin K. Atkinson, Rio W. Bonds, Janell M. Brush, Carolyn M. Enloe, Cameron B. Carter, Me lissa A. DeSa, and Zachariah C. Welch. I thank my parents, George H. and Linda S. Burgess, my brother Nathan H. Burgess, and my in laws, Kenneth C. and JoAnn M. Quiggle, for their love and encouragement through this journey. Thanks also to my many relativ es across the country for their appreciation
9 Last, but certainly not least, I thank my wonderful wife Brandy L. Burgess for all that she has done to encourage me and accommodate my pursuits of this degree. I certainly would not have been able to do thi s without her unconditional love and generous support
10 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF FIGURES ................................ ................................ ................................ ........ 13 LIST OF ABBREVIATIONS ................................ ................................ ........................... 18 ABS TRACT ................................ ................................ ................................ ................... 25 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ ..... 27 Primary Data Collection Methods in Natural Resources ................................ ......... 28 Boots on the Ground ................................ ................................ ........................ 28 Manned Aircraft ................................ ................................ ................................ 29 Space borne Satellites ................................ ................................ ..................... 31 Unmanned Aircraft Systems ................................ ................................ ............. 32 What are Unmanned Aircraft Systems? ................................ ................................ .. 33 The Unmanned Aircraft ................................ ................................ .................... 34 The Ground Control Station ................................ ................................ .............. 34 The Communication Linkages ................................ ................................ .......... 35 The Flight Team ................................ ................................ ............................... 36 UAS in Natural Resource Based Sciences ................................ ............................. 36 Foundations of the Research Topic ................................ ................................ ........ 38 2 INFERENCES OF HABITAT SELECTION BY PYGMY RABBITS ( Brachylagus idahoensis ) BASED ON NORMALIZED DIF FERENCE VEGETATIVE INDICES GENERATED FROM IMAGERY COLLECTED VIA SMALL UNMANNED AIRCRAFT SYSTEM ................................ ................................ .............................. 4 2 The Sagebrush Steppe Study Area s ................................ ................................ ...... 48 Magic Reservoir ................................ ................................ ............................... 48 Rocky Canyon ................................ ................................ ................................ .. 49 Methods Used for Assessing Pygmy Rabbit Habitats with sUAS derived NDVI Imagery Products ................................ ................................ ................................ 50 Results of Pygmy Rabbit Habitat Selection Using sUAS NDVI Imagery Products .. 56 Discussion of Pygmy Rabbit Habitat Selection Based on sUAS NDVI Imagery Products ................................ ................................ ................................ .............. 60 3 ESTIMATES OF AMERICAN WHITE PELICANS ( Pelecanus erythrorhynchos ) OBTAINED FROM PHOTOGRAMMETRIC PRODUCTS OF IMAGERY GATHERED USING A SMALL UNMANNED AIRCRAFT SYSTEM ....................... 76 American White Pelican Study Area ................................ ................................ ....... 80
11 Methods Used for Estimating Nesting American White Pelicans from sUAS derived Imagery Products ................................ ................................ .................... 82 Results of the sUAS Imagery Products for Assessing Estimates of American White Pelican Nesting ................................ ................................ .......................... 86 Discussion on the Use of sUAS I magery Products to Assess Estimates of American White Pelican Nesting ................................ ................................ ......... 89 4 AN INNOVATIVE DATA COLLECTION APPROACH FOR MANN ED AERIAL SURVEYS USING OPTICAL PAYLOADS DESIGNED FOR SMALL UNMANNED AIRCRAFT SYSTEMS ................................ ................................ .... 110 Study Areas ................................ ................................ ................................ .......... 113 Archie Carr National Wildlife Refuge ................................ .............................. 113 Water Conservation Area 2A ................................ ................................ .......... 114 Methods ................................ ................................ ................................ ................ 116 Results ................................ ................................ ................................ .................. 120 Discussion ................................ ................................ ................................ ............ 122 5 CONCLUSIONS ................................ ................................ ................................ .... 146 Pygmy Rabbit ( Brachylagus idahoensis ) Habitat Selection Inferences Determined by Normalized Difference Vegetative Index Computations of Aerial Imagery Products Generated from Data Col lected by a Small Unmanned Aircraft System ................................ ................................ ................ 147 Generating Estimates of Nesting American White Pelicans ( Pelecanus erythrorhynchos) on Spoil Islands in Minidoka National Wildlife Refuge (Idaho) from Aerial Imagery Products Produced from Data Collected by a Small Unmanned Aircraft System ................................ ................................ ...... 148 Use of High Resolution Optical Payloads Designed for Small Unmanned Aircraft Systems as an Innovative Approach for Aerial Data Collection from Manned Aircraft Flights ................................ ................................ ...................... 149 APPENDIX A THE HISTORY OF THE UNIVERSITY OF FLORIDA UNMANNED AIRCRAFT SYSTEMS RESEARCH PROGRAM (UFUASRP): 1999 2017 ............................ 151 The MLB Company FoldBat ................................ ................................ ............. 152 The UF Tadpole ................................ ................................ ................................ .... 161 The UF Nova 1 (Polaris) ................................ ................................ ....................... 168 The UF Nova 2 ................................ ................................ ................................ ..... 181 The UF Nova 2.1 (Mako) ................................ ................................ ...................... 192 The DJI Spreading Wings S1000+ ................................ ................................ ... 208 B A BRIEF HISTORY OF UNMANNED AIRCRAFT SYSTEM REGULATIONS IN THE UNITED STATES ................................ ................................ ......................... 217
12 C THE UNIVERSITY OF FLORIDA UNMANNED AIRCRAFT SYSTEMS RESEARCH PROGRAM (UFUASRP) FLIGHT CREW MODEL ........................... 226 Pilot in Command ................................ ................................ ........................... 226 Ground Station Operator ................................ ................................ ................ 227 Qualified Visual Observer ................................ ................................ ............... 228 D DIGITAL IMAGERY AND METADATA POST PROCESSING METHODOLOGY USED ................................ ................................ ................................ .................... 230 E SEVERAL CRITICAL LESSONS LEARNED BY THE UNIVERSITY OF FLORIDA UNMANNED AIRCRAFT SYSTEMS RESEARCH PROGRAM (UFUASRP): 1999 2017 ................................ ................................ ....................... 234 F THE UNIVERSITY OF FLORIDA UNMANNED AIRCRAFT SYSTEMS RESEACH PROGRAM (UFUASRP) BLANK FLIGHT DATA SHEET ................... 243 LIST OF REFERENCES ................................ ................................ ............................. 244 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 261
13 L IST OF FIGURES Figure page 1 1 Several of the most widely recognized drones used for unmanned military applications.. ................................ ................................ ................................ ....... 41 2 1 A pygmy rabbit ( Brachylagus idahoensis ). ................................ .......................... 64 2 2 Signs of recent pygmy rabbit ( Brachylagus idahoensis ) presence.. .................... 65 2 3 The extent of known pygmy rabbit ( Brachylagus idahoensis ) populations in the western United States. ................................ ................................ ................. 66 2 4 An orthophotomosaic aerial image showing mima mounds distributed across the sagebrush steppe landscape in the Lemhi Valley, Lemhi County, Idaho, USA. ................................ ................................ ................................ ................... 67 2 5 The geographic information system spatial modeling concept for generating a response surface based on individual predictor v ariable layers for pygmy rabbits ( Brachylagus idahoensis ). ................................ ................................ ....... 68 2 6 Brachylagus idahoensis ) data collection. ................................ ................................ ................................ ........... 69 2 7 The location of the Magic Reservoir study site for pygmy rabbits ( Brachylagus idahoensis ), in Blaine County, Idaho, USA. ................................ ........................ 70 2 8 The location of the Rocky C anyon study site for pygmy rabbits ( Brachylagus idahoensis ), in Lemhi County, Idaho, USA. ................................ ........................ 71 2 9 Locations and distribution of the n = 84 geolocated ground control points established at the Magic Reservoir study site during summer 2013 in Blaine County, Idaho, USA. ................................ ................................ ........................... 72 2 10 Locations and distribution of the n = 111 geolocated ground control points established at the Rocky Canyo n study site during summer 2014 in Lemhi County, Idaho, USA. ................................ ................................ ........................... 72 2 11 Selected mima mounds intensively ground truthed at the Magic Reservoir study site during summer 2013 in Blaine County, Id aho, USA. .......................... 73 2 12 An orthophotomosaic aerial image with a normalized difference vegetation index (NDVI) overlay of the Rocky Canyon study site during summer 2014 in Lemhi County, Idaho, USA. ................................ ................................ ................ 74
14 2 13 The southern two subareas of the Rocky Canyon study site shown with a normalized difference vegetative index (NDVI) overlay generated during summer 2014 in Lemhi County, Idaho, USA. ................................ ..................... 75 3 1 A colony of American White Pelicans ( Pelecanus erythrorhynchos ) nesting on Pelican Island, Minidoka National Wildlife Refuge, Cassia County, Idaho, USA. ................................ ................................ ................................ ................... 95 3 2 The location of the three spoil islands used for nesting by American White Pelic ans ( Pelecanus erythrorhynchos ) in Minidoka National Wildlife Refuge, Cassia County, Idaho, USA. ................................ ................................ ............... 96 3 3 The three spoil islands used for nesting by American White Pelicans ( Pelecanus erythrorhynchos ) in Minidoka National Wildlife Refuge, Cassia County, Idaho, USA. ................................ ................................ ........................... 97 3 4 The Virtual Cockpit preplanned flight path designed for autonomous waypoint navigation by the University of Florida Nova 2.1 small unmanned aircraft system over the three spoil islands used for nesting by American White Pelicans ( Pelecanus erythrorhynchos ) on 10 June 2014 in Minidoka National Wildlife Refuge, Cassia County, Idaho, USA. ................................ ....... 98 3 5 Examples of American White Pelican ( Pelecanus erythrorhynchos ) nesting locations on spoil islands in Minidoka National Wildlife Refuge, Cassia County, Idaho, USA. ................................ ................................ ........................... 99 3 6 Swirls that result from mosaicking still imagery of herbaceous vegetation blown around by ambient winds. ................................ ................................ ...... 100 3 7 Exaggerating the elevations of a digital elevation mo del in attempt to highlight ground nesting American White Pelicans ( Pelecanus erythrorhynchos ) on Gull Island, Minidoka National Wildlife Refuge, Cassia County, Idaho, USA. ... 101 3 8 Comparison of iTAG software for counting a subarea of nesting American White Pelicans ( Pelecanus erythrorhynchos ) on Gull Island, Minidoka National Wildlife Refuge, Cassia County, Idaho, USA. ................................ ..... 102 3 9 An orthophotomosaic generated by PhotoScan software from imagery collected using a small unmanned aircraft system over Pelican Island, Minidoka National Wildlife Refuge, Cassia County, Idaho, USA on 10 June 2014. ................................ ................................ ................................ ................ 103 3 10 An orthophotomosaic generated by PhotoScan software from imagery collected using a small unmanned aircraft system over Tern Island, Minidoka National Wildlife Refuge, Cassia County, Idaho, USA on 10 June 2014. ......... 104
15 3 11 An orthophotomosaic generated by PhotoScan software from imagery collected using a small unmanned aircraft system over Gull Island, Minidoka National Wildlife Refuge, Cassia County, Idaho, USA on 10 June 2014. ......... 105 3 12 Frequency histogram of observed number of American White Pelicans ( Pelecanus erythrorhynchos ) on Pelican Island, in Minidoka Natio nal Wildlife Refuge, Cassia County, Idaho, USA. ................................ ............................... 106 3 13 Frequency histogram of o bserved number of American White Pelicans ( Pelecanus erythrorhynchos ) on Tern Island, in Minidoka National Wildlife Refuge, Cassia County, Idaho, USA. ................................ ............................... 107 3 14 Frequency histogram of observed number of American White Pelicans ( Pelecanus erythrorhynchos ) on Gull Island, in Minidoka National Wildlife Refuge, Cassia County, Idaho, USA. ................................ ............................... 108 3 15 Abandoned or empty nests are visible among ground nesting American White P elicans ( Pelecanus erythrorhynchos ) in a small subarea of an aerial orthophotomosaic constructed from imagery collected using a small unmanned aircraft over Pelican Island, Minidoka National Wildlife Refuge, Cassia County, Idaho, USA on 10 June 2014. ................................ ................. 109 4 1 The location of the July 2013 external sensor pod study site over Archie Carr National Wildlife Refuge, Brevard County, Florida, USA. ................................ 129 4 2 The 21.0 kilometer stretch of beach that was aerially imaged in July 2013 with the external sensor pod o ver the Archie Carr National Wildlife Refuge, Brevard County, Florida, USA. ................................ ................................ ......... 130 4 3 The loca tion of the active restoration project sites in Water Conservation Area 2A of the Greater Florida Everglades, Palm Beach and Broward Counties, Florida, USA. ................................ ................................ .................... 131 4 4 The active restoration project sites in Water Conservation Area 2A of the Greater Florida Everglades, Palm Beach and Broward Counties, Florida, USA. ................................ ................................ ................................ ................. 132 4 5 The initial external sensor pod. ................................ ................................ ......... 133 4 6 The second generation external sensor pod. ................................ .................... 134 4 7 The third generation external sensor pod. ................................ ........................ 135 4 8 The third generation external sensor pod affixed to a Bell 407. ...................... 136 4 9 The second generation external sensor pod affixed to multiple manned aircraft types. ................................ ................................ ................................ .... 137
16 4 10 Image of the Cessna 172M Skyhawk equipped with the second generation external sensor pod flying a linear transect over the photographer on 23 July 2013 at Archie Carr National Wildlife Refuge, Brevard County, Florida, USA. 138 4 11 A resulting orthophotomosaic of selected imagery from a portion of a flight of the second generation external sensor pod attached to a Cessna 172M Skyhawk on 23 July 2013 over Archie Carr National Wildlife Refuge, Brevard County, Florida, USA. ................................ ................................ ......... 139 4 12 Enlarged portion of an orthophotomosaic generated from selected imagery collected during the flight of the second generation external sensor pod attached to a Cessna 172M Skyhawk on 23 July 2013 at kilometer 14.7 of Archie Carr National Wildlife Refuge, Brevard County, Florida, USA. .............. 140 4 13 Enlarged portion of an orthophotomosaic generated from selected imagery collected during the flight of the second generation external sensor pod attached to a Cessna 172M Skyhawk on 23 July 2013 at kilometer 14.8 of Archie Carr National Wildlife Refuge, Brevard County, Florida, USA. .............. 141 4 14 Enlarged portion of an orthophotomosaic generated from selected imagery collected during the flight of the second generation external sensor pod attached to a Cessna 172 M Skyhawk on 23 July 2013 at kilometer 14.9 of Archie Carr National Wildlife Refuge, Brevard County, Florida, USA. .............. 142 4 15 The paths of three separate flights with the second generation external sensor pod attached to a Bell 407 helicopter in August 2014 over the active restor ation projects in WCA 2A, Palm Beach and Broward Counties, Florida, USA. ................................ ................................ ................................ ................. 143 4 16 The two dimensional locati ons of imagery exposures from three separate flights with the second generation external sensor pod attached to a Bell WCA 2A, Broward County, Florida, USA. ................................ ......................... 144 4 17 The resulting orthophotomosaic of imagery from three separate flights with the second generation external sensor pod attached to a Bell 407 helicopter control plots in WCA 2A, Broward County, Florida, USA. ................................ ................................ ......... 145 A 1 Components of the University of Florida Nova 2.1 small unmanned aircraft system used to c ollect imagery and metadata of sagebrush steppe landscapes during summer 2013 and summer 2014 in Idaho, USA. ................ 215 A 2 The Olympus E 420 and the Canon EOS Rebel SL1 optical sensor payloads. ................................ ................................ ................................ .......... 216 E 1 An illustration showing the generalized process of proceeding from a starting point to an ending point. ................................ ................................ ................... 242
17 E 2 A recommended methodology for efficiently moving from the start of a scientific project in which a small unmanned aircraft system (sUAS) will be used for data collection to achieving the desired end products necessary for analyses. ................................ ................................ ................................ .......... 242
18 LIST OF ABBREVIATIONS Approximately By Degree < Fewer than, Less than Less than or Equal to > Greater than, More than Greater than or Equal to m Micrometer s Microsecond Negative Plus or Minus + Positive Â§ Section 2D Two Dimensional 3D Three Dimens ional A Ampere AC Advisory Circular ACNWR Archie Carr National Wildlife Refuge AGL Above Ground Level aka Also Known As AMA Academy of Model Aeronautics AMI Active Marsh Improvement ASL Above Sea Level ATV All Terrain Vehicle
19 AUVSI Association of Unmanned Vehicle Systems International AWP American White Pelican BOTG Boots on the Ground BSU Boise State University Copyright C Celsius CAD Computer Aided Design CCD Charge Coupled Device CFR Code of Federal Regulations CG Center of Gravity CHIP Cattail Habitat Improvement Project cm Centimeter cm/pix Centimeter per Pixel CMOS Complementary Metal Oxide Semiconductor CNC Computer Numerical Control COA Certificate of Waiver or Authorization COTS Commercial off the Shelf DEM Digital Eleva tion Model dSLR Digital Single Lens Reflex EAR Export Administration Regulations EM Electromagnetic EPS Expanded Polystyrene ESC Electronic Speed Controller ESP External Sensor Pod f/sec Frame per Second
20 FAA Federal Aviation Administration FFWCC F lorida Fish and Wildlife Conservation Commission FMRA Federal Aviation Administration Modernization and Reform Act of 2012 g Gram GB Gigabyte GCP Ground Control Point GCS Ground Control Station GEOM Geomatics Program GHz Gigahertz GIS Geographic In formation System GNSS Global Navigation Satellite System GPS Global Positioning System GPS/INS Global Positioning System/Inertial Navigation System GSO Ground Station Operator GUI Graphical User Interface ha Hectare HALE High Altitude, Long Enduranc e HFS Hybrid Flap/Spoiler hr Hour Hz Hertz IACUC Institutional Animal Care and Use Committee IDFG Idaho Department of Fish and Game IDPR Idaho Department of Parks and Recreation IFAS Institute of Food and Agricultural Sciences
21 INS Inertial Navigati on System IR Infrared ISR Intelligence, Surveillance, and Reconnaissance ITAR International Trade in Arms Regulations kg Kilogram km Kilometer KMIA Miami International Airport KMLB Melbourne International Airport KPBI Palm Beach International Airpo rt LiDAR Light Detection and Ranging LiPo Lithium Polymer LSCKNWR Lower Suwannee and Cedar Keys National Wildlife Refuge m Meter m 2 Square Meter m/sec Meter per Second mA Milliampere MAE Department of Mechanical and Aerospace Engineering mAh Milli ampere Hour MALE Medium Altitude, Long Endurance MAV Micro Aerial Vehicle MB Megabyte MEMS Microelectromechanical System mg milligram mg/kg Milligram per Kilogram MHz Megahertz
22 min Minute mm Millimeter mm 2 Square Millimeter MNWR Minidoka Nationa l Wildlife Refuge MOA Memorandum of Agreement MP Megapixel M.S. Master of Science mUAS Micro Unmanned Aircraft System mW Milliwatt n Number of Samples NAS National Airspace System NASA National Aeronautics and Space Administration NDVI Normalized Difference Vegetation Index NIR Near Infrared nm Nanometer NOAA National Oceanic and Atmospheric Administration NOTAM Notice To Airmen NPRM Notice of Proposed Rule Making NSF National Science Foundation NUASPO National Unmanned Aircraft Systems Proj ect Office OAS Office of Aviation Services OEM Original Equipment Manufacturer Ph.D. Doctor of Philosophy PIC Pilot in Command pix Pixel
23 QVO Qualified Visual Observer Registered Trademark RC Remote Control RGB Red Green Blue RMS Root Mean Squar e RPAS Remotely Piloted Aircraft System RTK Real Time Kinematic SD Secure Digital sec Second SfM Structure from Motion SFWMD South Florida Water Management District SOP Standard Operating Procedure sp. Species sUA Small Unmanned Aircraft sUAS S mall Unmanned Aircraft System TIR Thermal Infrared Unregistered Trademark UA Unmanned Aircraft UAS Unmanned Aircraft System UASIO Unmanned Aircraft Systems Integration Office UCF University of Central Florida UDP User Datagram Protocol UF Univers ity of Florida UFUASRP University of Florida Unmanned Aircraft Systems Research Program
24 UI University of Idaho US United States US$ United States Dollar USACE United States Army Corps of Engineers USB Universal Serial Bus USBLM United States Bureau of Land Management USBOR United States Bureau of Reclamation USDOC United States Department of Commerce USDOD United States Department of Defense USDOI United States Department of the Interior USDOS United States Department of State USDOT United Stat es Department of Transportation USFS United States Forest Service USFWS United States Fish and Wildlife Service USGS United States Geological Survey V Volt VLOS Visual Line of Sight VTOL Vertical Takeoff and Landing WCA 2A Water Conservation Area 2A WCA 3A Water Conservation Area 3A WEC Department of Wildlife Ecology and Conservation WGS84 World Geodetic System 1984 WSU Washington State University
25 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy SMALL UNMANNED AIRCRAFT SYSTEMS AND THEIR PAYLOADS AS AERIAL DATA COLLECTION PLATFORMS FOR NATURAL RESOURCE BASED APPLICATIONS By Matthew Alexander Burgess May 201 7 Chair: H. Franklin Percival Major: Wildlife Ecology and Conservation During the last decade, the interest in small unmanned aircraft systems for civilian purposes has grown exponentially, and investigations into whether these systems can be used as sci entific tools for field studies are actively taking place. The University of Florida Unmanned Aircraft Systems Research Program was created 18 years ago by a group of truly interdisciplinary researchers with the goals of developing low cost, autonomous, s mall unmanned aircraft systems and payloads specifically designed to address applications in natural resource based scientific disciplines. Advancements in technology have enabled small unmanned aircraft systems to become viable remote sensing platforms c apable of supplementing existing data collection methodologies and techniques used to assess, study, and monitor focal targets of environmental, natural resources, ecology, conservation, and management interests. Improvements in airframe designs and constr uction methods, coupled with implementation of directly georeferenced remote sensing capabilities into small unmanned aircraft system payloads, have transformed what were initially recreational hobby aircraft with a camera attached, into highly sophisticat ed spatiotemporal data
26 collection tools. Payloads are capable of obtaining high resolution aerial imagery and metadata, which can then be post processed with computer software to generate products such as three dimensional point clouds, digital elevation models, and georectified orthophotomosaics having actual spatial resolutions of less than 3.0 centimeters per pixel. The research highlighted in this document is but a fraction of a much larger effort involving many collaborators and years of testing and d evelopment. The two case studies presented show several of the advantages, and a few of the known limitations that small unmanned aircraft systems can provide to the natural resource professional for collecting data in the field. The third topic explores an alternative means of obtaining data from low altitude, slow airspeed manned aircraft flights using the high resolution directly georeferenced optical imagery payloads designed for small unmanned aircraft systems. The document also includes critical le ssons learned from conducting flights in many different environments with diverse objectives, and recommendations for future developments of small unmanned aircraft systems and their payloads for natural resource based data collection missions.
27 CHAPTER 1 INTRODUCTION The need for cost effective spatial and temporal monitoring, floral or faunal enumeration, or mapping of focal targets is a ubiquitous mission for natural resource scientists and managers to assess trends in attributes such as population den sity community composition, conservation needs and restoration efficacy (e.g., Ralph and Scott 1981 Wil liams et al. 2002 Nichols et al. 2004 Bart et al. 2010 ) Gathering baseline data permits tendencies in quantifiable measures to be analyzed over space and time, which allows researchers to provide guidance to decision makers in formulating policies, drafting regulations, or establishing management strategies based on research derived facts. When directly compared to research in other scientific di sciplines, most studies within the natural resources are at a disadvantage due to comparatively small er fiscal budgets under which natural resource based entities operate ( Giudice et al. 2010 Kobziar et al. 2015 Reynolds et al. 2016 ) Efforts to secure research funding for studies within natural reso urces are extremely competitive owing to the r elatively limited opportunities available for obtaining financial support Perhaps this has contributed in part to some methodologies, techniques, and tools used to gather scientific data in natural resources to remain relatively unchanged through the years F inding affordable ways to use emerging technological advances occurring in a myriad of scientific domains beyond natural resource s could facilitate critical conservation, management, and resource use stud ies that are potentially safer, more time efficient, and produce results that are statistically stronger and more accurate.
28 Primary Data Collection Methods in Natural Resources Three broad data collection methods are primarily exercised for conducting fiel d b oots on the techniques ; 2) manned aircraft surveys ; and 3 ) space borne satellites Within the last decade, a fourth data collection method has gained significant attention, and is the primary topic of this dissertation : t he use of small unmanned aircraft systems (sUAS) and their payloads as scientific data collection tool s in natural resource based research Brief overviews of each of these broad data collection methods are presented, several advan tages and disadvantages of each technique are discussed as are a few considerations that must be accounted for when using each methodology Boots on the Ground In terms of field studies for natural resource based sciences, both aquatic and terrestrial, mo st research data are collected at or below the surface of the Earth. These on the (BOTG) throughout this dissertation although the term is admittedly general F ield research ers often select BOTG methodologies for data collection in part because BOTG methods are : 1) often suitable for smaller scale, manageable investigations with a natural resource based budget; 2) a ble to be conducted in nearly all environmental conditions; 3 ) characteristically among the cheaper methods available for collectin g data; and 4) occasionally accomplished without the need for specialized equipment and/or training However, BOTG methods are typically some of the most time and labor intensive data c ollection options available to the natural resources scientific community, and can require researchers to potentially put their lives at risk due to the location or
29 proximity of their data collection site s to dangerous situations or focal targets ( Towler et al. 2012 Junda et al. 2015 Mulero Pzmny 2015 ) In certain BOTG data collection situations, researchers could either directly or inadvertently disturb the natural behavior or ecology of the focal object of study, yielding biased data sets that might go undetected ( Kendall et al. 2009 Williams et al. 2015 ) Complicating BOTG efforts even further are situations where foca l targets are mobile, cryptic, or dependent on the presence of specific environmental cues or conditions for researchers to observe them These are just a few of the many elements that may need to be addressed when planning, conducting, or making inferenc es from data collected using BOTG techniques or method ologie s for natural resource based research Manned Aircraft Aerial data collection from flights of manned aircraft over focal targets or regions of interest is a technique that has been use d for over a century. Riding as passengers onboard either fixed or rotary wing manned aircraft, visual observers generally collect data in real time by making notations on paper, electronic tablets, or onto voice recorders for transcription at a later date ( Kushlan 1979 Strong and Cowardin 1995 Wozencraft and Li llycrop 2003 ) Certain data can also be collected by hand held sensors by the observers, or from devices directly affixed to the aircraft itself. Use of manned aircraft also enables coverage of larger focal areas and collection of data at target field sites that have considerable distance or unforgiving terrain between them, which might restrict or inhibit BOTG methodologies. However, manned aircraft based data collection methodologies are commonly accompanied with some of the highest fiscal costs per f ield hour ( Urbanek et al. 2012 ) and the proximity and/or availability of aviation fueling locations relative to focal target
30 areas are logistical concerns th at must be accounted for before us ing manned aircraft for data collection. Manned aerial surveys often result in data that consists of a stack of paper data sheets containing a series of tick marks representing the number of target items observed per some prearranged unit of measure ( Tracey et al. 2008 Booth and Cox 2011 August et al. 2015 ) At the conclusion of a flight, the multiple visual observers tally their tick marks, compare their counts and apply various mathematical approaches to obtain a single number, or perhaps a set of counts, which then becom e the long term data available for analyses from that particular flight. Manned aerial surveys also require observers to identify and enumerate focal items quickly and accurately as the aircraft circumnavigates a target area or conducts linear transects. Factors such as observer experience, visual acuity, and fatigue must be accounted for when using manned aircraft methods for data collection ( Caughley 1977 Fleming and Tracey 2008 ) Visual observers are generally not provided much time to resolve possible doubts that may arise, nor are they given many opportunities to reassess areas in which the density o f focal targets may be particularly high and difficult to enumerate. Similar looking targets or rare species may be misidentified or overlooked completely which can generate biased data that may ultimately be used in part to determine larger management de cisions, set seasonal bag limits, or other natural resource policies ( Caughley 1974 Fewster et al. 2008 Pacifici et al. 2012 ) Finally, a significant concern when utilizing manned aircraft for data collection purposes is that conducting these flights typically requires aircraft to be flown at slow ai rspeeds and low altitudes presenting limited capability for successful recovery of flight should the flying of a manned aircraft become compromised. Sasse (2003 ) reported
31 that the leading cause of on the job mortality for wildlife biologists was death due to trauma sustained during crashes of manned aircraft performing slow airspeed, low altitude data collection surveys. While the Sasse (2003 ) study was focused specifically on wildlife biologists, and an update to the study is recommended it is hypothesized that a revised study with the latest data and extension of the subject matter to include additional natural resources professionals would reveal that slow and low manned aerial data collection flights would still be near the top of the list of leading causes of on the job mortality for natural resources professionals Space borne Satellites G athering natural resources data from an aerial perspective adds an additional layer of information that can be difficult to procure physically, financially, and/or repeatedly; however, these data can reveal details that would be e xtremely challenging or impossible to observe or document from BOTG methods alone. The addition of satellite acquired data to natural resource based research can be a powerful tool in generating plausible explanations for events or may spur additional sc ientific hypotheses for researchers to investigate Over the past 40 + years space borne satellite generated data suitable for natural resource based studies ha ve become an increasingly valuable re source ( Ozesmi and Bauer 2002 LaRue et al. 2015 Zweig and Newman 2015 ) With significant advancements in com puter ha rdware and software, along with parallel developments in satellite sensors, both obtaining and utilizing satellite data has become a much more conventional practice Yet even with these improvements factors such as finite data resolution, timing a nd location of satellite data collection, and atmospheric occlusion between the satellite sensors and the focal target of interest on or around the Earth, can hinder the reliance
32 of space borne satellites to provide continuous data sets that may be necessa ry for specific analyses ( Loarie et al. 2007 Gann et al. 2011 ) In such cases, manned aircraft surveys may be a more reliable choice because they can usually be conducted under overcast skies, and commonly provide data with higher resolution than those produced from affordable satellite data sources M any examples exist of space borne satellites that have exc orbit the Earth and provide very reliable data; even if the data resolution produced are coarse r than those available remain valuable, especially for natural resource based studies, because obtaining their data is largely cheaper than attaining similar data ( Gann and Richards 2009 Zweig and Newman 2015 ) tend to have lengthy historical records of archived data that can be used for a plethora of analyses. Unmanned Aircraft Systems Loca t ed between BOTG and manned aircraft flights in the hierarchy of data collection options is a technology driven methodology and emerging tool: sUAS. T he term unmanned aircraft system (UAS) is used to describe an unoccupied aircraft that is flown either r emotely by a human pilot or autonomously under the purview of a human pilot who can instantly attain manual control of the aircraft Perhaps the most currently this ter m has not been generally adopted in the unmanned systems vernacular ( Fahlstrom and Gleason 2012 Colomin a and Molina 2014 ) Use of the RPAS term has UAS are sometimes misconstrued as hobby aircraft toys or autonomous flying robots that have no method of intervention by a human operator Both of these UAS characterizations are
33 inaccurate for the systems that natural resource based researchers would use for data collection purposes Conversely, the term that has earned worldwide recognition primarily due to its lengthy history of usage an d widespread employment by the media and others is When the word drone is used, many people tend to conjure a mental picture of a large, military purposed aircraft designed to spy on someone or something to acquire intelligence, surveillance, and reconnaissance (ISR) data, or to provide strike capabilities to the warfighter in theater, which often invokes a negative connotation Because this dissertation is focused on civilian applications of UAS technology as a tool for data collection in natura l resource sciences, not UAS designed for covert spying or ordinance delivery, the use of the term is relegated to UAS that are designed specifically for military applications as their primary role Some of the mo re widely recognized drone s includ e the AeroVironment Raven the Boeing Insitu ScanEagle the AAI Corporation Shadow the Aeronautics Defense Systems Aerostar the General Atomics Predator and the Northrup Grumman Global Hawk ( Figure 1 1 ). What are Un manned Aircraft Systems? A UAS is truly a system, in that it consists of four primary components all working together to achieve takeoff, flight, and landing of an aircraft without a manned pilot onboard. The four key elements of a nonspecific UAS are: 1 ) the unmanned aircraft (UA) and its payload ; 2) the ground control station (GCS); 3) the communication linkages between the UA and the GCS; and 4) a fundamental item that does not always appear in the definition s of a UAS, but should likely be included: t he trained qualified, a nd experienced flight crew executing the mission Each of these four essential
34 components a re generally supplemented with various subsystems, equipment, and technical knowledge which assist in transforming what might be otherwise b e considered a basic remote control (RC) model or hobby aircraft into an unmanned airborne data collection platform. The Unmanned Aircraft The UA is strictly the flying component of the system; consisting of the airframe, its appendages, and the contents t hat leave the ground, fly in the air, and then return to the ground without a manned pilot onboard. Subsystems of an UA normally include elements such as a power source, a propulsion system, avionics, a comm unication linkage device, and an antenna. Other elements onboard might include flight specific payloads, an onboard data storage mechanism an autopilot component, a flight stabilization unit, a global positioning system (GPS), an inertial navigation system (INS), and additional antennae. The UA typic ally has either a fixed wing (airplane) or rotary wing (helicopter or multi rotor) configuration although other airframe options are available, e.g., ducted fan, flapping wing, and lighter than air UA ( Fahlstrom and Gleason 2012 ) Wingspans of UA can range from (c m) to meters (m) in length, and takeoff masses can range from 1 0 0 milligrams ( m g) to 1,400 kilograms (kg) ( Fahlstrom and Gleason 2012 ) Mass of the UA at takeoff is predominantly the determining factor used for the classification (and concomitantly the operatio nal regulations) of UAS worldwide. The Ground Control Station The second component of a UAS is the GCS; generally a computer based terminal with software that provides some level of telemetry and/ or UA status information to ground based personnel before, d uring, and after a UA flight. Depending
35 on the sophistication of the GCS, examples of additional features they may offer are a method of establishing and uploading of an autonomous flight plan for the UA to execute, ability to set failsafe protocols and w aypoints should communications between the UA and GCS unexpectedly become severed while the aircraft is in flight a method for allowing instantaneous manual command to control the UA, and perhaps present some level of visualization or control of the data being collected by the UA payload during a flight. A GCS can range in size from a smartphone up to a multi screened, multi terminal climate controlled semi trailer, which can accommodate a dozen personnel or more depending on the UA and its mission How ever, most GCS units are simply a laptop computer. The Communication Linkages The third component of a UAS is the comm unication linkages between the UA and the GCS. The comm unication linkages minimally consist of command and control which are bidirectiona l between the UA and the GCS, but may also include data feeds which can be either unidirectional or bidirectional from the UA to the GCS. Communication linkages between the UA and the GCS can be of many different types and specifications, e.g., analog or digital, un encrypted or encrypted, single frequency or frequency hopping spread spectrum, depending on the operational environment and type s of equipment used both on the airframe and on the ground. With increasing numbers of UAS users looking to occupy a finite area for operations of various applications limitations due to available communication frequencies are already an issue that will most likely get worse in the short term until a long term solution is devised.
36 The Flight Team The fourth component o f a UAS is the flight crew conducting the mission. A well trained, aviation knowledgeable, safety oriented, and collectively focused flight crew working as a team to conduct missions is an essential part of a UAS. Each individual is assigned a specific r ole and designated duties during all phases of a mission which helps distribute the individual workload for the entire team, and helps to ensure that no tasks are inadvertently overlooked before, during, or after a flight. By having flight crewmembers cr oss trained in the various roles on the team, tasks can be crosschecked and confirmed as being completed which incorporates an additional level of safety into the operation of the UAS. Further details about an endorsed flight crew member model are provided in Appendix C of this dissertation When all four elements are present and function ing together, a UAS can be a very safe and efficient aerial data collection platform Advantages in human safety, imagery resolution, frequency of flights, speed of mobilization, and the potential for long term cost savings are just a few of the reasons why UAS are garnering so much attention from natural resources users, and other civil consumers alike. UAS in Natural Resource B ased Sciences A numbe r of commercial companies produce large drones designed for high altitude, long endurance (HALE) military flights whose platforms and sensors can be adapted for scientific use; although the purchase price and the hourly operation and maintenance costs of t hese systems is for all intents and purposes beyond the scope of most natural resource budgets. The United States (US) National Aeronautics and Space Administration (NASA) Airborne Science Program has the means to own, fly and maintain an Ikhana UAS, which is a General Atomics Predator B drone system
37 that has been modified for Earth sciences research. Missions with the Ikhana UAS have included mutual aid for wildfire monitoring ( Ambrosia et al. 2011 ) sensor development (NASA 2015), and marine conservation initiatives ( Brooke et al. 2015 ) The US National Oceanic and Atmospheric Administration (NOAA) has also use d HALE, and medium altitude, long endurance (MALE), and sUAS (those whose mass at takeoff is 24.9 kg ) for natural resource based data collection. Several examples of NOAA UAS missions include ma pping forest fires, tracking sea pack ice ( Moreland et al. 2015 ) and identifying individual cetaceans based on distinctive external markings ( Mocklin et al. 2012 ) Based on the budgets and desired applications of most natural resource scientists routine scientific data collection airframes and sensors ha ve emerge d in the sUAS market Over the past nine years the interest in civil and commercial based sUAS uses has grown exponentially. O nly a short time ago a very limited number of commercial companies produced small and affordable airframes and/ or sensors tha t had the necessary flight features, imaging resolution, or georeferencing capabilit ies needed to generate scientific data to assist in answering questions based in natural resource disciplines Beginning as a hypothetical framework concept in 1999, then m ore formally structured by 2002, the University of Florida (UF) Unmanned Aircraft Systems Research Program (UFUASRP) was among the first academic based sUAS research programs in the US to custom design a sUAS explicitly purposed as a data collection tool t o assist natural resource professionals As a hallmark of the UFUASRP efforts to acquir e low altitude aerial data of focal targets in scientifically applicabl e manner s and keeping total
38 costs within natural resource scale budget s are just two of the ma jor themes which have driven the UF R esearch P rogram A fter nearly two decades of research and development funded in parts primarily by the United States Geological Survey (USGS), United States Fish and Wildlife Service (USFWS), Flori da Fish and Wildlife Conservation Commission (FFWCC), Idaho Department of Fish and Game (IDFG), United States Army Corps of Engineers (USACE), South Florida Water Management District (SFWMD), National Science Foundation (NSF), UF, and the UF Institute of F ood and Agricultural Sciences (IFAS), the UFUASRP is currently fielding sUAS operations with its fifth generation fixed wing platform, the UF Nova 2.1 sUAS, and its first generation rotary wing airframe, a modified DJI Spreading Wings S1000+ A brief hi story of the UFUASRP can be found in Appendix A of this dissertation. Through y ea rs of research, development, and experience, UFUASRP personnel have been conducting studies concerning all aspects of s UAS, e.g., airframes, avionics, sensors, payloads, flight planning, post processing, logistics, etc., and providing results, information, and guidance to others about the subject. A fundamental goal of the UF UASRP has been to provide examples to others considering fielding s UAS as a too l for natural resources research and data collection so that current and future s UAS users mistakes similar to those already realized by the UF UASRP Foundations of the Researc h Topic Emerging out of the high profile success and remarkable videos generated by various military drones and their payloads while in theater during the 1990s through the mid 2000s, the comparatively well fiscally funded US Department of Defense (USDOD)
39 and its contractors spurred various groups and individuals within both the public and private sectors to investigate ways to use sUAS technology to augment data collection options for a seemingly limitless number of non military applications including th ose in natural resources ( Boucher 2015 ) Integrating new technology into natural resource based research has potentially been protracted over time due in part to the characteristically high costs of obtaining and fielding newer technology; but in response to a recent push within the scientific community as a whole for increased multidisciplinary and integrated research efforts, scientists in natural resource fi elds of expertise are collaborating with researchers in other scientific disciplines, and gaining the benefits of utilizing new technology as tools and methodologies to improve their own science and research efforts ( Benson et al. 2010 Blickley et al. 2012 Galan Diaz et al. 2015 ) Developing ways to int egrate novel technology into natural resource based science can be complicated by multiple factors e.g. : 1. Working within the confines of particularly limited fiscal resources 2. Meeting the desires to obtain data/analyses/answers in very short periods of time 3. Focusing on specific applications in which novel technology can aptly assist researchers 4. Recognizing the diverse user skill levels and scopes of knowledge within the natural resource disciplines 5. Navigating the bureaucracy which generally accompanies new e ndeavors The goals of creating tools and techniques to gather data for natural resource based research more efficiently, accurately, and safely, by supplementing existing data collection methods with novel advancements in technology (such as sUAS and thei r payloads) are the foundations on which the UFUASRP was established. Fortuitously, the last decade has been a period in history where UAS for natural resource
40 applications ha ve progressed from a topic that would quite literally empty convention halls at professional meetings, to a subject that now often garners full day special sessions, symposia, and crowded presentation rooms at conferences. This dissertation consists of two chapters that document case studies of sUAS fielded as aerial data collection p latforms in natural resources, and a third chapter highlights an alternative method of using sUAS payloads to help address specific circumstances where sUAS may not be a suitable or legal data collection option The chapters show both benefits and limits of the technology that have been determined, and provide insight into both the short and long term prospects for sUAS and their payloads as data collection tools for the natural resource based scientific community. Because technology is always advancing, some of the material contained within these chapters is already outdated but it still provide s a historical account of where sUAS technology has been during ion The dissertation also contains a series of appendices that document s informatio n that could provide useful reference material to future sUAS endeavors. A detailed history of the UFUASRP in 1999 ( Appendix A ) a very brief history of UAS regulat ions in the US ( Appe ndix B ) and a flight crew model adopted by the UFUASRP that received abundant praise from the US Federal Aviation Administration (FAA) ( Appendix C ) are provided. Also included as appendices are an overview of the imagery data p ost processing methodologies used throughout this dissertation ( Appendix D ) a short commentary on some of the critical lessons learned by the UFUASRP ( Appendix E ) and a blank copy of the flight data log completed for each UFUASRP sUAS flight since 2009 ( Appendix F )
41 A B C D E F Figure 1 1. Several of the most widely recognized drones used for unmanned military applications. A) The Aer oVironment Raven B) T he Boeing Insitu ScanEagle C) T he AAI Corporation Shadow D) T he Aeronautics Defense Systems Aerostar E) T he General Atomics Predator F) T he Northrup Grumman Global Hawk www.northropgrumman.co m/Media/Resources/Photos/ www.ga asi.com/predator b M.A. Burgess M.A. Burgess M.A. Burgess M.A. Burgess
42 CHAPTER 2 INFERENCE S OF HABITAT SELECTION BY PYGMY RABBIT S ( Brachylagus idahoensis ) BASED ON NORMALIZED DIFFERENCE VEGETATIVE INDICES GENERAT ED FROM IMAGERY COLLECTED VIA SMALL UNMANNED AIRCRAFT SYSTEM With anthropomorphic changes alter ing ecosystems worldwide, there is an increasingly urgent nee d for researchers to identify characteristics of landscapes that are critical components in habitat selection preferences of wildlife ( Edgel et al. 2014 ) F ragm entation, degradation, and loss of critical habitat areas for wildlife results in decreased population s izes ( Francis et al. 2011 Hartter et al. 2015 McCleery et al. 2015 Zweig and Newman 2015 ) It is hoped that through prioritized research efforts predictive models of consequential changes to land scapes can be generated which will help decision makers m itigat e some of the negative effects of landscape change ( Rachlow and Svancara 2006 Anderson et al. 2010 Parsons et al. 2016 ) S agebrush obligate species face increased survival pressures prim arily in response to declines in available sagebrush steppe landscape due to land conversion for urbanization and agriculture, introduction of invasive species, energy development, unsustainable overgrazing by livestock, and increased fire frequency (Edgel 2013). The abundance and distribution of the pygmy rabbit ( Brachylagus idahoensis ) a sagebrush specialist, has been documented to be declining, but the rates of which are difficult to establish (e.g., Keinath and McGee 2004 Himes and Drohan 2007 Edgel 2013 Parsons et al. 2016 ) The pygmy rabbit is a small, cryptic semi fossorial, generally solitary leporid that inhabits the semi arid intermountain sagebrush steppe landscape year round ( e.g., Green and Flinders 1980a Rachlow et al. 2005 Himes and Drohan 2007 Lee 2008 Ulmschneider et al. 2008 Edgel 2013 ) ( Figure 2 1 ) With this life history the pygmy
43 rabbit is reliant on dense patches of sagebrush ( Artemisia sp.) for both protective cover and as a source of forage ( Green and Flinders 1980b ) The p ygmy rabbit is also unique in that it is one of only two North American lagomorphs known to excavate their own burrow systems that they use as shelter for avoiding predation and thermal extremes which occur throughout their range (e.g., Katzner and Parker 1997 Siegel Thines et al. 2004 Larrucea and Brussard 2009 Camp et al. 2013 Camp et al. 2015 Crowell et al. 2016 ) Interestingly, pygmy r abbit b urrow system entrances are nearly always found at the base of sagebrush plant s which makes the approximately ( ) 10 12 centimeter ( cm ) diameter burrow entrances difficult to locate for researchers and predators alike ( Green and Flinders 1979 ; 1980a Rachlow et al. 2005 Edgel et al. 2014 ) Sagebrush contains unusually high levels of toxic secondary compounds that ar e generally deterrents to herbivory; however, the pygmy rabbit has evolved m echanisms to consume and digest sagebrush (e.g., White et al. 1 982 Shipley et al. 2006 Shipley et al. 2009 Nobler et al. 2013 Ulappa et al. 2014 Fremgen 2015 ) During warm summer months the pygmy rabbit diet consists primarily of sagebrush ( 51% ) while grasses and forbs constitute a majorit y of the remainder ; however, in winter months when grasses and forbs are absent 99% of the pygmy rabbit diet consists of sagebrush (e.g., Green and Flinders 1980 b White et al. 1982 Shipley et al. 2006 Utz 2012 Olsoy 2013 ) A ctually spotting a pygmy rabbit among the sagebrush can be challenging therefore presence or absence is chiefly determined by looking for signs of recent pygmy rabbit activity, i.e., fresh 45 bite marks on sagebrush plants, ne w fecal pellet deposit s near burrow entrance s or forag ing sites active burrow system entryways, or leaping/landing pad snow compaction trails from burrow s during winter ( Green and
44 Flinders 1979 Rachlow and Witham 2006 Ulmschneider et al. 2008 Edge l et al. 2014 ) ( Figure 2 2 ). Based on occasional sightings of individuals, and mapping locations of recent signs of pygmy rabbit presence, their native distribution has been identified as patchy ( Severaid 1950 Green and Flinders 1980a Rachlow and Witham 2006 Sanchez and Rachlow 2008 ) ( Figure 2 3 ). The native pygmy rabbit distribution historically included parts of central Washington as well; however, a combination of fire s and large scale loss es and fragmentation of the sagebrush steppe landscape there extirpated the native population from that region ( Becker et al. 2011 ) The small size of adult pyg my rabbits [ mass : 375 500 grams ( g ) ; length : 23.5 29.5 cm ] (Becker et al. 2011) influences their thermal regulation, concealment from predators, and changes their locomotion strategies when compared to other lagomorphs Three other rabbit species have ran ges which overlap in part with that of the pygmy rabbit: the mountain cottontail rabbit ( Sylvilagus nuttallii ), white tail ed jackrabbit ( Lepus townsendii ), and black tail ed jackrabbit ( L californicus ) all hav e larger body sizes appendages, and home rang es than pygmy rabbits ( Shufeldt 1888 Katzner and Parker 1997 Ulmschneider et al. 2008 Nobler 2016 ) The se other rabbit species often utilize their sheer speed to escape predators; however, due to their dis proportionally short hind legs py gmy rabbits must use agility and precision cornering ability among vegetation en route back to a burrow as a means of eluding capture ( Green and Flinders 1980a ) It is hypothesized that these attributes may limit pygmy rabbits from venturing t oo far from a burrow to browse on forage ( Ulmschneider et al. 2008 Crowell et al. 2016 Utz et al. 2016 ) Additionally, cover provided by comparatively tall, dense sagebrush year round, and grasses and forbs during the
45 warmer months help pygmy rabbits minimize detection from both aerial and terrest rial predators ( Camp et al. 2012 Utz 2012 Olsoy et al. 2015 ) Pygmy rabbits inhabit microhabitats that have deep loose loamy soils, often with a clay component, capable of supporting their extensive burrow systems, yet soft enough for tunneling ( Gabler et al. 2000 Ulmschneider et al. 2008 ) W ith in Idaho, and perhaps in other parts of their range, pygmy rabbit burrow systems are largely found associated wit h mima mounds ; circular or oblong microtopograph ic feature s composed of loosely compacted mounded soil having an elevation 0.5 1.0 m eter (m) above the surrounding terrain and diameters 5 2 0 m The origins of mima mounds remain debated but they harbor sagebrush patches which are generally taller denser and have higher above ground biomass than mou interstitial areas where soils typically are more compacted and sagebrush plants are shorter thinner and have lower biomass (e.g., Weiss 1984 Roberts 2001 Ulmschneider et al. 2008 Estes Zumpf and Rachlow 2009 Cramer and Barger 2014 ) ( Figure 2 4 ) Through years of data collection via boots on the ground (BOTG) methodologies combin ed with laboratory based analyses of forage and non forage vegetation, feca l pellet analyse s, and other field collected samples researchers with the cooperative pygmy rabbit research group from the University of Idaho (UI), Boise State University (BSU), and Washington State University (WSU), ha ve identified many of the habitat f eatures that pygmy rabbits seemingly require, prefer, tolerate, or avoid, within their home range (J. L. Rachlow, UI personal communication; J S. Forbey, BSU, personal communication; L. A. Shipley, WSU, personal communication). Their global hypothesis i s that pygmy rabbit habitat selection is determined by areas having the most valuable
46 attributes available within the sagebrush steppe ecosystem, and therefore the locations of active burrows and signs of recent pygmy rabbit presence should theoretically b e prominent indicator s of regions containing preferred habitat features. T o test this hypothesis the cooperative pygmy rabbit research group looked to create a series of spatial response model s using geographic information system (GIS) techniques based on smaller scale, intensive BOTG efforts at multiple study locations primarily with in Idaho The GIS response models would consist of multiple layers of spatially mapped p redictor variable data that would includ e but not limited to the following : 1. Location density, size, spacing, above ground biomass and species of available vegetation 2. Nutritional quality, quantity, and energetics of forage and non forage items 3. Identification and concentrations of toxic secondary compounds found in browsed and unbrowsed v egetation 4. Availability and location of cover as refugia from predators and thermal extremes 5. Soil features mima mound sizes and locations 6. Locations of active, inactive, and abandoned pygmy rabbit burrow entrance s 7. Distances from burrows to roads, fences, o pen spaces, water, edge habitat, agriculture, and free ranging livestock operations 8. Identification and proximity of known rabbit predator sightings, tracks, scat, or other signs of predator presence Using GIS, the predictor variable data layers could then be stacked in various combinations result ing in spatial response model s revealing the geographic locations possessing the mo re preferred combination s of habitat features for pygmy rabbits in addition to locations that were noticeably avoided ( Figure 2 5 ) Should the spatial response models work for the smaller scale intensive BOTG pygmy rabbit study sites, the subsequent objective was to scale up to larger regions with the ultimate goal of achieving models at landscape leve l s
47 T o scale up, ex trapolat ion of remotely sensed data collected by space borne satellites, manned aircraft surveys, or perhaps emerging small unmanned aircraft systems (sUAS) was needed ( Figure 2 6 ) The resolution of affordable satellite data was too co ar se, and manned aerial surveys were not budgeted, t hus the cooperative pygmy rabbit research group approached the University of Florida Unmanned Aircraft Systems Research Program (UFUASRP) for assistance. The UFUASRP proposed to conduct low altitude, sUAS fixed wing flights over sagebrush steppe habitats in Idaho using both visible (RGB) and near infrared (NIR) wavelength optical sensors to address the following objectives : 1. Compare normalized difference vegetative index (NDVI) va lues between sagebrush plants located on and off mima mounds within each of two study sites 2. C ompare NDVI values between sagebrush plants that exhibited signs of recent pygmy rabbit activity to sagebrush plants not exhibiting signs of recent pygmy rabbit a ctivity within each of the two study sites 3. Compare the mean NDVI values of sagebrush plants that were growing on or off mima mounds, and showing or not showing signs of recent pygmy rabbit presence to sagebrush plants having the same disposition between th e two study sites At both study sites, we predicted that mima mounds, which harbor denser patches of sagebrush, would have higher mean NDVI values than those for sagebrush plants located off mima mounds in interstitial areas We also predicted that at bo th study sites, sagebrush plants associated with recent signs of pygmy rabbit activity would have higher mean NDVI values than those where signs of recent pygmy rabbit activity were absent Additionally, w e predicted that t he mean NDVI values for sagebrus h plants possessing the same disposition at both sites would be statistically different from each other because from ground level the vegetation distribution at the two sites present s a visual dissimilarity
48 The Sagebrush Steppe Study Areas Magic Reservoir The Magic Reservoir field site was located on the Camas Prairie 3.3 kilometers ( km ) southeast of Magic Reservoir and due west of Wedge Butte p eak on the west ern side of paved State Highway 75 in Blaine County, Idaho (WGS84 Datum: 43.241965 N orth ; 114.318094 W est ) ( Figure 2 7 ). The focal area of study covered 30 hectares (ha ) of sagebrush steppe habitat dominated by Wyoming big sagebrush ( Artemisia tridentata wyomingensis ), with a few small patches of low sagebrush ( A arbuscula ) and three tip sagebrush ( A. tripartita ) and several species of warm season grasses and forbs Site t opography was relatively flat with a slight downhill grade from the northeast [ elevation: 1,478 m above sea level (ASL) ] to the southwest (el evation: 1,470 m ASL) extents of the study site and the surface was defined by mima mounds harboring denser clusters of sagebrush with less dense sagebrush located off the mounds among areas of grasses, forbs, and some areas of bare ground. A n unnamed two rut dirt road ran north and south through the study site, and another two rut road ran west erly from State Highway 75 just north of the southern edge of the study site across the northern extent of a 1.2 ha dry borrow pit whi ch created the southeastern quadrat of the dirt road intersection A wire fence line with wooden fenceposts defined the northern limits of the study site A similar fence line delimited the eastern edge of the study site. National Oceanic and Atmospheric Administration (NOAA) climatology data for the locality classifie d it as semi arid with an average annual precipitation total of 20.2 cm per year ( NOAA 2016 ) Historical data showe d that for the month s of June, average daily air temperatures range from 5.5 24.5 Celsius (C), and rainfall is infrequent, w hile
4 9 in the month s of December, average daily air temperatures range from 13.8 0. 3 C, with infrequent snow showers ( NOAA 20 16 ) Rocky Canyon The Rocky Canyon field site was located in the Lemhi Valley 10.8 km east southeast of Leadore, on an alluvial fan of gravelly silt loam at the foothills of the Beaverhead Mountains in Lemhi County, Idaho (WGS84 Datum: 44.664152 North ; 113.225143 West) ( Figure 2 8 ). The focal area within the site encompassed 135 ha of sagebrush steppe habitat, dominated by Wyoming big sagebrush ( Artemisia tridentata wyomingensis ), with patches of three tip sagebrush ( A. t ripartita ) and several species of rabbitbrush ( Chrysothamnus sp.), and warm season grasses and forbs. Site topography was on a grade downhill from the northeast (elevation: 2,0 40 m ASL) corner of the study site to the southwest (elevation: 1,952 m ASL) co rner and the surface was defined by mima mounds harboring dense clusters of sagebrush with less dense but nearly contin uous sagebrush rabbitbrush, grasses, and forbs located between the mounds An unnamed two rut dirt road ran generally north and south just west of the eastern edge of the study site from Hawley Creek at southern end up the foothill northern terminus in Rocky Canyon proper. Hawley Creek Road (also a two rut dirt road) ran westerly along the mountain foothills just sou th of the northern extent of the study area from Rocky Canyon proper. Located just north of the south ern limits of the study site was Hawley Creek proper and access to the entire site required driving through Hawley Creek via a two rut dirt road which or iginate d perpendicularly o ff Forest Service Road 275, 2.0 km southwest of the Hawley Creek Lower Campground, US Forest Service (USFS) recreational area. Forest Service Road 275 was a graded
50 gravel and dirt road that was wide enough for two vehicles to pas s, and possessed ditches on each shoulder to direct water runoff during occasional rain showers NOAA climatology data for the Rocky Canyon site classifie d it as semi arid with an average annual precipitation total of 20 5 cm per year ( NOAA 2016 ) Historical data showe d that for the month s of June, averag e daily air temperatures range from 3. 2 22.3 C, and rainfall is infrequent while in the month s of December, average daily air temperatures range from 14.6 1.2 C, with infrequent snow showers ( NOAA 2016 ) Methods U sed for Assessing Pygmy Rabbit Habitats with sUAS derived NDVI Imagery Products The U niv ersity of F lorida (UF) Nova 2.1 sUAS ( detail ed in Appendix A ) was designed specifically as a platform for housing optical payloads capable of obtaining high resolution digital imagery and associated meta data to assist in answering n atural resource based scientific questions Emulating many principles of a glider to maximize flight time aloft, the 2.74 m wingspan fixed wing airframe was fabricated to offer hand launch and belly landing capabilities with a 6.0 kilogram (kg) maximum to tal mass at takeoff. The low altitude Idaho missions were conducted under Federal Aviation Administration (FAA) Certificate of Waiver or Authorization (COA) permits 2013 WSA 85 (Magic Reservoir study site) and 2013 WSA 89 (Rocky Canyon study site ) respe ctively At the time of the study, obtaining a COA was not trivial, requiring a huge investment of time over a period of three to six months if not longer. Flights conducted with the UF Nova 2.1 sUAS for the pygmy rabbit missions followed a standard opera ting procedure (SOP) developed at UF that met or exceeded FAA requirements for sUAS flight s The entire process from unpacking the UA transport case upon arriv al at the field site, to packing the UA away at the end of the
51 operational day, was co nducted by a qualified three person flight crew: 1) a pilot in command ( PIC ) ; 2) a ground station operator (G S O); and 3) a qualified visual observer (QVO). Each of the flight crew members had specific duties during all phases of the operation, and further detail about the UFUASRP flight crew model can be found in Appendix C O nce the UF Nova 2.1 was airborne it flew preprogrammed parallel transects under autopilot control over the target study area Transects were arranged to captu re individual images with sufficient overlap ( endlap and 50% sidelap) for complete ground coverage. At any time during a flight, manual remote control (RC) of the unmanned aircraft ( UA ) could be achieved by the PIC with the flip of a switch. After a target area had been thoroughly imaged by the sUAS payload or if the system battery voltage approached a threshold value the aircraft would be commanded by the GSO to proceed to a predetermined rally waypoint located downwind, then begin a controlled descent spiral, and autonomously land itself into the prevailing wind at a prearranged waypoint on the ground To achieve the intended research goals of obtaining both RGB and NIR digital imagery and its associated metadata to conduct NDVI analyses, a collaborative decision was made to gather NIR dat a only in the summer months when vegetation on the sagebrush steppe landscape was most active and not covered with snow. T wo separate summer field missions to Idaho were made with the UF Nova 2.1 sUAS the Olympus E 420 optical payloads, and a three per son flight crew. The first field mission for collecting imagery of pygmy rabbit habitat took place 24 2 6 June 2013 ( s ummer 2013) at the Magic Reservoir study site. To assist in
52 tightening the accuracy of the resultant post processed imagery end products, n = 84 ground control points (GCP) were spread throughout the entire site ( Figure 2 9 ) and were marked with 30 individually geolocated (latitude, longitude, and altitude) using a survey grade Top c on Hi P er V dual frequency global navigation satellite system (GNSS) receiver unit with base station. The entire study site w as subdivided into four smaller subareas east to west for imagery collection to maximize ground coverage of aerial transects that were oriented north and south with ambient winds having a southerly component on each of the flight days All of the four smaller subareas were flown once with the RGB spectrum payload, and once with the NIR optical sensor. The preprogrammed flight paths over each subarea were executed twice during each of the eight individual flights; an additional effort to maximize imagery of the target area for mosaicking The UFUASRP learned a tremendous amount about the UF Nova 2.1 sUAS air craft performance during the summer 2013 flights at Magic Reservoir as they were the first flown with that system from ground elevations considerably higher than sea level. Combining the altitude ASL (and the actual station pressure experienced at those altitudes) with the low dew points and uncharacteristically warm air temperatures experienced at the field site in June 2013 led to calculated density altitude values that effectively thinned the air to levels encountered at substantially higher altitudes. The effects of the density altitude were noted in aircraft performance ; especially evident in the decreased lift provided by the aircraft wings, and with decreased thrust due to the thinner air.
53 The second field mission for these research objectives took place 15 19 June 2014 ( s ummer 2014) at the Rocky Canyon study site. As was done in summer 2013 at the Magic Reservoir site, n = 111 GCP were spread throughout the entire Rocky Canyon site ( Figure 2 1 0 ) marked with coverboards, and geolocated in three dimensions ( 3D ) using a HiPer V GNSS receiver unit with base station. T he Rocky Canyon site w as subdivided into three subareas south to north to conduct east west flight transects in account of the easterly ambient winds encounte red Based on flights conducted in Idaho the previous summer, the UFUASRP outfitted the UF Nova 2.1 sUAS with a larger diameter propeller that also had an increased pitch in an attempt to more effectively operate in the higher altitudes ASL. The first tw o flights conducted at the Rocky Canyon site were executed with the RGB payload, and covered the two southernmost of the three sub areas After several days of poor weather, including rain, low clouds poor visibility, and even snow showers, the next two f lights were conducted with the NIR sensor payload over the same two southernmost Rocky Canyon sub areas previously flown with the RGB payload Each of the four flights that had been conducted up until that point during the s ummer 2014 mission had all experi enced particularly rough landings on the only suitable landing strip available at the Rocky Canyon site; the graded gravel and dirt Forest Service Road 275 south of the study are a Having l and ed on the same road during January 2014 flights with the same e quipment and personnel for other mission objective s (not featured in this dissertation ) landings at the Rocky Canyon site presented little problems for the UF Nova 2.1 sUAS because there was a n appreciable layer of snow and/or ice on the road surface that allowed the aircraft to skid to a stop
54 relatively smoothly. However, during the s ummer 2014 flights every landing on the Forest Service Road did various levels of damage to the airframes, avionics, or payloads. D uring summer 2014 slightly elevated expo sed faces of larger gravel and rocks imbedded in Forest Service Road 275 had a tendency to snag part s of the fuselage or main wing tips during belly landings causing the UA to either stop abruptly initiating a cartwheel or induce a pivot about the obstruc tion redirecting the aircraft into the roadway ditch containing larger exposed rocks or further off course into the adjacent sagebrush patches beyond the road shoulder All m inor repairs were conducted in the field but repairs that were more significant had to be conducted back at the local field station By the end of the fourth Rocky Canyon summer 2014 flight two of the three UA fuselages brought to Idaho were damaged beyond repair With future flights still planned, the UF UASRP flight team met with the cooperative pygmy rabbit research group to discuss their priorities for the ensuing flights. It was determined that completing the RGB spectrum data set for the remaining northernmost sub area at Rocky Canyon was a priority If the last RGB flight co uld be completed and landed without damage then a flight with the NIR payload would be executed to finish complete imagery coverage of the study site The last RGB flight was flown over the remaining northernmost subarea complet ing full coverage of the R ocky Canyon site with visible wavelength imagery U nfortunately the landing of that flight on Forest Service Road 275 damaged the fuselage, main wingset, and tail components beyond repair Data collection was
55 effectively terminated w ith no additional ai rframe s available for conducting further flights Data post processing for this study was conducted using the methodology delineated in Appendix D E ach of the eight flights at the Magic Reservoir study site were post processed ind ividually as were the five flights at the Rocky Canyon study site The final PhotoScan post processing step was to merge the individual RGB flights and NIR flights at each site into large digital elevation models ( DEMs ) and orthophotomosaic scene s for f urther analyses The NIR orthophotomosaic image was loaded into QGIS a free and open source GIS software program The image was saved as a raster file, and then a custom Python script was used to break down the NIR orthophotomosaic e, green, and red component bands ; which in reality we a small amount of green wavelength light + some NIR a small amount of red wavelength light + some NIR and a limited NIR based on the camera conversion process and user defined cut f ilter replacement specifications Using the raster calculator in QGIS the formula for computing NDVI : [ ( NIR red)/(NIR + red) ] was applied to the appropriate component bands and a resultant black and white NDVI raster layer was generated with a scale from black white (values: 1.0 + 1 0 ) 0. 2 ) are typically water, impenetrable surfaces, e tc., while l ow values ( 0.1 + 0.1) are generally associated with barren earth or rock. Moderate NDVI values (0.2 0.4) correspond with shrubs and grasses while h igh values (> 0.5 ) occur in dense areas of very green and healthy vegetation Healthy green p lants do not reflect much RGB light as they use these wavelengths for photosynthesis, but
56 less healthy plants tend to reflect much higher levels of RGB light. The b lack and white NDVI raster layers were re colored with a scale from red green, which enhance d visualization of variations that the black and white layer s d id not reveal as easily. Initial analyses were conducted on the Magic Reservoir site and the Rocky Canyon site individually, and then subsequently the information from both was use d. Using QGI S software, a standard basemap was projected, and the post processed orthophotomosaic products (RGB and NIR) were added as layers. An additional layer containing the 3D locations of the GCPs was added to ensure that all of the layer s lin ed up properly. Next, the NDVI raster layers were added to the GIS tree. A final layer consisting of 3D locations of specific sa gebrush plants that were identified ground truthed, and geo located b y the cooperative pygmy rabbit group was added (Figures 2 1 1 2 1 2 and 2 1 3 ) The NDVI values for an equal number of sagebrush plants on and off mima mounds were selected from the list of geolocated plants using a random number generator. The same system was used to randomly select sagebrush plants exhibiting signs of recent pygmy rabbit activity versus those that did not. The selected plants were quer i ed in the QGIS software for their NDVI values, the results were tabulated, and a series of stat istical tests were conducted to assess various combinations of factors Results of Pygmy Rabbit Habitat Selection Using sUAS NDVI Imagery Products The summer 2013 Magic Reservoir field mission resulted in eight flights (four with the RGB payload, four with the NIR payload). The three days of flying yielded a total flight time of 5 hours ( hr ) 42 minutes ( min ) and 8 seconds ( sec ) ; nearly 43 min per flight on average. During the eight flights, 7,678 total 10 .0 megapi xe l ( MP ) JPEG format images were capture d, occupying 24.4 gigabytes ( GB ) of digital drive space. The
57 steppe habitat, although the targeted area of study had an area of 29.3 ha. The difference between the two area calculations was attributed to ima ging an intentional buffer zone of significant extent around the target study area to maximize complete imagery coverage over the focal target area. Resultant ground resolution of the post processed imagery products had an average of 2.33 centimeters per pixel (cm/pix). The summer 2014 Rocky Canyon mission resulted in five flights (three with the RGB payload, and two with the NIR payload). The three days of flying yielded a total flight time of 3 hr, 7 min, and 28 sec; nearly 37 min per flight on average. During the five flights, 3,876 total 10 .0 MP JPEG format images were captured, occupying 16.0 GB steppe habitat, although the targeted area of study had an area of roughly 1 35 ha. The difference w as again attributed to imaging an intentional buffer zone of substantial area around the focal target study site to maximize imagery coverage over the target area of interest steppe habitat, although the area within the targeted area of study was roughly 95 ha. Resultant ground resolution of the post processed imagery products had an average of 2.88 cm/pix. The NDVI values computed for sagebrush plants within the greater sage brush steppe landscape from the sUAS imagery collected, independent of site resulted in values that ranged from 0.156 + 0.248 For the Magic Reservoir site, sagebrush plants located on mima mounds that exhibit ed signs of recent pygmy rabbit presence, and those lacking signs of recent pygmy rabbit presence had mean NDVI values of
58 0. 148 ( SE = 0 014 n = 35 ) and 0.0 74 (SE = 0.01 7 n = 35), respectively. Also at Magic Reservoir, sagebrush plants located off mima mounds that exhibited signs of recent pygmy rabbit presence, and those lacking signs of recent pygmy rabbit presence had mean NDVI values of 0.077 (SE = 0.01 6, n = 35) and 0.005 (SE = 0.019, n = 35) respectively. At the Rocky Canyon site, sagebrush plants located on mima mounds that showed signs of recent pygmy rabbit presence, and those lacking evidence of recent pygmy rabbit presence had mean NDVI values o f 0.141 (SE = 0.014, n = 35) and 0.078 (SE = 0.017, n = 35) respectively. Additionally at Rocky Canyon, sagebrush plants located off mima mounds that exhibit ed signs of recent pygmy rabbit presence, and those lacking signs of recent pygmy rabbit presenc e had mean NDVI values of 0.080 (SE = 0.017, n = 35) (SE = 0.018, n = 35) respectively. Within the Magic Reservoir site, tests of the mean NDVI values for sagebrush plants located on mima mounds exhibiting signs of recent pygmy rabbit presence were stati sti cally different compared to : 1) plants located on mounds showing no signs of recent pygmy rabbit presence ( t 66 = 3.29; P 0.001) ; 2) plants off mima mounds exhibiting signs of recent pygmy rabbit presence ( t 67 = 3.25; P 0.001) ; and 3) plants off mima mounds showing no signs of recent pygmy rabbit presence ( t 64 = 6.06; P 0.0 0 1) The mean NDVI values for sagebrush plants located off mima mounds exhibiting no signs of recent pygmy rabbit were statistically different compared to those of plants located on mima mounds showing no signs of recen t pygmy rabbit presence ( t 68 = 2.73; P = 0.0 04 ) and to those off mima mounds exhibiting signs of recent pygmy rabbit presence ( t 67 = 2.90; P = 0.0 02 ). However, the mean NDVI values for sagebrush
59 plants located on mima mounds that show ed no signs of recen t pygmy rabbit presence and those located off mima mounds exhibiting signs of recent pygmy rabbit presence were not statistically different ( t 68 P = 0.453). Within the Rocky Canyon site, tests of the mean NDVI values for sagebrush plants located on mima mounds showing signs of recent pygmy rabbit presence differed compared to : 1) plants located on mounds exhibiting no signs of recent pygm y rabbit presence ( t 66 = 2.77; P = 0.0 04 ) ; 2) plants off mima mounds showing signs of recent pygmy rabbit presence ( t 67 = 2.75 ; P = 0.00 4 ) ; and 3) plants off mima mounds showing no signs of recent pygmy rabbit presence ( t 6 5 = 6. 77 ; P 0.001). The mean ND VI values for sagebrush plants located off mima mounds exhibiting no signs of recent pygmy rabbit presence differed statistically compared to those of plants located on mima mounds showing no signs of recent pygmy rabbit presence ( t 68 = 3.69 ; P 0.0 0 1) an d to those off mima mounds exhibiting signs of recent pygmy rabbit presence ( t 67 = 3.84 ; P ). However, the mean NDVI values for sagebrush plants located on mima mounds that show ed no signs of recent pygmy rabbit presence and those located off mima mounds exhibiting signs of recent pygmy rabbit presence were not statistically different ( t 68 080 ; P = 0.4 68 ). Statistical c omparisons of mean NDVI values for sagebrush plants growing in similar microhabitats between the two study sites were all found not to differ statistically from each other Sagebrush plants at both sites growing o n mima mounds showing signs of recent pygmy rabbit presence had mean NDVI values that were not statistically different from each other ( t 6 8 = 0.344 ; P = 0. 366 ) while sa gebrush plants at both sites growing on mima mounds and not showing any signs of recent pygmy rabbit presence
60 also had mean NDVI values that were not statistically different from each other ( t 68 = 0.143; P = 0.443). Sagebrush plants at both sites not growing on mima mounds and exhibiting recent signs of pygmy rabbit presence had mean NDVI values that were not statistically different from each other ( t 68 P = 0.455), while sagebrush pla nts at both sites growing off mima mounds and not showing any signs of recent pygmy rabbit presences also had mean NDVI values that were not statistically different from each other ( t 68 = 0.750; P = 0.228). Discussion of Pygmy Rabbit Habitat Selection Base d on sUAS NDVI Imagery Products Determining the NDVI was just one option from a very lengthy list of vegetative indices that are available to isolat e specific spectra from composite imagery ( Zhang et al. 2015 ) Chlorophyll, t he green pigment in plants, strongly absorbs RGB light wavelengths as part of the photosynthetic process, and therefore the cells with in green leaves strongly reflect NIR wavelengths. By cr and capitalizing on the NIR reflective properties of vegetation, NDVI was designed to provide information specifically for green plants (e.g., Huang et al. 2002 Pettorelli et al. 2005 Agapiou et al. 2012 Dandois and Ellis 2013 Nijland et al. 2014 Boswell 2015 ) Within the published literature, manuscripts that utilize NDVI values for analyses are dominated by agricultural disciplines for vegetation such as planted row crops and turfgrass Comparing NDVI values reported with in th at literature to those of a sagebrush plant that has relatively small leaves and rather substantial woody s tructural component s is inequitable Although the computed NDVI values in this study were low and had a seemingly narrow spread of values, the use of NDVI was still appropriate for comparisons made among sagebrush steppe communities.
61 The mean NDVI values were the highest at both sites for sagebrush plants located on mima mounds that showed signs of recent pygmy rabbit presence as expected With the pygmy rabbit depending on sagebrush for food and cover year round, and knowing that mima mounds tend to harb or pygmy rabbit burrow systems, it was not surprising to observe burrows would be preferred by showing signs of recent pygmy rabbit presence Also not unexpected was that the lowest mean NDVI values at both study sites were associated with sagebrush plants off mima mounds that showed no signs of recent pygmy rabbit presence. Based on the behavior and ecology of pygmy rabbits, sagebrush plants in the interstitial areas between mima mounds are areas typic ally having less cover, and therefore lower mean NDVI values would indicate that those plants are less green; suggesting that the y do not provide as many green parts as forage or concealment cover to pygmy rabbits. This leads us to wonder if pygmy rabbit foraging and defecation influence sagebrush plants and cause them to potentially have higher NDVI values The determination that the mean NDVI values for sagebrush plants on mima mounds that did not exhibit sign s of recent pygmy rabbit presence were not statistically different from mean NDVI values of sagebrush plants off mima mounds that did show signs of recent pygmy rabbit presence was unpredicted Finding this to be the case at both study sites further reinf orced the result Perhaps this has to do with tradeoffs between forage quality and distance from burrows (e.g., Siegel Thines et a l. 2004 Shipley et al. 2009 Camp et al. 2013 Ulappa et al. 2014 Olsoy et al. 2015 Utz et al. 2016 ) In situations where pygmy rabbits may be forced to leave the security of a mima
62 mound due to age, sex, or territoriality issues, those ind ividual rabbit s likely select the greenest plants available in the interstitial areas to provide both cover and forage Conceivably the NDVI findings from this study may also be better explained and lead to additional justifications when the se data are added to other GIS layers and response model s are generated Determining that mean NDVI values between sites for sagebrush plants with similar location and signs of recent pygmy rabbit activity dispositions were all found to be not statistically different was also unanticipated It was thought that because the mima mound spacing and sagebrush densities appear visibly different between the two study sites at ground level the likelihood of the mean NDVI values for sagebrush having similar dispositions at bo th sites was expected to be low. A factor potentially contributing to these results was that the majority of the sagebrush plant s pecies at both study sites consist ed of Wyoming big sagebrush although their distribution was discrete. Sagebrush steppe s it es that are less homogeneous with Wyoming big sagebrush might reveal alternative result s Nevertheless, the se finding s certainly warrant further exploration through the testing of additional sites I f similar results are found elsewhere then the re sults of this study could theoretically be an important characterization in the conservation and management of critical pygmy rabbit habitat. When developing mechanistic models of herbivore habitats, it is important to consider that factors such as forage quali ty, density, cover, and biomass can all influence NDVI values and therefore may not singlehandedly identify any particular feature or element of a preferred area as ideal habitat For spatial analyses Anderson et al. (2010 ) suggest ed vegetation qua l ity and quantity are probably better predictors of
63 herbivore habitat preference than selected alternatives Additionally, t he monitoring of rangeland landscapes is difficult and particularly complicated due to the high degree of spatial and temporal variation in both vegetation and soils. An issue that presented problems in computing NDVI values as part of this study was the influence that plant shadows created wh en using high resolution sUAS imagery Several papers over the years have addressed shadows and the effects that they can have on vegetative indices (e.g., Ono et al. 2010 Park 2013 Gunasekara et al. 2015 Zhang et al. 2015 Cerra et al. 2016 ) T he time of day when imagery for NDVI analyses are obtained can change the overall impacts that shadows have on the resulting data. Further research is needed to account for the influence that shadows can i ntroduce to NDVI assessments especially with the high resolution imaging capabilities that sUAS provide This study focused on NDVI values generated from imagery collected by a sUAS over several sagebrush steppe field sites in Idaho A s part of a larger effort to scale up data collected at the plant and patch levels, to larger habitat scales and landscape levels the sUAS payload s were able to collect valuable data regarding pygmy rabbit habitat s election T his study found that remote sensing techniques of NDVI calculation via sUAS acquired imagery and metadata could be used as a tool to augment fiel d based sampling methodologies Utilizing the results of this work, theoretically an area of sagebrush steppe landscape could be overflown with a sUAS platf orm capable of generating NDVI data, and specific sagebrush locations within the resulting orthogeorectified scene could be identified as pygmy rabbits based on NDVI values.
64 Figure 2 1 A p ygmy rabbit ( Brachylagus idahoensis ). This individual was preparing to be transported to Washington State University as part of an Institutional Animal Care and Use Committee (IACUC) approved forage selection study by another researcher. M.A. Burgess
65 A B C D Figure 2 2. Signs of recent pygmy rabbit ( Brachylagus idahoensis ) presence A) Characteristic 45 angle, clean browse o f sagebrush ( Artemisia sp.) B) R epresentative scat deposits of jackrabbit ( Lepus sp.), mo untain cottontail rabbit ( Sylvilagus nuttallii ), and pygmy rabbit, respectively C) A m ima mound covered in dense sagebrush harboring a pygmy rabbit burrow system D) L eaping/landing pad snow compaction trails emanating from burrow systems M.A. Burgess M.A. Burgess www.rabbitweb.net/images/scat/ Rachlow and Witham 2006
66 Figu re 2 3. The extent of known pygmy rabbit ( Brachylagus idahoensis ) populations in the western U nited S tates The range incorporates portions of seven states although the actual distribution is patchy within this area M.A. Burgess N
67 Figure 2 4. An orthophoto mosaic aerial image showing mima mounds distributed across the sagebrush steppe landscape in the Lemhi Valley, Lemhi County, Idaho USA Darker green patches are mima mounds harboring denser concentrations of sagebrush ( Artemisia sp.) than interstitial ar eas having less dense sagebrush rabbitbrush ( Chrysothamnus sp.), grasses, and forbs The remnants of an old railroad grade possessing denser vegetation and a two rut dirt road are also visible traversing the top left of this orthophotomosaic. M.A. Burgess 0 100m
68 Figu re 2 5. The geographic information system spatial model ing concept for generating a response surface based on individual predictor variable layers for pygmy rabbits ( Brachylagus idahoensis ). The creation of a spatial response surface model should indicat e geographic locations exhibiting less preferred by pygmy rabbits J.R. Forbey
69 Figure 2 6. caling up Brachylagus idahoensis ) data collection Data gathered at the plant scale leads to information th at can be discerned at the patch scale, and then subsequently scaled up to the habitat level J.R. Forbey
70 Figure 2 7. The location of the Magic Reservoir study site for pygmy rabbit s ( Brachylagus idahoensis ) in Blaine County, Idaho USA The region shaded in red delineates the 29.3 hectare focal area of study during summer 2013 GoogleEarth
71 Figure 2 8. The location of the Rocky Canyon study site for pygmy rabbit s ( Brachylagus idahoensis ) in Lemhi County, Idaho USA The region shaded in red delineates the appro ximately 135 hectare focal area of study during summer 2014 GoogleEarth
72 Figure 2 9 Location s and distribution of the n = 84 geolocated ground control points established at the Magic Reservoir study site during summer 2013 in Blaine County, Idaho, USA Figure 2 1 0 Location s and distribution of the n = 111 geolocated ground control points established at the Rocky Canyon study site during summer 2014 in Lemhi County, Idaho, USA M.A. Burg ess 0 200m M.A. Burgess 0 200m
73 Figure 2 1 1 Selected mima mounds intensively ground truthed at the Magic Reservoir study site during summer 2013 in Blaine County, Idaho, USA. The dots are individual plants that were identified, measured and geolocated by the cooperative pygmy rabbit research group. The background image consists of an orthophotomo saic with a normalized difference vegetation index overlay both of which were constructed from individual images and metadata obtained using a small unmanned aircraft system.
74 Figure 2 1 2 An orthophotomosaic aerial image with a normalized differenc e vegetation index (NDVI) overlay of t he Rocky Canyon study site during summer 201 4 in Lemhi County, Idaho, USA. The red and green dots are individual plants that were identified, measured and geolocated by the cooperative pygmy rabbit research group. W hite dots indicate geolocated ground control points The orthophotomosaic and NDVI overlay both were constructed from individual images and metadata obtained using a small unmanned aircraft system. The northernmost portion of the study area was only imag ed with a visible wavelength payload; therefore, an orthophotomosaic was generated during post processing, but an NDVI layer could not be produced for that area
75 Figure 2 1 3 The southern two subareas of the Rocky Canyon study site shown with a norm alized difference vegetative index (NDVI) overlay generated during summer 2014 in Lemhi County, Idaho, USA. The red and green dots are individual plants that were identified, measured and geolocated by the cooperative pygmy rabbit research group ; red dot s are plants showing no signs of recent pygmy rabbit ( Brachylagus idahoensis ) presence, and green dots are plants exhibiting signs of recent pygmy rabbit presence The background image consists of a n NDVI layer constructed from individual images and metad ata obtained using a small unmanned aircraft system. The greener areas of the NDVI layer are locations with a higher density of green vegetation, while redder areas are locations with lower densities of green vegetation.
76 CHAPTER 3 ESTIMATES OF AMERICAN WHITE PELICAN S ( Pelecanus erythrorhynchos ) OBTAINED FROM PHOTOGRAMMETRIC PRODUCTS OF IMAGERY GATHERED USING A SMALL UNMANNED AIRCRAFT SYSTEM During the decade 2006 2017 civilian use of small unmanned aircraft systems (sUAS) became an increasingly popular means of conducting aerial surveys, including scientific studies with natural resources and conservation applications (e.g., Jones IV et al. 2006 Watts et al. 2010 Chabot 2014 Shahbazi et al. 2014 Johnson et al. 2015 Mulero Pzmny 2015 ) Significant developments in various components of sUAS, e.g., airframes, sensors, payloads, avionics, acquisition and operational costs, and lega l regulations, etc. have all contributed to heightened use of these systems worldwide (e.g., Hardin and Hardin 2010 Gordon et al. 2013 Stark et al. 2013 Cress et al. 2015 Pajares 2015 ) Use of any v isual data collection technique within natural resources research is subject to covariates such as observer biases, variations in sightings, discrepancies in training, observer fatigue, and detection probability issues that must be accounted for when gener ating assessments of focal targets (e.g., Link and Sauer 1997 Alldredge et al. 2006 Tracey et al. 2008 Walsh et al. 2011 Cook 2013 Beaver et al. 201 4 ) Quantification of focal items are also influenced by factors such as their heterogeneity, variability, behavior, physical features, and the timing of when the assessments are made (e.g., Magnusson et al. 1978 Kushlan and Frohring 1985 Packard et al. 1985 Frederick et al. 2003 Pollock et al. 2006 Langtimm et al. 2011 ) Whether data collection of focal targets is obtained from spac e borne satellites, on the (BOTG) methods], the selection and execution of the data collection techniques and their subsequent analyses requires thorough knowledge of the biases, limitations, and
77 underlying assumptions of the methods used ( Hines et al. 2010 Royle and Dorazio 2010 Thomas et al. 2010 ) Inappropriate application of data collection techniques or statistical analyses can produce inaccurate results of focal target abundance, their spatial and temporal trends, and pote ntially lead to inappropriate management decisions (e.g., Anderson 2001 Rosenstock et al. 2002 Williams et al. 2002 Tracey et al. 2008 Schmidt et al. 2012 Iknayan et al. 2014 ) Field researchers often default to BOTG methodologies because the data desired are collected at a finer resolution than affordable space borne satellite options can generate, or are logistically and fiscally easi er to obtain than manned aerial methods ( Berni et al. 2009 Madden et al. 2015 ) However, BOTG techniques can disrupt focal targets physically or behaviorally which introduces biases into the data collected, and BOTG methods are typically some of the most time and labor intensive data collection options available in natural resources (e.g., Ralph and Scott 1981 Kendall et al. 2009 Wang et al. 2010 Mulero Pzmny et al. 2014 Williams et al. 2015 ) Just like any tool, sUAS are by no means a magic bullet for all scientific data collection needs in natural res ources However, using an appropriate sUAS, payload, sampling methodology, and data post processing technique, many applications exist where sUAS can assist researchers by overcoming specific d eficiencies that accompany alternative data collection methods. within some defined unit of measure ( Chong and Evans 2011 ) Count data can be used to generate an index of relative abundance of focal targets, the total number or frequency of detections across sampling units, etc. ( Anderson 2003 Engeman 2003 ) ;
78 however in doing so, index counts often violate critical assumptions concerning uniform detectability ( Anderson 2001 Rosenstock et al. 2002 Stober and Smith 2010 ) On the oth generated using empirical models, and incorporates statistical parameters such as a coefficient of variance, confidence bounds, standard deviation, etc. ( Grenier et al. 2009 Chong and Evans 2011 Miller et al. 2011 ) The Am erican White Pelican [AWP, ( Pelecanus erythrorhynchos )] has generated particular interest over the years due to its and prominent bright white plumage B ecause of a relatively complex wildlife and fisheries management conflict it has generate d heightened attention from the Idaho Department of Fish and Game (IDFG) and other natural resource agencies ( IDFG 2016 ) ( Figure 3 1 ) Each spring, AWP migrate north from their warm winter ranges along marine coastlines of southern North America to nest in large colonies on specific inland remote islands within freshwater lakes and reservoirs of the western and north central United States ( US ) and southern Canada ( Knopf and Evans 2004 IDFG 2016 ) From the 1880s 1960s, AWP populations were deci mated by the damming of rivers for water storage and power generation, creation of reservoirs and artificial wetland habitat on former agricultural lands, draining of natural wetlands, and widespread spraying of organochlorine pesticides ( Knopf and Evans 2004 Keith 2005 ) After enactment of stringent environmental protection laws AWP populations have slowly imp roved ( King and Anderson 2005 ) An nual counts of the total AWP population have fluctuated considerably since the 1800s T he extensive breeding range, irregular timing of counts
79 and inconsistent estimation methodologies contributed to the un reliability of any single annual abundance estim ate up until the 1990s ( Keith 2005 ) In Idaho, AWP nesting colony populations have been systematically monitored by IDFG regional biologists since 2002 at Blackfo ot Reservoir and Minidoka Reservoir ( IDFG 2009 ) Survey s conducted using BOTG transects across each island resulted in a visual count of the number of nests ( IDFG 2009 ) These intrusive surveys were conducted in late May or early June during the late incubation/early nestling stage of AWP breeding ( IDFG 2009 ) Dramatic increases in AWP breeding populations in Idaho were noted from 2002 2008 at both the Blackfoot Reservoir and the Minidoka Reservoir locations where the counts of to to ( IDFG 2009 ; 20 16 ) The larger AWP population had negative impacts on native Yellowstone cutthroat trout ( Oncorhynchus clarkii bouvieri ), Bonneville cutthroat trout ( O c utah ), rainbow trout ( O. mykiss ), and other recreational fisheries within Idaho ( IDFG 2016 Meyer et al. 2016 ) Adult AWP and other piscivorous birds predominantly feed on abundant nongame fish species (e.g., Knopf and Evans 2004 IDFG 2009 Teuscher et al. 2015 IDFG 2016 Meyer et al. 2016 ) Due to declines in gamefish, the State of Idaho invested significant funding into fish hatchery programs to help recover native recreational fish populations ( Knopf and Evans 2004 Teuscher et al. 2015 Meyer et al. 2016 ) In 2009, IDFG generated a five year AWP management plan to balanc e viable AWP nesting populations with native and recreational fisheries interests using adaptive managem ent approaches. T rout stocking events were moved later in to t he year to reduce AWP depredation effects and AWP nesting areas at the Blackfoot Reservoir
80 location were intentionally decreased [with US Fish and Wildlife Service (USFWS) approval] Counts of AWP at the Blackfoot and Minidoka Reservoir locations have averaged 2,126 (range 724 3,174) breeding birds/year, and 3,600 (range 1,998 4,408) breeding birds/year, respectively ( IDFG 2016 ) I n 2012 AWP established a new breeding colony of at Island Park Reservoir Idaho ( IDFG 2016 ) As part of the five year AWP management plan, IDFG biologists looked to find potential methods of generating more accurate AWP population estimates ( IDFG 2009 ) In this study, our object ive s w ere to use the UF Nova 2.1 sUAS and its optical payload to : 1) determine if the sUAS could successfully fly autonomous transects over nesting AWP at an altitude that would produce sufficiently high resolution imagery without disturbing the nesting bi rds; 2) determine if the or thophotomosaics generated from the transect imagery could be used to create counts of nesting AWP; and 3) compare the latter counts with BOTG counts. American White Pelican Study Area This study was conducted o ver AWP nesting col onies in Minidoka Reservoir (Lake Walcott) within the USFWS Minidoka National Wildlife Refuge (MNWR), Cassia County, Idaho, USA ( Figure 3 2 ). The AWP breeding sites were located 2.75 kilometers (km) east of the Minidoka Dam, and consisted of three artificial spoil islands; Pelican (WGS84 Datum: 42.662538 West) Tern (WGS84 Datum: 42.663919 West) and Gull (WGS84 Datum: 4 2 662837 North; 11 3 450570 West) Islands ( Figure 3 3 ) Each spoil island had nearly flat topography .0 meters (m), and was composed principally of a light colored sandy mud substrate with interspersed gravel and larger rocks of non uniform s ize or shape. Essentially all rock surfaces which
81 projected above the waterline near each island were preferred perching, loafing, and feeding sites for AWP and other avifauna including Ring billed Gull ( Larus delawarensis ), California Gull ( L californic us ), and Double crested Cormorant ( Phalacrocorax auritus ) T he exposed rock surfaces were generally coated with white feces Sparse vegetation consisting of small shrubs and few trees were located irregularly along the perimeters of Pelican and Gull Isla nds, which supported nest s of Double crested Cormorant Based on US Bureau of Reclamation (USBOR) data, the Minidoka Reservoir was at 98.1% of its total active storage capacity during this study ( USBOR 2016 ) and t he southern and western most of the three spoil islands, Pelican Island, had an exposed area at water level of 0.17 hectares ( ha ), possessed the most total vegetation cover, and was the closest island to the mainland (59 m). Tern Island, the northern most and smallest ( 0.03 ha ) of the three spoil islands lacked any vegetation, had the least amount of available sandy nesting area for AWP, but had the most exposed rock surrounding the island, and was located the furthest distance from the mainland at 300 m. The eastern most and largest ( 0.30 ha ) of the three islands, Gull Island, exhibited some shrubby vegetation and tree cover, offer ed copious sandy substrate for AWP nesting, and was located 152 m from the mainland. The Minidoka Reservoir AWP nesting sites are at 1,295 m above sea level (ASL), and National Oceanic and Atmospheric Administration (NOAA) climatology data for the locality classifie d it as semi arid with an average annual precipitation total of 24.2 centimeters (cm) per year ( NOAA 2016 ) Historical data showe d that for the month of June when AWP are nesting, average daily air temperatures range from 7.8 25.6
82 Celsius (C), and rainfall is infrequent, while in the month of D ecember when AWP have migrated south to warmer latitudes, average daily air temperatures range from 8.9 + 2.2 C, and regular snows contribute the most to annual precipitation totals ( NOAA 2016 ) Methods Used for Estimating Nesting American White Pelican s from sUAS derived Imagery Products The UF Nova 2.1 sUAS equipped with an Olympus E 420 optical payload ( both detailed in Appe ndix A ) was flown over the Minidoka Reservoir spoil islands on 10 June 2014 under permits issued by the Federal Aviation Administration (FAA) Certificate of Waiver or Authorization (COA) #: 2014 WSA 42 COA ; and a USFWS Research and Monitoring Special Use Permit #: 14614 10 14. The mission was executed by a three person University of Florida Unmanned Aircraft Systems Research Program ( UFUASRP ) flight crew (detailed in Appendix C ) aboard an IDFG boat operated by the southeastern regi onal biologist The boat was positioned toward the middle of Minidoka Reservoir due north of the three spoil islands Personnel from the MNWR had already positioned a USFWS boat along the buoy line delineating the boundary of recreational boating activit ies t o observe the sUAS operations and to answer questions that might arise from curious boaters. P ersonnel from the USFWS, US Bureau of Land Management ( USBLM ) IDFG, and Idaho Department of Parks and Recreation (IDPR) were on hand with binoculars and ca meras along the southern reservoir shoreline to observe and note any bird reactions, and report any issues via radio to the IDFG regional biologist and flight team if disturbance were observed so the mission could be terminated
83 Meteorological conditions r ecorded during preflight checks by the ground station operator (GSO) at 0752 hours steady wind out of the west at 8.6 meters per second ( m/sec ) with occasional gusts to 10.0 m/sec, air temperature of 15.5 C, dew point 9.1 C, relative humidity 60.4 %, and a barometric pressure of 1012 hectopascals. At 0757, the UF Nova 2.1 sUAS was launched, ascended to the preprogrammed altitude of 225 m above ground level ( AGL ) and beg a n initial waypoint navigation through the preprogrammed flight path. The first three upwind passes (from east to west) were directly over Tern Island to evaluate if the UF Nova 2.1 sUAS overflights elicited any response from the AWP. The first upwind pass was conducted at 225 m AGL, and decreased by 50 m to 175 m AGL for the second pass and by another 50 m to 125 m AGL all directly over Tern Island N o changes in bird behavior were reported by any of the observers during the three trial passes, therefore at 0803, the aircraft was instructed to execute the entire preprogrammed flight path autonomously at a commanded altitude of 125 m AGL. The preprogrammed flight path (separate from the altitude testing flight path) consisted of 16 parallel linear transects ten exclusively upwind, and six exclusively downwind. A 30 .0 m spacing between parallel horizontal transects was calculated to be appropriate for sufficient imagery sidelap. All three islands were uniformly covered with upwind transects, which in response to the ambient winds aloft slowed the ground speed of the aircraft and therefore met or exceeded the desired imagery endlap calculated for the planned orthophotomosaic post processing. All downwind transects captured imagery as well; however, due to the sum of both the ambient wind speed aloft and the 1 5 .0 m/sec cruise airspeed of the aircraft, downwind imagery were captured
84 with the knowledge that they would have insufficient endlap for mosaicking purposes but they could provide additional imagery samples of the nesting avifauna With the westerly wind and location of the three focal islands taken into conside ration, a rectangular area with dimensions of 700 450 m would cover all three islands with parallel upwind transects and produce sufficient 30 .0 m sidelap spacing for post processing. However, due to the triangular geographic orientation of the three fo cal islands, the flight path was arranged so that the focal area of interest in which straight and level flight was required was reduced. Th e horizontal transects over Tern Island were shortened from 700 m in length to 300 m in length to save time and eff ort collecting imagery of open water in the areas directly west and north of Tern Island, which would not assist in meeting the specific objectives of this mission which was focused on the islands ( Figure 3 4 ) The entire flight p minutes ( min ) at which time everything checked out as nominal by the flight crew, so the aircraft was instructed to repeat the entire flight path With favorable time and conditions, additional exposure stations would add a larg er number of tie points for orthophotomosaic production during post processing. At 0829, the second pass of the entire flight plan was completed, and the aircraft was landed at 0832 at the landing waypoint. The total elapsed flight time was 35 min and 45 sec. Once back at the boat ramp, the onboard computer containing the imagery and metadata collected during the flight was backed up onto a secondary portable storage device for archival and reassurance purposes. P reliminary views of the imagery and
85 metad ata were all positive; imagery of the islands and the nesting AWP were clear, and the metadata file of telemetry data for the payload sensor system was complete. Data post processing for this study was conducted using the methodology delineated in Appendix D E ach of the three spoil islands were post processed individually and the final PhotoScan post processing step of merging the individual products was omitted as it was not necessary for the objectives of this study. S everal m ethods were us ed to experiment with various exported imagery products from the PhotoScan software to ultimately make counts of AWP on each of the three spoil island s The first method of analyses used Applied Imagery Quick Terrain Reader software to exa of the Gull Island digital elevation model ( DEM ) produce d using PhotoScan It was thought that inflating the elevation profiles of objects on the relatively flat topography of Gull I sland might assist in manually cou nting the number of AWP. A second method of analyses involved manually counting the AWP observed in the orthophotomosaic product of Gull Island using iT AG counting software ( Viquerat and van Neer 2015 ) The third method of counting AWP for comparative purposes was conducted using the Gull Island orthophotomosaic product generated in PhotoScan and a single nadir oriented individual JPEG image from the fligh t, in which a small subarea within the orthophotomosaic was also identified within the individual JPEG image. Using the iTAG software, the AWP were counted within the subarea from the orthophotomosaic, and then the number of AWP within the same subarea o n the individual JPEG flight image were enumerated by an observer counting visually Th e results of this analysis were designed to assess effects of AWP moving
86 between aerial transect overflights, which the PhotoScan software would have to address when c reating a single, static imagery product of Gull Island. The fourth analyses involved selecting a single method from the three outlined above that provided the most consistent counts using subsampling techniques of Gull Island U sing cr owdsourc ing (a meth od of achieving desired content through workforce multiplication in this case data in the form of AWP counts f rom a large pool of individuals) n = 30 volunteers with unknown histories of ornithology aerial observation, or focal object identification with in digital imagery were asked to conduct AWP counts for each of the three spoil island s in MNWR Each volunteer was shown n = 5 AWP individuals in each of the three orthophotomosaics loaded into the software as examples of the focal items in which to cou nt Volunteers took as much time as they needed to count individual AWP from the three orthophotomosaics, and could stop and resume as necessary. The final analyses w ere to c ompare sUAS estimates of AWP determined by the crowdsourced count data to IDFG n est counts that were obtained using BOTG transect methods within a week after the 10 June 2014 sUAS flight Results of th e sUAS Imagery Products for Assessing Estimates of American White Pelican Nesting The AWP and other birds near the target islands show e d no signs of disturbance by the fixed wing UF Nova 2.1 sUAS conducting transects over their nesting islands in Minidoka Reservoir at 225 m 175 m and 125 m altitude s AGL. Based on the optical parameters of the payload used along with the ambient wind sp eed and direction, the 125 m altitude for overflight of the islands was calculated to deliver sufficient endlap and optical resolution of the imagery captured during the flight; therefore, attempting transects at lower altitudes was not necessary.
87 With t he ir large bodies, orange bills, occasional black crest, and characteristic shadow signature s, detecting AWP within the post processed orthophotomosaics was particularly easy for pelicans nesting in open areas on the islands even with the light colored sandy mud substrate as a backdrop The detection of nesting AWP individuals adjacent to large rocks was more challenging because nearly all rock surfaces were generally coated in white feces Large rocks generated shadows that resembled the shadows of nesting AWP, and vic e versa. Additionally, detection of AWP was sometimes difficult in areas with overhanging vegetation. With nadir oriented imagery, individuals nesting close to tree trunks, or within the peripheral margins of dense brush, presented some dete ction issues. It was also noted that Double crested Cormorants nesting or roosting in trees or other elevated locations on Pelican and Gull Islands created particularly extensive areas of white feces on and under the supporting nesting structure, making detection of AWP near or under those localities quite difficult ( Figure 3 5 ) The w indy conditions during the data collection flight contributed to some orthophotomosaic blur which manifested itself as swirls in herbaceous vegetat ion waving in the wind but otherwise did n o t present any appreciable issues for the study goals of imaging AWP nesting on the islands ( Figure 3 6 ) To address other research questions, the post processed imagery products using t he PhotoScan software of Gull Island were selected as the testbed for various AWP enumeration techniques. The use of Quick Terrain Reader DEM was not effective at a factor of 5 .0 times the true elevation in helping to identify AWP within the Gull Island DEM. The DEM model elevation data was then exaggerated
88 by a factor of 10 .0 times the true elevation and similarly was not of any substantial assistance in identifying AWP within the scene ( Figure 3 7 ) Using the iTAG counting software, an arbitrary small subarea of the Gull Island orthophotomosaic yielded a count of 88 AWP individuals. Coincidentally, c ounting AWP within the same small subarea using iTAG software of a n individual JPEG image collected during the sUAS flight also yielded 88 individual AWPs ; however, several of the AWP individuals within the JPEG image were not in the same physical location within the orthophotomosaic, and vice versa ( Figure 3 8 ) Individual p elican movement s that occurred between imaging transects over the exact same target area were short, but these movements illustrate just how valuable direct georeferencing of imagery can be for identifying movement of featur ed targets between aerial transect passe s Based on the various software titles employed and subsampling of several arbitrary small sub areas of each of the three islands, using the iTAG counting software on the orthophotomosaic output from PhotoScan pro vided the most consistent and straightforward combination of data products and software for manually counting AWP on the three islands (Figures 3 9 3 10 and 3 11 ) T he crowd sourc ed data of AWP counts by each of the n = 30 volunteers using the iTAG software result ed in remarkably similar values for each of the three islands. For Pelican Island, the mean number of AWP counted was 571.5, with a standard deviation of 6.58 or 1. 2% The individual counts ranged from 559 582, and had a 95% confidence level of 2.46 ( Figure 3 12 ) For Tern Island, the mean number of AWP counted was 75.2, with a standard deviation of 2.62 or 3.5% The individual counts ranged from 70 80, and had a 95% confidence level of 0.98 ( Figure 3 13 ) For
89 Gull Island, the mean number of AWP counted was 1,217.1, with a standard deviation of 7.38 or 0.6% The individual counts ranged from 1,203 1,231, a nd had a 95% confidence level of 2.75 ( Figure 3 14 ) The mean total counts of crow dsourced AWP detect ions for all three spoil islands was 1,863.9 with a standard deviation of 15.8 or 0.8% The individual sums ranged from 1,83 5 1,890, and had a 95% confidence level of 5.91 The regional biologist with IDFG provided the following AWP nest counts based on a BOTG survey across each of the three islands less than a week after the sUAS overflight: 812, 90, and 1,230, AWP nests for Pelican, Tern, and Gull Islands, respectively. The sum total of AWP nests based on the BOTG survey of all three spoil islands was 2,132. Comparing crowdsourced mean total counts of individual AWP for Pelican, Tern, and Gull Island with counts generated b y a BOTG survey revealed that sUAS estimates were 70.4% 83.6% and 99.0 % ( P < 0.0 5 ) of BOTG nest counts respectively. Discussion on the Use of sUAS Imagery Products to Assess Estimates of American White Pelican Nesting T he UF Nova 2.1 sUAS fixed wing pla tform outfitted with an Olympus E 420 optical payload system was an effective combination of airframe and payload for the desired data collection needs. Flying at an altitude of 125 m AGL provided sufficient resolution for discriminating AWP from cormor ants and gulls, while not disturbing any of the birds. Using the UF Nova 2.1 sUAS airframe was particularly beneficial because transects over the target areas were conducted into the wind, which capitalized on the glider features of the airframe design by limit ing the use of the electric motor even with stiff ambient winds The resulting transects were flown nearly silent ly Spacing the parallel transects 25 .0 m apart rather than at the 30 .0 m spacing would have increased
90 the sidelap slightly and possibl y improved the post processing results. Repeating the entire flight plan a third total time would have capture d more imagery over the focal targets and would have likely enhanced the post processing results somewhat as well The use of manual counting te chniques of imagery for enumerating AWP did take time to accomplish, and was tedious work. Use of iTAG software made the process much easier for several reasons. The software left target, and attempting to mark another target very near to an existing mark would trigger a warning screen asking the observer to verify that the newly identified target was not the same target that had already been marked. This feature reduced t he chances of double counting a target contained within a si ngle orthophotomosaic to nearly zero. The software kept a running tally of the selected targets, so the observer making counts did not have to take their eyes off the computer screen to record physical counts by hand or in a separate spreadsheet, etc. Th is saved time and permitted the observer to stay focused on identifying target items on the screen Lastly, the ability to pause counting, save the file and re open it to resume counting later was also advantageous. Using automated detection computer rec ognition algorithms for estimating AWP seemingly would present several challenges due to their white plumage against a predominately light ly color ed substrate. Other issues include : 1) white feces coated rocks adjacent to nesting AWP ; 2) adult AWP on nest s were generally sitting, not standing ; and 3) detection of AWP nesting under dense, low elevation vegetation, or under trees containing large cormorant nests was difficult with only nadir oriented imagery.
91 In regards to the difference in numbers reported between the orthophotomosaics generated by the sUAS imagery and the BOTG manual ground surveys conducted across the islands, several important details were realized during the data analyses. First, and probably the most notable issue that somehow failed t o be recognized until the IDFG sent their BOTG ground count data to the UFUASRP, was that the UFUASRP was collecting data to count individual AWP on each of the three spoil islands in MNWR. The biologists at IDFG conduct their BOTG surveys by counting the number of nests observed on each island. T he target items being enumerated by the two entities were not one and the same. Understanding the source of discrepancy between the numbers generated by analyses of the sUAS data products and the data provided b y IDFG was reassuring. After identifying the primary source of error, additional investigation was conducted to analyze the Pelican Island orthophotomosaic even closer to try to establish why a difference of nearly 30% occurred between the total number of AWP counted on the island from a sUAS generated aerial orthophotomosaic and the number of nests counted using BOTG methods were considerably different Upon further inspection, there were several areas on Pelican Island that contained abandoned or empty nests that were not counted using the existing orthophotomosaic analys i s protocols established before the data was crowdsourced but those abandoned or empty nests were likely counted as part of the IDFG nest totals ( Figure 3 15 ) Taking this concept further, seeing that 99% of Gull Island aerial AWP counts from the orthophotomosaic image corresponded to IDFG BOTG nest survey numbers, careful investigation of the Gull Island orthophotomosaic that was generated revealed that there
92 were virtually no abandoned or empty nests detecta ble; therefore the counts of the respective targets were nearly identical on Gull Island B ecause of the repeatable, tightly georeferenced imagery that is capable of being collected using the optical paylo ads of the UF Nova 2.1 sUAS, existing and new avian t t + t + t + n spatial sampling assumptions and detection probability are accounted for, best effort counts of individual target s using dual observer techniques from a moving manned aircraft could possibly give way to more accurate population estimates of a specific t collected imagery Conceivably even more attractive is the possibility of generating estimates of seasonal t + t + possible if a routine sUAS flight schedule can be established and maintained throughout an entire reproductive season ( Williams et al. 2011 ) An advantageous reason for us ing a sUAS to conduct aerial surveys is that when outfitted with direct georeferencing payloads, the digita l imagery and associated metadata collec ted can be catalogued, archived, and stored in perpetuity as a photographic record of the flight. This feature permits a multitude of qualitative and quantitative analyses to be conducted in the future, especially t emporal based assessments of alterations that may be attributed to global climate change natural disaster impacts or other salient events in natural resource s and environmental disciplines. Unlike traditional visual observer manned aerial surveys where identification and enumeration of targets must be conducted immediately and quickly from the air as the aircraft makes transects over or circumnavigate s a target area,
93 sUAS collected raw data can be analyzed at any occasion after a flight concludes. Typi cal visual observer aerial surveys do not afford much time for observers to resolve potential doubts or reassess many areas while a flight is in progress, whereas sUAS data ca n be examined repeatedly Rare or difficult to identify targets are photographica lly documented and can be evaluated by numerous subject matter experts for their opinions, and sUAS data sets can be reanalyzed repeatedly into the future; benefits that are especially valuable as new computer software for imagery post processing emerges, innovative computer based feature recognition algorithms become available, and as statisticians develop novel methods to assist researchers in moving from typically providing counts of focal targets to generating estimates having testable bounds. If quest ions arise about the estimates of target items generated from a sUAS flight, digital imagery of the target area can be provided to help justify the reported results, rather than merely providing a stack of paper data sheets containing a series of tick mark s, which may be wrought with errors. Using increasingly sophisticated computer software which is progressively more prolific and affordable, directly georeferenced two dimensional (2D) imagery can be post processed using Structure from Motion ( SfM ) proce dures into three dimensional ( 3D ) point clouds, DEMs seamless orthophotomosaics, or other output types, permitting analyses that were historically both time and labor intensive to now be conducted relatively quickly and with improved precision. Utilizin g the same software, portions of larger orthophotomosaic scenes can be cut into smaller sampling areas with known dimensions, which allow detectable target density calculations, and other
94 computational inferences to be conducted with a level of accuracy th at is potentially superior to those constructed strictly from visual observers in manned aerial surveys. The count data accumulated from multiple observers of orthophotomosaics can later be used as input for more powerful statistical analyses to generate target item estimates having testable limits. Additionally, biases associated with multiple observers, imperfect detection, heterogeneity, timing, and other inherent limitations of aerial surveying can be accounted for and included as covar iates in statis tical analyses. It is important to emphasize that seamless orthophotomosaics are not necessarily the only or required product of all aerial imagery data collection efforts using a sUAS. A myriad of applications exist where smaller sampling units within po st processed raw imagery can provide appropriate data needed to use a number of powerful statistical techniques. Therefore using directly georeferenced imagery collected with a calibrated sUAS has the potential to improve the accuracy of the answers to th e oft asked questions: How many are there? How much is there? Although some will argue that sUAS have matured enough for routine scientific use, there is still much to be learned and disseminated about appropriate use of sUAS as a tool for scientifi c data collection. For example, selection and availability of an effective sensor and airframe combination, based on the scientific question at hand seems to be a relatively straightforward process on the surface H owever suitably making these decisions is often underemphasized There is still research that needs to be done especially since sUAS and their payloads have fostered new methodologies for data collection, and enabled modifications of existing techniques for studies using aerial surveys with greater efficiency, safety, and accuracy.
95 Figure 3 1 A colony of American White Pelicans ( Pelecanus erythrorhynchos ) nesting on Pelican Island, Minidoka National Wildlife Refuge, Cas s ia County, Idaho, USA. This image was taken from the southern sh ore of Minidoka Reservoir on 8 July 2011 during a preliminary site reconnaissance visit for potentially flying a small unmanned aircraft system over the colony to generate an estimate of the number of nesting pelicans. Double crested Cormorant ( Phalacroco rax auritus ) and gulls ( Larus sp ) are also seen using the spoil island. M.A. Burgess
96 Figure 3 2. The location of the three spoil islands used for nesting by American White Pelicans ( Pelecanus erythrorhynchos ) in Minidoka National Wildlife Refuge, Cassia County Idaho, USA. GoogleEarth
97 Figure 3 3. The three spoil islands used for nesting by American White Pelicans ( Pelecanus erythrorhynchos ) in Minidoka National Wildlife Refuge, Cassia County, Idaho, USA. Pelican Island (middle left), Tern Island (top center), an d Gull Island (middle right). GoogleEarth N
98 Figure 3 4. The Virtual Cockpit preplanned flight path designed for autonomous waypoint navigation by the University of Florida Nova 2.1 small unmanned aircraft system over the three spoil islands used for nesting b y American White Pelicans ( Pelecanus erythrorhynchos ) on 10 June 2014 in Minidoka National Wildlife Refuge, Cassia County, Idaho, USA. A mbient wind direction was from the west so upwind passes over the target islands from right to left (increasing waypoi nt numbers) were arranged The flight line spacing between parallel transects north to south was 30.0 m based on the parameters of the optical payload, the commanded flight altitude, and the calculated ground speed projection for the aircraft. All flight operations were conducted from a boat idling approximately 20.0 m waypoint at the top center of the left is the locat ion for which the aircraft loitered about after the hand launch takeoff from the stationary boat. T right is the location where the aircraft performed a controlled descent altitude to 40.0 m above ground (water) level before proceeding along the landing glideslope lin for an amphibious landing Virtual Cockpit N
99 A B C D Figure 3 5. Examples of American White Pelican ( Pelecanus erythrorhynchos ) nesting locations on spoil islands in Minidoka National Wildlife Refuge, Cassia Cou nty, Idaho, USA. Images are from o rthophotomosaic s generated with imagery and metadata collected from a small unmanned aircraft system flown over the islands on 10 June 2014. A) Pelicans nesting in open areas present very few detection issues B) N estin g adjacent to rocks covered in white feces can affect pelican detection C) N esting under trees or dense vegetation influences detection of pelicans D) N esting under Double crested Cormorant ( Phalacrocorax auritus ) nests or roosts complicate s ground nes ting pelican detection. M.A. Burgess M.A. Burgess M.A. Burgess M.A. Burgess
100 Figure 3 6. Swirls that result from mosaicking still imagery of herbaceous vegetation blown around by ambient wind s Imagery and metadata were collected 10 June 2014 using a small unmanned aircraft system and its optical pa yload flown over Pelican Island, Minidoka National Wildlife Refuge, Cassia County, Idaho, USA. M.A. Burgess
101 A B C Figure 3 7. Exaggerating the elevations of a digital elevation model in attempt to highlight ground nesting American Wh ite Pelicans ( Pelecanus erythrorhynchos ) on Gull Island, Minidoka National Wildlife Refuge, Cassia County, Idaho, USA. A) Native digital elevation model showing the true elevations without exaggeration B) E levations exaggerated by a factor of 5 .0 times the true elevation C) E levations exaggerated by a factor of 10 .0 times the true elevation Quick Terrain Reader Quick Terrain Reader Quick Terrain Reader
102 A B Figure 3 8. Comparison of iTAG software for counting a subarea of nesting American White Pelicans ( Pelecanus erythrorhynchos ) on Gull Island, Minidoka National Wildlife Refuge, Cassia County, Idaho, USA. A) Locations of n = 88 individual pelicans (red dots) from a portion of a digital orthophotomosaic product B) L ocations of n = 88 individual pelicans (red dots and red hollo w squares indicating targets in alternate locations ) from a portion of a single image captured during the flight of a small unmanned aircraft system over the island on 10 June 2014 Pelican movement between imaging passes over the target area can affect c ounts using either data source. M.A. Burgess M.A. Burgess
103 A B Figure 3 9. An ortho photomosaic generated by PhotoScan software from imagery collected using a small unmanned aircraft system over Pelican Island, Minidoka National Wildlife Refuge, Cass ia County, Idaho, USA on 10 June 2014 The orthophotomosaic was constructed to enumerate the nesting American White Pelicans ( Pelecanus erythrorhynchos ) on the island. A) The native orthophotomosaic product B) T he orthophotomosaic with individual pelic ans marked with red dots using iTAG counting software A random area is magnified 250% to show the marking process. M.A. Burgess iTAG
104 A B Figure 3 10. An orthophotomosaic generated by PhotoScan software from imagery collected using a small unmanned aircraft sys tem over Tern Island, Minidoka National Wildlife Refuge, Cassia County, Idaho, USA on 10 June 2014. The orthophotomosaic was constructed to enumerate the nesting American White Pelicans ( Pelecanus erythrorhynchos ) on the island. Distinguishing pelicans f rom large rocks coated in white feces can present challenges A) The native orthophotomosaic product B) T he orthophotomosaic with individual pelicans marked with red dots using iTAG counting software. A random area is magnified 125% to show the markin g process. M.A. Burgess iTAG
105 A B Figure 3 11. An orthophotomosaic generated by PhotoScan software from imagery collected using a small unmanned aircraft system over Gull Island, Minidoka National Wildlife Refuge, Cassia County, Idaho, USA o n 10 June 2014. The orthophotomosaic was constructed to enumerate the nesting American White Pelicans ( Pelecanus erythrorhynchos ) on the island. A) The native orthophotomosaic product B) T he orthophotomosaic with individual pelicans marked with red dot s using iTAG counting software. A random area is magnified 300% to show the marking process. M.A. Burgess iTAG
106 Figure 3 12 Frequency histogram of observed number of American White Pelican s ( Pelecanus erythrorhynchos ) on Pelican Island, in Minidoka National Wildlife Refuge, Cassia County, Idaho, USA. Visual count s ( = 571.5, SE = 1.20) were made using iTAG software of a digital orthophotomosaic of the island constructed from imagery collected with a small unmanned aircraft flown on 10 June 2014. C rowdsourced volu nteers ( n = 30) with an unknown ornithology, aerial observation, or foca l object identification history generated the pelican count data. 0 1 2 3 4 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 Frequency Observed Number of Pelicans
107 Figure 3 13 Frequency histogram of observed number of American White Pelicans ( Pelecanus erythrorhynchos ) on T ern Island, in Minidoka National Wildlife Refuge, Cassia County, Idaho, USA. Visual counts ( = 75.2, SE = 0.48) were made using iTAG software of a digital orthophotomosaic of the island constructed from imagery collected with a small unmanned aircraft f lown on 10 June 2014. Crowdsourced volunteers ( n = 30) with an unknown ornithology, aerial observation, or focal object identification history generated the pelican count data. 0 1 2 3 4 5 6 69 70 71 72 73 74 75 76 77 78 79 80 81 Frequency Observed Number of Pelicans
108 Figure 3 14 Frequency histogram of observed number of American White P elicans ( Pelecanus erythrorhynchos ) on Gull Island, in Minidoka National Wildlife Refuge, Cassia County, Idaho, USA. Visual counts ( = 1217.1, SE = 1.35) were made using iTAG software of a digital orthophotomosaic of the island constructed from imagery collected with a small unmanned aircraft flown on 10 June 2014. Crowdsourced volunteers ( n = 30) with an unknown ornithology, aerial observation, or focal object identification history generated the pelican count data. 0 1 2 3 4 5 6 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 Frequency Observed Number of Pelicans
109 Figure 3 15. Abandoned or empty nests are visible among ground nesting American White Pelicans ( Pelecanus erythrorhynchos ) in a small subarea of an aerial orthophotomosaic constructed from imagery collected using a small unmanned aircraft o ver Pelican Island, Mi nidoka National Wildlife Refuge, Cassia County, Idaho, USA on 10 June 2014 A visual count of 21 pelicans are observed in the image, however, a t least 25 abandoned or empty nests are counted in the image as well. A nest census conducted by walking across the island would overestimate the number of pelicans on nests M.A. Burgess
110 CHAPTER 4 AN INNOVATIVE DATA COLLECTION APP ROACH FOR MANNED AERIAL SURVEYS USING OPTICAL PAYLOADS DESIGNED FOR SMALL UNMANNED AIRCRAFT SYSTEM S One of the most often asked question s of natura l resource and environmental professionals is How much is there? which for example could refer to the population size of some specific flora or fauna, or to the extent of some habitat type. The answers provided by experts are cr itical and often influence resource management decisions. Anderson (2001 ) indicates, data collection methodologies are not trivial, and statistically valid estimates must ac count for many sources of error, e.g., detection probability, observer bias, timing of data collection, etc. An aerial perspective of n atural resource targets of interest can reveal important features which may be challenging or unattainable using ground based data collection methods alone From a historical standpoint, g athering remotely sensed information from an aerial perspective pre cedes manned flight by quite some time ( Verhoeven 2009 Colomina and Molina 2014 ) Since 1900, both p hotography and aviation have advanced tremendously. Black and white film still frame photographs requiring darkroom development have progressed to high resolution digital sensors recording visible color (RGB) and other wavelengths to electronic storage me dia capable of immediate viewing ( Verhoeven 2010 ) Aircraft design construction methods and materials, propulsion systems, and aircraft reliability have s imilarly evolved with facilitating science and technology ( Crouch 2004 ) By the end of World War II, aerial imagery taken from manned aircraft for mapping, intellige nce, surveillance, and reconnaissance (ISR) purposes were a primary application of flight ( Wolf et al. 2014 ) Now, orbiting satellites capture and transmit
111 remo tely sensed imagery and data, and enabl e three dimensional (3D) positional determination and advanced navigational capabilities through g lobal p ositioning s ystems (GPS) and i nertial n avigation s ystems (INS) ( Kuenzer and Dech 2013 Finkl and Makowski 2014 Wolf et al. 2014 ) Within the l ast decade, exponential growth in civilian based small unmanned aircraft systems (sUAS), their sensor payloads, and applications, have occurred from technology initially developed for military purposes. As costs have decreased for sUAS, their availability has increased; yet in the United States (US), Federal Aviation Administration (FAA) regulations integratin g sUAS into the National Airspace System (NAS) lagged behind the technology imped ing routine sUAS flight (e. g., Rango and Laliberte 2010 Watts et al. 2010 Hardin and Jensen 2011 Elias 2016 ) As the number of sUAS flights has increased worldwide, the use of sUAS as a tool for collecting high resolution aerial imagery has opened countless new doors for obtaining aerial imagery of focal targets ( Richards 2013 ) Nearly all sUAS are equipped with an autopilot unit capable of making numerous simultaneous flight adjustments per second, maintaining altitude, airspe ed, and course ( Christiansen 2004 Linchant et al. 2015 ) Existing methods of aerial surveying and monitoring of natural resource targets using visual observers from manned aircraft are primarily accomplished at slow airspeeds and low altitudes in fixed or rotary wing aircraft which generally lack cost prohibitive autopilot units. Consequently, the r eliance on pilot skill and experience in conducting prolonged slow and low flight is paramount for the safety and success of such missions. Sasse (2003 ) reporte d aviation accidents were the most prevalent form of on the job mortality for wildlife biologists The safety
112 benefit that sUAS offer as an aerial data collection tool is just one reason that interests in civilian based uses of sUAS technology have flouri shed. However, sUAS are not necessarily a solution for all aerial data collection activities in natural resources. In fact, fielding sUAS may be more expensive or even impractical than using manned aircraft for aerial data collection due to logistics co sts, and regulations that accompany sUAS operations. The use of sUAS can assist in data collection for certain applications, but not all. For example, when: 1) e xtreme weather events occur; 2) field sites are separated by significant geographic distances; 3) focal targets encompass large landscapes; 4) operational field sites are difficult to access by ground or water based transportation methods; or 5) field sites are located in areas prohibited to sUAS flight; each of these situations can render sUAS da ta collection methods impossible, impractical, or illegal. C apitaliz ing on advantages that sUAS payload s offer the following objectives are addressed herein: 1) determine the legal ity of attachment of a small external sensor pod (ESP) containing affordabl e high resolution digital camera payloads designed for sUAS to manned aircraft commonly used in natural resource based applications; 2) determine if sufficient imagery resolution could be obtained from optical payloads in ESP equipped manned aircraft safel y flying at slow airspeeds and low altitudes; and 3) determine if imagery and metadata collected from payloads contained within ESPs can be post processed into ortho geo rectified imagery products. T his chapter highlights two ESP applications : 1) identificat ion and enumeration of sea turtle nesting crawls along a 21 .0 kilometer (km) stretch of the Archie Carr National Wildlife Refuge (ACNWR) on the Atlantic coastline of central Florida in which existing
113 FAA regulations would not permit use of sUAS because of proximity to housing, people on the beach, and a nearby airport, and logistically collecting imagery of an area of such size in a short amount of time was impractical with a sUAS; and 2) developing methods for maximizing the efficiency of ESP flights over several different vegetation based targets in Water Conservation Area (WCA ) 2A of the Everglades in south Florida where use of sUAS was impractical owing to distances between plots and illegal at the time due to existing FAA regulations surrounding Miami International Airport (KMIA) Study Areas Archie Carr National Wildlife Refuge The imagery and metadata collection for the first part of this study were conducted over a stretch Atlantic Ocean coastline that is located south of Melbourne Beach, in southern Brevard County, Florida, USA ( Figure 4 1 ). A 21.0 km portion of beach that began at the northern limits of the ACNWR (WGS84 Datum: 28.039419 West), and terminated approximately ( ) 1.0 km north of Sebastian Inlet (WGS84 Datum: 27.861850 West) was the focal area of study ( Figure 4 2 ). The University of Central Florida ( UCF ) m arin e t urtle r esearch g roup mark ed the 21.0 km stretch of beach every 0.1 km with wooden stake s in the dunes starting at the northern end of ACNWR and proceeding south to Sebastian Inlet to provide general turtle crawl locations along the focal survey area T he white sandy beaches found along the coastline in this part of Florida are characteristic of the high energy near shore coastal systems found along many parts of the eastern coast of Florida, and are preferred nesting habitat for three species of sea tur tles: leatherback ( Dermochelys coriacea ), loggerhead ( Caretta caretta ), and green turtles ( Chelonia
114 mydas ). Female sea turtles show high site fidelity for nesting beaches, lay several clutches per season, and nest every 2 3 years ( Weishampel et al. 2003 ) The ACNWR has an elevation of 2 .0 meters ( m ) above sea level ( ASL ) depending on the location and National Oceanic and Atmospheric Administration ( NOAA ) climatology data for the locality classified it as subtropical with an average annual precipitation total of 126.0 centimeters ( cm ) per year ( NOAA 2016 ) Historical data showed that for the month of June, average daily air temperatures ranged from 22.3 31.4 C elsius (C) and rainfall was frequent, while in the month of December, average daily air temperatures ranged from 12.1 22.9 C, and rainfall was irregular ( NOAA 2016 ) Water Conservation Area 2A Imagery and metadata collection f or the second part of this study were conducted over several specific experimental and control plots located within northeastern WCA 2A in the Greater Florida Everglades, incorporating the southern Palm Beach and northern Broward Count y line Florida USA ( Figure 4 3 ). Located 4.0 km west of Parkland, Florida, and 4.0 km south southwest of the L 39 L evee / Hillsboro Canal, the primary focal area where these experimental rehabilitation efforts are occurring lies within an region that occupies a rectangular area of 6.0 3.25 km [ 2,000 hectares ( ha) ] that extends southwest erly A n additional 200 ha region that is also part of the study is located southwest of the primary focal area (WGS84 Datum: 26 326737 80 369351 West) In the mid 2000s, the South Florida Water Manag ement District ( SFWMD ) established a series of six 250 250 m Cattail Habitat Improvement Project (CHIP) experimental plots and adjacent control s in WCA 2A Various treatment s were used to
115 clear the experimental CHIP plots of dense cattail ( Typha dominge nsis ) vegetation including herbicides, prescribed fire, and manual vegetation removal The experimental plots are tests of management practices to control dense cattail establishment resulting from anthropogenic alterations upstream of WCA 2A. The CHIP tr eatment plots and their adjacent controls are located in two groups of three for replication purposes. The northernmost group of three plots per kilogram (mg/kg ) The m levels 401 999 mg/kg. The southernmost group of control 400 mg/kg ( Newman 2016 ) Also created in the primary focal area were a pair of larger Active Marsh Improvement (AMI) plots having dimensions of 500 750 m The AMI plots were also cleared of vegetation to evaluate changes in ruderal community composition and spatial hydrological effects for comparison to Figure 4 4 ). The WCA 2A CHIP plots are inundate d having both submerged and emergent wetland vegetation possess an elevation of 6 .0 m ASL, and NOAA climatology data for the locality classifie d it as sub tropical with an average annual precipitation total of 145.5 cm per year ( NOAA 2016 ) Historical data showe d that for the month of June, average daily air temperatures range d from 23.2 32.7 C, and rainfall was frequent, while in the month of December, average daily air temperatures range d from 16.3 26 1 C, and rainfall wa s fairly infrequent ( NOAA 2016 )
116 Methods The initial ESP design was a box constructed from folded and riveted sheet aluminum with a hole cut in the bottom over which the optical payload lens was positioned ( Figure 4 5 ). The maiden ESP flights were conducted to examine the resultant imagery for effects from aircraft engine vibrations, ambient airflow buffeting, and to inspect the imagery metadata files for interference potentially generat ed by the manned aircraft engine electrical or communication systems. These proof of concept flights of the initial ESP design provided important information revealing that the ESP methodology was viable, and with refinements warranted further explorati on. A second generation ESP design possessing a larger internal volume to facilitate simultaneous use of multiple payload systems was created using computer aided design (CAD). The CAD diagrams served as input files to mill sheets of thicker aluminum allo y into ESP structural components via computer numerical control methods. The resulting second generation ESP was assembled with FAA approved hardware, and provided capacity for two complete optical payloads, a single lithium polymer ( LiPo ) system battery, and the ability to host alternative payloads as technology allowed ( Fig ure 4 6 ). To improve the success of obtaining imagery of the focal targets below the ESP affixed to a manned aircraft, a nadir oriented GoPro Hero video cam era was added to the second generation ESP payload in early 2014 (which was after the flights over ACNWR, but before the WCA 2A flights) which transmitted a wireless stream of RGB video to an Apple device inside the manned aircraft providing a situationa l awareness view of what the nadir oriented still frame sensors in the ESP were capturing. Used primarily for ensuring that the focal targets of interest were located within the field of
117 view of the still imagery being collected by the digital single lens reflex ( dSLR ) sensors in the ESP, the GoPro video was ultimately a useful aid for future ESP flights. In early spring 2015 a third generation ESP design having an even more voluminous capacity for payloads was built out of aluminum alloy and assembled w ith high strength welded seams for a reduction in ESP hardware and improved water resistance. A 3D printed tapered nosecone was designed and attached to help make the entire third generation ESP more aerodynamic ( Fig ure 4 7 ). Th e design of the third generation ESP incorporated details such as : 1) an identical aircraft mounting bolt configuration as the second generation ESP ; 2) a compressible neoprene watertight seal between the ESP body and lid ; 3) a sizeable glass pane on its v entral surface over which optical sensors could be arranged in countless nadir configurations ; and 4) features that simplified accessibility to the pod contents while the unit was affixed to an aircraft. This pod wa s the heaviest of the ESP models, and th erefore was best suited for operations aboard rotary wing aircraft having a wide entry/exit step mounting surface area to distribute the add itional mass The d ata collection flights for this dissertation chapter had been completed before the third generat ion pod was put into service, and it has only been affixed to a SFWMD rotary wing aircraft through 2017 ( Figure 4 8 ) Data collection flights for both parts of this dissertation chapter utilized the second generation ESP outfitted with a n commercial off the shelf (COTS) Olympus E 420 optical payload system (detailed in Appendix A ) that was designed by the University of Florida Unmanned Aircraft Systems Research Program (UFUASRP) for use primarily in the UF Nova 2.1 sUAS ( also detailed in Appendix A ). The d ata collection for the ACNWR sea turtle nesting crawl survey s was initially proposed to be
118 conducted with the UF Nova 2.1 sUAS over a smaller portion of the ACNWR ; however, the FAA was reluctant to issue a Certificate of Waiver or Authorization (COA) for the proposed low altitude sUAS flights due to : 1) the proximity of the study site to housing along the beach ; 2) potential gathering of onlookers o n the beach to observe the sUAS fl ight transect s ; and 3) the vicinity of the Melbourne International Airport (KMLB) to the study site With the FAA hesitan t to issue a COA and the window of opportunity to capture aerial imagery of sea turtle nesting crawls during 2013 coming to a close, d ata collection for the sea turtle nesting survey over ACNWR were conducted using an ESP attached to a Cessna 172M Skyhawk fixed wing aircraft on 23 July 2013 from the Valkaria Airport in Valkaria, Florida. The flight took off at 0627 hours (shortly afte r civil twilight) and began beach transects at 0631 hours. Four complete transects of the 21.0 km of beach were imaged, and four partial passes of concentrated areas within the 21.0 km stretch were also imaged by 0801 hours. By this time, human foot traf fic on the beach was picking up, and nesting tracks left in the sand by sea turtles were beginning to be inadvertently disturbed by the human traffic. The aircraft landed safely back at Valkaria Airport a short time later Average altitude of the beach s urvey transect s was 196.9 m ASL with an aver age ground speed of 35.8 meters per second ( m/sec ) Flights for the AMI and CHIP plots in WCA 2A took place 13 15 August 2014 with the ESP affixed to a SFWMD Aviation Unit Bell 407 helicopter based out of West Palm Beach International Airport (KPBI). On 13 August, the aircraft departed KPBI at 0906 hours and conducted imagery transects over both AMI and all primary focal area CHIP plots and their controls with
119 four horizontal (east to west, or west to east) rence only The aircraft arrived back at KPBI at 1122 hours collecting 2, 271 images total with an average altitude of 203.1 m ASL and an average ground speed of 21.2 m/sec The following morning, 14 August 2014, the aircraft departed KPBI at 0844 hours and conducted imagery transects over all of the CHIP plots and their controls each and p ach. The aircraft arrived safely back at KPBI at 1042 hours, collecting 2, 026 images total with an average altitude of 194.7 m ASL, and an average ground speed of 19.6 m/sec On 15 August, the SFWMD Bell 407 helicopter with ESP attached departed KPBI at 0839 hours, and conducted vertical (north to south or south to north) transects over the two AMI plots Nine transects were made horizon before t he flight arrived back at KPBI at 1039 hours, collecting 1,505 total images with an average altitude of 200.3 m ASL and an average ground speed of 21.9 m/sec Data post processing for b oth parts of this study were conducted using the methodology delineated in Appendix D The July 2013 flight over the 2 1.0 km focal stretch of ACNWR beach for sea turtle nesting crawl s urvey was broken down into smaller and manageab le chunks ( each chunk for areas that had the most imagery overlap occurring over the sandy beach focal target area Similarly, e ach
120 of the primary focal CHIP experimental and adjacent control plots were post processed individually by pooling imagery collected during the three consecutive morning flights conducted with the second generation ESP affixed to a SFWMD Bell 407 helicopter and the AMI plots w ere processed independently with imagery amassed from the days in which those pl ots were included as focal target area s Results Since the inaugural ESP flight test in December 2009, > 40 total ESP flights have been conducted. The second generation ESP design has been the workhorse for testing natural resource based applications of t he methodology ESP flight planning procedures novel payload testing, and mission execution techniques. The engineering of the second generation ESP and its mounting configuration incorporated details which necessitated only slight modifications to an ex isting external component of the manned aircraft by an FAA authorized mechanic, e.g.: drilling several small holes for mounting bolt attachment into a fueling, inspection, or entry/exit step, etc. Once the ESP was affixed on an individual airframe, an FAA field approval for the airframe modification could then be requested The second generation ESP design has received FAA authorization as an approved aircraft modification when affixed to five separate manned airframes: two Cessna 172 fixed wing aircraft a Bell 206 and a pair of Bell 407 rotary wing aircraft ( Figure 4 9 ). For the sea turtle nesting crawl survey flight over ACNWR successful generation of imagery post products for portions of the 21.0 km target area that were a ble to be aligned and composed into orthophotomosaics were indications that suitable flight parameters were indeed able to be attained ( Figure 4 10 ). As mentioned, t he sea turtle nesting flight was accomplished before a GoPro H ero video camera was added to the
121 ESP payload to provide a situational awareness view in the cockpit of what the nadir oriented still frame sensor in the ESP w as capturing. Although the manned pilot and co pilot tried their very best to keep the ESP over the target area of sandy beach from the waterline to the base of the sand dunes portions of the parallel flight transects recorded imagery over the surf zone, or locations west of the dune s which hindered the construction of a single seamless orthophotom osaic of the 21.0 km of sandy beach ( Figure 4 11 ). The ability to distinguish fresh sea turtle crawls along the beach from older crawls that had been traversed by an all terrain vehicle (ATV) indicating that the crawl had been p reviously counted the day before (or even earlier) was also used a measure of success ( Figure 4 12 ) In Figure 4 12 t he narrower crawl with an alternating gait was deposited by a loggerhead, while a g reen turtle left the other crawl with a noticeable tail drag. Further c lassifying each fresh sea turtle crawl as a successful nesting or a false nesting crawl was another form of ESP evaluation used in this study ( Figure 4 13 ) In Figure 4 13 t he top crawl pattern was a false nesting crawl by a green turtle, and the other crawl was a successful nesting crawl by another green turtle. Determination of individual sea turtle species as revealed by the cr awl deposited in the sand was utilized as a final assessment of the ESP methodology for this specific application ( Figure 4 14 ) As indica ted in the previous two figures and in Figure 4 14 which shows two fresh successful nesting crawls by green turtles, sea turtle species identification via crawl examination was straightforward using the imagery collected with the ESP technique. For the ESP flights over the CHIP and AMI vegetation plots, the primary an alysis was to post process each experimental plot individually, and identify the plots that were
122 able to produce nearly complete and seamless orthophotomosaics as indicat ors that adequate flight parameters and ESP flight methodologies were achieved for thi s specific application ( Figure 4 15 ). The distribution of the imagery exposures in remote sensing is extremely critical, and maintaining altitude, airspeed, course, and attitude from a moving aircraft without the aid of an autop ilot unit can be quite challenging for even the most experienced manned pilots It is imperative that ample imagery overlap ( 60% endlap and sidelap) is achieved to attain full target area coverage if that is the end product needed to answer the scientific question ( Figure 4 16 ). Creating an orthophotomosaic from imagery that lacks suff icient overlap across the entire scene will leave areas where the post processing software is unable to match enough tie points to generate a full seamless orthophoto mosaic ( Figure 4 17 ). However, a tremendous amount of informat ion can still be garnered from the remaining intact portions of such products, or from individual images that have been orthogeorectified Discussion F urnishing a manned aircraft able to conduct slow airspeed and low altitude manned visual surveys with an ESP containing optical payload system s capable of producing directly georeferenced near nadir imagery permits multiple data sets of the target area to be obtained during a single flight. By limiting the need for substantial physical alterations to the man ned aircraft or considerable changes to existing manned survey flight plan methodologies, the benefits of utilizing an ESP affixed to a manned aircraft outweigh potential drawbacks. A well piloted ESP equipped aircraft can be systematically flown over the focal area at a predetermined altitude, air speed, and along parallel flight paths based on ESP payload specifics, aircraft performance characteristics and the target item of interest
123 An advantageous reason for using a sUAS payload in an ESP to conduct a erial surveys is that when outfitted with direct georeferencing payloads, the digital imagery and associated metadata collect ed can be catalogued, archived, and stored in perpetuity as a photographic record of the flight. This feature permits a multitude of qualitative and quantitative analyses to be conducted in the future, especially temporal based assessments of alterations that may be attributed to global climate change, natural disaster impacts, or other salient events in natural resources and environ mental disciplines. Unlike traditional visual observer manned aerial surveys where identification and enumeration of targets must be conducted immediately and quickly from the air as the aircraft makes transects over, or circumnavigate s a target area, ESP s containing sensors designed for sUAS can typically collect raw data that can be analyzed at any occasion after a flight concludes. Typical visual observer aerial surveys do not afford much time for observers to resolve potential doubts or reassess many areas while a flight is in progress, whereas most ESP sensor data can be examined repeatedly Rare or difficult to identify targets are photographically documented and can be evaluated by numerous subject matter experts for their opinions These benefits are especially valuable as new computer software for imagery post processing emerges, innovative computer based feature recognition algorithms become available, and as statisticians develop novel methods to assist researchers in moving from providing count s of focal targets to generating estimates having testable bounds. In situations where target areas for aerial data collection are separated by extensive distances, or isolated by terrain that is difficult to traverse via ground or boat
124 (such as in the Eve rglades), the ability of a manned aircraft to ferry an ESP from a specific survey area to the next is considerably more time efficient than conducting surveys reliant on surface based transportation methodologies. Consequently, a much greater sampling are a containing targets for aerial data collection can be accomplished during an operational day. Equipping manned flights for surveys with an ESP using slow airspeed and low altitude methodologies still put pilots and their manned aircraft in potentially ri sky situations. However, with the currently rapid advances in optical sensor and payload technologies predicted to continue, soon optical sensors should provide improved resolution and payload systems ought to become quicker, permitting ESP flights by man ned aircraft to be conducted at higher flight altitudes and at greater airspeeds, therefore increasing the overall safety of such missions. Williams et al. (2015 ) use d digital video acquisition from four video cameras simultaneously affixed to a fixed wing aircraft as part of their data collection of mid Atlantic Outer Continental Shelf fauna, and their study highlighted several of the s ame benefits that are mentioned in this study regarding the gains that optical sensors provide as data collection methods over traditional manned visual aerial surveys alone. Using the ESP and methodology offers advantages in that the use of a high a ccuracy GPS/INS in synchronization with still frame optical sensors in a nadir orientation are able to generate directly georeferenced imagery products during post processing. In addition, thi s approach uses increasingly higher resolution optical sensors that improve the ability of observers to routinely identify individual focal targets to the species level. While conducting ongoing ESP research, there has been interest in generating a quantitative based preflight software program to accompany an ESP that could enable
125 natural resources or environmental professionals to input basic details of the ESP optical payloads, the desired resultant imagery end products, and the aircraft hosting the ESP, whereby the software w ould generate an optimum altitude, flight path spacing for transects, and ideal ground speed for successfully capturing the desired imagery and associated metadata from the sensors contained within the ESP. Theoretically, effective development of preplanning software could permit a researcher to design and upload a flight plan into a GPS unit for a manned aircraft pilot to execute; eliminating the necessity of having a natural resources professional onboard the aircraft, and thereby reducing the number of souls at risk being airborne at slow airs peeds and low altitudes. For certain natural resource applications, e.g.: documenting animal behavior, rudimentary presence/absence determinations, or in cases where directly georeferenced imagery is not required to generate the desired end product, etc., the video data collected by a GoPro camera is generally sufficient for such purposes (e.g., Anderson et al. 2014 Mulero Pzmny et al. 2014 Bevan et al. 2015 Pomeroy et al. 2015 van Ge mert et al. 2015 ) As pixel resolution of optical sensors and number of frames captured per s ec continu e to advance with COTS digital video camera technology, the ability to obtain individual high quality still frame images from video recordings have gr eatly improved. Perhaps the greatest drawback to aerial video data collection is that unless a network of surveyed ground control points appear within the resultant video, retroactively attempting to georeference target items from recorded video can be a difficult and time intensive task (e.g., Jones IV et al. 2006 Wilkinson 2007 Eugster and Nebiker 2008 Sarda Palomera et al. 2012 )
126 It is important to emphasize that seamless orthophotomosaics are not necessarily the only end product of al l aerial imagery data collection efforts using an ESP or a UAS. A myriad of applications exist where smaller sampling units within orthorectified post processed raw imagery can provide appropriate data needed to address scientific questions Therefore us ing directly georeferenced imagery collected with an ESP has the potential to improve the accuracy of the answers to the oft asked questions: How much is there? Beyond providing a valuable tool for wildlife conservation, the ins ights gained during this study can be extended to the conservation and management of other species, thus increasing the value of the ESP payloads as data collection tools for research in other natural resources and environmental disciplines While the cost s for manned aircraft flight time are not inexpensive, a widespread misconception exists that the fielding of sUAS is considerably cheaper than using manned aircraft ( Hardin and Hardin 2010 Hugenholtz et al. 2013 Wing et al. 2013 ) The purchase prices of sUAS continue to drop, however when the total ex pense s of using the technology for scientific applications are compiled, most novice sUAS users soon discover that substantial investments beyond the cost of the sUAS are required to ultimately obtain data fit to answer scientific questions Additional co sts include e.g. : 1. FAA sUAS aircraft registration 2. Repair and replacement sUAS parts 3. Maintenance of airframes and payloads 4. Multiple payload sensor packages 5. Pilot in command (PIC) training and airframe practice 6. FAA sUAS PIC testing and certification 7. Maintena nce of PIC currency and proficiency 8. Obtaining landowner permission and other permits for flight operations at the selected field site 9. Compliance with institutional, local, state, and federal provisions
127 10. Payload testing, adjustment, and calibration 11. Wages of the PIC and the flight team 12. Transportation of the flight crew and sUAS to and from the field site 13. Miscellaneous other expenses The list does not even address an y data post processing, which can be substantial investments in time and money. Ultimately, to achieve sUAS data collection costs that are equivalent to or cheaper than those incurred using a manned aircraft necessitates routine sUAS flights conducted by an experienced flight crew. Sporadic use of a sUAS is generally fiscally inefficient. Visual d ata collection techniques using ground aerial or satellite based methods are all subject to observer biases, variations in sightings, and detection probability concerns that must be accounted for when making counts or assessments of focal targets (e.g., Link and Sauer 1997 Williams et al. 2002 Alldredg e et al. 2006 Walsh et al. 2011 Cook 2013 Beaver et al. 2014 ) Whether addressing presence/a bsence, abundance, density, etc. of targets, it can be difficult to accurately state the true number of target items in uncontrolled conditions when even slight variations in data collection methodologies or techniques can alter reported values ( Ralph and Scott 1981 Temple and Wiens 1989 ) Therefore, statistical approaches involving many independent observer s providing counts of detectable focal targets within an area of known size permits the opportunity to generate considerably more accurate estimates of targets with testable bounds. Planning to continue to design, develop, and test new methods of collectin g aerial data safer, more accurately, and with the natural resource the UFUASRP will pursue further manned aircraft options as well. T he ESP offers a novel approach that has worldwide applicability for gathering high resolution imag ery
128 and metadata that cannot be affordably obtained from satellite imagery or in situations where using a sUAS is impossible, impractical, or illegal The use of ESP methods have the potential to improve aerial data collection and produce more accurate e stimates of target items that can assist decision makers by generating more informed choices. While both convenience sampling and reliance on index values are simplified means for generating data, Anderson (2001 ) reminds researchers that valid inferences about target items are only dependable when data are a ppropriately collected and measures of their precision can be reliably justified. Due primarily to the attention that civilian sUAS and their payloads have received over the last decade, there are some who feel that the technology has already reached matur ity; however, there is still much to learn about sUAS, their sensor payloads, and appropriately utilizing them as tools for aerial data collection in support of answering scientific based questions. By enacting rules specifically directed towards sUAS in fall 2016, the FAA has made advances that will enable increased testing of sUAS and their payloads in the US. In the meantime, the small, yet affordable, high resolution directly georeferenced and calibrated optical payloads currently available for use in an ESP affixed to a manned aircraft appears to be a viable alternative method for collecting aerial data for many applications ; especially for natural resources and environmental research
129 Figure 4 1 The location of the July 2013 external sensor pod study site over Archie Carr National Wildlife Refuge, Brevard County, Florida, USA. The red box encompasses the 21 .0 kilometers of sandy beach that was the focal area of sea turtle nesting crawls for this study. GoogleEarth
130 Figure 4 2 The 21 .0 kilometer stretch of beach that was aerially imaged in July 2013 with the external sensor pod over the Archie Carr National Wildlife Refuge, Brevard County, Florida, USA. The yellow star shows the proximity of Melbourne In ternational Airport to the study area. GoogleEarth
131 Figure 4 3 The location of the active restoration project sites in Water Conservation Area 2A of the Greater Florida Everglades, Palm Beach and Broward Counties, Florida, USA. The red polygon encompasses the primary study areas for the Cattail Habitat Improvement Project and the Active Marsh Improvement sites. Area above the green line is in Palm Beach County, while area below the green line is in Broward County. GoogleEarth
132 Figure 4 4 The active restoration project sites in Water Conservation Area 2A of the Greater Florida Everglades, Palm Beach and Br ow ard Counties, Florida, USA. The primary Cattail Habitat Improvement Project (CHIP) experimental plots are outlined in solid white, and the ir control plots are outlined in broken white lines. Three control short distance southwest of this image. The two A ctive M arsh I mprovement (AMI) experimental plots are outlined in red. GoogleEarth
133 Figure 4 5 The initial external sensor pod. Constructed from folded and riveted sheet aluminum, the initial external sensor pod had a hole cut in the bottom over which the optical payload lens was positioned. The power switch es and global positioning system/inertial navigation system antenna u tilized the holes in the lid. M.A. Burgess
134 Figure 4 6 The second generation external sensor pod. This design possessed a larger internal volume to facilitate sim ultaneous use of multiple payload systems. With the added volume, the pod was fabricated out of milled sheets of thicker aluminum alloy to support the additional mass. M.A. Burgess
135 Figure 4 7 The third generation external sensor pod. Ev en more voluminous, this external sensor pod was also built out of aluminum alloy, but was assembled with high strength welded seams for a reduction in hardware and improved water resistance. A thre e dimensional printed tapered nosecone was custom designe d and attached to help make the entire pod more aerodynamic M.A. Burgess
136 Figure 4 8 The third generation external sensor pod affixed to a Bell 407. The third generation pod used the same mounting plate configuration as the second gene ration pod. In addition, the third generation pod featured a much larger glass window over which to position multiple nadir oriented optical sensors. M.A. Burgess M.A. Burgess
137 C D Figure 4 9 The second generation external sensor pod affixed to m ultiple manned aircraft types. A) A Cessna 172M Skyhawk fixed wing B) A Bell 206B 3 JetRanger III rotary wing C) A Bell 407 rotary wing D) A Cessna 172P Skyhawk fixed wing A F ederal A viation A dministration approval for aircraft modification was received for each of these sensor pod attachments. M.A. Burgess M.A. Burge ss M.A. Burgess W. Auer B A
138 Figure 4 1 0 Image of the Cessna 172M Skyhawk equipped with the second generation external sensor pod flying a linear transect over the photographer on 23 July 2013 at Archie Carr National Wildlife Refuge, Brevard County, Florida, USA. M.A. Burgess
139 Figure 4 1 1 A resulting orthophotomosaic of selected imagery from a portion of a flight of the second generation external sensor pod attached to a Cessna 17 2M Skyhawk on 23 July 2013 over Archie Carr National Wildlife Refuge, Brevard County, Florida, USA. This orthophotomosaic covers kilometers 14.5 15.0 (delineated from north to south) of the Archie Carr National Wildlife Refuge. When this orthophotomosai c is enlarged the sea turtle crawls deposited in the sand are remarkably distinct. M.A. Burgess
140 Figure 4 1 2 Enlarged portion of an orthophotomosaic generated from selected imagery collected during the flight of the second generation exte rnal sensor pod attached to a Cessna 172M Skyhawk on 23 July 2013 at kilometer 14.7 of Archie Carr National Wildlife Refuge, Brevard County, Florida, USA. This enlarged uppermost portion of the orthophotomosaic in Figure 4 1 1 shows all terrain vehicle tracks over sea turtle crawls that have already been counted, as well as two fresh nesting crawls that crossed over each other near the bottom part of this image. The narrower crawl with an alternating gait was deposited by a loggerhead, while a green turtle left the other crawl with a noticeable tail drag. Note the differences in the crawl patterns. M.A. Burgess
141 Figure 4 1 3 Enlarged portion of an orthophotomosaic generated from selected imagery collected dur ing the flight of the second generation external sensor pod attached to a Cessna 172M Skyhawk on 23 July 2013 at kilometer 14.8 of Archie Carr National Wildlife Refuge, Brevard County, Florida, USA. This enlarged portion of the middle part of the orthop hotomosaic in Figure 4 1 1 shows two fresh sea turtle nesting crawls that crossed over each other near the middle of this image. The top crawl was a false nesting crawl from a green turtle, and the other a successful nesting c rawl by another green turtle. M.A. Burgess
142 Figure 4 1 4 Enlarged portion of an orthophotomosaic generated from selected imagery collected during the flight of the second generation external sensor pod attached to a Cessna 172M Skyhawk on 23 July 2013 at kilometer 14.9 of Archie Carr National Wildlife Refuge, Brevard County, Florida, USA. This enlarged portion from the lower part of the orthophotomosaic in Figure 4 1 1 shows two fresh sea turtle nesting crawls ; both successful by green turtles. M.A. Burgess
143 Figure 4 1 5 The paths of three separate flights with the second generation external sensor pod attached to a Bell 407 helicopter in August 2014 over the active restoration projects in WCA 2A, Palm Beach and Broward Counties, Florida, USA. The three flights were to attempt different data collection techniques and assess the methodologies based on the resulting data products Blue dots are camera exposures from the flight on 13 August ora nge dots are camera exposures from the flight on 14 August, and green dots are camera exposures from the flight on 15 August. GoogleEarth
144 Figure 4 1 6 The two dimensional locations of imagery exposures from three separate flights with the second generation external sensor pod attached to a Bell 407 helicopter in August 2014 over in WCA 2A, Broward Count y Florida, USA. The three flights all made overhead Blue dots are camer a exposures from the flight on 13 August, orange dots are camera exposures from the flight on 14 August, and green dots are camera exposures from the flight on 15 August. GoogleEarth
145 Figure 4 1 7 The resulting orthophotomosaic of imagery from three separate flights with the second generation external sensor pod attached to a Bell 407 helicopter in August 2014 over experimental and control plots in WCA 2A, Broward Count y Florida, USA. The generation of complete orthophotomosaics requires sufficient endlap and sidelap between transects to achieve full target area coverage. Gaps due to insufficient imagery plot in upper right of this image. M.A. Burgess
146 CHAPTER 5 CO NCLUSIONS Throughout the last decade, the interest and desire to use small unmanned aircraft systems (sUAS) and their payload systems for civilian uses has grown exponentially. In what is now predicted to be a multi billion US$ industry around the world, sUAS as payload platforms for a seemingly endless number of applications only continues to grow. H aving been immersed in the technology before and during this recent upsurge, it is safe to say that sUAS are not going away anytime in the near future. Work ing with sUAS and their payloads, the remote sensing industry and others have already seen major shifts in how data are acqui red on a routine basis For the natural resource based user, both sUAS and their high resolution yet affordable payloads present many advantageous features that have the potential to supplement existing data collection methodologies and techniques in positive ways. T he University of Florida (UF) Unmanned Aircraft Systems Research Program (UFUASRP) will continue t o provide valuable aerial data collection tool s through further research and development for the natural resource based user while remaining mindful of the limited fiscal resources available, and diverse knowledge base that environmental research professionals and scientist s possess Within the published literature, t here has been a noticeable increase in the number of manuscripts highlighting sUAS as a tool for natural resources and environmental professionals as an aerial survey platform A majority of th e se articles have focused on the sUAS rather than data or results from conducting quantitative studies using said aircraft and payload s Without doubt, sUAS are data collection tool s that grab significant attention, and the execution of proof of concept applications are
147 essential to the development of any novel technique or methodology. However, it is imperative to stress that the unmanned aircraft ( UA ) is just a platform or an avenue for positioning a sensor or a sui te of sensors over a specific target area to collect aerial data. The real value of sUAS as a scientific tool lies within the quality and methodologies used to obtain and post process the data once it has been collected. We anticipated that soon research manuscripts in which sUAS w ere tools used to obtain scientific data should begin to H ey, look what we can see from a U sing a n suitable sUAS airframe outfitted with a calibrated data collection These papers will h ighlight the results garnered from the ae rial data collected rather than focusing so much on the data collection device itself The vast utility of sUAS as aerial data collection tool s in the fields of natural resources, wildlife, forestry, fisheries, ecology, conservation, management, etc. is particularly exciting; especially for researchers who have interdisciplinary expertise in appropriately combining the critical components of the various sciences with the capabilities of the emerging technology Pygmy Rabbit ( Brachylagus idahoensis ) Habita t Selection Inferences Determined by Normalized Difference Vegetative Index Computations of Aerial Imagery Products Generated from Data Collected by a Small Unmanned Aircraft System Based on the results of this study conducted at two separate sagebrush ste ppe field sites during two consecutive summers in Idaho, we determined that computation of normalized difference vegetative indices (NDVI) of individual sagebrush plants could be used to successfully identify plants having a higher likelihood of harboring active pygmy rabbits within a post processed scene. Using this methodology, a sUAS with a payload capable of generating directly georectified NDVI data valu es could theoretically be flown over an area of sagebrush steppe that has never been ground truthed or was ground
148 truthed some time ago. Through data post processing and NDVI calculations, field on the then be directed to specific sagebrush plant locations with the highest likelihood of harboring signs of recent pygmy rabbit activity via global positioning systems (GPSs) The technique could save substantial time for BOTG surveyors from having to systematically traverse large areas of sagebrush steppe landscape looking for signs of recent pygmy rabbit acti vity by guiding the surveyors immediately to plants having the highest NDVI values within a post processed scene. Generating Estimates of Nesting American White Pelicans ( Pelecanus erythrorhynchos ) on Spoil Islands in Minidoka National Wildlife Refuge (Ida ho) from Aerial Imagery Products Produced from Data Collected by a Small Unmanned Aircraft System Olympus E 420 optical payload system was able to successfully conduct parallel transects at 125 meters (m) above gr ound level (AGL) over three spoil islands in the Minidoka National Wildlife Refuge that harbor colonies of nesting American White Pelicans (AWP) without disturbing them Using the airframe and payload combination selected, two dimensional (2D) digital ima gery and metadata captured during the flight was able to post processed into high resolution three dimensional (3D) imagery end products which facilitated manual counts of AWP on each of the islands by means of computer software Using crowdsourcing metho ds of obtaining AWP counts, the pooled count data for each island w ere able to be statistically tested to generate estimates of total AWP individuals detected on each of the three islands. When the estimates of AWPs on the islands were compared to data pr ovided by Idaho Department of Fish and Game (IDFG) biologists using BOTG techniques, it was realized that we were developing estimates of individual AWP birds
149 using the post processed imagery products while the IDFG was generating total nest counts with B OTG methodolog ies Ultimately enumerating two different targets made direct comparisons inappropriate; however, upon further review of the sUAS obtained data products it appears that using the sUAS data collection method s which w ere unobtrusive to the nes ting avifauna, w ere time and labor efficient, and could be conducted on a regular sampling schedule in the future to assess nesting turnover rates through the entire ty of the reproductive season, was able to generate reliable estimates that could be used f or future AWP nesting surveys. Use of High Resolution Optical Payloads Designed for Small Unmanned Aircraft Systems as an Innovative Approach for Aerial Data Collection f rom Manned Aircraft Flights While the use of sUAS and their payload systems for civil applications have grown exponentially during the last decade, there are still situations where using sUAS data collection methods may be impossible, impractical, or illegal T herefore utilizing manned aircraft equipped with a small external sensor pod (E SP) containing high resolution optical payloads and sensors designed for sUAS applications may be a better choice logistically, fiscally, and legally for obtaining aerial imagery and metadata for certain applications Circumstances such as when sampling s ites have extensive distances or inhospitable terrain for ground or water based transportation located between them, or when a targeted area of study is large in size and collecting aerial data must be accomplished in a short period of time, or if sUAS us e is prohibited due to proximity of property, people, or airports, the use of an ESP affixed to a manned aircraft can often be a viable solution. The ESP technique was conducted over sea turtle nesting crawls on a 21.0 kilometer (km) stretch of beach in 2 013, and was tested again in 2014 over aquatic vegetation targets in the Greater Florida Everglades for additional
150 feasibility analyses. In both scientific trial applications, specific refinements of the methodology were identified, and modifications to f uture ESP flight plans and techniques established. At the conclusion of the ESP trials, we determined that the methodology appears to be a feasible aerial data collection alternative to coarse resolution satellite imagery data, traditional double observer manned visual surveys, and for study sites where sUAS flights are impossible, impractical, or illegal.
15 1 APPENDIX A THE HISTORY OF THE U NIVERSITY OF FLORIDA UNMANNED AIRCRAFT SYSTEMS RESEARCH PRO GRAM (UFUASRP) : 1999 2017 In the late 1990s, researchers wit hin three independent departments at the University of Florida ( UF ) were actively addressing key components of what would ultimately become a pioneering academic based unmanned aircraft systems ( UAS ) research program in the United States ( US ) Wildlife bi ologists and ecologists in UFs on the borne imagery to gather data pertaining to a multitude of natural resource based q uestions. Researchers within the UF Department of Mechanical and Aerospace Engineering (MAE) were developing novel micro aerial vehicle (MAV) airframes out of composite materials with wingspans as small as 10 centimeters (cm), total masses of approximatel y ( ) 40 g rams (g) flight endurance of up to 15 minutes (min), and having both manual and autonomous control and imagery capturing capabilities. Meanwhile, Geomatics (GEOM) Program professors in the UF Department of Civil and Coastal Engineering (the GEOM P rogram is now located within the School of Forest Resources and Conservation) were investigating ways to improve and augment natural resource focused remote sensing data obtained from orbiting satellites, manned aerial flights, and BOTG methodologies. In J une 1999, two UF WEC professors attended an afternoon seminar on the UF campus about a prototype 1.5 m eter (m) wingspan, 4.3 kilogram ( kg ) maximum takeoff mass, nitromethane powered pusher configuration FoldBat UAS that the MLB Company was in the process of developing. After attending the seminar, the WEC professors expressed interest to each other about the possibilities of perhaps trying a
152 FoldBat system as a method of imagery data collection for scientific questions within natural resources if they c ould procur 35,000), and secure funding to support a Master of Science (M S ) level graduate student for several y ea rs who would be interested in taking on such a novel project. Concurrently in the UF MAE MAV Laboratory, a team of graduate and undergraduate students were preparing for competing in their first national MAV competition; an event in which they would earn top honors, and retain that title over the next eight y ea rs. Also at that time, another UF WEC research scientis t with a team of biological field technicians were annually enumerating predominately long legged wading bird species (Ciconiiformes and their allies) colonially nesting on selected tree islands in the Greater Florida Everglades via BOTG and dual observer manned aircraft survey methodologies. Information spread quickly in UF WEC that an effort was coming together to test the ability of using a small unmanned aircraft system ( sUAS ) as a natural resource purposed data collection platform. The WEC avian ecol ogist expressed interested in experimenting with the technology to collect imagery over several of the nesting bird colonies to compare, contrast, and potentially supplement existing enumeration techniques that were in use The MLB Company FoldBat In the fall of 2000, UF WEC researchers contacted the MLB Company and expressed their desire for purchasing a prototype FoldBat system as a test platform for acquisition of low altitude aerial imagery over several areas of Florida that contained natural resour ce targets of interest. In June 2001, a prototype FoldBat system arrived at UF from the MLB Company in California. The aircraft fuselage was constructed of Kevlar carbon fiber, balsa wood, and aluminum components, and had its unique
153 Dacron covered a ccordion folding main wingset and tail surfaces for which the platform derived its named ( Jones IV 2003 ) Additional components also arrived including the gr ound control station ( GCS ) laptop computer containing proprietary MLB Company software featuring a moving basemap, telemetry, and up to 15 pr eprogrammed waypoint capability a directional 2.4 gigahertz (GHz) video downlink antenna, and other essentials su ch as the standard 8 channel 72 megahertz (MHz) pulse code modulation remote control (RC) aircraft radio controller used for traditional hand launch takeoffs, manual flight ability, and belly sliding landings, interface cabling, and video recording equipme nt ( Jones IV 2003 Jones IV et al. 2006 ) The FoldBat was natively furnished with two underwing senso r pods, each containing a nadir oriented ( downward oriented ) complementary metal oxide semiconductor (CMOS) analog interlaced video chip camera with 330 horizontal line resolution and a fixed focal length 50 millimeter ( mm ) lens; one sensor pod capturing i magery in the visible range of the electromagnetic (EM) spectrum [the visible range is often referred to by the red green blue (RGB) additive color model in which various quantities of red, green, and blue wavelength light are added together to form a broa d range of colors that the normal human eye is capable of viewing ], and the second sensor pod capturing imagery in the near infrared (NIR) wavelength range of the EM spectrum. The interlaced video imagery feeds were downlinked with a pair of 2.4 GHz trans mitters aboard the aircraft to a directional antenna located at the GCS where the feeds were recorded onto digital Sony GV D800 Hi8 video cassette recorders in the field for future viewing and analysis back in the laboratory ( Jones IV 2003 Jones IV et al. 2006 )
154 At that time, interlaced video format was the industry standard for affordable high speed movi e recording. By capturing 50% of an image frame (every other horizontal line) at a rate of 50 frames per second (f/s ec ), a 0.02 second (s ec ) time delay occurred between each half frame recorded using the interlaced format. Depending on the altitude and g round speed of the interlaced imagery sensor at the time of half frame captures, a noticeable gap ( or shift ) within the sequential captures could be observed when the video stream was reviewed frame by frame. This drawback made extracting photographically clear still imagery from the interlaced video format recordings problematic. Most of the video captured with the CMOS chip cameras in the FoldBat underwing sensor pods produced blurry still imagery and difficulties in reliably capturing images of target s with a physical size smaller than the gap induced by the movement of the imaging sensor between half frame exposures. A UF WEC professor suggested that professors in the UF GEOM Program be consulted about suggestions for improving the optical payload an d remote sensing data collected by the FoldBat platform. After a brief meeting among the group in mid 2002, a unanimous decision was made by the UF UAS collaboration to move away from the native interlaced video format sensors included with the FoldBat and transition to newly emerging, and increasingly more affordable, progressive scan video sensors which was capable of taking a full frame image every 0.033 s ec In late 1999, Canon introduced the ELURA2 digital video camcorder to the commercial off th e shelf (COTS) market as a small and affordable digital progressive scan video camcorders. The ELURA2 featured 525 horizontal line RGB video captured with a 6.35 square millimeter (mm 2 ) progressive scan charge coupled device
155 (CCD) imaging sensor having 0 .36 megapixel (MP) resolution, its own battery power supply, and onboard MiniDV 8 mm videotape storage in a relatively small physical and lightweight (460 g) package. The ELURA2 camcorder was inserted into the FoldBat fuselage; however, the added mass introduced by the larger payload required that the two external underwing pods be removed, and the nitromethane fuel tank only be filled to a maximum of halfway before flight to keep the aircraft below the critical takeoff mass. Fuel reduction limited the flight duration, and the high wing loading induced by the ratio of the relatively small main wing area to the substantial overall mass of the aircraft contributed to general flight instability and many failed takeoffs. To further reduce mass, the downlin k video modems were removed from the airframe, and the progressive scan imagery was stored directly to MiniDV 8 mm video cassette tapes onboard the aircraft. Overall, the decision to move away from interlaced video to the progressive scan format made cap turing clearer still images from video a reality, and therefore target identification contained within the still images was considerably improved From mid 2001 through spring 2003, the FoldBat flew > 30 total flights for the UF UAS collaboration at altit udes from 100 150 m above ground level (AGL) and .0 16 .0 meters per second (m/s ec ) to investigate the strengths and weaknesses of using the platform and payloads over various natural resource based targets of interest within the S tate of Florida. At Go odwin Waterfowl Management Area, a wetland impoundment located near Melbourne, the FoldBat was flown several times to collect imagery over a series of nesting wading bird species [W hite Ibis ( Eudocimus albus ), egrets ( Egretta sp.), and Wood Stork ( Mycteri a americana )] ( Jones IV 2003 )
156 Other missions using the FoldBat included flights to collect imagery over: 1) agricultural land southwest of Archer, for both planted crops and grazing livestock; 2) a n island used extensively for nesting by birds at Seahorse Key [ part of the Lower Suwannee and Cedar Keys National Wildlife Refuge (LSCKNWR ) ] for additional avifauna imagery; 3) a mangrove estuary (Pine Island) for vegetation classification purposes; 4) a eutrophic freshwater lake for partially submerged American alligator ( Alligator mississippiensis ) decoys and emergent wetland vegetation at Lake Alice; and 5) a Florida manatee ( Trichechus manatus latirostris ) warm water winter aggregation site north of Apollo Beach ( Jones IV 2003 Jones IV et al. 2006 ) ( H. F. Perciv al University of Florida, pers onal comm unication ). The FoldBat skid style landings on its fuselage belly ( Jones IV 2003 ) Finding a suitable landing site in the field meeting these criteria were a difficult task, and on average only 25% of takeoff launches were successful with the FoldBat ; a low percentage for an unmanned aircraft (UA) that did not ship with many spare parts, and all major structural repairs required sending the unit back to manufacturer ( Jones IV 2003 ) From personal conversation with WEC researchers, the prototype FoldBat may have accrued more flight time aboard a FedEx manned cargo aircraft making trips back and forth between Florida and California for repairs than it did flying operational missions ( H. F. Percival pers onal comm unication ). The operation of the FoldBat required a two person flight team: an individual to hand launch the aircraft (which required significant arm strength and general athleticism), who then observed the aircraft in flight while monitoring the telemetry on the GCS, and then manually landed the aircraft; while a se cond person
157 had to aim the directional antenna panels toward the UA at all times to maintain data and communication linkages between the aircraft and the GCS. An added difficulty encountered when fielding the FoldBat was its nitromethane fueled small eng ine which was ultimately undersized for the aircraft due to the modifications of the optical payload, was messy to refuel, and required a considerable amount of fine adjustment via needle valve tinkering due to it being tuned to run smoothly for ambient co nditions encountered at the MLB Company in California, but operational flight environments in Florida were substantially different (e.g., higher temperatures, humidity, and air density, etc.) ( Jones IV 2003 Jones IV et al. 2006 ) It was an engine failure during a Florida manatee survey mission near Apollo Beach in spring 2003 which ultimately landed the prototype FoldBat into Tampa Bay; its final flight for the UF UAS collaboration ( Jones IV et al. 2006 ) To assess whether a wildlife biologist without any pri or RC flight training could quickly learn enough to become proficient in successfully launching, flying, and landing comparatively inexpensive gas powered, Styrofoam constructed, fixed wing model aircraft purchased as a surrogate trainer for the considerably more expensive FoldBat UAS. As part of the RC aircraft club membership, eac h UF UAS collaboration pilot in command (PIC) had to join the Academy of Model Aeronautics (AMA), a national non profit RC model flight organization for hobbyists that looks out for the best interests of model aircraft operators at a national level, and en courages the public to explore the fun and excitement of RC model aviation in a safe and responsible manner. The AMA
158 developed a recommended set of standard ized flight safety guidelines for RC model aircraft hobbyists over 70 y ea rs ago to keep the flights of model aviation hobbyists safely separated from those of manned flight operations. The AMA regularly revisits its safety code with updates, and an annual AMA membership insures its members with a liability policy that covers costs incurred if property is damaged or people are injured while operating an RC model aircraft within the parameters specified by the AMA safety code. Back in early 2002, the UF UAS collaboration was contacted by the Federal Aviation Administration (FAA), and made aware of a polic y that at that juncture was not and flight must be kept within a 1. 85 km visual line of sig ht (VLOS) radius from the PIC at all times in the interest of safety. Up until this instance, the UF UAS collaboration was under the impression that the flights that they had been conducting with the FoldBat based model aircraft operating standards contained in FAA Advisory Circular (AC) 91 57 which was issued in 1981, and made no mention of limitations in horizontal flying distance from the PIC ( H. F. Percival pers onal comm unication ) It took until fall of 2015 for the FAA to issue AC 91 57A, the first model/hobby aircraft operating standard update in 34 y ea rs. The FAA policy for keeping UAS flights within a 1.85 km VLOS radius of the PIC at all times presented a significant obstacle that ultimately shelved several upcoming field missions that the UF UAS collaboration had intended to conduct from the levee system surrounding Lake Okeechobee (in southern Florida) to collect imagery of
159 floating and emergent vegetation. The plans that h ad to be tabled were to hand launch the FoldBat 5.6 km from shore to the target area, conduct a series of linear transects over the selected area of vegetation, and then have the aircraft fly back to the levee where it could be manually belly landed. It was at this time that the UF UAS collaboration came to the realization that to fulfill the 1.85 km VLOS radius operational guideline imposed by the FAA, and meet the data collection needs of many natur al resource scientists, especially those conducting wetlands research, would require a s UAS airframe that could be hand launched from a boat, and recovered on the water surface similar to a floatplane. Unfortunately at the time, no publicly available mode l aircraft existed that could repeatedly land on the surface of a water body so the UF UAS collaboration was obligated to brainstorm alternative solutions while it continued to fly the MLB Company FoldBat UAS within the 1.85 km VLOS radius for other natu ral resource based missions. As part of the efforts to find ways to comply with the FAA policy while meeting the data collection needs of natural resource based users, the UF UAS collaboration assembled a list of criteria or features that at the time they believed would make for an based questions. The list was established from the experiences that the UF collaboration had realized while fielding the FoldBat UAS, and many of the desired items listed were believed to be achievable simply with engineering, changes in FAA policy, and continued advances in technology over time. In late spring 2002, the UF UAS collaboration met with the UF MAE MAV Laboratory engineers, and showed them the prototype FoldBat ( H F. Percival and P.
160 G. Ifju University of Florida, pers onal comm unication ). The MAE contingent recognized right away that the prototype UA had some design and construction flaws that were contributing to making routine flight a chall enge; however, consi dering that it was just a prototype and not a more polished final version, they were quite impressed with several features of the FoldBat including its autopilot system and GCS software which were particularly advanced for their time ( P. G. Ifju pers onal comm unication ). The UF UAS collaborative briefed the UF MAV Laboratory personnel about several of the specific issues that had been encountered that made fielding the FoldBat system perhaps more difficult, expensive, and less reliable than had been init ially anticipated; including the FAA 1.85 km VLOS radius operational restriction. It was then that the UF UAS collaborative asked the UF MAE MAV Laboratory if they might be able to develop something MAV like, yet larger, that would be simple enough for no vice sUAS operators to fly, save time and money by being constructed (and ultimately repaired when needed) locally rather than shipping the equipment across the US for service, and meet as many of the desired focused by the challenge, the UF MAE MAV Laboratory was onboard for research and development of a natural resource based sUAS to succeed the FoldBat Although not knowing it at the time, it was at that late spring 2002 meeting that the Universi ty of Florida Unmanned Aircraft Systems Research Program ( UFUASRP ) was essentially established for all extents and purposes. Before that meeting, the WEC/end user component, and the GEOM/photogrammetry connection had been built as the UF UAS collaboration ; however, with the addition of the MAE experts to the fold, the foundations of knowledge needed for a successful academic based UAS research
161 program with a concentration on natural resource based applications were officially in place. What started as an idea after an afternoon seminar in 1999 had come full circle and developed into what would ultimately become a pioneering academic based sUAS research programs in the US, and perhaps the first sUAS research program in the world to focus specifically on nat ural resource based applications. The UF Tadpole In the fall of 2002, design and fabrication of a new sUAS specifically for natural resource focused applications was completed at UF. The system was constructed to meet as many of the desired items on the l ist of criteria as possible, and the resulting From personal conversations with UFUASRP personnel, the folding main wingset of the prototype MLB Company FoldBat was probably the weakest physical component of that airframe ( H. F. Percival and P. G. Ifju, pers onal comm unication ); a situation that was exacerbated by emphasizing the need for portability of the platform which ultimately was lost in the durability that a rigid main wingset would have provided. Therefore the UF MAE team invested a considerable amount of time and effort into the design and construction of the main wingset of the UF Tadpole. The additional attention to wing durability was beneficial as the wings only ever received minor damage during its lifetime of flights. The resulting 2.0 m wingspan, 3.37 kg takeoff mass, completely electric powered fixed wing airframe was hand constructed from high density foam, Kevlar fiberglass, balsa wood, and carbon fiber materials for high strength and the low est mass possible. The UF Tadpole had a 0.47 square meter (m 2 ) main wing surface area, and used a high aspect ratio LHK2411 airfoil, which improved overall lift ( Lee 2004 ) The propulsion system for the Tadpole consisted of a n AXI 2028/10 brushless motor with a twin blade
162 propeller in a conventional tractor configuration, a Jeti Advance 40 Plus brushless electronic speed controller (ESC), and power was provided by a Thunder Power 2050 4S4P 14.8 volt (V) 7,600 milliampere hour (mAh) lithium polymer (LiPo) battery which gave the aircraft a 22 .0 m/s ec ( Lee 2004 ) The Tadpole was outfitted with a Procerus Technologies Kestrel v.2.2 autopilot featuring global positioning system ( GPS ) waypoint navigation, a temperature compensated microelectromechanical system (MEMS) i nertial navigation system ( INS ) a barometric pressure altimeter, and a pitot tube airspeed sensor which were all monitored on the ground with Procerus Technologies Virtual Cockpit GCS software. An E2TEK 12.0 V 1,300 mAh LiPo battery provided electrica l power to almost all of the non propulsion system electronics in the Tadpole airframe including the autopilot, control surface servos, and the modems ( Lee 2004 ) According to Lee (2004 ) the Tadpole UAS used a Furuno GH 80 16 channel GPS receiver which provided positional information for waypoint navigation to the Kestrel autopilot unit in the aircraft, and a pair of Aerocomm AC4490 spread spectrum 900 MHz modems ( a unit onboard the aircraft and a matching modem at the GCS) to provide a communications and telemetry linkage between the aircraft and the flight personnel on t he ground. In late 2001, Canon introduced the ELURA20 progressive scan camcorder to the consumer market as a successor to the ELURA2 model which had been retrofitted into the MLB Company FoldBat by UF personnel and provided superior imagery to the sma ller and lighter CMOS interlaced sensors that were shipped with the FoldBat system. The Canon ELURA20 possessed the same optical and physical specifications as the ELURA2 having 525 horizontal line RGB video captured with a
163 6.35 mm 2 progressive scan CCD imaging sensor producing 0.36 MP resolution and onboard MiniDV 8 mm videotape storage capability, and due to its general success as the primary imagery collection sensor for the FoldBat it was therefore selected as the primary optical sensor for the UF Tadpole airframe. The ELURA payloads use d their original equipment manufacturer (OEM) compact rechargeable batteries for power. The Tadpole was equipped with a U NAV PDC1200 GPS video overlay device that imprinted the GPS positional information fr om the Ketstrel autopilot to the video feed before it was recorded to the onboard cassette tape ( Lee 2004 ) Accuracy analyses conducted later found that the GPS p ositional information recorded on the video was better than not having any information at all; however, a highly significant amount of error was determined to be present in the positional data, and therefore reliance upon the on screen positional coordinat es for survey grade accuracy purposes was deemed in appropriate ( Wilkinson 2007 ) The UF Tadpole also carried a small side mounted Super Circuit KPC S900C camera which featured a 4.3 mm fixed focal length pinhole lens and a Sony 6.35 mm 2 super hole accumulation diode CCD sensor producing a 380 line interlaced video stream ( Lee 2004 ) The Canon ELURA20 imagery was recorded to the onboard MiniDV 8 mm videotape at all times during flight, but a Teledyne Relays ER412D video feed switch allowed flight personnel to downlink either the ELURA20 or the side looking KPC S 900C video stream using a Black Widow AV 1,000 milliwatt (mW) 2.4 GHz video transmitter on the aircraft to a RF Link SDX 22LP 2.4 GHz video receiver on the ground at the GCS for near real time video monitoring ( Lee 2004 ) The KPC S900C camera was affixed to the left side of the fuselage under the main wing at an
164 angle 20 below the horizon so that if a target of interest to the researchers were to be encountered during a flight while monitoring the ELURA20 video feed at the GCS, the aircraft could be immediately commanded to loiter about the current UA GPS location in a left bank so that researchers could have an extended look at the target ( Lee 2004 ) Once the fuselage, main wingset, and tail components were fabricated and compiled for the UF Tadpole, the propulsion system, avionics, and payload elements were then integra ted into the aircraft. It became apparent as the pieces were being assembled that preliminary estimations for component mass were underestimated and lever arm calculations for the placement of components were also affected. As a result, some mass would h ave to be removed for the aircraft to achieve flight, and selected elements would have to be rearranged within the fuselage to attain a n appropriate center of gravity (CG) location ( Lee 2004 ) Lee (2004 ) determined that the easiest way to accomplish these tasks was to carefully disassemble the COTS Canon ELURA20 camcorder l eaving only its OEM battery and parts that were essential for imagery capture and videotape recording, and movement of components within the fuselage would resolve the CG issues encountered during system assembly. The UF Tadpole flew missions from spring 2 003 through fall 2006 for the UFUASRP, with most missions flown at an altitude of 150 m AGL, but ranged from 100 400 m AGL depending on the target subject of interest ( Wilkinson 2007 ) Flights with the Tadpole were conducted over a series of natural resource based targets across the S tate of Florida, as well in Idaho and Montana. Within Florida, flights occurred primarily over areas south of Lake Okeechobee for imagery of nesting wading bird colonies, aquatic vegetation, and another effort over manatee aggregations. Professors in UF
165 WEC and GEOM, members of the UFUASRP, developed algorithms and software from both wading bird and aquatic vegetation data coll ected using the sensor payload system use d aboard the UF Tadpole which was published in several peer reviewed papers and presented at professional meetings ( A bd Elrahman et al. 2000 Abd Elrahman et al. 2001 Abd Elrahman et al. 2005 ) In the late spring of 2005, through a partnership between the UFUASRP and the Idaho Department of Fish and Game ( IDFG ) the UF Tadpole was flown over several leks of Greater Sage Grouse, Centrocercus urophasianus Among many results realized from that mission, conducting unmanned flights with the UF Tadpole UAS at h igh relative altitudes (400 m AGL) did not seem to cause disturbance to the birds; however the resulting video imagery resolution of the ELURA20 payload was extremely poor at relative flight altitudes greater than 200 m AGL ( H. F. Percival pers onal comm u nication ). Incremental relative flight altitude reductions of the flight line transects over the grouse leks in Idaho also revea these particular birds would show a behavioral reaction to the Tadpole UA passing overhead, and would temporarily cease to exhibit their reproductive rituals ( P. E. Zager, Idaho Department of Fish and Game, pers onal comm unication ). As part of the annual meeting of The Wildlife Society held in Madison, Wisconsin butor. The symposium presentations, a question and answer session, and displays of three different wingspan UA: the 2.0 m UF Tadpole, and a 60 cm and a 10 cm UF MAV. Those in attendance
166 showed great interest in the use of sUAS to their particular fields of study, with future development and implementation of this novel tool anxiously awaited. The UF Tadpole was also flown over four field sites in the National Bison Range located near Moiese, Montana, in the f all of 2005 as a collaborative effort between the UFUASRP and the US Fish and Wildlife Service ( USFWS ) The National Bison Range flights were designed to capture sUAS imagery of American bison ( Bison bison ) herds for population estimates which were at the time conducted exclusively by a combination of dual observer manned aircraft surveys and ground based visual surveys by the USFWS ( Wilkinson 2007 ) In the spring of 2006, the UF Tadpole sUAS was flown at the Florida Panther National Wildlife Refuge east of Naples, Florida, to collect imagery for assessing vegetative cover and vegetative change in response to intensive physical removal of willows ( Salix sp.) from a pair of ponds as part of an ongoing wetland restoration project. In the fall of 2006, the Tadpole UAS flew several flights at the Tom Yawkey Wildlife Center Heritage Preserve, near Georgetown, South Carolina, to census wading birds in several coast al marsh impoundments. As flights were concluding in South Carolina with the UF Tadpole sUAS, the UFUASRP was putting the final touches on a multi y ea r contractual agreement with the US Army Corps of Engineers ( USACE ) Jacksonville District, for generating a fleet of sUAS that would potentially assist the USACE efforts at an operational level in invasive vegetation mapping and herbicide treatment efficacy, and other infrastructure monitoring applications. The USACE proposal would fund several UFUASRP gradua te students and support a full
167 for collaboration with the USACE as cooperative funders and sUAS end users, and coordinate the day to day interdisciplinary actions of the various departments and peop le involved in the rapidly growing UFUASRP. During the y ea rs in which the Tadpole fixed wing sUAS was the airframe of focus of the UFUASRP, a plethora of undergraduate students from various disciplines on the UF campus became interested in, and/or voluntee red with the UFUASRP laboratories gaining hands on training and experience with the various aspects of the budding Research Program. Several of these volunteers and undergraduate hourly employees went on to become graduate students in the UFUASRP, as thei r advisors were able to secure funding sources for research topics specifically related to the UFUASRP objectives. A M S degree seeking GEOM student completed thesis work on developing imagery georeferencing techniques for UAS using video data collected from the 2005 National Bison Range flights of the UF Tadpole ( Wilkinson 2007 ) Wilkinson (2007 ) indicated that the UF Tadpole sUAS possessed a number of valuable attributes, but was still a bit limited in its utility as an aerial imagery sensor platform because the resultant optical payload video suffered from non uniform zig zag motion along the preplanned flight lines most likely attributed to the inherent positional accuracy errors of the Furuno GH 80 receiver (15 m horizontal ; 22 m vertical), imperfect tuning of the autopilot units, and perhaps most import antly, the U NAV PDC1200 GPS video overlay device which imprinted geolocation data on the recorded videos was found to be fairly unreliable due to sporadic dropouts and irregular time lags. These issues made accurate reconstruction of the scene geometry during imagery post processing an arduous task ( Wilkinson 2007 )
168 Several other shortcomings discovered during the fielding of the UF Tadpole motivated the UFUASRP to pursue the development of a subsequent UA platform that retained the best features of the Tadpole design, yet implemented newer technology and novel solutions to address some of the deficits ( Wilkinson 2007 Bowman 2008 ) For example, the realization that the main wing s et of the UF Tadpole was notably overbuilt and consequently much heavier than initially drawn up on paper that led to a change in design and construction Recall that wing durability was a major point of emphasis during the design phase of the UF Tadpole s UAS as a successor to the prototype FoldBat ; however, the resulting product was overbuilt Overall, the UF Tadpole was much more successful for the UFUASRP during field deployments for natural resource based applications, in that the fully electric design of the Tadpole was remarkably sturdier during flight operations, recorded video data having substantially less effect from vibration in the absence of a liquid fueled engine, and presented a much lower level of disturbance to noise sensitive targets on the ground ( Wilkinson 2007 ) A few of the other findings with the design, development, and fielding of the UF Tadpole sUAS included advancements in airframe component prototyping and composite layup, flight planning techniques for v arious floral and faunal targets, autonomous aircraft flight control tuning, data post processing, and logistics to be considered when traveling and transporting a sUAS and its crew across the country. The UF Nova 1 (Polaris) In the second half of 2006, th e UF MAE portion of the UFUASRP spent an abundant amount of time developing the successor to the UF Tadpole sUAS to rectify some of the flaws discovered during its tenure. The resulting sUAS airframe was mounted
169 forward facing single white light emitting diode. A significant push was made to develop the UF Polaris optical payload system so that it could collect imagery that might lead to directly georeferenced end products ( Bowman 2008 ) According to Mohamed and Wilkinson (2009 ) direct georeferencing is defined as the computation of the transformation parameters between an arbitrary input coordinate system [such as that of an airborne integrated GPS/INS unit], and a global or mapping coordinate system. The implementation of direct georeferencing can lead to photogrammetric end products that do not rely on having ground control points (GCP) with measured three dimensional (3D) position information within the individual images or the re sulting scene, and can assist in locating specific ground based focal targets repeatedly over time ( Perry 2009 ) Direct georeferencing is a significant advantag e for natural resource based aerial imagery data collection applications in that many of the targeted areas of study are not easily accessible by BOTG to place and geolocate GCP s or doing so may undesirably disturb the target subject. Lisein et al. (2013 ) describe direct georeferencing as the act of measuring and recording the exterior orientation parameters, i.e., the 3D position in relation to the Earth and t he attitude (angles of pitch, roll, and yaw) of a remote sensing optical sensor at the moment of imagery capture, which improves the precision of the resulting post processed photogrammetric products. Direct georeferencing enables the alignment of aerial imagery products with a two dimensional (2D) or perhaps 3D surface constructed using a terrestrial based mapping coordinate system. Built on a modified Tadpole fuselage design, the UF Polaris sUAS fixed wing airframe had a 2.44 m wingspan, 4.80 kg takeoff mass, and was fully electric powered. The airframe was constructed using Kevlar carbon fiber, fiberglass, expanded
170 polystyrene (EPS) foam, high strength epoxy based resin, and epoxy doped balsa and spruce wood materials ( Bowman 2008 ) Polaris was constructed with water resistance for its contents including creation of airtight enclosures for specifi c avionic and electronic components. A NACA2313 and NACA0012 airfoil design were us ed for the main wingset and tail sections, respectively. The propulsion system of the UF Polaris consisted of an E flite Power 46 brushless outrunner motor, a three blad e 25.4 cm diameter propeller in a conventional tractor configuration, and a Jeti Advance 70 Pro Opto brushless ESC. A single 10,000 mAh, 18.4 V LiPo battery powered all elements of the Polaris UA through a custom designed power distribution circuit boar d. Typical flight time for an operationa .0 m/s ec and the stall airspeed was 10 .0 m/s ec ( Bow man 2008 ) A Procerus Technologies Kestrel v.2.23 autopilot system was used in the UF Polaris sUAS which featured a barometric pressure altimeter, a pitot tube airspeed sensor, a temperature compensated MEMS three axis gyroscope/accelerometer INS, and GPS waypoint navigation capability via a three axis magnetometer and a Furuno GH 81 16 channel GPS receiver. The aircraft and GCS communication linkages were achieved using Aerocomm AC4490 900 MHz spread spectrum modems which provided bidirectional con trol and telemetry feeds that were monitored on the ground with Procerus Technologies Virtual Cockpit GCS software. Y earning to make incremental improvements from the UF Tadpole to the airframe, optical payloads, and onboard computing of the new Polaris sUAS, several details about the optical payload pushed the UFUASRP in new directions. First, Canon ended the production of the ELURA20 camcorder, which made their
171 availability low and replacements difficult to obtain, even on the used market. Second, t he ELURA20 only permitted collection of video imagery data in the RGB wavelengths of the EM spectrum, and there was increasing demand for imagery capture in other spectral regions for certain natural resource based targets of interest. Finally, the UFUAS RP had realized that directly georeferencing still images obtained from video recordings was an extremely difficult endeavor due to inherent errors in isolating the specific 3D position and attitude metadata information at the instantaneous time of exposur e ( Wilkinson et al. 2009 ) In response to these challenges, the UFUASRP transitioned to a still frame COTS optical camera with the Polaris sUAS, which reduce d overall payload mass, increased optical sensor resolution, provided a means to capture imagery in various EM wavelength regions, and improved the ability to synchronize optical exposure with georeferencing metadata. The primary optical payload for the UF Polaris sUAS changed several times during the first few months of testing and tuning. Initially a COTS Canon PowerShot SD600 point and shoot still frame RGB camera with a 4.3 5.8 mm CCD optical sensor producing 6.0 MP images was selected. Shortly th ereafter, the primary optical payload was upgraded to a 10.0 MP COTS Canon PowerShot A640 point and shoot still frame RGB camera having a 5.3 7.2 mm CCD imaging sensor. By summer 2007, the primary optical payload was further upgraded to a COTS Canon PowerShot A650 point and shoot still frame RGB camera having a larger (5.7 7.6 mm) CCD imaging sensor producing even higher (12.1 MP) resolution imagery ( Bowman 2008 ) A method of field calibration for the primary optical payload before each flight was devised to provide a fast and repeatable method of standardizing the resulting
172 imagery and metadata from flight to flight. The COTS PowerShot point and shoot still frame cameras were selected by the UFUASRP due to their small form factor, increasingly larger sensor sizes and higher sensor resolutions, while maintaining their relatively low costs. Bec ause these COTS cameras were not designed or produced by Canon as mapping grade metric imaging devices, the calibration process before each flight was particularly necessary to capture imagery whose pixels could be use d as scientific data rather than just capturing high of focal targets or areas ( Bowman 2008 ) ( H. F. Percival pers onal comm unication ). A custom built onboard payload control computer was developed and integrated into the UF Polaris airframe. This custom computer served as the hub of activity for the success of the newly designed payload system. Through a rever se engineering process of the Canon camera firmware code, the onboard payload control computer was able to generate simulated electronic signals of button presses on the camera to d be operated from the remot ely located GCS providing a mechanism to change optical payload settings if desired while in flight. The onboard payload control computer also synchronized the captured imagery from the primary optical sensor with GPS/INS metadata generated by the Proceru s Technologies Kestrel v.2.23 autopilot system. Finally, the onboard payload control computer was equipped with a pair of DOSonChip Secure Digital (SD ) memory card slots: one each for recording the imagery data, and corresponding GPS/INS metadata. By carefully considering the location of the SD card slots during the design of the onboard computer such that they would be in an accessible location within the UA
173 fuselage, time spent on the ground between flights was minimized. Once a flight was complet ed, the discharged LiPo system battery could be re moved and replaced with a fully recharged pack, and the two 250 megabyte (MB) SD data cards could be removed from the onboard computer and replaced with formatted empty cards for the next flight. In a sim ilar fashion to the Canon ELURA20 camcorder optical payload used in the UF Tadpole, the UFUASRP looked to reduce unneeded mass within the Polaris UA, so the COTS Canon PowerShot point and shoot still frame cameras were delicately disassembled to remove any components determined to be unnecessary for capturing imagery while aboard the sUAS platform (e.g., flash, battery holder, liquid crystal display, external plastic casing, etc. ). After disassembly, the primary optical before being installing into the sUAS ( Bowman 2008 ) The UF Polaris sUAS was also equipped with a secondary camera capturing RGB video located in the nose of the aircraft oriented forward in the direction of flight. This camera was strictly use d for general situational awareness. The small KX 141 circuit board based video camera featured a 2.8 mm fixed focal length lens and a 480 horizontal line resolution CCD sensor. The KX 141 camera configuration generated an of view whose inter laced video output stream was downlinked via a 1,000 mW, 2.4 GHz transmitter to a complementary receiver at the GCS. The resulting video stream possessed inherent delay, and was not intended or used as a means of first person flight; however, the video st ream did provide an additional measure of increased situational awareness to the ground based flight crew, and was especially helpful during the tuning process of the autopilot unit for observing minor changes in attitude as the
174 gains were adjusted. The s econdary camera video feed could be viewed at the GCS and on an additional monitor, television, or computer projector, and provided a the UA which was a marvelous feat ure to most spectators at flight demonstrations, especially to young people, the media, funding managers, or executives who were ordinarily office or desk based. In addition to the custom hardware (power distribution circuit board, and onboard payload con trol computer), two custom software programs, each with a graphical user interface (GUI) were written to help support the utility of the UF Polaris UAS. The first w hile in the field. The use of the PolarisLink software GUI and window showing the downlinked video stream necessitated that the GCS be equipped with a second touchscreen monitor so that the Virtual Cockpit GCS software interface could occupy the entire p rimary computer screen in the field. An additional reason for the creation of the PolarisLink software was as a video stream switching mechanism for the anticipated integration of a FLIR Photon 320 uncooled microbolometer thermal infrared (TIR) camera i nto the payload options for the UF Polaris sUAS. The Photon 320 featured a 324 256 line sensor array of 38 38 micron pixels, equipped with a 19 mm fixed focal length lens providing a 36 horizontal field of view. At that time, TIR cameras were being produced almost exclusively for the military, and the sensor units were large and liquid cooled for resolution improvements which added mass and were for the most part fiscally unattainable with natural resource scale budgets. The FLIR Photon 320 was perhaps the first TIR unit available on the commercial market that could even be
175 considered as a payload option for smaller UA and was still extremely expensive. With the help and guidance of FLIR sales representatives in selecting an appropriate sensor and lens combination for the UFUASRP desired applications, the representatives indicated that the TIR camera system ultimately purchased by the UFUASRP would be a fine product for the anticipated end products. As the UFUASRP discovered after several test flights with the Photon 320, the relatively low spatial resolution and lack of a sufficiently fine range in temperature sensitivity of the uncooled sensor array, were ultimately not satisfactory for the s UAS applications of natural resources users with a UF Polaris UA. Polaris sUAS flight was completed, the resulting imagery and metadata saved to SD cards from the primary optical sensor could be pseudo registered directly into a Google Earth viewer to provide a flat Earth model of the resultant imagery footprints and their relative coverage ( Bowman 2008 ) The results of the PolarisView software were not ortho photo mosaics, and multiple image geometry was not fully integrated into the software; however, as a first attempt at direct georeferencing for the UFUASRP by projecting each of the UF Polaris captured still images to a flat surface, the UF Research Program was on the right path toward achieving direct georeferencing capabilities from sUAS captured imagery ( Perry 2009 ) The culmination of all the UF Polaris flights provided sufficient proof of concept for obtaining outside funding to keep the UFUASRP moving forward and expanding into the coming y ea rs. The US ACE Jacksonville Dist rict sUAS administrators had many goals for the UFUASRP in funding the research and development of the UF Polaris sUAS. That
176 being said, a goal that seemingly overshadowed all others was for the UFUASRP to produce a sUAS that would be fully operational by USACE personnel so that they could fly routine missions independently of UFUASRP assistance as soon as possible ( A. C. Watts, University of Florida, pers onal comm unication ). To achieve this goal, many of the other objectives had to be completed first, fo r example : 1. Producing a pair of fully outfitted UF Polaris airframes with optical payloads and an additional airframe for RC flight training containing a simulated payload 2. Enhancing the reliability and ease of operation of the new airframe and payload desig ns 3. Training of USACE personnel how to fly RC and then sUAS fixed wing aircraft 4. Developing flight planning technique manuals for the optical payloads 5. Writing operational user manuals and troubleshooting manuals 6. Teaching basic field repairs 7. Finding an effici ent method for post processing the collected data to achieve, but the USACE Jacksonville District sUAS administrators pushed the UFUASRP extremely relentlessly so that the Jac ksonville District could become operational in just a fraction of that time. Most UF Polaris sUAS flights were for airframe and payload testing and tuning in late 2006 and early 2007, followed by a series of USACE demonstration flights throughout the remai nder of 2007 and into spring of 2008. Missions of the Polaris sUAS were almost all conducted within Florida, and included flights over agricultural fields near Archer, flights over water and emergent vegetation at Lake Santa Fe near Melrose, and USACE Jac ksonville District flight demonstrations arranged by Jacksonville District sUAS administrators, but conducted by UFUASRP personnel, were flown at Gateway RC Airpark in northwest Jacksonville, a levee demonstration near Clewiston, a flight demonstration and USACE Jacksonville District official acceptance of an operational
177 UF Polaris sUAS platform at Camp Blanding Joint Military Training Center northeast of Starke in February 2008, followed by a demonstration flight in Vicksburg, Mississippi at the US Army En gineer Research and Development Center, and a final demonstration flight at John Stretch Park near South Bay, Florida in March 2008. Ten days before the February 2008 USACE Jacksonville District product acceptance flight demonstration at Camp Blanding, a USACE Jacksonville District sUAS administrator who had been training independently to operate as a PIC, successfully assumed PIC duties of a UF Polaris sUAS during a flight initiated by UFUASRP personnel, and manually landed the aircraft without incident. During a second flight that day, the same administrator serving as PIC preceded to auger the Polaris UA into the ground, totaling that airframe and payload. Using the second of two UF Polaris airframes and pay loads constructed for the USACE Jacksonville District for the official acceptance flight demonstration at Camp Blanding, the demonstration in Vicksburg, Mississippi, and the demonstration near South Bay, which were attended by many high ranking USACE personnel, the UFUASRP fully executed those demons tration flights without any issues. After the USACE dignitaries left the South Bay flight demonstration, the same USACE Jacksonville District sUAS administrator who destroyed the first UF Polaris airframe and payload seven weeks prior insisted that USACE personnel conduct a full flight on their own with UFUASRP personnel looking on but not actively participating so that the USACE could be operational The UFUASRP personnel on hand repeatedly questioned whether the sUAS administrator really fe lt ready for the challenge and provided numerous opportunities for the flight to be called off; but as the liaison for the funding agency, the administrator was determined to be independently operational, and
178 the UFUASRP personnel on hand felt pressured to comply. With UFUASRP personnel standing aside, the USACE Jacksonville District personnel conducted the preflight checklists, and shortly thereafter, the aircraft was hand launched, and flown directly by the sUAS administrator serving as PIC into the rock y aggregate composing the landside slope of the levee upon takeoff, inflicting major damage to both the airframe and payload ( A. C. Watts pers onal communication ). Fortunately in the fall of 2007, as testing and tuning of the USACE Jacksonville UF Polaris airframes and payloads were being conducted, the UFUASRP was able to pool non USACE funding from the Florida Fish and Wildlife Conservation Commission ( FFWCC ) the South Florida Water Management District ( SFWMD ) and the UF Institute of Food and Agricultural Sciences ( IFAS ) to construct an additional UF Polaris UA airframe and payload for several scientific research missions in Florida over the next y ea r. These missions included flights off Cedar Key for a series of floral and faunal targets, Ho neymoon Island west of Dunedin for shorebird surveys, Snake Bight near Flamingo in Everglades National Park for additional shorebird aggregation surveys, the Greater Everglades Water Conservation Area 3A (WCA 3A ) over tree islands harboring colonies of nes ting wading birds, Seahorse Key for vegetation and shorebird surveys, and flights over agricultural land for a series of aeronautical and optical payload tests. Although limited primarily by the low metadata update rate from the autopilot to the onboard pa yload control computer, and also by the inherent error in the autopilot caliber GPS/INS components which were accurate enough for navigating the UF Polaris UA through a flight plan, but not nearly as exact as required for photogrammetry or
179 mapping grade s tandards, the resulting Polaris captured imagery did enable the horizontal georegistration of targets within individual frames to a 67.6 m root mean square (RMS) error level ( Bowman 2008 ) Even in 2008, the 67.6 m RMS error was knowingly large; however, it revealed several key elements for the UFUASRP to consider as the Research Program moved forward. First, calibration of the optical sensors was critical for comparing data products. Second, use of the autopilot GPS/INS as the source of metadata for direct georeferencing of sUAS imagery was clearly an inferior method for t he desired generation of high ly accurate, georectified photogrammetric end products. Third, a higher end, higher accuracy, GPS/INS unit, totally independent from the autopilot, would need to be obtained and integrated as part of the optical payload to enable the required rates of met adata update and data transfer, and provide the increased accuracy needed for direct georeferencing of end products. Lastly, the suggested higher end GPS/INS unit would be most effective if it were affixed directly to the imaging sensor device (camera) to both account for any motion of the imagery sensor within the fuselage of the aircraft, and to reduce the cumulative error introduced by the boresight and lever arm misalignment effects of separation between the optical sensor and the GPS/INS unit. The GEO M personnel of the UFUASRP were satisfied with the resolution that the sensors in the later COTS Canon PowerShot point and shoot still frame camera models produced; however, the relatively poor lens optics of those cameras led the GEOM team to recommend that the next primary optical payload for the UFUASRP feature a COTS digital single lens reflex (dSLR) camera providing the capability of
180 interchangeable lenses for producing appreciably superior optics, and improved photogrammetric end products. Perry (2009 ) indicated that the nave projection of UF Polaris obtained imagery to a flat surface as produced by the PolarisView software compared poorly to photogrammetr ically adjusted direct georeferenced solutions generated by fully calibrated metric optical sensors accompanied with higher accuracy GPS/INS systems. These professional units were routinely flown on traditional manned aerial survey platforms where their l arge volume and mass were not nearly as restrictive as payloads for sUAS, and had been shown to attain horizontal RMS accuracies on the order of c entimeters ( Cram er et al. 2000 ) The UFUASRP proposed an optical payload design improvement for the next UF sUAS platform with the goals of direct georeferencing the captured dSLR imagery, attempting to approach the accuracy of existing manned aircraft imaging systems using primarily COTS components, and to accomplish these goals within the confines of a natural resource scale budget ( Perry 2009 ) To incorporate the proposed l arger dSLR imaging device, and potentially have payload space to fly a second dSLR simultaneously in the future, the existing UF Polaris sUAS did not have sufficient planform wing surface area to lift such mass nor space in the fuselage for carrying such v olume. Additionally, the MAE personnel of the UFUASRP had accumulated several airframe design improvements to assimilate into the next UFUASRP fixed wing sUAS platform, e.g., moving from a conventional tail to a t tail configuration to keep the horizontal stabilizer and its electronic actuator above the waterline after an amphibious landing, designing a larger diameter fuselage to tail composite tube that would be stronger and provide a larger grip location under the UA
181 CG enabling easier hand launching, e levating the motor mounting location to the highest position on the fuselage to accommodate the largest diameter propeller possible, and reverting to a pusher configuration for the motor and propeller to increase the airflow over the tail control surfaces at low airspeed/high rotations per min experienced during takeoff conditions, and to accommodate mounting the pitot tube as high above the waterline without airflow interference introduced by other airframe components or propeller wash. The UF Nova 2 In la te 2007, a new airframe design was drawn up and developed based on estimated dimensions and masses of the optical payload, onboard computer, system battery, avionics components, and the anticipated approximate mass of the airframe materials and hardware. Throughout the spring and summer of 2008, the new airframe was fabricated, and by fall 2008, initial RC flight testing With a full slate of missions planned to begin in the coming months and throughout the following y ea r, after several RC test flights of the prototype aircraft, the UFUASRP and its funding partners were satisfied with the resulting UA design to proceed with mass production of the seven additional airframes needed to meet the contractual obligations of the various funding groups. As the UF Nova 2 UA prototype was being fabricated and while its initial flight tests were taking place, the development of a new optical p a yload system with a COTS dSLR camera and an independent higher end GPS/I NS was concurrently taking place. The UF Nova 2 sUAS fixed wing airframe had a 2.51 m wingspan, which provided 0.67 m 2 of main wing area, a 6.21 kg maximum takeoff mass, and was hand constructed from high density foam, carbon fiber, and fiberglass. The pr opulsion
182 system for the UF Nova 2 consisted of a Scorpion Power Systems 4025/12 brushless outrunner motor with a twin blade 43.2 cm diameter propeller in a pusher configuration, a Castle Creations Phoenix 80 brushless ESC, and power for all UA component s was provided by a single 18.5 V 10,000 mAh MaxAmps 5 cell LiPo battery which gave the aircraft a 22 .0 m/s ec cruising for flights conducted with ground elevations near sea level The UF Nova 2 UA used a Procerus Technologies Kestrel v.2.3 autopilot featuring GPS waypoint navigation, a temperature compensated MEMS INS, a barometric pressure altimeter, and pitot tube airspeed sensor, which were all monitored at the GCS through Procerus Technologies Virtual Cockpit software. A UBlox LEA 5H code solution single frequency GPS module directly mounted to a copper ground pla ne antenna provided positional information for waypoint navigation to the Kestrel autopilot onboard the aircraft, and Maxstream 9XTend 900 MHz spread spectrum modems were used to provide bidirectional communication and telemetry linkages between the air craft and the GCS. A preplanned flight path was designed and uploaded before each aircraft launch and autonomously executed by the UA with a level of precision exceeding that of a human pilot. As with previous Procerus Technologies autopilots, this vers ion allowed instantaneous flight plan changes, a user friendly GUI, and an abundance of failsafes to ensure positive aircraft control and promote the safety of other objects in the air and on the ground. The standard optical payload for the UF Nova 2 sUAS consisted of a COTS 10.0 MP Olympus E 420 dSLR still frame camera with a fixed focal length 25 mm Olympus Zuiko Digital ultra equipped with a Hoya HMC
183 ultraviolet filter The E 420 natively captured RGB wavelength imager y with its 17.3 13.0 mm Micro Four Olympus camera produced high quality imagery and had the dynamic sensitivity range of a CCD array, but through advancements in technology, it only drew the power of a typical CMOS sensor. The Olympus E 420 was selected because at the time it was the smallest and lightest dSLR on the market, and offered a software development kit that allowed control of the camera settings and a method of digital exposure t riggering which was essential for the continued efforts of the UFUASRP to capture directly georeferenced imagery. The optical payload was outfitted with an independent, higher accuracy GPS/INS unit; a commercially available Xsens MTi G that was hypothes ized to provide major improvements in 3D position, attitude, and data update rates over the comparatively coarse devices use d by the UA autopilot system. All of the imagery from the optical sensor, and the telemetry metadata from the MTi G were synchroni zed through a custom circuit board device, which both initiated a shutter exposure, and timestamped each captured image with a telemetry data packet at the exact moment of maximum shutter aperture. The custom synchronization board was dubbed the Burr it by the UFUASRP, due to initial versions being fairly large and wrapped for protection in white rubberized foam; having an appearance much like the tortilla it faster, and looke d less like the original component; however, it persisted Imagery and the telemetry data were stored together aboard the aircraft on a VIA Technologies EPIA P700 10L Pico ITX form factor x86 computer running Microsoft Windows XP with a 1.0 GHz VIA C7 processor, 1.0 gigabytes (GB) of
184 system memory, and an 80 GB solid state hard drive. The entire payload system was integrated with a series of universal serial bus (USB) 2.0 interfaces. With the Nova 2 sUAS conducting most of its flight s over wetland environments, efforts were made to try to keep as many of the electronic components within the fuselage resistant to water. The autopilot unit aboard the aircraft was outfitted with its own sealed composite box to protect it from moisture; meanwhile the ESC, modem, and onboard computer were all positioned relatively high within the fuselage as a means to try to protect them from small amounts of water that might enter the fuselage during amphibious landings. As an additional means to keep w ater out of the fuselage, the fully loaded aircraft was designed with an increased distance from the waterline when floating on the water surface to the main payload hatch. All of these approaches assisted in trying to keep the sensitive electronic compon ents contained within the UF Nova 2 from failing due to small amounts of water entering the fuselage. In an endeavor to facilitate the potential need to obtain replacement parts for the UA and/or its optical payload while in remote locations (places where natural resource sUAS users often operate), extra consideration was imparted into the design and construction of the UF Nova 2 sUAS. With the exception of the hand fabricated composite airframe elements, and the custom timing synchronization circuit board the remainder of the UF Nova 2 was built from items that could be obtained commercially with relative ease. By adhering to this model, when the Nova 2 needed a replacement part, a field based flight crewmember would have a spare on hand, could potential ly find a replacement in the nearest town/city, or go online to a distributer and place an order to have items overnighted to the closest parcel delivery location; thereby losing as little
185 fieldwork time as possible, and essentially eliminating the possibi lity of having to completely terminate a mission due to unavailability of obtaining replacement parts. For UF Nova 2 custom made elements, a minimum of at least a spare of each item was brought into the field for every mission, and for extended missions, multiple spares accompanied the flight crew. In keeping with the ease of replacement model, every effort was made to be able to directly swap a faulty component with a replacement without extensive modifications that took expertise, training, and signific ant time, e.g., board, soldering connectors to specialty wire connectors, etc., which had been a constituent of UFUASRP systems before the Nova 2 sUAS. A typical consumer wh o purchases a COTS dSLR is looking to capture imagery within the RGB wavelengths of the EM spectrum. The cones within the retina of the normal huma optical sensors within consumer digital cameras are composed of millions of photodiodes that are design ed to measure and record the intensities of EM wavelengths within the RGB spectrum. Human bodies are subjected to various EM wavelengths in regions making up the entire EM spectrum, e.g., ultraviolet rays, microwaves, radio waves, etc.; however the human wavelengths. According to ( Stark and Chen 2014 ) the sun produces most of its energy within the RGB ran ge of the EM spectrum; however, the sun emits slightly more NIR than RGB wavelength light. Therefore, nearly all COTS dSLR cameras are equipped
186 eliminates NIR wavelengths of the EM spectrum from reaching the sensitive optical sensor sites as they measure and record RGB light intensities. If this NIR filter were carefully removed from the camera optical sensor, the resulting imagery would appear overexposed due to the noise introduced by NIR wavelengths overwhelming the optical sensor photodiodes with both RGB and NIR wavelengths. However, if a calibrated filter blocking a range of wavelengths in a portion of the RGB spectrum is installed where the is reduced to a manageable level for the photodiode sensor sites. The UFUASRP sent off two of its Olympus E 420 dSLR camera units to SA) who specialize in converting COTS natively RGB optical sensors into sensors that can capture imagery in portions of the NIR spe use s the cut filter removal method, and replaces them with an ser defined portion of the NIR spectrum to be captured by the photodiodes of the optical sensor in lieu of a user defined range of wavelengths in the RGB portion of the EM spectrum. For example, individual photodiodes that make up a typical COTS digital c amera optical sensor are sensitive to red, green, and blue wavelengths. During a n RGB camera exposure, each individual photodiode making up the optical sensor simultaneously measures the intensity of red, green, and blue light at a sensor site, whereby a composite of all the sites can be assembled forming an RGB image. However, the same COTS digital camera optical sensor that has undergone the most common NIR filter exchange modification will no longer be exposed to violet and blue visible color wavelengt hs; instead what was NIR, red, and green to the normal human eye s will be captured and reproduced as red,
187 green, and blue, respectively, by the modified camera sensor. Light wavelengths in the NIR spectrum are absorbed and reflected differently by objects on Earth; and when this information is captured and used appropriately, it can help to determine features such as soil moisture, plant health, and other factors that are not visible to normal human eyes. Due to the co llaborative work with the USACE Jackso nville District, the UFUASRP was able to earn a Certificate of Aircraft Airworthiness in October 2008 from the Department of the Army at Redstone Arsenal who certify the entire US Army drone inventory. With the Certificate of Aircraft Airworthiness, and t hrough a Memorandum of Agreement (MOA) between the US Department of Defense (USDOD) and the FAA, the UFUASRP was able to achieve clearance to fly low altitude ( 366 m AGL ) missions with the UF Nova 2 sUAS throughout large portions of the National Airspace System (NAS). Some of the earliest UF Nova 2 sUAS flights were achieved at a nearby National Guard training site, and subsequent missions with the USACE in an d around Lake Okeechobee, the Greater Florida Everglades, and nearshore coastal areas of the LSCKNWR which were ideal places to conduct low altitude aerial missions because they contained significant ecological targets, were essentially uninhabited by peop le, and were nearly uniform in elevation. The Nova 2 sUAS completed 17 missions within Florida using a combination of RGB and NIR optical sensors for a total of 53 flights from fall 2008 through summer 2009. Missions included: 1) conducting research and d evelopment of optical sensor calibration techniques by integrating real time kinematic (RTK) surveyed GCPs and imagery metadata provided by the higher end GPS/INS unit in north central Florida; 2)
188 evaluation of advantages and disadvantages of several COTS imagery post processing software programs; 3) surveying and delineating wetland vegetation and the efficacy of herbicide treatments on invasive vegetation [primarily water lettuce ( Pistia stratiotes ), water hyacinth ( Eichornia crassipes ), and Tropical Amer ican water grass ( Luziola subintegra )] in and around Lake Okeechobee; 4) assessing the condition, health, safety, and construction of water control structures in the Greater Everglades; and 5) appraising the abundance and distribution of nesting colonial w ading birds and shorebirds in several active tree island rookeries around the state including WCA 3A, Loxahatchee National Wildlife Refuge, and the LSCKNWR. The resulting imagery using the COTS Olympus E 420 dSLR camera with any of the interchangeable Ol ympus Zuiko Digital fixed focal length lenses (the UFUASRP us ed 25 mm and 35 mm options for data collection missions, and tested a 50 mm version that was ultimately too heavy and had a field of view that was too narrow to permit imagery overlap for low a ltitude aerial surveys by sUAS) equipped with a Hoya HMC ultraviolet filter, was noticeably improved over the optics and imagery delivered by the point and shoot cameras use d in the previous UFUASRP airframe and optical payload iteration. Although the dS LR optical payloads were inherently heavier than previous models, the UF Nova 2 airframe was designed with a larger main wing planform, and a substantially larger fuselage having a greater volume available for the optical payload components, avionics, and a larger system battery. Additionally, the Olympus E 420 payload was able to record an image every 2.7 s ec ; a bottleneck result of the maximum transfer speeds of USB 2.0 technology, which in combination with the relatively high 22 .0 m/s ec cruise airspee d of the Nova 2 sUAS ultimately
189 influenced the imagery footprint overlap that resulted from flights with the Nova 2 UA airframe and optical payload combination. The independent higher accuracy Xsens MTi G GPS/INS unit directly affixed to the E 420 dSLR camera body provided an incredible improvement in the collection of imagery metadata which facilitated directly georeferenced images, and subsequent imagery end products, confirming the UFUASRP hypotheses about its use. The UFUASRP was able to routinely achieve a .0 m in spatially registered imagery products, and actual spatial resolution of c m with the UF Nova 2 sUAS and its optical payloads ( Perry 2009 ) O bstacles that were encountered during the fielding of the Nova 2 sUAS included situations where small amounts of water would enter the fuselage during an amphibious landing. Generally a small collection of water would gathe r in each of the fuselage skid channels after a water landing which did not cause any problems until the airframe was lifted off the water surface, and the pooled liquid in the skid channels would tend to move about, often leaving water droplets on the cam era lens filter or on the interior surface of the through fuselage glass panel over which the camera and lens were positione d. When this occurred, almost all of the fuselage contents would have to be carefully removed and appropriately dried in the field so that subsequent flights could take place without potential electronics damage or inhibited optical clarity. After many dunk tests in the laboratory, it finally was determined where the water was entering the fuselage. Two areas, which were difficult to necessarily recreate: 1) at adequate composite materials. At the location where the tail boom connected to the
190 fuselage, minute conduits would form between the applied caulking and the fuselage skin due to dynamic stresses induced at this location as the grip for hand launching the airframe. Routine water testing in the laboratory was generally conducted with airframes that had been freshly re caulked at the fuselage an d tail boom connection; therefore, they had not been subjected to the dynamic stresses of a hand launch which was causing the opportunity for water entry observed in the field. Additional laboratory testing also revealed that due to the mass production of eight airframes in a short period not all of the fuselages were laid up during hand fabrication with equal layers of carbon fiber fabric or epoxy resin. Closer attention to detail and documentation of each step completed during each fuselage layup would have prevented this situation from occurring in the first place. Another shortcoming of the Nova 2 sUAS was that in the efforts to make the fuselage as waterproof as possible, there was essentially no airflow within the fuselage. As the UFUASRP learned q uite quickly during an initial flight for a mission in south Florida, when all of the UF Nova 2 UA electronic equipment was running, a tremendous amount of byproduct heat was produce d. Coupling the derivative heat generated by the electronics in a sealed fuselage with the ambient temperatures of south Florida, led to audible warnings by both the ESC and the autopilot, that those components were approaching the upper limits of their operational temperature ranges. Upon hearing the audible warnings, the air craft was safely landed, and once the main payload hatch was opened, the component temperatures rapidly cooled. To remedy the ESC temperature issue, a small hole in the fuselage was made over the ESC attachment location and an aluminum heatsink was extern ally affixed. The ESC unit itself was internally mounted to
191 the external heatsink using thermal epoxy. Additionally, a conscious effort was made as part of the standard operating procedures (SOPs) for the Nova 2 sUAS to keep the fuselage shaded during pe riods of non flight, i.e., rep ositioning transit, pre flight, and post flight which helped keep the electronic components contained within the fuselage, especially the autopilot unit inside its watertight composite box, from overheating while on the ground or on the deck of a boat. Ultimately the UF Nova 2 sUAS ended up becoming practically too heavy for its main wing design, and complicating these issues was the fact that the airframe was becoming more and more nose heavy from efforts to keep water out of the fuselage by applying various sealants excessively throughout the fuselage shifting the CG location further and further forward. While flight with the UF Nova 2 was still achievable even with its increasing mass, maintaining straight and level flight c onditions was difficult under manual control, and the general success of hand launching the progressively heavier aircraft began to decline. Finally, as a result of the UF Nova 2 airframes being constructed primarily out of carbon fiber due to its relativ e ease and speed of layup, and since carbon fiber is an electrical conductor u ltimately did cause radio frequency flight testing. Repositioning the various antennae onbo ard the aircraft resolved these issues. While the UFUASRP was able to complete a number of valuable missions with the UF Nova 2 sUAS, and continue its hallmark of steady learning, identification of operational limits, and evolution of various sUAS techniqu es for natural resource based applications, it became quite apparent that perpetual attempts at retrofitting the UF
192 Nova 2 airframe design to keep water out of the fuselage were simply ineffective. As electronic components began to succumb to issues resul ting from water entering the fuselage, the entire UFUASRP gathered together with its cooperators and collectively decided that it was in the best interest of all involved to design another airframe to reliably carry the optical payload components of the UF Nova 2 over focal targets of natural resource based sUAS applications. Allocating additional time, effort, and money, into makeshift solutions to inherent flaws in the airframe design and construction of the Nova 2 UA was simply going to exhaust all rema ining funding without fulfilling the contracted deliverables. The UF MAE team provided some rudimentary sketches of a proposed airframe that could theoretically resolve many of the UF Nova 2 issues. Those sketches would ultimately lead to the next fixed wing airframe design in the UFUASRP lineage. The UF Nova 2.1 (Mako) As mentioned, the UF Nova 2 UA airframe had several fundamental problems that could not be retroactively engineered out while staying under budget and with collecting overlapping still ima gery of natural resource based targets as the primary UAS application, maximizing the runtime and ground area coverage of each flight was extremely important. With this in mind, the UFUASRP retained the best features of the UF Nova 2 airframe, including m ost of the optical payload, and designed an aerial platform around the calculated mass and volumes of the avionics, optical payload, system battery, and anticipated airframe materials and hardware into an airframe which emulated many aeronautical features of a glider; having a main wingset with a large planform surface area, general thinning and decambering of the wing from the root toward the tips, and polyhedral, taper, and sweep to generate maximum lift and
193 minimize parasitic drag due to laminar flow sep aration. During the fall of 2009, the UFUASRP produced the prototype UF Nova 2.1 sUAS (aka: Mako). The fully loaded 6.0 kg at takeoff, fixed wing airframe had a 2.74 m wingspan, which provided a 0.82 m 2 main wing area, and was outfitted with a v tail rud dervator to minimize drag that other empennage configurations would have eclipsed. The fuselage was constructed using a Kevlar and epoxy resin laminate with strategic carbon fiber reinforcements that provided high strength and radio frequency transparenc y. Significant features incorporated into the fuselage design included a single deep basin lower fuselage section that composed most of the total fuselage surface including a contoured area under the gasketed main wing to fuselage attachment hatch to assi st hand launching techniques, a single upper fuselage section on which the interior surface served as a location for affixing many of the electrical components, and a joggle seam (similar to the junction that holds two halves of a plastic Easter egg togeth er) which projected .0 cm from the interior ventral portion of the upper fuselage section creating a strong seam with ample surface area for a continuous bead of adhesive to bond and seal the upper and lower fuselage sections together. The fu selage seam was purposely located extremely high above the waterline to reduce potential conduits for water entry. The wings and the tails of the UF Nova 2.1 were constructed with dense extruded Styrofoam cores that were cut in house from customized compu ter aided design (CAD) files on a laptop computer connected to a computer numerical control (CNC) hot wire cutting machine to produce an SD7032 airfoil for the main wingset which tapered toward the tips, and a progressively blended HT14 to HT12 airfoil for the v tail
194 members. The custom cut wing cores were then reinforced with a thin strip of aramid fiber along the leading edge for integrity, and then covered with several layers of fiberglass and epoxy resin to create a skin for lamination using vacuum bag ging and heat curing composite material layup methodologies. The main wing was divided into three sections for portability during transport but retained horizontal rigidity with minimal lift induced bending moments at the wing joints when assembled via em bedded aluminum joiner tubes and small anti rotation pins. In the design of the ruddervator, efforts were made to make it both easily attachable and removable. Additionally, rotational drive systems affixed to the control surface actuators were used thro ughout the airframe wings and tails to move the control surfaces of the UF Nova 2.1; a design feature which prevented any electrical connections from having to bridge removable segments of the wing or tail, further improving the water resistance of the air craft and reducing the opportunity for amphibious landings to damage electrical components due to shorts in wire connectors bridging wing segments. The larger area of the control surfaces aided in achieving and maintaining straight and level flight by min imizing the amount of actuator throw input needed to initiate changes in the aircraft attitude and/or altitude. After several y ea rs of field experience working from the relatively cramped quarters of an airboat deck to assemble and preflight previous UF s UAS airframes, the Nova 2.1 sUAS was designed to be as tool free as possible by us ing knurled thumb screws for main wing attachment and forward payload hatch closure allowing for quick access to components within the fuselage, and for rapid field assembly and breakdown. The propulsion system of the UF Nova 2.1 consisted of a NeuMotors 1509 1.5Y 6.7:1 geared and fanned brushless electric motor, spinning an Aeronaut CAM 45.2 cm
195 diameter folding propeller in a conventional tractor configuration. When the mo tor would be commanded to stop spinning during routine gliding flight, or especially during the later stages of a landing sequence, the blades of the propeller would fold flat anteriorly against the fuselage, reducing induced drag while in flight, and prot ecting the propeller blades from damage during landings. A Castle Creations Phoenix Ice 100 brushless ESC was selected to regulate the motor, and a single 18.5 V 10,000 mAh MaxAmps 5 cell LiPo battery powered all onboard systems. The aircraft had a 15 .0 m/s ec cruise airspeed, an 11 .0 m/s ec min for flights conducted with ground elevations near sea level The UF Nova 2.1 UA used a Procerus Technologies Kestrel v.2.4 autopilot featuring a barometric pr essure altimeter, a pitot tube airspeed sensor, GPS waypoint navigation, and a temperature compensated MEMS INS, which were all monitored at the GCS through Procerus Technologies Virtual Cockpit software. A UBlox LEA 5H code solution single frequency GPS module directly mounted to a copper ground plane antenna provided positional information for waypoint navigation to the Kestrel autopilot onboard the aircraft, and Microhard Nano 900 MHz spread spectrum modems were used to provide bidirectional comm unications and telemetry linkages between the aircraft and the GCS. As with previous UFUASRP fixed wing airframes, the UF Nova 2.1 sUAS was designed to be hand launched to eliminate the need for hauling specialized launching equipment, and was fabricated f or belly landings on relatively flat terrestrial terrain or on the surface of a water body. Operational reliability was enhanced for the Nova 2.1 UA by utilizing the autonomous takeoff and landing flight modes incorporated into the
196 Kestrel autopilot unit During each preflight, the autopilot was programmed with a user designed flight path based on specific flight goals, payload capabilities, and ambient environmental conditions. Additionally during preflight, the onboard autopilot unit was uploaded with detailed failsafe procedures so that if bidirectional communications were lost while the aircraft was in flight, the UA would transition seamlessly into a sequence of failsafe protocols allowing adequate time for communications to be reestablished while p roactively making its way into a controlled loiter about the GCS and aircraft operator. If bidirectional communication reestablishment failed to occur after a user defined period of time, the autopilot unit would autonomously initiate a controlled descent and landing approach sequence for the UA at a user defined location as part of the failsafe protocols uploaded before each flight. Reliable sUAS operations in wetland areas necessitated a waterproof platform, so the UF Nova 2.1 fuselage use d numerous wate rproofing features. The electric motor generated a significant amount of byproduct heat; therefore, a motor compartment in the nose of the airframe was isolated from the rest of the main fuselage with an interior watertight composite bulkhead. The motor compartment assisted in reducing motor generated heat from adding to the total heat budget within the main fuselage, and allowed the strategic placement of several small openings to facilitate ambient air cooling of the fanned motor, the motor compartment, and a method of draining any water droplets that might enter the motor compartment. During aircraft fabrication, the smooth metallic exterior can of the geared electric motor was inserted into a custom milled sleeve portion of an aluminum block heatsink that bridged the interior of the
197 motor compartment with ambient air moving over the fuselage. The exterior surface of the motor heatsink was flat and nearly flush with the fuselage which provided very little surface area for heat transfer; however, the mo tor block heatsink was equipped with six verticall y drilled holes for accepting up to six Thermacore copper heat pipe temperature dissipation tubes which increased surface area, thermal conductance, and substantially wicked away heat generated by the moto r. All of these methods simultaneously reduced the negative effects of thermal damage on the electric motor used for propulsion of UF Nova 2.1 aircraft; therefore increasing the overall motor efficiency and lengthening its service life. The sealed main fu selage of the UF Nova 2.1 UA contained most of the onboard electronic components including the optical payload, and wa s outfitted with advantageously placed heatsinks and heat dissipation devices An externally mounted aluminum heatsink with substantial s urface area was thermally epoxied directly to the ESC unit. Also in the main fuselage, the autopilot unit was placed within a custom milled aluminum box having waterproof connectors for both wiring and tubing interfaces to provide an added layer of water protection and promote heat t ransfer away from the thermally sensitive autopilot unit itself. The onboard computer natively supported an onboard brushless system fan to direct cooling ambient air over heatsinks affixed to critical onboard computer chips. Upon further investigation it was determined that the system fan of the onboard computer provided essential air movement and thermal benefits to all components within the sealed main f uselage of the UF Nova 2.1 sUAS ( Figure A 1 ).
198 Retaining the standard optical payload that was used for the UF Nova 2 sUAS, the UF Nova 2.1 continued to use a COTS 10.0 MP Olympus E 420 dSLR still frame camera with a fixed focal length 25 mm Olympus Zuiko Digital ultra compact f/2.8 ns equipped with a Hoya HMC ultraviolet filter The E 420 natively captured RGB wavelength imagery with its 17.3 13.0 mm Micro Four Thirds Live MOS imaging sensor. Two additional COTS Olympus E 420 dSLR RGB cameras were & Photo for low pass IR filter replacement that enabled the cameras to collect imagery in a portion of the NIR spectrum in the place of visible violet and blue wavelengths. All of the optical payloads were outfitted with independent high accuracy Xsens M Ti G combination GPS/INS units and antennae that provided enhanced 3D position, attitude, and data update rates compared to similar devices use d by the UA autopilot system. All of the imagery from the optical sensor, and the telemetry metadata from the M Ti G were synchronized through a custom circuit board that both initiated a shutter exposure and timestamped each captured image with a telemetry data packet at the exact moment of maximum shutter aperture. Imagery and the telemetry data were stored toge ther aboard the aircraft on a VIA Technologies EPIA P700 10L Pico ITX form factor x86 computer running Microsoft Windows XP with a 1.0 GHz VIA C7 processor, 1.0 GB of system memory, and an 160 GB solid state hard drive. The entire payload system was integrated with a series of USB 2.0 interfaces which allowed imagery to be captured every 2.7 s ec Due to the time lapse between successive imagery exposures, imagery payloads were oriented with the longer axis of the sensors parallel to the direction of flight. By positioning the sensors in this configuration, sequential images possessed a higher
199 endlap percentage than if the imagery were collected with the shorter axis of the sensor parallel to the direction of flight. Tradeoffs between resultant image ry resolution and imagery overlap for generating the desired end products had to be decid ed prior to each flight The UFUASRP has use d the fixed wing UF Nova 2.1 sUAS for an extended period of time (eight y ea rs to date the longest of any single UF design ) due to its rugged construction, general utility, and ability to meet the needs of many of the natural resource based data collection applications a fixed wing sUAS platform of its size can deliver. However, just because the UFUASRP has not designed a ne w fixed wing airframe in quite some time, the Research Program has continued to examine, develop, and test ways to use technological advances in many areas pertaining to sUAS for natural resource based data collection such as optical sensors, payload archi tecture, avionics, aircraft controls, mission planning, data post processing, end products, etc., through the support of its funding cooperators. With interest in sUAS technology for commercial and industrial interests increasing at exponential rates, and with a surplus of companies looking to provide products to meet those demands, many of the electrical components used in the initial UF Nova 2.1 airframes were soon being replaced on the COTS electronics market with newer, faster, and smaller devices. For example, the Olympus E 420 dSLR was discontinued shortly after the UF Nova 2.1 began operational missions, so the UFUASRP proactively bought up a few of the remaining units that were available and still affordable while researching newer dSLR options. Finding a dSLR RGB (and potential NIR) successor on the COTS market that offered a software development kit
200 was more difficult than initially anticipated as most consumer camera companies ceased access to their proprietary firmware codes. Xsens announced that the MTi G model GPS/INS would soon be discontinued and replaced with a similar, slightly smaller but more expensive version; therefore questions about the output format and timing of the metadata packets of the new GPS/INS unit and its compliance wi th existing Burr it o synchronization hardware and software code became concerns to the UFUASRP. As an academic based Research Program, the UFUASRP was accustomed to having students at all levels join the Program for various lengths of time; some volunteer ing for a semester or two to gain hands on experience, others working hourly on a regular schedule seeking letters of recommendation before they graduate, and still others had lengthy tenures with the Program as undergraduate and/or graduate students befor e moving on to new opportunities as they became available. Generally speaking, rates of ingress and egress of student personnel had been nearly equivalent and predictable for the UFUASRP through 2011; however in summer 2012, about the time that operationa l field missions with the UF Nova 2.1 sUAS were scheduled to increase, and the need to update and integrate new electrical components into the UA was taking place, a significant change in student personnel occurred that temporarily impeded progress. A new cohort of students were brought into the UFUASRP to replace the exiting personnel, however the amount of time to mentor the incoming individuals with the institutional knowledge on how to build, troubleshoot, and repair UF Nova 2.1 sUAS airframes, existing payloads, and conduct flights under established SOPs was short. With the introduction of new students into the Research Program came fresh thoughts
201 and ideas, which historically had been an advantageous characteristic of the UFUASRP in keeping the Resear ch Program on the cutting edge of UAS technology; especially about applications pertaining to natural resource based questions. The incoming group of new personnel into the UFUASRP during this particular transition was no different; i.e., regarding the on board computer there was particular interest in 1) moving from a Microsoft Windows operating system to an open source Linux based Ubuntu operating system; 2) progressing from an x86 computer board to a smaller and lighter ODROID U3 ; and 3) writing new computer code to accommodate the implementation of the two proposed changes to the onboard computer system. Several COTS dSLR cameras were purchased for bench testing and payload integration assessments as potential RGB optical sensor replacements into th e novel optical payload framework that had been proposed. The UFUASRP experimented with an 18.0 MP Canon EOS M an 18.0 MP Canon EOS Rebel T2i a 24.3 MP Sony and a 36 .0 MP Sony from being a viable option as an RGB (and NIR) sensor for the updated UFUASRP optical payload. The most recently tested alternative COTS dSLR camera en ded up becoming the sensor of choice as a successor to the discontinued Olympus E 420 : a Canon EOS Rebel SL1 18.0 MP dSLR with a fixed focal length Canon EF 40 mm f/2.8 Stepping Motor lens equipped with a Hoya HMC ultraviolet filter The COTS SL1 captured RGB wavelength still frame imagery with its 22.3 14.9 mm CMOS sensor. Several SL1 pass IR filter over the optical sensor exchanged for a filter that allowed the optical sensor to
202 col lect visible green, red, and a portion of the NIR spectrum wavelengths as had been done previously with other UFUASRP COTS dSLR optical sensors. The camera bodies were affixed with independent high accuracy Xsens MTi G 700 GPS/INS unit s to provide 3D p osition and attitude metadata of the optical sensor with accuracy superior to that of the devices used to navigate the aircraft from waypoint to waypoint. As with previous UFUASRP optical payloads, all of the imagery from the optical sensors and the telem etry metadata from the MTi G units were synchronized through a custom circuit board device that both initiated a shutter exposure, and timestamped each captured image with a telemetry data packet at the exact moment of maximum shutter aperture. The image ry and telemetry data were stored together aboard the aircraft on an ODROID U3 Linux based Ubuntu operating system with a 1.7 GHz quad core processor, 2.0 GB of system memory, and data storage on a 128 GB Class 10 MicroSD card. The payload was integrat ed with USB 2.0 interfaces, and was able to collect imagery every 2.5 s ec ( Figure A 2 ) Just as it had done with the Nova 2 sUAS in October 2008, the UFUASRP was able to secure a Certificate of Aircraft Airworthiness from the De partment of the Army at Redstone Arsenal for the UF Nova 2.1 sUAS in March 2010. With the Certificate of Aircraft Airworthiness, and through the MOA between the USDOD and the FAA, the UFUASRP was able to achieve clearance to fly low altitude ( 366 m) missions with the UF Nova 2.1 sUAS throughout large portions of the NAS. In spring 2011, the UFUASRP filed for, and acquired its own FAA Certificate of Waiver or Authorization (COA) to fly the UF Nova 2.1 sUAS for missions in the Big Bend coastal region of Florida, independent of the USACE military agreement. Since that time, the UFUASRP
203 has been granted 2 9 additional COAs by the FAA, which has permitted the Research Program to fly the UF Nova 2.1 in areas for which most other sUAS entities were prohibited from legally operating until 2015. Experience in working with the FAA as they developed and refined the COA process for sUAS helped the UFUASRP acquire FAA authorizations to fly the UF Nova 2.1 in many locations. In 2014 (and again in 2015), t he UFUASRP confirmed its good standing with the FAA when the Research Program hangar at the In & Expo in Lakeland, Florida. A Southeastern Region FAA UAS administrator advised the UFUASRP t he way that your Program [ the UFUASRP ] conducts and fields small unmanned aircr aft systems is the way that we [ the FAA ] wo K. Wilson, Federal Aviation Administration, pers onal comm unication ). The UFUASRP has flown the UF Nova 2.1 on over 1 30 flights; predominantly over grassy fields, lakes, wetlands, inshore coastal areas, fo rest stands, and agricultural fields throughout Florida; but has also flown the sUAS over six different sites across Idaho as well. Missions with the UF Nova 2.1 sUAS have produced high resolution imagery and very dense directly georeferenced imagery prod ucts. Individual images can be post processed into end products that allow data to be extracted from individual pixels to provide evidence to address ecological questions. For example, RGB imagery of a specific Lake Okeechobee 25 .0 hectare (ha ) target ar ea, Fisheating Bay, with emergent vegetation known to contain areas of invasive vegetation were collected during a single flight by the UF Nova 2.1 on a day that for the sake of convenience will
204 t the aircraft was recovered off the water surface, wetland vegetation experts ground truthed a series of random geographic locations within the target area with 2D GPS positions to identify the primary vegetation composition at each location. The day afte r the UF Nova 2.1 sUAS flight, USACE contracted airboat crews went into the target area and spot treated invasive t was post processed into a large georeferenced orthop hotomosaic image, which consisted of n = 248 individual images. Using the ground truthed GPS data points in combination with the vegetation identifications at the random locations within the target area, the georeferenced orthophotomosaic image was classi fied into seven macro level vegetative categories based on computer recognition algorithms utilizing spectral signatures. Six weeks after the spot treatment of invasive vegetation with herbicide inside the defined target area of interest at Fisheating Bay a day that will be referred to t + flew the exact same 25.0 ha target area with the UF t + treatment RGB georeferenced orthophotomosaic was constructed back in the laboratory of the exact same area, and temporal change was clearly evident and able to be quantified. Missions with the UF Nova 2.1 have also recorded imagery of avifauna of many different species at various locations, e.g., White Ibis ( Eudocimus albus ) i n the Loxahatchee National Wildlife Refuge and on Seahorse Key in the LSCKNWR, Brown Pelican ( Pelecanus occidentalis ) in the LSCKNWR, American White Pelican ( Pelecanus erythrorhynchos ) in the LSCKNWR and on spoil islands in several reservoirs in Idaho, Dou ble crested Cormorant ( Phalacrocorax auritus ) on spoil islands in Idaho reservoirs,
205 and a number of gull species ( Laridae ) in the LSCKNWR and on spoil islands in Idaho reservoirs. Because of the repeatable, tightly georeferenced imagery that is capable of being collected using the optical payloads of the UF Nova 2.1 sUAS, existing and new t t + t + t + n spatial sampling assumptions and detection probability are accounted for, best effort counts of individual targets using dual observer techniques from a moving manned aircraft could possibly give way to more accurate population estimates of a specific t collected imagery; b ut perhaps even more inspiring is the possibility of generating estimates of seasonal production of t + t + sUAS flight schedule can be established and maintained throu ghout an entire reproductive season ( Williams et al. 2011 ) While the example detailed above is focused on avifauna, a similar methodology using sUAS could be applied to a tremendous number of other faunal targets with relative ease. Other missions conducted in Florida utilizing the UF Nova 2.1 sUAS included flights over water control structures for the USACE Jacksonville District in the Greater Everglades and Picayune Strand, a mission for investigating statistical techniques accounting for the distribution of hidden objects from sUAS derived imagery ( Martin et al. 201 2 ) and missions over various small islands in the LSCK N WR to collect baseline data and imagery for ongoing and future climate change and sea level rise studies. Other UF Nova 2.1 missions to date include flights over Greater Everglades wetland vegetati on in WCA 3A to examine the ability of the airframe and payload to delineate fine scale vegetation types ( Zweig et al. 2015 ) and a series of three separate miss ions
206 to collect data over three discrete terrestrial sites in Idaho during subsequent summers and a winter field season to investigate the possibility of using the UF Nova 2.1 sUAS to assist in minimizing time spent with BOTG methodologies looking for pref erred pygmy rabbit ( Brachylagus idahoensis ) habitat in the sagebrush steppe ecosystem. The Nova 2.1 sUAS also conducted a discrete autumn mission in the lower Clearwater River of Idaho to evaluate the potential for using the fixed wing UA platform to esti mate the number of Chinook salmon ( Oncorhynchus tshawytscha ) redds in portions of the river as an alternative to the existing method of using dual observer biologists in low altitude manned helicopters (which have tragically taken the lives of multiple bio logists and pilots ). Additional vegetation delineation missions were conducted with the UF Nova 2.1 over Orange Lake in north central Florida, agricultural fields near Archer, Myakka, and Balm, Florida, and over the Ocklawaha Prairie also located in north central Florida. While the UF Nova 2.1 sUAS fixed wing design has been the most reliable UFUASRP platform to date, it still has room for improvement ( Balmori 20 14 ) Perhaps the biggest drawback with the design is ironically a product of overachieving success in meeting a few of the fundamental design objectives; once airborne in straight and level flight, and at or above the 15 .0 m/s ec cruising airspeed, the U F Nova 2.1 does not rapidly scrub speed due to an airframe design built to minimize total drag. The aircraft has a tendency to remain airborne during a landing sequence in which the motor is off and the sinking rate of the airframe is particularly shallow due to the high lift to drag ratio of the main wingset, especially when descending below the altitude of the ground effect cushion generated by the wings. Due to most of its flights occurring in or over wetland environments which requires belly landings, and the desire to maximize flight
207 time to systematically cover a user defined target area of interest by capturing overlapping imagery and metadata, the need for landing gear which adds drag and additional mass to an airframe was intentionally omitted for the UF Nova 2.1 sUAS ( Balmori 2014 ) A number of COTS fixed wing sUAS without landing gear use either a deep stall landing method, or the use of a parachute to land the airframe at the conclusion of a flight. However, the UFUASRP is not a proponent of these methods because once either technique is initiated, there is essentially no way to abort a landing should the need arise to do so. Deep stall methods cau se the airframe to hit the ground at relatively high rates of speed, imparting undue stress on the UA and its payloads, while parachutes must be folded and stowed properly within the fuselage and once deployed are subject to ambient winds aloft which make landing at a specific location on the ground difficult at best. As natively designed, the glideslope ratio of the UF Nova 2.1 sUAS is 10:1; therefore, landing the aircraft requires a fairly lengthy strip of relatively unobstructed terrain or water surfac e, particularly when descending through the final 5 .0 m AGL. The UFUASRP MAE contingency spent time in 2010 and 2011 working on developing a set of plain flaps (12.7 50.8 cm) to the trailing edge of the root span of the main wing on each side of the fus elage that could be deployed 30 downward during landing to: 1) 3) induce parasitic drag to help shorten the glideslope ratio ( Evers 2011 ) The initial plain flap efforts by Evers (2011 ) were ultimately able to reduce the glideslope ratio by a substantial 55%.
208 However, the UFUASRP fielded applications for natural resource based flights with the UF Nova 2.1 sUAS where an even shorter glideslope ratio was desired. As documented in Balmori (2014 ) the UFUASRP MAE group took the plain flaps (deployed 30 downward) on the trailing edge of the root span of the main wing and combined them with spoilerons (essentially ailerons deployed upwards 20 which still permitted limited aileron roll control even when deployed) but unfortunately this combination created too much instability and induced a crash during field testing. Balmori (2014 ) reported that the trailing edge plain flaps deployed at 30 downward were then combined with 2.5 15.2 cm Schempp Hirth airbrakes, which extended vertically up out of the root span of the main wing section on each side o f the fuselage when the flaps were deployed downward, and reduced the glideslope ratio to an improved 3.6:1. Ultimately the third option tested, a novel hybrid flap/spoiler (HFS) design reduced the glideslope ratio to 3.3:1 ( Balmori 2014 ) The DJI Spreading Wings S1000+ The UFUASRP had always wanted to dabble in the rotary wing sUAS market, but since most natural resource based questions required the payload capacity and flight endurance that a fixed wing platform could deliver, the Research Program focused on learning as much as possible and refining details encountered as it fielded fixed wing sUAS platforms for natural resource applications. In August 2014 personnel from the US Geological Survey ( USGS ) National UAS Project Office (NUASPO) came to UF to discuss with the UFUASRP a proposal to construct a rotary wing sUAS that could meet certain criteria for natural resource based applications, including inte grating a series of new optical sensors, and conducting operat ions primarily in the western US where vertical takeoff and landing (VTOL) mu lti rotor capability is advantageous over fixed wing
209 airframe design s due to rugged terrain limiting feasible runway l anding options The UFUASRP determined that it could build a custom multirotor sUAS for the USGS NUASPO that would meet nearly all the desired requirements at a very reasonable cost; but with the limitations that the US Department of the Interior (USDOI) Office of Aviation Services (OAS) instilled in regulating UAS used by the USGS, the UFUASRP suggested that going with a COTS airframe and avionics as most of the rotary wing platform would most likely be a more advisable route for the collaboration to purs ue. The thinking behind this was that obtaining a fairly elusive certificate of airworthiness which was a barrier that had immobilized USDOI UAS acquisitions up until that time, and since COTS rotary wing platform availability was in the process of expone ntial growth, and certain COTS companies were emerging as clear leaders in rotary wing platforms and avionics, the UFUASRP thought that the USDOI OAS logically would be more amenable to approving the operation of an airframe and avionics from a company who se products were logging hundreds of thousands of operational hours worldwide, rather than attempting to gain approval of a custom designed and constructed airframe. Additionally, by going with a COTS platform as most of the airframe opposed to a custom b uilt design or prototype would make obtaining replacement parts much easier for the USGS NUASPO in the long run (a lesson learned by the UFUASRP in fielding the MLB Company FoldBat UAS many y ea rs earlier). Optical payload components were scoped out based on specific criteria prioritized by the USGS NUASPO. With the range of EM wavelength spectra that natural resource scientists currently use to collect data to assist in assessing and monitoring focal targets from satellites, manned aircraft, and BOTG met hodologies,
210 obtaining a single sensor package that would encompass the entire EM spectral ranges desired by the NUASPO could potentially be available on the market; however, the total cost of an all in one sensor package was most likely well beyond the ava ilable fiscal budget available to the USGS NUASPO, and the mass and volume of such a device was almost certainly too great for a rotary wing UAS with in the size class the NUASPO was looking to field. The UFUASRP had also realized through its y ea rs of expe rience that only sensors collecting data pertinent to the scientific question at hand should leave the ground during any sUAS flight. Collecting data for the sake of collecting data was generally a waste of time, money, and effort if it was not helping le ad to answering the scientific question necessitating the sUAS flight in the first place. Putting all of the available payload budget into a single all in one sensor package was extremely risky because if something caused the airborne UA to fail and crash there would be little to no funding available to repair or replace the damaged items. Another reason that the UFUASRP suggest ed only flying the sensors needed to answer a scientific question aboard a sUAS is that the least amount of mass that has to be flown around during a flight generally leads to longer flight durations over the target area of interest and less time spent on the ground where no aerial data collection takes place. A multitude of additional reasons exist suggesting not to fly sensors a board a sUAS unless they are actively collecting data leading to answering the question at hand, but a final reason advocated by the UFUASRP is that if a single sensor included in the all in one package should fail, then executing the flight and/or subsequ ent flights of a mission may have to be delayed, postponed, or ultimately cancelled so that the failed sensor c ould be repaired. This situation is unfortunate because time dedicated to repairing a failed
211 sensor in the all in one package essentially ground s the remainder of the sensors from collecting valuable data even though they may be fully operational. A number of physically small, lightweight, and predominantly affordable sensors with appropriate optical specifications were selected to meet the highes t priority EM spectral ranges chosen by the USGS NUASPO. Once the sensors were identified, the UFUASRP got to work acquiring them, and made tentative plans on how the various sensors could be grouped into at least two separate, but interchangeable, gimbal ed payload bays to reduce the total mass of any single flight to increase runtime of an electric powered rotary wing platform which inherently has a reduced runtime compared to fixed wing designs, and to assemble sensor combinations that would most often b e running together to address natural resource based sUAS applications. The UFUASRP discussed several sensor grouping options with the NUASPO, and the collaborative group decided that a gimbaled payload bay should include the dSLR RGB still frame camera, the TIR camera, and the digital video camera, while a second payload bay should include the hyperspectral camera, and the digital video camera. Regrettably, the first sensor cut from the initial USGS NUASPO proposal due to budget limitations was a light d etection and ranging (LiDAR) unit. However, the UFUASRP was able to procure a LiDAR unit independently, which the Research Program was able to ultimately test as a potential payload option for the USGS NUASPO. Through continued collaboration, the USGS NU ASPO was able to purchase a LiDAR unit of their own, and the UFUASRP is poised to integrat e it into the gimbaled payload options. The following sensors were selected for the USGS NUASPO octocopter project: 1) a COTS Canon EOS Rebel SL1 18.0 MP dSLR image r with a fixed focal length
212 Canon EF 40 mm f/2.8 S tepping Motor lens equipped with a Hoya HMC ultraviolet filter for RGB wavelength still frame imagery; 2) a FLIR A65 SC TIR focal plane array uncooled microbolometer with a 7.5 hertz (Hz) streaming res olution of 640 512 pixel (pix) digital video, a spectral range of 7.5 13.0 micrometer (m), a 25 + 135 Celsius (C) object temperature range, and a f/1.25 fixed focal length 13 mm lens; 3) a GoPro HERO 3+ Silver E dition 1080p RGB video camera producing 1,920 1,080 pix screen resolution digital video files; 4) a Rikola Snapshot hyperspectral camera with a 12 bit CMOS image sensor capable of recording spectral images with 1,010 1,010 pix resolution, up to 380 user programmable spectral bands in the range from 500 900 nm and 5) a Velodyne VLP 16 Puck LiDAR sensor featuring 16 channels of dual returns from a Class 1 eye safe 903 nm wavelength laser, a range of up to 100 m with an accuracy of 3 cm, a vertical field of view of 30 with an angular resolution of 2, a hori zontal/azimuth field of view of 360 with an angular resolution of 0.1 0.4, and user datagram protocol (UDP) output packets containing distances, calibrated reflectivities, rotational angles, and microsecond (s) resolution synchronized time stamps. Based on the calculated masses and volumes of the payload gimbal mounts, the payload bays, and their contents, an approximation of the total payload physical size was used to seek out an appropriate COTS airframe capable of reliably situating the payload sensor s over the target areas of interest. Several manufacturers offered products that were potential options, however, in keeping with the plan to select a COTS airframe brand that was emerging as a leader in rotary wing design, sales, and service, the UFUASRP went with DJI maker of the Phantom platforms, which are
213 perhaps the top selling commercially available rotary wing sUAS to date. The DJI Spreading Wings S1000+ octocopter sUAS with a DJI WooKong M autopilot and DJI Lightbridge 2 video downlink w ere chosen as the rotary wing platform for the UFUASRP / USGS NUASPO collaboration. With an airframe and avionics mass of 4.2 kg, and a maximum takeoff mass of 11.0 kg, the S1000+ is capable of flying 6.8 kg of total battery, gimbal pod, and payload mass The 1.05 m maximum width airframe has eight 40 ampere (A) ESCs which each regulate a DJI 4114 Pro Motor spinning DJI 38.1 cm folding propellers. A 22.2 V Tru Power 30,000 mAh 15 cell LiPo battery having a mass of 3.8 kg powers all systems onboard th e S1000+ When the airframe is fully loaded with its heaviest payload and gimbal combination, the DJI S1000+ is at altitudes near sea level When the gimbal is 25 min is possible at altitudes near sea level The fabrication of the payload gimbals, the payload bays, and the integration of the appropriate sensors with GPS/INS data, imagery storage, and other details have taken some time to accomplish. However, while these activities were ongoing, the UFUASRP has been able to tune the aircraft and its autopilot, learn the DJI GCS software, and conduct several missions. The first missions were completed at the UF Ordway Swisher Biological Station near Melrose, Florida, to initially calibrate the optical sensors, and to experiment with various flight line spacing, altitudes AGL, and airspeed/groundspeed effects on sensor resolutions, imagery overlap, and gimbal sensitivity. Another mission included conducting multiple flights at the CM Stripling Irrigation Research Park near Camilla, G eorgia, over peanut fields and corn crops
214 grown under differing irrigation schedules. This data will be used to augment ongoing research that could potentially improve agricultural irrigation schedules for farmers in the southeastern US where water use an d water conservation tradeoffs are critical issues.
215 A B C D Figure A 1 Components of the University of Florida Nova 2.1 small unmanned aircraft system used to collect imagery and metadata of sagebrush steppe landscap es during summer 2013 and summer 2014 in Idaho, USA. A) The Nova 2.1 small unmanned aircraft. B) The ground control station, bidirectional communications box, and manual flight controller. C) The three person trained, qualified, and experienced flight t eam. D) The optical payload system composed of a system battery (bottom left, black), onboard computer (right, white), power switch (top left, black), MTi G combination global positioning system and inertial navigation system (center, orange) with antenn (center, red), universal serial bus 2.0 interfaces (center, black), all affixed to the back of a nadir oriented Olympus E 420 digital single lens reflex camera equipped with a 25 millimet C.R. Milling M.A. Burgess M.A. Burgess M.A. Burgess
216 Figure A 2 The Olympus E 420 (top) and the Canon EOS Rebel SL1 (bottom) optical sensor payloads. The power switch (a), onboard computer (b), system battery (c), universal serial bus 2.0 interfaces (d) payload synchronization device (e), MTi G global positioning system/inertial navigation system (f), and the MTi G antenna (g). M.A. Burgess M.A. Burgess
217 APPENDIX B A BRIEF HISTO R Y OF U NMANNED AIRCRAFT S YSTEM REGULATIONS IN THE U NITED S TATES Putting together information on the current regulatory environment in which unmanned aircraft systems ( UAS ) must operate in the United States ( US ) is difficult because it is a complex arrangement that up until fairly recently was frequently changing. This appendix provides a brief look at t he history of UAS regulation in the US, and is provided to help explain why the University of Florida (UF) Unmanned Aircraft Systems Research Program ( UFUASRP ) conducted small unmanned aircraft system ( sUAS ) flights in the manner described as it was based on the regulatory environment at the time. I n the US, the Federal Aviation Administration ( FAA ) is charged with supervising the safety and regulations of any contrivance invented, used, or designed to navigate, or fly in, the air (Title 49 US Code Section (Â§) 40102 and Â§ 40103; Title 14 Cod e of Federal Regulations ( CFR ) Â§ 1.1; and Title 14 CFR Â§ 91). The air located outdoors, over land, water, or territory of the US, from the surface of the Earth skyward is known as the N ational A irspace S ystem ( NAS ) The US military has worked cooperatively with the FAA since its inception to temporarily restrict air traffic in certain parts of the NAS from time to time to conduct testing and training exercises of both manned and unmanned aircraft in designated military o perations areas across the country. When the US military requests that the FAA close a specific military operations area to air traffic and the FAA approves the request, the military then assumes control, responsibility, and regulatory oversight for all a ircraft operations within that designated area for a predetermined period of time.
218 Because UAS are legally considered aircraft within the US (Public Law 112 95 Â§ 331 and Â§ 336), the FAA is responsible for ensuring the safety and regulation of unmanned syst ems in the same manner that they oversee manned aircraft operations in the NAS. Up until the mid 2000s, UAS in the US were primarily model remote controlled ( RC ) aircraft built and flown manually for recreation/hobby purposes, or military drones that were either preparing to be deployed or returning from theater overseas. At that time, the model aviation industry and the manned aircraft industry in the US were for the most part able to coexist amicably as they had done for decades prior with few documente d confrontations. The preeminent model aviation oversight organization in the US, the Academy of Model Aeronautics (AMA), issued safety guidelines and incentives for aviation modelers to operate safely which were largely followed, and the manned aircraft industry continued to function as it had with FAA oversight. However, by the mid 2000s, three major events occurred nearly simultaneously that altered the congruence between the manned aircraft, hobby aircraft, and UAS interests: 1) the desire of the US m ilitary to fly many of its large drones back to the US from theater and through the NAS to various facilities across the country for scheduled overhaul, maintenance, and repairs was not permitted by the FAA; 2) the initially slow rise in the numbers of mod el aviation hobbyists beginning to experiment with sUAS having autopilots and optical payloads quickly grew at an exponential rate; and 3) the general lack of the FAA to act on repeated expert recommendations to proactively prepare for an unparalleled upsu rge in sUAS technology that was just beginning, but had been forecasted to require a substantial number of additional FAA
219 personnel for integration and management, and the development of a regulatory framework catered specifically to sUAS. The FAA is faced with the unfortunate reality that the NAS is the busiest and most complicated airspace in the world, and with the safety of all the existing NAS users and the safety of people and property on the ground as its primary mandates, the FAA is exceptionally me thodical in generating new regulations; especially those that may somehow jeopardize the existing levels of safety. Initially, the FAA attempted to regulate all UAS as if they were manned aircraft, which led to a series of legal challenges, loopholes, and semantics. Realizing that unmanned aircraft ( UA ) were generally not compliant with many of t he various sections of rules and regulations for manned aircraft in the NAS which are found in Title 14 CFR, in September 2005 the FAA produced several alternativ e compliance methods for UA, including the Certificate of Waiver or Authorization (COA) methodology, and a set of interim UA regulations (AFS 400 UAS Policy 05 solutions imposed by the FAA affected all unmanned flight activities; including those being tested for natural resource based purposes The FAA and their parent organization, the US Department of Transportation (USDOT), were burdened with additional litigation, objections, challenges, and pus hback from model aviation enthusiasts and AMA administrators, legal teams from the Association of Unmanned Vehicle Systems International (AUVSI) the undisputed international society for all things unmanned, the US military, from Congress and the public. In response to the uproar, the FAA and USDOT further acknowledged that UAS were indeed intrinsically different from manned aircraft; stating that a set of rules and regulations specifically for
220 UAS operations in the NAS would need to be developed, but it w ould take some time to accomplish these tasks. To facilitate the need for UAS specific regulatory and NAS integration policies, the FAA created the Unmanned Aircraft Systems Integration Office (UASIO), and then followed that by releasing a document titled 01 This document further refined the 2005 interim framework in which UAS operations would be conducted in the NAS until the FAA UASIO c ould draft a finalized set of rules. The unmanned industry was reminded by the FAA that once a proposed set of final rules was drafted, it would then be available for a 60 day period of public comment, revised by the FAA and its legal staff, then the rev ised draft would be submitted to the FAA authorization board and then the USDOT authorization board for approval s. Once both authorization boards had given their approvals, the draft would be released back to the public, and then shortly thereafter writte n into the Federal Register, and finally 60 days later the rules would actually go into effect a lengthy process by any standards. It was about this period in time that sUAS for civil and commercial applications were gaining traction and transformation from merely abstract ideas to actually being constructed and in some cases being fielded as potential tool s for various applications such as law enforcement, telecommunications, surveying/ mapping and natural resources. In 2008, the FAA UASIO said publicly that the proposed set of final rules for UAS would be coming out within the next two years; however, 18 months after that announcement was made it became quite apparent that the final rules for UAS
221 integration into the NAS were still a ways away from bec oming a reality. In the meantime, the FAA UASIO continued to use the COA system as a method of regulating UAS operations in the NAS, and COAs that were approved were only issued to governmental agencies or public academic institutions. The UFUASRP worked with the FAA to obtain nearly 30 COAs, and was a UAS Research Program that earned FAA respect to legally fly sUAS within a tight set of regulations and operational guidelines that provided the UF Program to make significant strides in all pha ses of civilian sUAS research and development; especially for natural resource based applications of the technology. Businesses or commercial entities at that time were not authorized to use the COA process, and were forced to use an alternative method th at took considerable time and effort. The inability for US businesses or commercial entities to use the COA system ultimately put the US at a significant disadvantage behind most other nations developing sUAS from a business standpoint. On 14 February 20 12, seeing that the FAA UASIO was continuing to struggle to release a proposed set of final rules for routine UAS flights in the NAS, the FAA Modernization and Reform Act of 2012 (FMRA), (Public Law 112 95) was passed by Congress and signed by the Presiden t of the US. The FMRA mandated that the FAA UASIO develop a comprehensive plan for safely integrating UAS into the NAS by 30 September 2015. During the period between the issuance of the FMRA and the mandated deadline, the FAA UASIO sporadically would ame nd its interim UAS regulatory guidelines to allow certain low risk sUAS flights to take place in the NAS while it continued to draft a proposed set of final rules for sUAS. By introducing several key legislative initiatives during this period such as: 1) producing a multi year UAS
222 Integration Roadmap document in November 2013; 2) establishing six FAA UAS test sites across the US in December 2013; 3) issuing the first Public Law 112 95 Â§ 333 exemptions in June 2014 to several commercial movie and television production companies; and 4) releasing a proposed framework of regulations, or a notice of days of public comment on 23 February 2015 that outlined h ow the FAA proposed to integrate certain sUAS operations to fly routinely in the NAS; the FAA UASIO was able to somewhat pacify the budding sUAS industry and other critics in the US by shifting focus away from the looming congressionally mandated deadline, and focus on what the UASIO had been able to do for sUAS to date. As the 30 September 2015 deadline approached and passed without the FAA UASIO producing a set of final rules for sUAS integration, US sUAS industry leaders, investors, and others presented a letter to the administrator of the FAA, urging him to finalize a set of rules for sUAS as soon as possible, citing the significant benefits that the technology offered to such a wide variety of commercial, public, and academic efforts in the US. The le tter also indicated that the US was in a position to lose additional jobs and billions of US$ in sales to foreign entities the longer the country waited to break into the sUAS market with routine sUAS flights. Perhaps as an additional attempt to shift focu s away from the missed deadline, in December 2015 the FAA UASIO revisited the fact that small unmanned aircraft (sUA) were indeed legally considered aircraft just like their manned counterparts; therefore, in the interest of safety, sUA should be register ed in the same method that manned aircraft are registered as of February 2016 (Public Law 112 95 Â§ 331 and Â§ 336). By
223 registering an aircraft and marking it with an assigned FAA registration number, a sUA would be linked to the contact information for an individual via a nationally searchable database. Recently, the FAA mandated that all UAS be registered through an online portal, with the exception of micro unmanned aircraft systems (mUAS) which are defined below. Should a sUA be observed flying in a re ckless manner and the FAA registration number be obtained or illegally possessed an FAA registration number, the registered party could face legal consequences from the FAA and/or law enforcement ag encies. In the spring of 2016, the FAA UASIO clarified a series of definitions of UA classes based on the aircraft mass at takeoff, and elucidated interim guidelines for operating UAS in the NAS depending on the UA classification. According to the FAA UAS IO, an autonomous grams ( g ) at takeoff was classified as a micro UAS kilograms ( kg ) ed most UAS in th e US; therefore, the FAA UASIO prioritized establishing regulations for operating mUAS and sUAS in the NAS as their first priority under the FMRA Any UA platforms with a mass > 24.9 kg at takeoff in the US, were the FAA UASIO wo uld address regulatory matters for these larger unmanned systems upon completion of final rulemaking for the mUAS and sUAS categories. As of late spring 2016, the FAA UASIO had yet to produce a final rulemaking document for sUAS; stating that it was still reviewing each of the > 4,600 public comments that were submitted during the 60 day comment period from spring 2015, and wanted the issuance of the final rules to be appropriately restrictive, yet not too
224 overbearing to cease advancement of US based sUAS v entures. The FAA UASIO looked to incorporate suitable accommodations in the final sUAS rules to adapt to the rapid advancements in technology, in the hopes that they would not have to amend the final rules for sUAS immediately upon their release. Finding a suitable balance between restriction and accommodation continued to stall the release of final rules for sUAS [also known as (aka ) ] to the public. The final version of Part 107 was written into the US Federal Register a week later, making 29 August 2016 the date the Small Unmanned Aircraft Rules officially went into effect in the US. Some of the basic pol icies of Part 107 remain essentially unchanged from existing guidelines concerning sUAS flights in the NAS, e.g., the takeoff mass of the UA line of sight (VLOS) radius of the pilot in c ommand ( PIC ) at all times, the UA must give way to manned aircraft operations, and the PIC can only operate a single UA at a time, etc. Some of the changes incorporated into Part 107 included removing the stipulation that the PIC be a licensed and current manned aircraft pilot holding a Class 2 medical clearance, the elimination of the provision requiring a pilot to provide the FAA with a certificate of aircraft airworthiness, and Part 107 established that all sUAS flight must remain below 121.9 meters ( m ) above ground level ( AGL ) unless the flight is within a 121.9 m radius of a structure (such as a bridge, stack, silo, radio tower, etc.). With the changes mentioned above, the FAA also establish ed : 1) a remote pilot airman certificate requirement and pla ced the medical onus decision on the PIC; 2) that the
225 PIC inspect the aircraft for airworthiness and declare its condition as safe for operation before each flight; and 3) a clause which eliminated the need for certain sUAS flights to obtain a COA or other FAA exemption before flight as long as specific provisions within a detailed set of rules were met. Prudently, Part 107 was constructed with a framework based on the recognition that sUAS technology is evolving and maturing at a rapid pace, and therefore the rules were composed in such a way to best accommodate the anticipated changes.
226 APPENDIX C THE UNIVERSITY OF FLORIDA UNMANNED AIRCRAFT SYSTEMS RESEARCH PROGRAM ( UFUASRP ) FLIGHT CREW MODEL (UA) opera tions regarding the piloting requirements of small unmanned aircraft systems ( sUAS ) the University of Florida Unmanned Aircraft Systems Research Program ( UFUASRP ) utilized a flight crew model that was highly regarded by the Federal Aviation Administration (FAA). Even since the Part 107 changes have been instituted, the UFUASRP continues to requirements for sUAS pilots outlined in Part 107 to the existing qualifications necess ary for the flight crewmember roles. The UFUASRP flight crew model consists of a minimum of three individuals, each with a specific set of duties, qualifications, and responsibilities. All three crewmembers of a UFUASRP flight team have completed and pass ed an FAA accredited ground school curriculum that provides considerable knowledge of the basics of aviation in the National Airspace System (NAS). Since the implementation of the Part 107 regulations, the flight team members have also achieved FAA sUAS p ilot certificates as well. The three individuals that make up a UFUASRP flight crew consist of the following: 1) a pilot in command (PIC) 2) a ground station operator (GSO) and 3) a qualified visual observer (QVO) Pilot in Command The P IC minimally hol ds a Title 14 Code of Federal Regulations (CFR) Part 61 certification as a licensed and current manned private pilot with a Class 2 Medical Certificate. The PIC is responsible for keeping visual contact on the UA at all times during a flight. While keepi ng eyes on the UA, the PIC holds a traditional two stick
227 remote control (RC) controller or gamepad device, and has the ability to instantaneously regain full manual RC control of the aircraft with the flip of a single toggle switch or the press of a button should the need arise. The PIC is ultimately the legal party responsible for the flight of the aircraft (both under autonomous or manual control), even though a team of individuals is involved in conducting a flight. Ground Station Operator The GSO recor ds a written flight log for each flight attempt, and monitor s the ground control station (GCS) and telemetry data throughout the entirety of a flight. The GSO scans the GCS and oversees items such as : power and connectivity strength of the bidirectional c ommunication signals, system battery voltage, indicated airspeed, number of global positioning system (GPS) satellite signals the aircraft is locked onto, the pitch, roll, and yaw of the aircraft, the total flight time, winds aloft speed and direction, UA ground speed, and a moving map of the aircraft position showing the location, range, and bearing of the UA from the GCS and the preloaded flight path and failsafe protocols. During the flight, the GSO is also responsible for announcing upcoming UA turns, manual changes to the flight path, etc. so that the other flight crew personnel are aware of impending aircraft actions before observing them. This helps the PIC recognize a commanded alteration in the flight of the aircraft that could otherwise be misint erpreted as a flight anomaly cueing the PIC to assume manual RC control. The GSO reports if any of the observed parameters shown on the GCS extend beyond their nominal ranges so that the entire flight team can determine an appropriate course of action. A s required, the GSO may deviate the aircraft from the preprogrammed flight path (i.e.: have the aircraft loiter about its current position, manually increase or decrease the commanded airspeed or altitude, etc.), or switch
228 between the various flight modes (e.g.: takeoff climb, waypoint navigation, loiter now, return to home, initiate a landing sequence, etc.) depending on the flight as it evolves. Qualified Visual Observer A QVO also holds an FAA Class 2 Medical Certificate, prepares the airframe and payloa d for flight, launches the aircraft, and then is responsible for scanning the airspace within and beyond the FAA mandated 1.85 kilometer (km) radius visual line of sight (VLOS) sUAS operational area looking for potential hazards. Examples of hazards inclu de any other aircraft, flocks of birds, approaching weather, etc. Should a potential hazard present itself during a flight, the QVO announces the identification and re flight team can collectively decide on a course of action to mitigate a potential incursion. In some cases, the UFUASRP will have several QVOs available for a flight, which is advantageous in that more eyes and ears are available for scanning the opera tional area for potential hazards. The division of labor delineated above helps to ensure that: 1) safety is never compromised from flight to flight; 2) crucial steps in preparation, execution, and termination of an unmanned flight are performed orderly an d completely which helps curtail an inadvertent omission of an essential task; and 3) with routine practice together, the flight crew becomes a cohesive team which conducts missions with effective coordination of the many critical elements necessary to ach ieve success. whereby the three flight crewmembers are able to communicate with each other using an ordinary speaking volume, conversation is only in regards to the flight b eing conducted, and any non crewmembers nearby are asked to withhold questions or
229 comments until after a flight is completed, or asked to move to a safe and remote location before takeoff so that the sterile cockpit conditions can be maintained until the a ircraft is safely back on the ground after a flight. A final component of the UFUASRP flight crew model that also promotes safety is that the Research Program operates with a policy that at any time before, during, or after a flight, if any member of the f light crew feels that conducting an upcoming flight is not a sensible decision, or if the success of a flight is in doubt, all a crewmember has to do is speak up and say something to the rest of the crew. At that point, a flight will not be initiated, or a flight in progress will be safely terminated as quickly as possible without question. This policy provides each flight crewmember an equal voice within the team, and aids in the overall safety of the flight operations.
230 APPENDIX D DIGITAL IMAGERY AND M ETA DATA POST PROCESSING METHODOLOGY USED All imagery and its associated metadata were preprocessed so that they could be input directly into advanced imagery processing software. The MTi G global positioning system/inertial navigation system ( GPS/INS ) da ta log file which recorded multitudes of data parameters per second ( sec ) during the flight was parsed via a custom Python script to generate a geotag text file which contained delimited image identification, three dimensional ( 3D ) positional information, and roll, pitch, and yaw data of the imagery sensor at the exact moment of maximum shutter aperture for each image captured. Using ArcMap software, the geotag text file metadata of the exposure stations was added to a basemap as a shapefile layer. Casu al visual analyses of the exposure station distributions provided a rough estimate of how well the focal target area was covered with imagery during the flights. selected to run the ad collective amount of total system memory, clock speed, and number of core processors available on the system and video hardware boards), study site subareas were required to be broken into small er and more manageable chunks with the ArcMap software to avoid potentially crashing the image processing software. Multiple flights over a specific target area, or independent flights could be post processed individually d epending on the particular stud y Advanced imagery post processing software titles many designed especially for small unmanned aircraft system ( sUAS ) imagery have rapidly improved over the last five years. Until fairly recently, p ost processing software had comp utational operations
231 that would often take days to complete; however, through advances in computing technology, the ability for software manufacturers to update their products and instantly users, and cutting edge pho togrammetric and computer vision based algorithms used in newer software titles, were able to accomplish even more complex tasks than their predecessors, with steps measured in hours (hr) or min utes (min) rather than days. Several specific software titles have excelled in their ability to create 3D imagery products suitable for planimetric measurements from two dimensional (2D) imagery from (e. g., Turner et al. 2012 Westoby et al. 2012 Fonstad et al. 2013 Lucieer et al. 2014 Burns et al. 2015 Madden et al. 2015 ) A popular SfM software title was used for imagery post processing, Agisoft LLC PhotoScan Professional Edition, to produce the desired imagery products necessary to address the scientific questions in this dissertation Using PhotoScan the imagery from a particular chunk was loaded, the camera calibration p arameters were entered, and the delimited exposure station geotag data were added. Next, the images were aligned to create a digital sparse point cloud using a three step automated image correlation process of detecting tie points, selecting pairs of imag es, and matching the tie points. This activity would repeat itself until all of the tie points and image pair combinations had been assessed. Images that could not be automatically aligned were selectively disabled, and the alignment procedure was conduc ted again to create an updated and more precise digital sparse point cloud. The 3D metadata for the ground control points ( GCPs ) was added during the next phase of post processing. After the GCP metadata was entered, the software
232 would identify specific f light images that contained a GCP, and would ask the user to fine tune the placement of the GCP at the center of each marker. The next post processing step performed a photogrammetric bundle adjustment by optimizing the imagery alignment. This procedure helped account for lens distortions based on the camera focal length, the principal point of the lens, and a portion of the radial distortion coefficients; all of which helped reduce the root mean square (RMS) error, and provided a nominal indication of th e accuracy of the digital sparse point cloud model at the current stage of imaging processing. Through a series of three separate refining processes, the geometry of the model being constructed via modified aerotriangulation techniques was improved by eli minating points collectively distorting the level of reconstruction uncertainty, the level of projection accuracy, and the level of reprojection error. Also by incorporating the remaining camera alignment and distortion parameters, and tightening the accu racy threshold of the tie points, the RMS error of the digital sparse point cloud model continued to improve after each successive step of the process. The next phase of imagery post processing using PhotoScan was the creation of a dense point cloud model based on information derived during the refinements of the digital sparse point cloud. Once the dense point cloud was constructed, a 3D polygon mesh of the model was generated which helped account for variations in topographic relief, and an image overla y of texture for the model was built. After these stages of post processing had been completed, the creation of a digital elevation model (DEM) in a desired coordinate system and projection was conducted, and an overlay of isometric contour lines at user specified elevation intervals was provided as an option. Next was
233 the creation of an orthophotomosaic via orthorectification methods in a user desired coordinate system and projection. The final post processing step was to merge any chunks that were post processed independently into much larger DEMs and orthophotomosaics of the entire scene if necessary for subsequent analyses. A plethora of products that were generated during the PhotoScan imagery post processing workflow were then available for export in various file formats for visual display, or to be used as input files into other software titles for additional analyses or investigations.
234 APPENDIX E SEVERAL CRITICAL LESSONS LEARNED BY THE UNIVERSITY OF FLORIDA UNMANNED AIRCRAFT SYSTEMS RESEARCH PRO GRAM (UFUASRP): 1999 2017 T he University of Florida ( UF ) Nova 2.1 fixed wing small unmanned aircraft system ( sUAS ) was specifically designed as a precision natural resources data collection tool. The airframe of the Nova 2.1 was designed around the imagin g sensor and payload components necessary to complete missions that specifically address natural resource based questions. This is a distinctly different approach than that which is often used by other entities. Many who seek to collect aerial data for n atural resource based applications using sUAS begin by acquiring an airframe and then attempt to find a sensor or a suite of sensors that will fit into, or attach onto the fuselage. In learning by doing over the last 18 years, the UFUASRP has determined t hat the most efficient approach to maximizing sUAS utility for multiple natural resources applications is to follow a course of action based on answers to questions raised in a decision matrix outlined below. A critical lesson was realized by the UFUASRP j ust before the design of the UF Nova 2.1 sUAS; a lesson that applied to all unmanned aircraft system ( UAS ) applications, not just those primarily focused on natural resources. Although it took the UFUASRP nearly a decade of airframes, optical payloads, an d experience fielding sUAS to recognize the gaffe that had been repeatedly made, once the Research Program identified that to achieve the utmost efficiency and utility out of a sUAS airframe and payload, adhering as closely as possible to the methodology f or sUAS project success detailed below was frankly indispensable. In any project /study/endeavor and series of details/processes/particulars located in between which must be completed
235 or accounted for to Figure E 1 ). This concept is not novel, any project because obscurities in either endpoint lead to ineffic iencies in progressing stretching the scarce resources that are avail able. The UFUASRP has found that abiding by the following methodology provides a remarkably efficient means for garnering the most utility out of a sUAS for practically all applications. This methodology is described in a manner as if a single objective i s the endpoint of a sUAS data collection project; however, this is solely for the sake of simplicity of conveying the critical steps in the process. It has been found that the methodology is valid for much more complicated efforts and objectives as well. As overstressed when considering a sUAS as an aerial sensor platform for data collection purposes. It all begins with pinpointing the specific scientific question to be answered, specifics that can be used to narrow down a generally broad problem into a focused query, the more resourceful and productive the use of this methodology is for s UAS projects. Once the scientific question has been thoroughly defined and framed along with amount of data necessary to collect to assist in answering the scientific questi on at hand. Critical sub steps in determining the types of data to be collected include
236 electromagnetic (EM) spectra range, minimum and maximum useable resolution of sensor products, frequency of data collection, season of data collection, metadata accura cy, etc., which then leads to the selection of the appropriate sensor. The selection of the payload sensor depends on a series of factors primarily driven by the find ing a sensor unit that has a relatively small mass and volume as well; that being said, at least in most natural resource based sUAS applications, the sensor selection is often ultimately driven by the available budget. The UFUASRP has realized over the ye ars, nearly every type of sensor that a researcher could realistically want to include on a sUAS has probably already been developed and used by the military for intelligence, surveillance, or reconnaissance ( ISR ) or other purposes on drones in theater A dditionally, many of those very sensors could be used to assist in answering some of the most critical and salient scientific questions faced by researchers today. Unfortunately, the sensors used by the military in their existing form factor are commonly not small enough in mass or volume to serve as payload items for sUAS in the size range of most civilian users. Furthermore, those military sensors that are physically small enough to be payload items on civilian use sUAS are largely cost prohibitive and/ or subject to laws restricting their sale and usage to entities outside of the military. In nearly all cases, a US$ 1,500 sensor is better than a US$ 150 sensor, and a US$ 15,000 sensor is considerably better than a US$ 150 sensor. Users of sUAS need to b e aware of import and export laws about sensitive items that may be included as part of a sUAS purchase. The US Department of State
237 (USDOS) has a list of items that are restricted by International Traffic in Arms Regulations (ITAR), and the US Department of Commerce (USDOC) has a similar list of items that are restricted by Export Administration Regulations (EAR). Both of these lists of items specify who, what, where, when, how, and why, a person has access to individual components and/or their schematics imagery, or collected data because they be used against the US in a nefarious manner. Items such as higher end global positioning system/inertial navigation system ( GPS/INS ) units, autopilots, and high resolution sensors, etc., are just a few of the items that may appear on an ITAR or EAR list. The responsibility of complying with ITAR and EAR laws falls on the PIC to ensure that no laws are broken, and that all per sonnel having access to controlled items have been vetted in some manner by qualified authorities. Both commercial and industrial sUAS manufacturers abound who are so focused on making sales to turn a profit, that they may very willingly sell a sUAS that contains ITAR or EAR controlled items to a customer without indicating that said items are part of the purchase. It is up to the buyer to make certain that the products being purchased do or do not include restricted items, and are handled appropriately. Penalties and consequences of violating Federal import or export regulations can lead to incarceration and serious fiscal punishments. The UFUASRP has witnessed over the last decade, both patience and persistence have generally been rewarded with advance s in technology that have improved virtually all EM spectra sensors. With increasing demand for smaller, lighter, and lower power consuming sensor units, additional players are now in the market
238 initiating competition into what was predominantly a monopol y held by a handful of corporations who catered primarily to the military. Increased competition has resulted in lower acquisition costs for nearly all sensors, especially for users with reduced fiscal budgets such as those found in natural resources and environmental fields of study. After the sensor has been selected, the next step in the methodology is to find an airframe that will appropriately position the sensor over the target area of interest. The UFUASRP has found that four primary questions abou t the anticipated sUAS flights generally dictate the airframe selection; however, each of these primary questions has a series of sub questions and tradeoffs that may alter the final airframe decision. The four primary questions are : 1. How much ground area needs to be covered in a single day? 2. How long must the airframe stay aloft for the payload sensor to capture the data required to answer the scientific question? 3. What are the total mass, volume, and power requirements of the optical payload components? 4. Wha t are the environmental conditions in which the sUAS will be expected to operate? The first question determines whether a fixed wing or a rotary wing sUAS airframe is the most appropriate platform design for the intended mission. If a significant area o f interest [ > 300 hectares ( ha) ] needs to be flown per day, then a fixed wing platform is highly recommended. If the target area of ground coverage is smaller, and wing platform is likely a better choice. Other considerations that should be addressed about the selection of a sUAS platform design, especially for natural resource based users include focal and non focal target noise tolerance levels resemblance of the airframe to
239 a predator of the focal target, and flight crew training/capabilities of flying either a fixed or rotary wing sUAS. The second question determines the sUAS airframe primary source of power, e.g., battery, or liquid fuel, that is most appropriate for the in tended mission. Maximum flight time for most battery powered rotary wing platforms is < 25 minutes ( min ) while maximum flight time for most battery powered fixed wing platforms is < 75 min. Liquid fueled airframes vary in flight time capabilities based on many factors; however nearly all liquid fueled airframes can run for > 60 min before having to land for refueling. Considerations that play a part in determining the airframe source of power, especially for natural resource users, are sensitivities of the focal and non focal targets to noise, obtaining or transporting liquid fuels may present problems for travel or during fieldwork, determining how much data the optical payload can store with increasingly larger data sets and lengthy airframe runtimes, and with the 1.85 kilometer ( km ) visual line of sight (VLOS) radius restriction in place, a user can only cover an operational circle of just less than 1,100 ha before repeating coverage or landing and moving the ground control station ( GCS ) to a new 1.85 km radius VLOS circle center location. Coverage of 1,100 ha is a large area, and while many researchers aim to collect as much data as possible, a factor that must be kept in mind is that all of the data collected will need to be post processed in some fa shion; otherwise, time and resources will be wasted. The third question helps determine the physical size of the sUAS airframe. If a micro air vehicle ( MAV ) is all that is needed to gather the desired data, there is no need to arrange for a Northrup Grumm an Global Hawk UAS to serve as the aerial sensor platform. Ideally, an airframe that is sufficiently rated for the anticipated payload mass
240 and volume in addition to the mass and volume of the airframe, avionics, and propulsion system would be recommend ed. Other considerations, especially for natural resource based users include airframe portability, the need to use and consequently tote specialty launching equipment, the cost per hour for contracting the airframe (if that route of data collection is se lected), or the cost per h ou r of the contracted flight crew (if that route is selected). The final question to ask in regards to airframe selection is to account for the effects of the anticipated environmental conditions in which the small unmanned aircra ft ( sUA ) airframe will be operated. Airframes purchased commercial off the shelf ( COTS ) and certainly those that are hand fabricated, have not necessarily been tested or rated as airworthy for the environmental conditions in which they may be selected fo r routine operations. Just a few of the environmental items of the flight operational area that need to be considered when selecting a sUAS airframe, particularly for natural resource based users of sUAS include : 1. Ambient temperature, humidity, and dew poi nt 2. Ambient wind speed and direction 3. Typical weather patterns for the season 4. Hours of operational daylight available 5. Elevation of the ground above sea level (ASL) 6. Air density and density altitude calculations 7. Cloud ceilings 8. Terrain and appropriate takeoff, landing, and emergency landing locations 9. Ability to maintain a 1.85 km VLOS radius of the aircraft at all times 10. Icing of the wing surfaces and/or pitot tube 11. Overheating of electronic components 12. Performance of the propulsion system components in the ambient conditions 13. Stability of autonomous and/or manual flight over intense thermal changes, i.e., fire, volcanic activity, etc. 14. Communication reliability based on known sources of nearby interference
241 As mentioned, this is not an exhaustive list, however select ing an effective s UA airframe depends on a plethora of factors; and for these reasons, just like sensor selection, no single airframe is appropriate for all types of applications. Once the airframe has been selected, there are a substantial number of addit ional details that need attention in order to successfully and efficiently ending up at However, for the sak e of simplicity, following the process of first defining the scientific question and desired end products next determining the data collection needs, followed by acquiring the appropriate sensor, then selecting the suitable airframe based on design, power source, physical size, and environmental conditions needed for the application, has shown to be an exceptionally efficient method of getting from point Figure E 2 ).
242 Figure E 1. An illustration showing the generalized process of proceeding from a starting point to an ending point. Once the endpoints are clearly established, a series of steps must be completed in between to move from left to right. The order in which these steps are conducted can dramatically affect the efficiency of the whole process of moving from the start to the end. Figure E 2. A recommended methodology fo r efficiently moving from the start of a scientific project in which a small unmanned aircraft system (sUAS) will be used for data collection to achieving the desired end products necessary for analyses. Always start by defining the scientific question to be answered, and the desired end products to be generated as specifically as possible at the onset. Through experience, it has been found that starting a project with an airframe, or a sensor first, and then trying to develop a methodology to get to the desired endpoint can be extremely inefficient. However, moving from left to right using the methodology contained within the green box has repeatedly shown to be an exceptionally efficient technique for selecting and employing the appropriate sUAS compone nts to meet the desired end products. M.A. Burgess M.A. Burgess
243 APPENDIX F THE UNIVERSITY OF FLORIDA UNMANNED AIRCRAFT SYSTEM S RESEACH PROGRAM (UFUASRP) BLANK FLIGHT DATA SHEET
244 LIST OF REFERENCES Abd Elrahman, A. H., L. G. Pearlstine, B. A. Dewitt, and S. E. Smith. 2001. Detection of positional errors in systems utilizing small format digital aerial imagery and navigation sensors using area based matching techniques. Photogrammetric Engineering & Remote Sensing 67:825 831. Abd Elrahman, A. H., L. G. Pearl stine, and H. F. Percival. 2005. Development of pattern recognition algorithm for automatic bird detection from unmanned aerial vehicle imagery. Surveying and Land Information Science 65:37 45. Abd Elrahman, A. H., L. G. Pearlstine, S. E. Smith, and P. Pr incz. 2000. Use of small format digital aerial images for classification of satellite images, Page 6 in XVI IMEKO World Congress. 25 28 September 2000. Agapiou, A., D. G. Hadjimitsis, and D. D. Alexakis. 2012. Evaluation of broadband and narrowband veget ation indices for the identification of archaeological crop marks. Remote Sensing 4:3892 3919. Alldredge, M. W., K. H. Pollock, and T. R. Simons. 2006. Estimating detection probabilities from multiple observer point counts. The Auk 123:1172 1182. Ambrosi a, V. G., S. S. Wegener, T. Zajkowski, D. V. Sullivan, S. W. Buechel, F. Enomoto, B. M. Lobitz, S. Johan, J. A. Brass, and E. Hinkley. 2011. The Ikhana unmanned airborne system (UAS) western states fire imaging missions: from concept to reality (2006 2010) Geocarto International 26:85 101. Anderson, D. R. 2001. The need to get the basics right in wildlife field studies. Wildlife Society Bulletin 29:1294 1297. Anderson, D. R. 2003. Response to Engeman: index values rarely constitute reliable information. Wildlife Society Bulletin 31:288 291. Anderson, R. C., P. C. Shanks, L. A. Kritis, and M. G. Trani. 2014. Supporting remote sensing research with small unmanned aerial systems, Pages 51 56 in The International Archives of the Photogrammetry, Remote Sensin g and Spatial Information Sciences. 17 20 November 2014. Anderson, T. M., J. G. C. Hopcraft, S. Eby, M. Ritchie, J. B. Grace, and H. Olff. 2010. Landscape scale analyses suggest both nutrient and antipredator advantages to Serengeti herbivore hotspots. E cology 91:1519 1529. August, T., M. Harvey, P. Lightfoot, D. Kilbey, T. Papadopoulos, and P. Jepson. 2015. Emerging technologies for biological recording. Biological Journal of the Linnean Society 115:731 749.
245 Balmori, A. 2014. Design of secondary contro l surfaces on unmanned aircraft systems to achieve low glide slope ratios. Master's Thesis, University of Florida, Gainesville, Florida, USA. Bart, J., S. Droege, P. Geissler, B. Peterjohn, and C. J. Ralph. 2010. Density estimation in wildlife surveys. Wi ldlife Society Bulletin 32:1242 1247. Beaver, J. T., C. A. Harper, R. E. Kissell Jr., L. I. Muller, P. S. Basinger, M. J. Goode, F. T. Van Manen, W. Winton, and M. L. Kennedy. 2014. Aerial vertical looking infrared imagery to evaluate bias of distance sam pling techniques for white tailed deer. Wildlife Society Bulletin 38:419 427. Becker, P. A., D. W. Hays, and R. D. Sayler. 2011. 2011 Columbia Basin pygmy rabbit ( Brachylagus idahoensis ) reintroduction and genetic management plan: addendum to Washington S tate recovery plan for the pygmy rabbit (1995). Washington Department of Fish and Wildlife. Olympia, Washington, USA. Benson, B. J., B. J. Bond, M. P. Hamilton, R. K. Monson, and R. Han. 2010. Perspectives on next generation technology for environmental sensor networks. Frontiers in Ecology and the Environment 8:193 200. Berni, J. A. J., P. J. Zarco Tejada, L. Suarez, V. Gonzalez Dugo, and E. Fereres. 2009. Remote sensing of vegetation from UAV platforms using lightweight multispectral and thermal imagin g sensors, Page 6 in The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. C. Heipke, K. Jacobsen, S. Muller, and U. Sorgel, editors. 2 5 June 2009. Bevan, E., T. Wibbels, B. M. Z. Najera, M. A. C. Martinez, L. A. S. Martinez, F. I. Martinez, J. M. Cuevas, T. Anderson, A. Bonka, M. H. Hernandez, L. J. Pena, and P. M. Burchfield. 2015. Unmanned aerial vehicles (UAVs) for monitoring sea turtles in near shore waters. Marine Turtle Newsletter 145:19 22. Blickley, J L., K. Deiner, K. Garbach, I. Lacher, M. H. Meek, L. M. Porensky, M. L. Wilkerson, E. M. Winford, and M. W. Schwartz. 2012. Graduate student's guide to necessary skills for nonacademic conservation careers. Conservation Biology 27:24 34. Booth, D. T., a nd S. E. Cox. 2011. Art to science: tools for greater objectivity in resource monitoring. Rangelands 33:27 34. Boswell, A. K. 2015. Rangeland monitoring using remote sensing: an assessment of vegetation cover comparing field based sampling and image analy sis techniques. Master's Thesis, Brigham Young University, Provo, Utah, USA. Boucher, P. 2015. Domesticating the drone: the demilitarisation of unmanned aircraft for civil markets. Science and Engineering Ethics 21:1393 1412.
246 Bowman, W. S. 2008. Design a nd validation of autonomous rapid mapping system using small UAV. Master's Thesis, University of Florida, Gainesville, Florida, USA. Brooke, S., D. Graham, T. Jacobs, C. Littnan, M. Manuel, and R. O'Conner. 2015. Testing marine conservation applications o f unmanned aerial systems (UAS) in a remote marine protected area. Journal of Unmanned Vehicle Systems 3:237 251. Burns, J. H. R., D. Delparte, R. D. Gates, and M. Takabayashi. 2015. Integrating structure from motion photogrammetry with geospatial softwar e as a novel technique for quantifying 3D ecological characteristics of coral reefs. PeerJ 3:e1077. Camp, M. J., J. L. Rachlow, B. A. Woods, T. R. Johnson, and L. A. Shipley. 2012. When to run and when to hide: the influence of concealment, visibility, an d proximity to refugia on perceptions of risk. Ethology 118:1010 1017. Camp, M. J., J. L. Rachlow, B. A. Woods, T. R. Johnson, and L. A. Shipley. 2013. Examining functional components of cover: the relationship between concealment and visibility in shrub steppe habitat. Ecosphere 4:1 14. Camp, M. J., L. A. Shipley, T. R. Johnson, J. S. Forbey, J. L. Rachlow, and M. M. Crowell. 2015. Modeling trade offs between plant fiber and toxins: a framework for quantifying risks perceived by foraging herbivores. Ecol ogy 96:3292 3302. Caughley, G. J. 1974. Bias in aerial survey. The Journal of Wildlife Management 38:921 933. Caughley, G. J. 1977. Sampling in aerial survey. The Journal of Wildlife Management 41:605 615. Cerra, D., J. Bieniarz, F. Beyer, J. Tian, R. M uller, T. Jarmer, and P. Reinartz. 2016. Cloud removal in image time series through sparse reconstruction from random measurements. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 9:3615 3628. Chabot, D. M. 2014. The rise of unmanned aircraft in wildlife science: a review of potential contributions and their application to waterbird research. Doctoral Dissertation, McGill University, Montreal, Canada. Chong, S. C., and K. K. Evans. 2011. Distributed versus focused attentio n (count vs estimate). WIREs Cognitive Science 2:634 638.
247 Christiansen, R. S. 2004. Design of an autopilot for small unmanned aerial vehicles. Master's Thesis, Brigham Young University, Provo, Utah, USA. Colomina, I., and P. Molina. 2014. Unmanned aerial systems for photogrammetry and remote sensing: a review. ISPRS Journal of Photogrammetry and Remote Sensing 92:79 97. Cook, M. I. 2013. South Florida wading bird report. Volume 19. South Florida Water Management District. West Palm Beach, Florida, USA. Cramer, M. D., and N. N. Barger. 2014. Are mima like mounds the consequence of long term stability of vegetation spatial patterning? Palaeogeography, Palaeoclimatology, Palaeoecology 409:72 83. Cramer, M. D., D. Stallmann, and N. Haala. 2000. Direct geor eferencing using GPS/inertial exterior orientations for photogrammetric applications, Pages 198 205 in International Archives of Photogrammetry and Remote Sensing. Cress, J. J., M. E. Hutt, J. L. Sloan, M. Bauer, M. Feller, and S. Goplen. 2015. U.S. Geolo gical Survey unmanned aircraft systems (UAS) roadmap 2014. Open File Report 2015 1032. U.S. Geological Survey. Reston, Virginia, USA. Crouch, T. D. 2004. Wings: a history of aviation from kites to the space age. Reprinted. W.W. Norton & Company, New York New York, USA. Crowell, M. M., L. A. Shipley, M. J. Camp, J. L. Rachlow, J. S. Forbey, and T. R. Johnson. 2016. Selection of food patches by sympatric herbivores in response to concealment and distance from a refuge. Ecology and Evolution 6:2865 2876. Dandois, J. P., and E. C. Ellis. 2013. High spatial resolution three dimensional mapping of vegetation spectral dynamics using computer vision. Remote Sensing of Environment 136:259 276. Edgel, R. J. 2013. Habitat selection and response to disturbance by pygmy rabbits in Utah. Master's Thesis, Brigham Young University, Provo, Utah, USA. Edgel, R. J., J. L. Pierce, and R. T. Larsen. 2014. Pygmy rabbit ( Brachylagus idahoensis ) habitat selection: does sagebrush ( Artemisia spp.) age influence selection? Weste rn North American Naturalist 74:145 154. Elias, B. 2016. Unmanned aircraft operations in domestic airspace: U.S. policy perspectives and the regulatory landscape. R44352. Congressional Research Service. Engeman, R. M. 2003. More on the need to get the b asics right: population indices. Wildlife Society Bulletin 31:286 287.
248 Estes Zumpf, W. A., and J. L. Rachlow. 2009. Natal dispersal by pygmy rabbits ( Brachylagus idahoensis ). Journal of Mammalogy 90:363 372. Eugster, H., and S. Nebiker. 2008. UAV based a ugmented monitoring -real time georeferencing and integration of video imagery with virtual globes, Pages 1229 1236 in The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Evers, B. S. 2011. Improving operatio nal efficiency of small unmanned aircraft systems for remote sensing applications. Master's Thesis, University of Florida, Gainesville, Florida, USA. Fahlstrom, P. G., and T. J. Gleason. 2012. Introduction to UAV systems. Fourth edition. John Wiley & Sons Ltd., Chichester, West Sussex, United Kingdom. Fewster, R. M., C. Southwell, D. L. Borchers, S. T. Buckland, and A. R. Pople. 2008. The influence of animal mobility on the assumption of uniform distances in aerial line transect surveys. Wildlife Researc h 35:275 288. Finkl, C. W., and C. Makowski, editors. 2014. Remote sensing and modeling: advances in coastal and marine resources. Volume 9. Springer. Fleming, P. J. S., and J. P. Tracey. 2008. Some human, aircraft and animal factors affecting aerial sur veys: how to enumerate animals from the air. Wildlife Research 35:258 267. Fonstad, M. A., J. T. Dietrich, B. C. Courville, J. L. R. Jensen, and P. E. Carbonneau. 2013. Topographic structure from motion: a new development in photogrammetric measurement. E arth Surface Processes and Landforms 38:421 430. Francis, C. D., J. Paritsis, C. P. Ortega, and A. Cruz. 2011. Landscape patterns of avian habitat use and nest success are affected by chronic gas well compressor noise. Landscape Ecology 26:1269 1280. Fre derick, P. C., B. Hylton, J. A. Heath, and M. Ruane. 2003. Accuracy and variation in estimates of large numbers of birds by individual observers using an aerial survey simulator. Journal of Field Ornithology 74:281 287. Fremgen, M. R. 2015. Plant toxins i nfluence diet selection and intestinal parasites in a specialist herbivore. Master's Thesis, Boise State University, Boise, Idaho, USA. Gabler, K. I., J. W. Laundre, and L. T. Heady. 2000. Predicting the suitability of habitat in southeast Idaho for pygmy rabbits. The Journal of Wildlife Management 64:759 764.
249 Galan Diaz, C., P. Edwards, J. D. Nelson, and R. van der Wal. 2015. Digital innovation through partnership between nature conservation organisations and academia: a qualitative impact assessment. Am bio 44:S538 S549. Gann, D., and J. H. Richards. 2009. Determine the effectiveness of plant communities classification from satellite imagery for the greater Everglades freshwater wetlands & community abundance, distribution and hydroperiod analysis for WC A 2A. Final Report. Florida Internationl University. Miami, Florida, USA. Gann, D., J. H. Richards, and H. Biswas. 2011. Consulting services to determine the effectiveness of vegetation classification using WorldView 2 satellite data for the greater Ever glades. Synthesis Report. Florida International University. Miami, Florida, USA. Giudice, J. H., J. R. Fieberg, M. C. Zicus, D. P. Rave, and R. G. Wright. 2010. Cost and precision functions for aerial quadrat surveys: a case study of Ring necked Ducks in Minnesota. The Journal of Wildlife Management 74:342 349. Gordon, C., M. Kujawa, J. Luttrell, D. K. MacArthur, J. Robinson Willmott, and C. Thaxter. 2013. High resolution aerial imaging surveys of marine birds, mammals, and turtles on the US Atlantic out er Continental Shelf -utility assessment, methodology recommendations, and implementation tools. BOEM 2013 01130. Normandeau Associates, Inc. Herndon, Virginia, USA. Green, J. S., and J. T. Flinders. 1979. Techniques for capturing pygmy rabbits. The Murrelet 60:112 113. Green, J. S., and J. T. Flinders. 1980a. Brachylagus idahoensis Mammalian Species 125:1 4. Green, J. S., and J. T. Flinders. 1980b. Habitat and dietary relationships of the pygmy rabbit. Journal of Range Manageme nt 33:136 142. Grenier, M. B., S. W. Buskirk, and R. Anderson Sprecher. 2009. Population indices versus correlated density estimates of black footed ferret abundance. The Journal of Wildlife Management 73:669 676. Gunasekara, N. K., M. M. Al Wardy, G. A. Al Rawas, and Y. Charabi. 2015. Applicability of VI in arid vegetation delineation using shadow affected SPOT imagery. Environmental Monitoring and Assessment 187:454 464. Hardin, P. J., and T. J. Hardin. 2010. Small scale remotely piloted vehicles in en vironmental research. Geography Compass 4:1297 1311.
250 Hardin, P. J., and R. R. Jensen. 2011. Small scale unmanned aerial vehicles in environmental remote sensing: challenges and opportunities. GIScience & Remote Sensing 48:99 111. Hartter, J., S. J. Ryan, C. A. MacKenzie, A. Goldman, N. Dowhaniuk, M. Palace, J. E. Diem, and C. A. Chapman. 2015. Now there is no land: a story of ethnic migration in a protected area landscape in western Uganda. Population and Environment 36:452 479. Himes, J. G., and P. J. D rohan. 2007. Distribution and habitat selection of the pygmy rabbit, Brachylagus idahoensis in Nevada (USA). Journal of Arid Environments 68:371 382. Hines, J. E., J. D. Nichols, J. A. Royle, D. I. MacKenzie, A. M. Gopalaswamy, N. S. Kumar, and K. U. Kar anth. 2010. Tigers on trails: occupancy modeling for cluster sampling. Ecological Applications 20:1456 1466. Huang, C., B. Wylie, L. Yang, C. Homer, and G. Zylstra. 2002. Derivation of a tasselled cap transformation based on Landsat 7 at satellite reflect ance. International Journal of Remote Sensing 23:1741 1748. Hugenholtz, C. H., K. Whitehead, O. W. Brown, T. E. Barchyn, B. J. Moorman, A. LeClair, K. Riddell, and T. Hamilton. 2013. Geomorphological mapping with a small unmanned aircraft system (sUAS): f eature detection and accuracy assessment of a photogrammetrically derived digital terrain model. Geomorphology 194:16 24. IDFG. 2009. Management of American White Pelicans in Idaho: a five year plan (2009 2013) to balance American White Pelican and native cutthroat trout conservation needs and manage impacts to recreational fisheries in southeast Idaho. Idaho Department of Fish and Game. Boise, Idaho, USA. IDFG. 2016. Management plan for the conservation of American White Pelicans in Idaho 2016 2025. Id aho Department of Fish and Game. Boise, Idaho, USA. Iknayan, K. J., M. W. Tingley, B. J. Furnas, and S. R. Beissinger. 2014. Detecting diversity: emerging methods to estimate species diversity. Trends in Ecology & Evolution 29:97 106. Johnson, R., K. Smi th, and K. Wescott. 2015. Unmanned aircraft system (UAS) application to land and natural resource management. Environmental Practice 17:170 177. Jones IV, G. P. 2003. The feasibility of using small unmanned aerial vehicles for wildlife research. Master's Thesis, University of Florida, Gainesville, Florida, USA.
251 Jones IV, G. P., L. G. Pearlstine, and H. F. Percival. 2006. An assessment of small unmanned aerial vehicles for wildlife research. Wildlife Society Bulletin 34:750 758. Junda, J., E. Greene, and D. M. Bird. 2015. Proper flight technique for using a small rotary winged drone aircraft to safely, quickly, and accurately survey raptor nests. Journal of Unmanned Vehicle Systems 3:222 236. Katzner, T. E., and K. L. Parker. 1997. Vegetative characterist ics and size of home ranges used by pygmy rabbits ( Brachylagus idahoensis ) during winter. Journal of Mammalogy 78:1063 1072. Keinath, D. A., and M. McGee. 2004. Species assessment for pygmy rabbit ( Brachylagus idahoensis ) in Wyoming. United States Depart ment of the Interior, Bureau of Land Management, Wyoming State Office. Cheyenne, Wyoming, USA. Keith, J. O. 2005. An overview of the American White Pelican. Waterbirds: The International Journal of Waterbird Biology 28:9 17. Kendall, W. L., S. J. Convers e, P. F. Doherty Jr., M. B. Naughton, A. Anders, J. E. Hines, and E. Flint. 2009. Sampling design considerations for demographic studies: a case of colonial seabirds. Ecological Applications 19:55 68. King, D. T., and D. W. Anderson. 2005. Recent populati on status of the American White Pelican: a continental perspective. Waterbirds: The International Journal of Waterbird Biology 28:48 54. Knopf, F. L., and R. M. Evans. 2004. American White Pelican ( Pelecanus erythrorhynchos ). In P. G. Rodewald, editor. T he Birds of North America Online. The Cornell Lab of Ornithology, Ithaca, New York, USA. . Accessed Dec 2016. Kobziar, L. N., D. Godwin, L. Taylor, and A. C. Watts. 2015. Perspectives on trends, eff ectiveness, and impediments to prescribed burning in the southern U.S. Forests 6:561 580. Kuenzer, C., and S. Dech, editors. 2013. Thermal infrared remote sensing: sensors, methods, applications. Volume 17. Springer, New York, New York, USA. Kushlan, J. A. 1979. Effects of helicopter censuses on wading bird colonies. The Journal of Wildlife Management 43:756 760. Kushlan, J. A., and P. C. Frohring. 1985. Decreases in the Brown Pelican population in southern Florida. Colonial Waterbirds 8:83 95.
252 Langtimm C. A., R. M. Dorazio, B. M. Stith, and T. J. Doyle. 2011. New aerial survey and hierarchical model to estimate manatee abundance. The Journal of Wildlife Management 75:399 412. Larrucea, E. S., and P. F. Brussard. 2009. Diel and seasonal activity patter ns of pygmy rabbits ( Brachylagus idahoensis ). Journal of Mammalogy 90:1176 1183. LaRue, M. A., S. Stapleton, C. Porter, S. Atkinson, T. Atwood, M. Dyck, and N. Lecomte. 2015. Testing methods for using high resolution satellite imagery to monitor polar bea r abundance and distribution. Wildlife Society Bulletin 39:772 779. Lee, J. E. 2008. Pygmy rabbit ( Brachylagus idahoensis ) h abitat use, activity patterns and conservation in relationship to habitat treatments. Master's Thesis, Brigham Young University, Provo, Utah, USA. Lee, K. 2004. Development of unmanned aerial vehicle (UAV) for wildlife surveillance. Master's Thesis, Unive rsity of Florida, Gainesville, Florida, USA. Linchant, J., J. Lisein, J. Semeki, P. Lejeune, and C. Vermeulen. 2015. Are unmanned aircraft systems (UASs) the future of wildlife monitoring? A review of accomplishments and challenges. Mammal Review 45:239 252. Link, W. A., and J. R. Sauer. 1997. Estimation of population trajectories from count data. Biometrics 53:488 497. Lisein, J., J. Linchant, P. Lejeune, P. Bouche, and C. Vermeulen. 2013. Aerial surveys using an unmanned aerial system (UAS): compariso n of different methods for estimating the surface area of sampling strips. Tropical Conservation Science 6:506 520. Loarie, S. R., L. N. Joppa, and S. L. Pimm. 2007. Satellites miss environmental priorities. Trends in Ecology & Evolution 22:630 632. Luci eer, A., S. M. de Jong, and D. Turner. 2014. Mapping landslide displacements using Structure from Motion (SfM) and image correlation of multi temporal UAV photography. Progress in Physical Geography 38:97 116. Madden, M. M., T. Jordan, S. Bernardes, D. L. Cotten, N. O'Hare, and A. Pasqua. 2015. Unmanned aerial systems and structure from motion revolutionize wetlands mapping. Pages 195 220 In R. W. Tiner Jr., M. W. Lang, and V. V. Klemas, editors. Remote Sensing of Wetlands: Applications and Advances. CRC P ress. Magnusson, W. E., G. J. Caughley, and G. C. Grigg. 1978. A double survey estimate of population size from incomplete counts. The Journal of Wildlife Management 42:174 176.
253 Martin, J., H. H. Edwards, M. A. Burgess, H. F. Percival, D. E. Fagan, B. E. Gardner, J. G. Ortega Ortiz, P. G. Ifju, B. S. Evers, and T. J. Rambo. 2012. Estimating distribution of hidden objects with drones: from tennis balls to manatees. PLoS ONE 7:e38882. McCleery, R. A., A. R. Holdorf, L. L. Hubbard, and B. D. Peer. 2015. Max imizing the wildlife conservation value of road right of ways in an agriculturally dominated landscape. PLoS ONE 10:e0120375. Meyer, K. A., C. L. Sullivan, P. Kennedy, D. J. Schill, D. M. Teuscher, A. F. Brimmer, and D. T. King. 2016. Predation by America n White Pelicans and Double crested Cormorants on catchable sized hatchery rainbow trout in select Idaho lentic waters. North American Journal of Fisheries Management 36:294 308. Miller, M. W., E. V. Pearlstine, R. M. Dorazio, and F. J. Mazzotti. 2011. Oc cupancy and abundance of wintering birds in a dynamic agricultural landscape. The Journal of Wildlife Management 75:836 847. Mocklin, J. A., D. J. Rugh, S. E. Moore, and R. P. Angliss. 2012. Using aerial photography to investigate evidence of feeding by b owhead whales. Marine Mammal Science 28:602 619. Mohamed, A. H., and B. E. Wilkinson. 2009. Direct georeferencing of stationary LiDAR. Remote Sensing 1:1321 1337. Moreland, E. E., M. F. Cameron, R. P. Angliss, and P. L. Boveng. 2015. Evaluation of a ship based unoccupied aircraft system (UAS) for surveys of spotted and ribbon seals in the Bering Sea pack ice. Journal of Unmanned Vehicle Systems 3:114 122. Mulero Pzmny, M. 2015. Unmanned aerial systems in conservation biology. Doctoral Dissertation, Uni versity of Seville, Seville, Spain. Mulero Pzmny, M., R. Stolper, L. D. van Essen, J. J. Negro, and T. Sassen. 2014. Remotely piloted aircraft systems as a rhinoceros anti poaching tool in Africa. PLoS ONE 9:e83873. Newman, S. 2016. Update on rehabilit ating impacted regions of Water Conservation Area (WCA) 2A: the Active Marsh Improvement (AMI) project. South Florida Water Management District, West Palm Beach, Florida, USA. Nichols, J. D., W. L. Kendall, and M. C. Runge. 2004. Estimating survival and movement. Pages 119 139 In W. J. Sutherland, I. Newton, and R. E. Green, editors. Bird Ecology and Conservation: A Handbook of Techniques. Oxford University Press, Oxford, United Kingdom.
254 Nijland, W., R. de Jong, S. M. de Jong, M. A. Wulder, C. W. Bater, and N. C. Coops. 2014. Monitoring plant condition and phenology using inf rared sensitive consumer grade digital cameras. Agricultural and Forest Meteorology 184:98 106. NOAA. 2016. Historical climate data online. In National Oceanic and Atmospheric Administration, National Centers for Environmental Information, Asheville, No rth Carolina, USA. Accessed Dec 2016. Nobler, J. D. 2016. Risky business: tradeoffs between nutrition, toxicity, and predation by a specialist mammalian herbivore. Master's Thesis, Boise State University, Boise, Idaho, USA. Nobler, J. D., P. J. Olsoy, and J. S. Forbey. 2013. Spectroscopy as a tool to measure the foodscapes of vertebrate herbivores. Boise State University. Boise, Idaho, USA. Olsoy, P. J. 2013. Determination of sagebrush aboveground biomass using terrestrial laser scanning (TLS). Master's Thesis, Idaho State University, Pocatello, Idaho, USA. Olsoy, P. J., J. S. Forbey, J. L. Rachlow, J. D. Nobler, N. F. Glenn, and L. A. Shipley. 2015. Fe arscapes: mapping functional properties of cover for prey with terrestrial LiDAR. BioScience 65:74 80. Ono, A., K. Kajiwara, Y. Honda, and A. Ono. 2010. Development of new vegetation indexes, shadow index (SI) and water stress trend (WST), Pages 710 714 i n The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. C. Jun, editor. Ozesmi, S. L., and M. E. Bauer. 2002. Satellite remote sensing of wetlands. Wetlands Ecology and Management 10:381 402. Pacifici, K., R. M. Dorazio, and M. J. Conroy. 2012. A two phase sampling design for increasing detections of rare species in occupancy surveys. Methods in Ecology and Evolution 3:721 730. Packard, J. M., R. C. Summers, and L. B. Barnes. 1985. Variation of visibility bias during aerial surveys of manatees. The Journal of Wildlife Management 49:347 351. Pajares, G. 2015. Overview and current status of remote sensing applicaitons based on unmanned aerial vehicles (UAVs). Photogrammetric Engineering & Remote Sensing 81:281 3 29.
255 Park, S. 2013. Cloud and cloud shadow effects on the MODIS vegetation index composites of the Korean Peninsula. International Journal of Remote Sensing 34:1234 1247. Parsons, M. A., T. C. Barkley, J. L. Rachlow, J. L. Johnson Maynard, T. R. Johnson, C. R. Milling, J. E. Hammel, and I. Leslie. 2016. Cumulative effects of an herbivorous ecosystem engineer in a heterogeneous landscape. Ecosphere 7:e01334. Perry, J. H. 2009. A synthesized directly georeferenced remote sensing technique for small unmanned aerial vehicles. Master's Thesis, University of Florida, Gainesville, Florida, USA. Pettorelli, N., J. O. Vik, A. Mysterud, J. M. Gaillard, C. J. Tucker, and N. C. Stenseth. 2005. Using the satellite derived NDVI to assess ecological responses to environ mental change. Trends in Ecology & Evolution 20:503 510. Pollock, K. H., H. D. Marsh, I. R. Lawler, and M. W. Alldredge. 2006. Estimating animal abundance in heterogeneous environments: an application to aerial surveys for dugongs. The Journal of Wildlife Management 70:255 262. Pomeroy, P., L. O'Connor, and P. Davies. 2015. Assessing use of and reaction to unmanned aerial systems in gray and harbor seals during breeding and molt in the UK. Journal of Unmanned Vehicle Systems 3:102 113. Rachlow, J. L., D. M. Sanchez, and W. A. Estes Zumpf. 2005. Natural burrows and nests of free ranging pygmy rabbits ( Brachylagus idahoensis ). Western North American Naturalist 65:136 139. Rachlow, J. L., and L. K. Svancara. 2006. Prioritizing habitat for surveys of an unco mmon mammal: a modeling approach applied to pygmy rabbits. Journal of Mammalogy 87:827 833. Rachlow, J. L., and J. Witham. 2006. Aerial surveys for pygmy rabbits in the BLM Shoshone Field Office area: results and an evaluation of the technique. DLF050581. University of Idaho. Moscow, Idaho, USA. Ralph, C. J., and J. M. Scott, editors. 1981. Estimating numbers of terrestrial birds. Volume 6. Cooper Ornithological Society and Allen Press, Inc., Lawrence, Kansas, USA. Rango, A., and A. S. Laliberte. 2010. Impact of flight regulations on effective use of unmanned aircraft systems for natural resources applications. Journal of Applied Remote Sensing 4:1 12.
256 Reynolds, J. H., M. G. Knutson, K. B. Newman, E. D. Silverman, and W. L. Thompson. 2016. A road map fo r designing and implementing a biological monitoring program. Environmental Monitoring and Assessment 188:399 423. Richards, J. A., editor. 2013. Remote sensing digital image analysis: an introduction. Fifth edition. Springer Verlag, New York, New York, U SA. Roberts, H. B. 2001. Survey of pygmy rabbit distribution, numbers and habitat use in Lemhi and Custer Counties, Idaho. Technical Bulletin 01 11. Rosenstock, S. S., D. R. Anderson, K. M. Giesen, T. Leukering, and M. F. Carter. 2002. Landbird counting techniques: current practices and an alternative. The Auk 119:46 53. Royle, J. A., and R. M. Dorazio. 2010. Parameter expanded data augmentation for Bayesian analysis of capture recapture models. Journal of Ornithology 152:S521 S537. Sanchez, D. M., and J. L. Rachlow. 2008. Spatio temporal factors shaping diurnal space use by pygmy rabbits. The Journal of Wildlife Management 72:1304 1310. Sarda Palomera, F., G. Bota, C. Vinolo, O. Pallares, V. Sazatornil, L. Brotons, S. Gomariz, and F. Sarda. 2012. Fine scale bird monitoring from light unmanned aircraft systems. IBIS: The International Journal of Avian Science 154:177 183. Sasse, D. B. 2003. Job related mortality of wildlife workers in the United States, 1937 2000. Wildlife Society Bulletin 31:1015 1020. Schmidt, J. H., K. L. Rattenbury, J. P. Lawler, and M. C. Maccluskie. 2012. Using distance sampling and hierarchical models to improve estimates of Dall's sheep abundance. The Journal of Wildlife Management 76:317 327. Severaid, J. H. 1950. The pigmy ra bbit ( Sylvilagus idahoensis ) in Mono County, California. Journal of Mammalogy 31:1 4. Shahbazi, M., J. Thau, and P. Mnard. 2014. Recent applications of unmanned aerial imagery in natural resource management. GIScience & Remote Sensing 51:339 365. Shipl ey, L. A., T. B. Davila, N. J. Thines, and B. A. Elias. 2006. Nutritional requirements and diet choices of the pygmy rabbit ( Brachylagus idahoensis ): a sagebrush specialist. Journal of Chemical Ecology 32:2455 2474. Shipley, L. A., J. S. Forbey, and B. D. Moore. 2009. Revisiting the dietary niche: when is a mammalian herbivore a specialist? Integrative and Comparative Biology 49:274 290.
257 Shufeldt, R. W. 1888. The leaps of Lepus Nature 37:247. Siegel Thines, N. J., L. A. Shipley, and R. D. Sayler. 2004. Effects of cattle grazing on ecology and habitat of Columbia Basin pygmy rabbits ( Brachylagus idahoensis ). Biological Conservation 119:525 534. Stark, B., and Y. Q. Chen. 2014. Optimal collection of high resolution aerial imagery with unmanned aerial syst ems, Pages 89 94 in 2014 International Conference on Unmanned Aircraft Systems (ICUAS). 27 30 May 2014. Stark, B., C. Coopmans, and Y. Q. Chen. 2013. Concept of operations for personal remote sensing unmanned aerial systems. Journal of Intelligent & Robo tic Systems 69:5 20. Stober, J. M., and L. L. Smith. 2010. Total counts versus line transects for estimating abundance of small gopher tortoise populations. The Journal of Wildlife Management 74:1595 1600. Strong, L. L., and L. M. Cowardin. 1995. Improvi ng prairie pond counts with aerial video and global positioning systems. The Journal of Wildlife Management 59:708 719. Temple, S. A., and J. A. Wiens. 1989. Bird populations and environmental changes: can birds be bio indicators? American Birds 43:260 27 0. Teuscher, D. M., M. T. Green, D. J. Schill, A. F. Brimmer, and R. W. Hillyard. 2015. Predation by American White Pelicans on Yellowstone cutthroat trout in the Blackfoot River drainage, Idaho. North American Journal of Fisheries Management 35:454 463. Thomas, L., S. T. Buckland, E. A. Rexstad, J. L. Laake, S. Strindberg, S. L. Hedley, J. R. B. Bishop, T. A. Marques, and K. P. Burnham. 2010. Distance software: design and analysis of distance sampling surveys for estimating population size. Journal of Ap plied Ecology 47:5 14. Towler, J., B. Krawiec, and K. Kochersberger. 2012. Radiation mapping in post disaster environments using an autonomous helicopter. Remote Sensing 4:1995 2015. Tracey, J. P., P. J. S. Fleming, and G. J. Melville. 2008. Accuracy of some aerial survey estimators: contrasts with known numbers. Wildlife Research 35:377 384. Turner, D., A. Lucieer, and C. Watson. 2012. An automated technique for generating georectified mosaics from ultra high resolution unmanned aerial vehicle (UAV) ima gery, based on structure from motion (SfM) point clouds. Remote Sensing 4:1392 1410.
258 Ulappa, A. C., R. G. Kelsey, G. G. Frye, J. L. Rachlow, L. A. Shipley, L. Bond, X. Pu, and J. S. Forbey. 2014. Plant protein and secondary metabolites influence diet sele ction in a mammalian specialist herbivore. Journal of Mammalogy 95:834 842. Ulmschneider, H., D. Hays, H. B. Roberts, J. L. Rachlow, T. Forbes, J. Himes, E. Sequin, M. Haworth, T. E. Katzner, A. Kozlowski, R. Rauscher, and P. Lauridson. 2008. Surveying fo r pygmy rabbits ( Brachylagus idahoensis ). I daho P ygmy R abbit W orking Group. Urbanek, R. E., C. K. Nielsen, T. S. Preuss, and G. A. Glowacki. 2012. Comparison of aerial surveys and pellet based distance sampling methods for estimating deer density. Wildlife Society Bulle tin 36:100 106. USBOR. 2016. Historical data for MIN Minidoka Dam and Lake Walcott, ID. In US Bureau of Reclamation, Pacific Northwest Region, Boise, Idaho, USA. Accessed Dec 2016. Utz, J. L. 2012. Understanding the tradeoff between safety and food quality in a mammalian herbivore specialist, the pygmy rabbit. Master's Thesis, Boise State University, Boise, Idaho, USA. Utz, J. L., L. A. Shipley, J. L. Rachlow, T. Johnstone Yellin, M. J. Camp, and J. Sorensen Forbey. 2016. Understanding tradeoffs between food and predation risks in a specialist mammalian herbivore. Wildlife Biology 22:167 173. van Gemert, J. C., C. R. Verschoor, P. Mettes, K. Epema, L. P. Koh, and S. A. Wich. 2015. Nature conservation drones for automatic localization and counting of animals. Pages 255 270 In L. Agapito, M. M. Bronstein, and C. Rother, editors. Computer Vision ECCV 2014 Workshops. Springer Interna tional Publishing, Zurich, Switzerland. Verhoeven, G. J. J. 2009. Providing an archaeological bird's eye view an overall picture of ground based means to execute low altitude aerial photography (LAAP) in archaeology. Archaeological Prospection 16:233 24 9. Verhoeven, G. J. J. 2010. It's all about the format unleashing the power of RAW aerial photography. International Journal of Remote Sensing 31:2009 2042. Viquerat, S., and A. van Neer. 2015. iTag image tagging software manual. Walsh, D. P., H. Camp a III, D. E. Beyer Jr., and S. R. Winterstein. 2011. Measurement error and survey design in sightability model development. The Journal of Wildlife Management 75:1228 1235.
259 Wang, K., S. E. Franklin, X. Guo, and M. Cattet. 2010. Remote sensing of ecology, biodiversity and conservation: a review from the perspective of remote sensing specialists. Sensors 10:9647 9667. Watts, A. C., J. H. Perry, S. E. Smith, M. A. Burgess, B. E. Wilkinson, Z. Szantoi, P. G. Ifju, and H. F. Percival. 2010. Small unmanned airc raft systems for low altitude aerial surveys. The Journal of Wildlife Management 74:1614 1619. Weishampel, J. F., D. A. Bagley, L. M. Ehrhart, and B. L. Rodenbeck. 2003. Spatiotemporal patterns of annual sea turtle nesting behaviors along an east central Florida beach. Biological Conservation 110:295 303. Weiss, N. T. 1984. Habitat and distribution of pygmy rabbits ( Sylvilagus idahoensis ) in Oregon. Master's Thesis, Oregon State University, Corvallis, Oregon, USA. Westoby, M. J., J. Brasington, N. F. Gla sser, M. J. Hambrey, and J. M. Reynolds. 2012. 'Structure from Motion' photogrammetry: a low cost, effective tool for geoscience applications. Geomorphology 179:300 314. White, S. M., J. T. Flinders, and B. L. Welch. 1982. Preference of pygmy rabbits ( Bra chylagus idahoensis ) for various populations of big sagebrush ( Artemisia tridentata ). Journal of Range Management 35:724 726. Wilkinson, B. E. 2007. The design of georeferencing techniques for unmanned autonomous aerial vehicle video for use with wildlife inventory surveys: a case study of the National Bison Range, Montana. Master's Thesis, University of Florida, Gainesville, Florida, USA. Wilkinson, B. E., B. A. Dewitt, A. C. Watts, A. H. Mohamed, and M. A. Burgess. 2009. A new approach for pass point ge neration from aerial video imagery. Photogrammetric Engineering & Remote Sensing 75:1415 1423. Williams, B. K., J. D. Nichols, and M. J. Conroy. 2002. Analysis and management of animal populations. Academic Press, San Diego, California, USA. Williams, K. A., E. E. Connelly, S. M. Johnson, and I. J. Stenhouse, editors. 2015. Wildlife densities and habitat use across temporal and spatial scales on the Mid Atlantic Outer Continental Shelf (2012 2014): final report to the Department of Energy EERE Wind & Wate r Power Technologies Office. Biodiversity Research Institute, Portland, Maine, USA. Williams, K. A., P. C. Frederick, and J. D. Nichols. 2011. Use of the superpopulation approach to estimate breeding population size: an example in asynchronously breeding birds. Ecology 92:821 828.
260 Wing, M. G., J. Burnett, J. Sessions, J. Brungardt, V. Cordell, D. Dobler, and D. Wilson. 2013. Eyes in the sky: remote sensing technology development using small unmanned aircraft systems. Journal of Forestry 111:341 347. Wolf P. R., B. A. Dewitt, and B. E. Wilkinson. 2014. Elements of photogrammetry with applications in GIS. Fourth edition. McGraw Hill Education, New York, New York, USA. Wozencraft, J. M., and W. J. Lillycrop. 2003. SHOALS airborne coastal mapping: past, pre sent, and future. Journal of Coastal Research:207 215. Zhang, L., X. Sun, T. Wu, and H. Zhang. 2015. An analysis of shadow effects on spectral vegetation indexes using a ground based imaging spectrometer. IEEE Geoscience and Remote Sensing Letters 12:2188 2192. Zweig, C. L., M. A. Burgess, H. F. Percival, and W. M. Kitchens. 2015. Use of unmanned aircraft systems to delineate fine scale wetland vegetation communitites. Wetlands 35:303 309. Zweig, C. L., and S. Newman. 2015. Using landscape context to map invasive species with medium resolution satellite imagery. Restoration Ecology 23:524 530.
261 BIOGRAPHICAL SKETCH Matthew Alexander Burgess was born in July 1977, in Gainesville, Florida. Growing up, his routine participation in fieldwork with his fath er, annual family vacations to Crescent Beach, and summers sailing on the Hudson River and Long Island Sound helped foster his love of science and the environment. In June 1995, Matthew graduated from Gainesville High School. After beginning his collegia te career at the University of Florida (UF) in c ivil e ngineering, he realized that his desire for working outdoors was stronger than computing integrals, so he changed his major to zoology. In August 1999, he earned his Bachelor of Science degree from UF and was hired full time as a laboratory ass istant and international field technician in the UF Zoology Department Matthew accepted a five year term appointment in late 2001 as a fisheries biologist for the US G eological S urvey in Gainesville. During th at time, he began a at UF in interdisciplinary ecology. When the USGS appointment ended, he worked part time for the Florida Program for Shark Research at the Florida Museum of Natural History, which provided time to complete his thesis Quantification and Ecological Role of Snag Habitat in the Apalachicola River, Florida In August 2008, Matthew earned his Master of Science degree from UF. Never thinking his hobby of racing remote control cars would ever end up on his CV, nor be a fact or in landing a job, Matthew was hired in July 2008 as the Program Coordinator of the UF Unmanned Aircraft Systems Research Program, administered out of the Florida Cooperative Fish and Wildlife Research Unit. S erving in that role full time and slowly bec oming a subject matter expert in UAS for natural resource applications, his boss and mentor Dr. Percival encouraged Matthew to pursue a Ph.D. in the UF Department of Wildlife Ecology and Conservation, which he began in August 2011 and completed in May 201 7