This item has the following downloads:
1 THE ECOLOGY AND CONSERVATION OF CALLOPHRYS IRUS GODART: THE ROLE OF FIRE AND MICROHABITAT By MATTHEW D. THOM A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREME NTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2013
2 2013 Matthew D. Thom
3 To Mom, Dad, and Neil
4 ACKNOWLEDGMENTS I want to thank my mom and dad for their unconditional support th roug h everything, always Thanks also goes to Lisa, who has been so supportive and understanding, doing everything right to make this possible. I would like to thank my advisor Jaret and committee (Marc, Leda, Rob, and Heather) for being a part of this project and this process, and providing just the right amount of help when I needed it. also like to thank the support of assistantships and research funds from UF, the Entomology and Nematology Department, Jaret Daniels, Marc Branham, Dan Hahn, and Andrea Lu cky. The staff of the Entomology and Nematology Department has been critical to my success, including those who retired during my graduate school journey. Special t hanks to Dr. Dan Hahn, always an open ear and so free with his time and resources. Andrea Lu cky also provided a similar open ear, good advice, and encouragement. IFAS stats consulting by James Colee was critical at several points. Thanks also to McGuire Center folks such as Jackie Miller, Charlie Covell, Andrei Sourakov, Keith Willmott, and Akito Kawahara for always providing support and advice. Further thanks to Curation and Collections Manager Andy Warren for taking the time to talk my project and connect me to others. Matt Standridge, Lukaz Barszczak, and Marissa Streifield all provided a huge amount of their time, sweat, and positivity to field work that was often grueling. An even greater amount of thanks goes out to a person person fundamental to my success Jon Colburn, a partner on a sometimes endless journey t hrough the trials, tribulation s, tantrums, and endless analysis in the field, lab, and in the code. Thanks also to Sandy Koi for collaboration on rearing Eumaeus atala Finally, thanks to the staff at the Butterfly Rainforest at the Flordia Museum of Natural History for being flexible and helpful with E. atala rearing.
5 I must also thank Jennifer Hart and Brian Camposano who so graciously allowed me several years of permitting at Ralph E. Simmons Memorial State Forest. Thanks also to the folks at the Apalachicola National Forest for thei r permitting. And final ly, Steve Coates, his staff, the Nature Conservancy and the St. Johns River Water Management District folks that allowed me ample space and opportunity to conduct experiments during their prescribed burning. Thanks also to Jon Calh oun, Dean and Sally Jue, Mary Ann Friedman, and Bill Berthet, talking to me about the natural history and ecology of Florida Lepidoptera, in particular the frosted elfin. Without your interest and enthusiasm, I would not have nearly as clear a picture as I now have. Finally I want to thank all my Entomology department friends for their great attitudes and interest in all things science. Their support and the role of EN SO was a key part of my sanity, and gave me a family. Post docs including Giancarlo Lopez, Caroline Williams, Jen Hamel, and Charlotte Germain were very valuable to talk with, discuss problems, and develop methods to get what I wanted done, so thank you all. A last thank you to Deb and Mike and family, for being my home away from home!
6 TABLE O F CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 9 LIST OF FIGURES ................................ ................................ ................................ ........ 12 ABSTRACT ................................ ................................ ................................ ................... 15 CHAPTER 1 CONSERVATION PROFILE OF THE FROSTED ELFIN BUTTERFLY, CALLOPHRYS IRUS GODART (LEPIDOPTERA: LYCAENIDAE) : A COMPARISON OF DIFFERENCES THROUGHOUT A LOCALLY RARE YET WIDE RANGING IMPERILED INSECT ................................ ................................ .. 17 Introduction ................................ ................................ ................................ ............. 17 Taxonomic History ................................ ................................ ................................ .. 18 Species Description and Life History ................................ ................................ ...... 19 Geographic Distribution ................................ ................................ .......................... 21 Habitat Affinity and Associations ................................ ................................ ............. 21 Popula tion Structure ................................ ................................ ............................... 24 Larval Host Plant Use ................................ ................................ ............................. 26 L. perennis Feeding Behavior ................................ ................................ ................. 29 Phenology ................................ ................................ ................................ ............... 30 Adult Flight Time ................................ ................................ .............................. 30 Larval Development Time ................................ ................................ ................. 31 Conservation Status ................................ ................................ ............................... 31 Efforts To Reduce The Decline Of C. Irus ................................ ............................... 33 Habitat Management ................................ ................................ ........................ 33 Prescribed fire ................................ ................................ ............................ 33 Tree removal ................................ ................................ .............................. 34 Deer Population Management ................................ ................................ .......... 35 Implications Of C. irus Conservation ................................ ................................ ....... 35 L. melissa samuelis Spillover ................................ ................................ ........... 35 Community and Biodive rsity Preservation ................................ ........................ 36 Scientific Discovery ................................ ................................ .......................... 37 Future Outlook ................................ ................................ ................................ ........ 37 2 PATTERNS OF MICROHABITAT AND LARVAL HOST PLANT USE BY CALLOPHRYS IRUS GODART AT RALPH E. SIMMONS MEMORIAL STATE FOREST ................................ ................................ ................................ ................. 56 Introduction ................................ ................................ ................................ ............. 56 Methods ................................ ................................ ................................ .................. 59
7 Study Site ................................ ................................ ................................ ......... 59 Study Organisms ................................ ................................ .............................. 60 Focal Fo llow and Markomg of C. irus ................................ .............................. 62 Microhabitat Surveys ................................ ................................ ........................ 63 2010 ................................ ................................ ................................ ........... 63 2012 ................................ ................................ ................................ ........... 64 L. perennis and C. irus Census ................................ ................................ ........ 65 Data Analysis ................................ ................................ ................................ ... 66 2010 microhabitat survey. ................................ ................................ .......... 66 2012 microhabitat survey ................................ ................................ ........... 67 L. perennis census ................................ ................................ ..................... 68 Results ................................ ................................ ................................ .................... 69 Focal Follow and Mark of C. irus ................................ ................................ ...... 69 Microhabitat Surveys ................................ ................................ ........................ 70 2010 ................................ ................................ ................................ ........... 70 2012 ................................ ................................ ................................ ........... 71 L. Perennis and C. irus Census ................................ ................................ ........ 73 Discussion ................................ ................................ ................................ .............. 73 3 CONSEQUENCES FOR MORTALITY BY FIRE: THE EFFECT OF PUPATION LOCATION OF THE FROSTED ELFIN, CALLOPHRYS IRUS GODART (LEPIDOPTERA: LYCAENIDAE) ................................ ................................ ............ 96 Introduction ................................ ................................ ................................ ............. 96 Methods ................................ ................................ ................................ ................ 102 Study Sites ................................ ................................ ................................ ..... 102 Study Organisms ................................ ................................ ............................ 104 Pupal Depth Measurements of C. irus ................................ ............................ 106 Heat Tolerance of Butterfly Pupae ................................ ................................ 109 Pupal Survival Following Prescribed Burning ................................ ................. 112 Data Analysis ................................ ................................ ................................ 115 Results ................................ ................................ ................................ .................. 118 Field Observations of Depth of Pupae of C. irus ................................ ............ 118 Heat Pulse in Soil of Prescribed Burns in Sandhill at Ordway Swisher Biological Station, Putnam County, Florida, July, 2012 ............................... 120 Heat Tolerance of E. atala Pupae Following Laboratory Water Bath Treatment ................................ ................................ ................................ .... 120 Survival of Buried and Unburied E. atala Pupae to Successful Eclosion Following Prescribed Burning ................................ ................................ ...... 121 Factors Correlated to Survival of E. atala Pupae to Successful Adult Emergence Following Water Bath Heating ................................ .................. 122 Factors Correlated to Survival of E. atala Pupae to Successful Adult Emergence Following Prescribed Burning ................................ .................. 122 Factors Correlated to the Heat Pulse During Prescribed Burning .................. 123 Discussion ................................ ................................ ................................ ............ 123
8 4 STATUS OF THE FROSTED ELFIN, CALLOPHR YS IRUS GODART (LEPIDOPTERA: LYCAENIDAE) IN FLORIDA ................................ ..................... 154 Introduction ................................ ................................ ................................ ........... 154 Methods ................................ ................................ ................................ ................ 156 Study Organisms ................................ ................................ ............................ 156 Species Record Collection ................................ ................................ ............. 156 C. irus Monitoring Surveys ................................ ................................ ............. 157 Niche Modeling ................................ ................................ ............................... 158 Results ................................ ................................ ................................ .................. 161 Species Record Collection ................................ ................................ ............. 161 C. irus Monitoring Surveys ................................ ................................ ............. 162 MaxEnt Niche Modeling ................................ ................................ ................. 162 Discussion ................................ ................................ ................................ ............ 163 5 CONCLUSIONS AND SYNTHESIS ................................ ................................ ...... 180 General Importance ................................ ................................ .............................. 180 Need for Communication ................................ ................................ ...................... 181 Influence of Current Thinking ................................ ................................ ................ 182 LIST OF REFERENCES ................................ ................................ ............................. 183 BIOGRAPHIC AL SKETCH ................................ ................................ .......................... 19 0
9 LIST OF TABLES Table page 1 1 Regional overview of habitat affinities of extant populations of C. irus .. ............. 40 1 2 Conservation status of C. irus from NatureServe as of April 2013. ..................... 40 1 3 Conservation status of L. perennis from USDA PLANTS database as of April 201 3. ................................ ................................ ................................ .................. 41 2 1 Observed behaviors from Callophrys irus adult focal follow. .............................. 77 2 2 C. irus microhabitat sampling dates and features re corded during 2010. ........... 78 2 3 C. irus microhabitat features sampled over an 11 day period March April 2012. ................................ ................................ ................................ .................. 79 2 4 Hypotheses for m odel selection. ................................ ................................ ......... 80 2 5 Summary of individual size and environmental conditions at capture of C. irus ................................ ................................ ................................ ..................... 81 2 6 Summary statistics of the 2010 C. irus microhabitat survey ............................... 82 2 7 Output summary of the best generalized linear regression model of C. irus eggs on L. perennis 2012 ................................ ................................ .................. 83 2 8 Output summary of the best logistic regression model of C. irus feeding damage on L. perenni s 2010 ................................ ................................ ............. 83 2 9 Summary statistics of the C. irus microhabitat survey 2012. .............................. 84 2 10 Output summary of generalized linear regression models of sub hypotheses ... 85 2 11 Output summary of generalized linear r egression final conceptual models ........ 86 2 12 Model fit (pseudo R 2 ) of the most likely generalized regression models ............ 87 2 13 Model a veraged estimates from AICc model selection. ................................ ...... 87 3 1 Experimental treatment setup for the follow up laboratory water bath heat tolerance experiment ................................ ................................ ........................ 132 3 2 Fire weather data for experimental prescribed fires. ................................ ......... 132 3 3 Experimental treatment numbers for controlled burn experiments. .................. 133 3 4 Fire experiment treatment and replication ................................ ........................ 133
10 3 5 Regression model setup for laboratory and prescribed burning experiments ... 133 3 6 Logistic regression model selection of survival of E. atala pupae to successful adult eclosion as a function of heat, ................................ ................ 134 3 7 Output summary of the best logi stic regression model of survival of E. atala pupa to successful adult eclosion as a function of heat ................................ .... 134 3 8 Logistic regression model selection of survival of E. atala pupae to successful ad ult eclosion as a function of peak temperature ............................ 134 3 9 Output summary of the best logistic regression model of survival of E. atala pupa to successful adult eclosion as a function of peak temp erature. .............. 134 3 10 Logistic regression model selection of survival of E. atala pupae to successful adult eclosion as a function of heat ................................ ................. 134 3 11 Output summary of the best logistic regression model of survival of E. atala pupa to successful adult eclosion as a function of heat ................................ .... 135 3 12 Logistic regression model selectio n of survival of E. atala pupae to successful adult eclosion as a function of peak temperature ............................ 135 3 13 Output summary of the best logistic regression model of survival of E. atala pupa to suc cessful adult eclosion as a function of heat \ ................................ .. 135 3 14 Logistic regression model selection of survival of E. atala pupae to successful adult eclosion as a function of burial depth ................................ ..... 135 3 15 Output summary of the best logistic regression model of survival of E. atala pupa to successful adult eclosion as a function of burial depth ........................ 136 3 16 Linear regression model selection of burial depth as a function of heat, peak temperature, and time to peak temperature. ................................ .................... 136 3 17 Output summary of the best linear regressi on model of burial depth. ............... 136 4 1 Climate variable conceptual models for maximum entropy species distribution modeling of L. perennis in Florida. ................................ ................. 168 4 2 Climate variable conceptual models for maximum entropy species distribution modeling of C. irus in Florida.. ................................ ........................ 168 4 3 Vouchered Florida L. perennis specimen records. ................................ ........... 169 4 4 Florida records of C. irus prior to 2010 from Butterflie s and Moths of America Database ................................ ................................ ................................ .......... 170 4 5 Florida C. irus surveys 2 010 2012 by NABA and M. Thom. ............................. 170
11 4 6 MaxEnt model output for L. perennis ................................ ................................ 171 4 7 MaxEnt model output for C. irus ................................ ................................ ....... 171
12 LIST OF FIGURES Figure page 1 1 Adults of Callophrys irus (Lepidoptera: Lycaenidae) ................................ .......... 42 1 2 P lacement of eggs of C. irus on, L. perennis ................................ ..................... 43 1 3 Larvae feeding behavior, C. irus on L. perennis ................................ ............... 44 1 4 Late instar larva o f C. irus ................................ ................................ ................... 45 1 5 Pupa of C. irus discovered at the forest floor. ................................ ..................... 46 1 6 State and province distribution of C. irus ................................ ........................... 47 1 7 USA county records for C. irus ................................ ................................ .......... 47 1 8 C. irus habitat series following a February controlled burn. ................................ 48 1 9 U. reversalis on L. perennis ................................ ................................ ................ 49 1 10 Lepidopteran larvae on L. perennis ................................ ................................ .... 50 1 11 Blister be etle Epicauta sp. feeding on L. perennis ................................ .............. 51 1 12 Florida records for C. irus ................................ ................................ .................. 52 1 13 Southeastern USA county records for C. iru s ................................ .................... 53 1 14 State and province records for L.perennis and B. tinctoria ............................... 54 1 15 Larval host plant use by C. irus ................................ ................................ .......... 54 1 16 Simultaneous feeding on L. perennis by Lepidopteran larvae ............................ 55 2 1 Location of Ralph E. Simmons Memorial State Forest and the Lupinus pe rennis and Callophrys irus in Nassau County, Florida ................................ ... 88 2 2 Eggs and larvae of C. irus on L. perennis ................................ ........................... 89 2 3 L. perenni s ................................ ................................ ................................ .......... 90 2 4 Most commonly encountered organisms in close proximity to L. perennis ......... 91 2 5 Survey areas at Ralph E. Simmons Memorial State Fo rest, Nassau County, Florida. ................................ ................................ ................................ ............... 92
13 2 6 Plot of logistic regression of presence of C. irus feeding damage on L. perennis 2010 ................................ ................................ ................................ .... 93 2 7 Spline correlogram of the 2012 best regression model ................................ ...... 94 2 8 1m resolution 1:3549 color digital ortho image and density overlays from a census conducted March May, 2012. ................................ ................................ 95 3 1 Datum setup and reference point diagram for field excavation of C. irus pupae ................................ ................................ ................................ ............... 137 3 2 Diagram of pupae locations within prescribed fire treat ment units. ................... 138 3 3 Depth from top of soil of C irus pupae excavated ................................ ............. 139 3 4 Temperature at ground surface as measured ................................ .................. 140 3 5 Soil surface temperatures as estimated using a non contact infrared thermom eter during controlled burning ................................ ............................ 142 3 6 Soil temperatu res as measured with iButton thermocrons July 5th .................. 143 3 7 Soil temperatures as measured with iButton thermocrons July 6th ................. 144 3 8 Soil temperatures as measured with iButton thermocrons July 26th ................ 145 3 9 Plot of eclosion of E. atala eclosing follow ing heated water bath treatment .... 146 3 10 Proportion of E. atala pupae successfully eclosing follow ing heated water bath treatment ................................ ................................ ................................ 147 3 11 Proportion of E. atala pupae successfully eclosing followin g prescribed burning ................................ ................................ ................................ ............. 148 3 12 Plot of survival of E. atala pupae to successful adult eclosion as a function of heat from the laboratory water bath experiment. ................................ .............. 149 3 13 Plot of survival of E. atala pupae to successful adult eclosion as a function of time to peak temperature from the laboratory water bath experiment,. ............ 149 3 14 Plot of survival of E. atala pupae to successful adult eclosion as a function of peak temperature from the laboratory water bath experiment .......................... 150 3 15 Plot of survival of E. atala pupae t o successful adult eclosion as a function of heat from prescribed burning. ................................ ................................ ........... 150 3 16 Plot of survival of E. atala pupae to successful adult eclosion as a function of peak temp erature from presc ribed burning ................................ ...................... 151
14 3 17 Plot of survival of E. atala pupae to successful adult eclosion as a function of time to peak temperature from prescribed burning. ................................ .......... 151 3 18 Plot of survival of E. atala pupae to successful adult eclosion as a function of buri al depth from prescribed burning ................................ ................................ 152 3 19 Plot of heat and burial depth as related to survival of E. atala pupae to eclosion following prescribed burning ................................ ............................... 152 3 20 Plot of peak temperature and burial depth as related to survival of E. atala pupae to eclosion follo wing prescribed burning ................................ ................ 153 4 1 C.rus (A, B) and L. perennis var. gracilis ................................ ......................... 172 4 2 Geo referenced records of Florida L. perenni s from vouchered specimens .... 173 4 3 Records of C. irus in Florida prior to 2010 ................................ ........................ 174 4 4 Ecological niche model of L. perennis us ing the maximum entropy method. ... 175 4 5 Ecological niche model of C. irus with the L. perennis layer using the maximum entropy method ................................ ................................ ................ 176 4 6 Ecological niche model of C. irus without the L. perennis layer using the maximum entropy method ................................ ................................ ................ 177 4 7 L. perennis at Munson Hills, Apalachicola National Forest, Leon County, F lorida. ................................ ................................ ................................ ............. 178 4 8 USA county records of C. irus ................................ ................................ .......... 179
15 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 THE ECOLOGY AND CONSERVATION OF CALLOPHRYS IRUS GODART: THE ROLE OF FIRE AND MICROHABITAT By Matthew D. Thom August 2013 Chair: Jaret C. Daniels Major: Entomolo gy and Nematology The disturbance caused by fire is highly influential in shaping, promoting, and sustaining certain successional stages, and is a major contributor to the dynamic nature of most terrestrial ecosystems. The frosted elfin, Callophrys irus i s a rare and imperiled butterfly that inhabits oak pine barrens and savannahs in North America where fire is a major disturbance force. The ability of C. irus to avoid the lethal or sub lethal effects of fire was investigated in both lab and field studies using Eumaus atala as an experimental surrogate. Survival of at least 50% to successful adult eclosion occurred at peak temperatures between 42 44 C. The results suggest management using fire should only occur in a portion of an area, be rotated between years, and be a fast moving fire. Habitat quality is fundamentally important to the persistence of a given species. For herbivorous insects, habitat quality is a function of a highly specific combination of complementary resources that includes host plants or nectar sources, and physical features. A study was conducted to investigate the role of microhabitat in a population C. irus in northern Florid a. Field surveys were conducted across three years at a state
16 forest that was actively managed with prescribed fire. C. irus eggs were correlated to larger larval host plants without the presence of other herbivory, areas of some accumulation of duff, smal l size of surrounding vegetation. The results suggest female oviposition preference is focused on host plant size and location, specifically in areas with at least some amount of accumulated duff and shade. Recommendations for land management for this impe riled insect include maintenance of a heterogenous landscape that includes both open and shaded areas, which support both the insect and the larval host plant L. perennis. A final study was conducted to document C. irus in Florida, and included museum and collection records, monitoring, and the creation of species distribution models. C. irus is historically rare, was rarely encountered over the study period, but is predicted to be present in a large geographic area of northern Florida.
17 CHAPTER 1 CONSERVA TION PROFILE OF THE FROSTED ELFIN BUTTER FLY, CALLOPHRYS IRUS GODART (LEPIDOPTERA: LYCAENIDAE) : A COMPARISON OF DIFFERENCES THROUGHO UT A LOCALLY RARE YE T WIDE RANGING IMPERILED INSECT Introduction Callophrys irus Godart is a hairstreak butterfly in the fami ly Lycaenidae, widespread yet locally rare in its historical distribution, ranging throughout the eastern third of North America Specific to regularly disturbed habitats such as barrens, savannahs, and other similar areas where its larval host plants grow C. irus is in decline throughout most of its range, and is considered globally imperiled by NatureServe. Several characteristics including population structure, life history, and larval host plant specialization serve to make it particularly vulnerable t o habitat loss and other negative side effects of anthropogenic influence. C. irus may benefit from association with species that have received a large amount of attention in the conservation and restoration community, in particular the Karner blue butterf ly, Lycaeides melissa samuelis Nabokov a federally endangered species. This chapter will focus on detailing the biology and ecology of C. irus and its larval host plants, beginning with a brief section on the taxonomic history of C. irus a species and li fe history description, geographic range and habitat affinity, and current conservation status. New information from Florida populations is included, and contrasted to a summary of the species distribution wide. The final section will include reasons for a nd solutions to the decline of C. irus as well as some solutions that have potential to spillover to the community or habitat that this insect is part of.
18 Taxonomic History Though now recognized by the scientific name Callophrys irus this member of the L epidoptera family Lycaenidae has undergone numerous name changes since first described by Jean Pierre Godart in 1824. The original description by Godart was as Polyommatus irus and was so named from a specimen from the Museum Nationale e in Paris, France (Godart, 1824). The type locality is given as Jersey (Gatrelle, 1999). C. irus was next described as Thecla irus (Boisduval and LeConte, 1833). Later, the genus name Incisalia was assigned (Scudder, 1872). Gatrelle suggested the new genus Deciduphagus and was finally described as Callophrys irus by Pelham (Gatrelle 1999; Pelham, 2008). For this current study, the name Incisalia irus was commonly encoun tered in various state profiles and publications, as the renaming by Pelham is relatively recent. Three subspecies are currently recognized : C. irus irus (Godart, 1824), C. irus arsace (Boisduval and LeConte 1835), and C. irus hadros (Cook and Watson 1909) (Pelham, 2008). C. irus arsace refers to some isolated populations in South Carolina, and there has been some doubt expressed as to the type specimens, which were described from some drawings that were questionable in their taxonomic quality (Calhoun, 200 4). The hadros subspecies (Cook and Watson, 1909) is found west of the Mississippi River in Texas, Oklahoma, Arkansas, Louisiana, and probably Kansas; additionally this subspecies has been argued to be distinct enough to form its own species (Klots, 1951; Schweitzer, et al. 2011) Most discussion and comparison will be regarding C. irus irus as this subspecies contains most of the known populations, is
19 widely dispersed geographically, and has considerable variation in larval host plant choice, larval host plant feeding behavior, phenology, and colony connectedness Species Description and Life History What follows in this description is a compilation of information from peer reviewed journal articles, field guides, natural history publications, online data bases, and field observation. A full and highly detailed description of Callophrys (Incisalia) irus is given by Scudder (1889), but conflicts strongly with descriptions from every other source. Additionally, Scudder gives Callophrys (Incisalia) irus the co Callophrys (Incisalia) polios the hoary elfin. In general, C. irus adults and larvae are always found in close proximity to their larval host plants. Males are territorial, often perching on vegetatio n nearby host plant patches, and engage in vertical aerial combat flights and horizontal chasing flights (Albanese, et al., 2007). Females are easily recognized by larval host plant searching, probing, and egg laying (oviposition), discerning features for sex identification. Overall, C. irus is a drab brown grey appearance and is slightly larger than other species within Callophrys (Figure 1 1 ) The b ody is highly pubescent, brown on dorsal surface, lighte ning to grey ventral, with light grey to white l egs. Antennae are distinctly white and black patterned, black near the clubbed ti p with lighter brown at the apex. The wingspan ranges from 2.3 to 2.8cm, with 1 2 short tails posteriorly on the hind wing. Distal to the tails the hind wing is scalloped inwa rds. The dorsal wing surface is an unremarkable uniform brown, though the ventral surface shows quite the opposite in coloration and patterning. At the base, the ventral wing surface is dark brown, lighting towards the apex, with some checkering of darker and lighter colors. Olive iridescence is present on the leading edge of the forewing, and a distinct jagged line that is dark
20 brown and white patterned provides a border between the darker proximal and lighter distal regions. The jagged line is what distin guishes C. irus from the similar appearing C. henrici where the line is straight rather than jagged. Along with the fore wing, the hind wing also has sections of such a line, but a light dusting of white scales approximately midway in the wing continuing to the distal edge is the standout feature that gives C. irus its common name, the frosted elfin. Some sexual dimorphism is present in the adults, as male andreconia are seen best from the dorsal view (Schweitzer, et al., 2011). Finally, when at rest perch ing on vegetation, C. irus moves its hind wings forward and backward in the manner characteristic to most hairstreaks. Eggs are a pale green upon first being laid, drying to almost white before hatching (Figure 1 2). Larvae appearance is typical for the fa mily Lycaenidae, as they are onisciform (sowbug like), pubescent with many short setae, with prolegs and head hidden from view. The head becomes more visible in later instars when feeding, as the feeding mode changes from scraping of leaf or flower surface s to consumption of whole tissues. Early instar larvae are yellow to pale green, deepening to a light green in later instars (Figure 1 3 B C). Weak white chevron patterning is seen on abdominal segments in some populations, though in others this is absent. The final larval instar develops a distinct humpbacked appearance as it matures, commonly recognized as 1 4). Like other lycaenid larvae, C. irus larvae have a dorsal nectary organ present on the A7 segment. Ants of var ious species have been observed to be focusing their attention on this area in final instar larvae, a behavior observed throughout the wide geographic range C. irus inhabits (Albanese, et al., 2007;
21 Schweitzer, et al., 2011; M. Thom obs.). The degree to wh ich C. irus interacts with ants is not particularly understood, and should be the focus of further work. Pupae of C. irus are obtect, with legs, wingpads, and eyespots visible yet held close to the body by the pupal cuticle (Figure 1 5). The pupal case is a medium to dark brown in color, with darker spots occasionally present. As with the larvae, pupae are pubescent, but it is largely restricted to the abdomen. Pupae are found in the leaf litter or soil near the base of host plants, and are the diapause or overwintering stage (Barnes, 2003). When provided loose sand, captive reared mature C. irus larvae burrowed down to 1cm in the sand prior to pupation (Schweitzer et al., 2011). Pupae are also described in some locations to spin a loose cocoon silk and leaf debris (Cook, 1906). Finally, as the pupa is the overwintering stage and that this species is univoltine, the majority of the life cycle is spent as a pupa, some 9 10 months of the year. Geographic Distribution Historically, C. irus has been d ocumented f rom Ontario, Canada to Northern Flori da, from North Carolina west to Wisconsin and Texas for a total of 32 out of 50 states plus Washington D.C., and Ontario, Canada (Figure 1 6). Currently C. irus is thought to be extirpated in Ontario, Illinois, and Mai ne (NatureServe, 2012). A more detailed look at the geographic occurrence of the frosted elfin within the USA reveals a mix of scattered and cl ustered county records (Figure 1 7 ). Habitat Affinity and Associations In general, C. irus can be found in freque ntly disturbed habitats such as oak pine barrens, oak savannahs, upland pine or sandhill (Table 1 1). Such habitats share in common an open understory and a heterogeneous mix of open and closed canopy and edges. Examples include state and federally managed forests such as Ralph E.
22 Simmons Memorial State Forest (RESMSF) and Apalachicola National Forest (ANF) in Florida, and the state conservation areas such as the Rome Sand Plains and Albany Pine Barrens in New York and Kitty Todd State Nature Preserve in no rthwestern Ohio. Regularly disturbed areas such as edges are also frequented by C. irus such as roadsides and power line right of ways. C. irus is most often found in direct association to its larval host plants, which only occur in the above habitat typ es. Larvae can be found feeding on one of two primary larval hosts, sundial lupine, Lupinus perennis L., and wild indigo, Baptisia tinctoria (L.) R. Br., both from the plant family Fabaceae. These two plants also are the best place to encounter adults, and are a major identifying feature to separate them C. henrici and hoary elfin, C. polios Adult C. irus are reported to nectar on heaths such as blueberry (Vaccinium spp.) and huckleberry ( Gaylussiaci a sp.), as well as the flowers of L. perennis (Schweitzer et al., 2011). In the Midwest and Northeast C. irus is often associated with the Karner blue butterfly, Lycaeides melissa samuelis Nabokov (Lepidoptera: Lycaenidae) which feeds entirely on L. per ennis (Klots, 1951; Shapiro, 1974). L. melissa samuelis was listed as endangered under the Endangered Species Act in 1992, and is now part of many state restoration programs. The statewide Karner blue habitat conservation plan for Wisconsin gives a great o verview of the biology of L. m elissa samuelis and includes a wealth of information on habitat and species associations (Wisconsin Statewide Karner Blue Habitat Conservation Plan, 2010: Appendix B). In brief, L. melissa samuelis is a bivoltine lycaenid but terfly in the subfamily Polyommatinae, and becomes active in April
23 from overwintering eggs. The first adult flight follows larval and pupal development, and begins in late May (Swengel and Swengel, 1996; WI DNR, 2010). As of the recent 5 year USFWS review of the Karner blue recovery efforts, a large number of management units across 3 states have shown enough of an increase in population sizes to recommend reclassification or delisting (USFWS, 2012). For further discussion on the conservation of L. melissa samuelis see the conservation status section below. In Florida, C. irus is associated with sandhill pine and oak uplands where the larval host plant, L. perennis is found (Figure 1 8). These habitats feature a mosaic of pines such as longleaf pine, Pinus palustris Mill., and slash pine, Pinus ellioti Englem., along with a number of hardwood trees and shrubs such as turkey oak, Quercus laevis dwarf live oak, Quercus minima common persimmon, Diospyros virginiana L ., and wooly paw paw, Asimina incana (W. B artram) Exell. The understory typically consists of low lying vegetation, dominated by clumps of grasses such as wire grass, Aristida stricta M ichx., with occasional clumps of blueberry, Vaccinum sp., prickly pear cactus, Opuntia sp., and saw palmetto, Se renoa repens ( Bartram ) J.K.Small Other understory plants that are encountered include several milkweeds such as the pinewoods milkweed Asclepias humistrata Walter, tread softly, Cnidoscolus stimulosus (Michx.) Engelm. & Gray and gopherweed, Baptisia lanc eolata (Walter) Elliott Several insects aside from C. irus are occasionally found consuming L. perennis during the same period of time as C. irus The most commonly encountered at RESMSF is the pyralid moth, Uresephita reversalis Guenee. Multiple larvae o f this moth are typically found on a single L. perennis genet, and hatch from an egg cluster deposited on a leaf (Figure 9 A). Larvae of U. reversalis are highly distinctive, brightly patterned
24 with yellow and black, and are reported to be aposematic, sequ estering quinolizidine alkaloids that are present in consumed Lupinus tissues ( Bernays & Montllor, 1989 ) (Figure 1 9 B D). Other Lepidopteran larvae have been observed on L. perennis during the early spring in Florida, including several unidentified larvae from the Geometridae and Arctiidae families (Figure 1 10). Finally, a late spring to early summer insect herbivore frequently seen are beetles of the genus Epicauta (Coleoptera: Meloidae), at or after the time C. irus larvae have wandered off to pupate (F igure 1 11). Population Structure As was mentioned earlier, C. irus is relatively widespread in eastern North America, but several factors serve to designate it as rare. The rarity of an organism is often considered to be a function of its local population size, habitat specificity, and geographic range (Rabinowitz et al. 1986). For many animals such as mammals, local population size data are lacking, and is often not included in describing rarity. In insects, local populations can be highly variable acros s years, even across seasons, and so local population size may also not be a good measure of rarity. Instead, local population extent or area is a good replacement that still gets at the local scale dynamics of rarity (Pagel et al. 1991). Another factor t hat is impo rtant to consider is dispersal or migratory ability. C. irus itself exhibits limited dispersal ability, coloni zing patches around 2 km away (Schweitzer, et al. 2011). This is important to consider when thinking about colony connectedness, as in teraction (e.g. mating) between colonies separated by distances greater than the dispersal ability of the organism will be inhibited. C. irus is known to typically exist in small colonies, and these colonies often exist in a single isolated county in a num ber of states (Figure 1 7 ). There is a general trend of increasing isolation the further south and away from the coast, and at the very south of
25 the range of C. irus this isolation is quite apparent. For example, C. irus was first discovered in Florida in 1990 by Tom Neal at a power line right of way near Middleburg in Clay County (Gatrelle, 1991) C. irus was also recorded in another location in Clay County, near the small town of Penney Farms. Later, a C. irus colony was found approximately 80km north at Ralph E. Simmons Memorial State Forest in Nassau County. No other colonies were found between these locations, or in nearby counties The next nearest colony f ound was slightly south of Tallahassee at Munson Hills in the Apalachicola National Forest, this time over 230km west of the original Clay County Location. Additional colonies were found in several counties that spanned ANF in the year or two following. Right about the same time, C. irus w as observed in Blackwater River State forest, about 200 230km northwest of the ANF colonies. Plotting all of these records on a map demonstrates the large distances and uninhabited space between each colony (Figure 1 12). The record nearest to any of the Florida colonies is north of Tallahassee, in Thomas County Geo rgia, a county approximately 60km north of the Munson Hills record (Figure 1 13) Some populations of C. irus are connected, such as the colonies in southeastern Massachusetts, where metapopulation dynamics are likely (Albanese, et al., 2007). Given this, it is certainly possible that the known colonies at ANF in Florida are also part of a metapopulation network, as several colonies are in relative proximity to each other, and may be within the ~2km maximum dispersal distance reported (Schweitzer, 2011) (Fi gure 1 12). Some colonies are also quite large, and can be considered stable, such those in southern New Jersey (Golden and Pettigrew, 2005). Despite the size of large colonies, connectivity and local extinction probabilities remain important. Population
26 d ynamics of individual C. irus colonies throughout its range are especially important, and represent a knowledge gap in many areas, certainly in the lesser studied southern and western boundaries. Larval Host Plant Use C. irus specializes exclusively on leg uminous plants in the f amily Fabaceae. From a search of HOSTS (The Natural History Museum) database, there are four larval host plants listed, but two comprise of all the occurrences in a search of the C. irus literature: wild indigo, Baptisia tinctoria a nd sundial lupine, Lupinus perennis (Gatrelle, 1991; Schweitzer, 1992; Albanese, et al. 2007; Albanese, et al. 2008; Pfitsch & Williams, 2009). The other two plants from the HOSTS database, B australis and Crotalaria sagittalis may be only occasionally used, or may be used at the very western edge of the butterflies range in Louisiana, Texas, Kansas, Arkansas, and Oklahoma. These populations of the frosted elfin are the previously mentioned subspecies C irus hadros different from the C irus irus subsp ecies elsewhere or the proposed South Carolina subspecies C irus arsace For the majority of cited literature records C. irus has two larval hos t plants throughout its range: B. tinctoria and L. perennis (Figure 1 14 ). Typically one or the other host pl ant is utilized by larvae in a population, but not both in the same population (Schweitzer, 1992; Schweitzer, et al. 2011). A brief note about L. perennis : there are essentially two subspecies, gracilis (sundial lupine) and perennis (blue lupine). Sundial lupine is found in southern states, and blue lupine is found in the northeast and Midwest. Sundial lupine tends to be smaller and more prostrate than blue lupine, with smaller leaves and a diminutive appearance in comparison to the larger blue lupine. Bey ond appearance, it is unknown if there are any other differences between the two
27 subspecies, so for the purposes of this review, they will be considered both as L. perennis and assumed interchangeable. B. tinctoria and L. perennis occur in similar areas, preferring similar types of habitat, and so the areas they inhabit include a large amount of overlap (USDA, NRCS, 2013). Both species also serve as the larval host plant for a number of other insect species, including the federally endangered Karner blue, L m elissa samuelis as mentioned in the previous section Competition for resources is likely avoided by differences in flight season, as the frosted elfin is active early in the spring compared to the late spring and early summer active L m elissa samuel is in the Albany Pine Bush (Barnes, 2003). Utilization of larval host plants by C. irus varies between states, and even between some counties in the northeast US. Some geographic patterns are apparent, and have been suggested by Gatrelle and Schweitzer, i n that Eastern Coastal Plain populations utilize B. tinctoria and the remaining populations use L. perennis (Figure 1 15 ). The situation becomes complex in several northeastern states as both L. perennis and B. tinctoria grow in the same areas and are bo th utilized by C. irus County level differences have been observed in Connecticut, Massachusetts, and New York, and the same behavior may be present in nearby eastern coastal plain states such as Rhode Island and the more distant Delaware (Pfitsch & Willi ams, 2009; Bried, et al. 2012). A single population in western Maryland was reported to feed on B. tinctoria yet the more numerous coastal populations were reported to feed only on L. perennis despite B. tinctoria being more common (Frye, 2012; Schweitz er, 1992). Exclusive L. perennis feeding despite the presence of B. tinctoria was also reported by the same authorities in New Hampshire and New York, with the opposite being true (exclusive B. tinctoria
28 feeding) in New Jersey and Pennsylvania. C. irus typ ically use s L. perennis through the Midwest, though B. tinctoria is also present in many of these states (Figure 1 14) At the southern distributional boundary in northern Florida, C. irus larvae feed exclusively on L. perennis Other lupine species can b e found in the same areas where C. irus is present including L. villosus and L. diffusus but frosted elfins have not been observed using these plants as hosts. L. villosus grows in the same location as L. perennis at Ralph E. Simmons Memorial State Fores t (Nassau County) and Apalachicola National Forest (Leon, Franklin, and Wakulla Counties) as does L. diffusus at a site in the Blackwater River State Forest in Okaloosa County. A survey of all three Lupinus species at the BRSF site in 2011 revealed frosted elfin eggs only on L. perennis Additional surveys of L. villosus at RESMSF, ANF, and at a single site in Clay County, Florida, did not uncover any C. irus eggs, larvae, or feeding damage. Wild indigo, B. tinctoria has not been recorded in Florida, but s everal other Baptisia species including B. lanceolata and B. lecontei are present in areas C. irus inhabits, such as at RESMSF and ANF. At the RESMSF site, no evidence of egg laying or larval feeding damage by the frosted elfin has been observed on either Baptisia species, but other L epidoptera n larvae have been observed feeding such as the Pyralid moth Uresephita reversalis (Lepidoptera: Pyralidae). The caterpillars of this moth are also often observed feeding on L. perennis throughout the active C. irus s eason, and there is likely some amount of competition for lupine host plants by it and other species (Figure 1 12 A: a geometrid larvae and C. irus from ANF; Figure 1 12 B: U. reversalis and C. irus at RESMSF). One C. irus larva was observed to be dead on a plant containing active and alive U. reversalis larvae at RESMSF in 2012, and may be a
29 result of interspecific aggression (Figure 1 12 C D). A variety of other Lepidopterans were observed feeding on L. perennis at RESMSF, including geometrid and arctiid larvae and adult blister beetles (Coleoptera: Meloidae) of the genus Epicauta L. perennis F eeding B ehavior In Florida, adult C. irus lay eggs on the new growth of the L. perennis plant, the join between two leaflets, the growing flower stalk, or on openin g or mature flowers (Figure 1 2 ). Up to 7 eggs have been recorded on a single plant, in contrast to the B. tinctoria feeding C. irus of Massachusetts that lay eggs singly on plants (Nelson, 2002; Albanese, et al. 2007a; Albanese, et al. 2007b). Early ins tar L. perennis feeding larvae typically can be found on the underside of leaves where they eat away the bottom and inner tissues leaving behind a series of opaque holes (Figure 1 3 A B). Feeding on leaves continues as the larvae mature, proceeding to compl ete consumption of leaf tissues (Figure 1 3 C). Flowers and early seed pods are sometimes eaten by larvae, but not developing flower heads (Figure 1 3 D). In contrast, flower and seed pod feeding is reported most often in more northern areas, including the A lbany Pine Bush of New York (Barnes, 2003). This is likely due to differences in larval host plant and C. irus phenology, discussed in the following section. L. perennis feeding C. irus favor different parts of the plant throughout their range. Northern p opulations primarily feed on flowers and seed pods, while Florida populations feed on leaves for the most part, with occasional flower feeding observed. This difference in plant tissue feeding is likely due to what is available to the larvae once they hatc h from their eggs. L. perennis is one of the first green plants to emerge in the winter in sandhill forests of Florida, and at RESMSF, L. perennis was observed to be up at the beginning of February some years (2011, 2012). It is possible it might be up
30 bef ore then; some L. perennis may even survive the sh ort and mild winter that happen some years, given enough moisture. L. perennis has been observed through mid November, and so this is a likely possibility. Combining a February (or earlier) emergence of pla nts with a mid February emergence of C. irus results in early plants to be fed upon by larvae. L. perennis plant surveys at RESMSF at the beginning of the adult flight and during the initial feeding of the larvae reveal that the majority of plants generall y do not have flowers, or are just beginning to develop an inflorescence. This is in contrast to the C. irus and L. perennis phenology farther north, where C. irus tends to emerge later in the spring, after flowering of the L. perennis has occurred. L. per ennis begins to sprout from rhizomes in March, flowering as early as April in the Carolinas to May or June farther north (Wofford, 1989; Dirig, 1994; Grigore, et al. 1996). Typical C. irus flight for the northern range as mentioned previously is mid April to early June, which corresponds to the flowering time for L. perennis Phenology Adult Flight T ime The re are clea r differences in phenology throughout the range of C. irus specifically concerning the timing of the adult flight and larval development tim e. Winters are typically short in northern Florida, characterized by dry and mild conditions, with only a few days of freezing or slightly below freezing temperatures. As a result, the growing season begins early, and accordingly, the C. irus adult flight season also begins early. C. irus adults emerge during the early part of spring, which often begins mid to late February but there is often considerable variation. For example, in 2010 a late cold spell delayed the adult flight to begin in late March to mid April. Even with delays due to late cold spells, the adult flight begins much earlier in the year than in
31 areas such as Massachusetts and New York, where the adult flight typically ranges from late April to early June, peaking around mid May (Albanese, et al. 2007; Pfitsch & Williams, 2009). Larval Development T ime Larval development time (hatchling to pupation) is shorter in the southern edge of the C. irus distributional range, as would be expected by the typically warmer climate at lower latitudes From 2010 to 2012 a t the RESMSF site in Florida, larval development was observed to be 4 5 weeks. In contrast, C. irus larvae in Massachusetts are reported to develop in 6 weeks (Albanese, et al. 2007). Despite an earlier and quicker development time in t he southern edge of its range, C. irus is completely univoltine. There have been some suggestions of a bimodal emergence of adults or even two short but distinct adult flights in Florida (J. Daniels, pers. comm.), but this this was not observed from 2010 2 012. Conservation Status C. irus is currently known or thought to be extirpated in se veral areas, including Ontario, Maine, Illinois, and Washington D.C., with populations declining throughout the rest of its range. The result of all these population decli nes is the listing of the frosted elfin as S1 critically imperiled or S2 imperiled in 20 out of the 32 states or provinces it has been documented to inhabit ( Table 1 2; NatureServe, 201 3 ). Many factors are contributing to the decline of C. irus including aspects related to habitat loss and direct mortality. Fire or disturbance suppression, land development, and local extinction of larval host plants are the most common examples of habitat loss cited (Albanese, et al., 2007; Swengel and Swengel, 2007; Pfit sch and Williams, 2009; S chweitzer, et al., 2011). Browsing of flower heads of L. perennis by white tailed deer,
32 Odocoileus virginianus Zimmerman, has been linked to reduced seed set and the consumption of C. irus eggs or larvae that are present (Schweitze r, et al., 2011; Frye, 2012; Jon Shuey, pers. comm). Finally, direct death or decline from anthopogenic disturbance has been suggested, particularly the improper timing, frequency, or extent of said disturbance leading to the attrition of populations (Swen gel, 2001; Swengel and Swengel, 2007). These factors are likely to have such a strong effect on C. irus populations because of the main features of its life history: locally rare, small and isolated populations (or colonies), larval host plant specializat ion, and habitat specialization In particular, it is widely recognized that small populations are strongly influenced by demographic stochasticity, particularly genetic drift and sex ratio shifts. It can be argued that C. irus populations are especially b alanced between persistence and extinction, and any perturbation, whether natural or anthropogenic in nature, will have a great effect on a given population. In addition to the decline of C. irus one of the larval host plants is a lso in decline L. perenn is is becoming quite rare in many states in the northeast US, states where the C. irus utilize it as a larval host (Table 1 3) As of 2003, L. perennis was thought to have been reduced to ten percent of what it had been 15 20 years prior in the Albany Pine Bush of New York (Barnes, 2003). In addition, according to the USDA PLANTS database, L. perennis is either considered threatened or endangered in several additional states where C. irus feed s upon it, including Maryland, New Hampshire, Rhode Island, and V ermont. In Maine, L. perennis is presumed extirpated, and is certainly a major contributing factor to the extirpation of C. irus there. What is clear is that there is a decline in L. perennis in the northeast US, which presents several
33 dilemmas for the C. irus populations that feed exclusively upon it. First is that the L. perennis feeding populations will go extinct as their larval host plant declines and goes extinct. This lack of a food source might force a shift to a different host plant, and in severa l of these states B. tinctoria is present, which currently acts as a food source for other frosted elfin populations. If no shift is possible, either due to lack of availability of B. tinctoria or lack of a response by previously L. perennis feeding popula tions, then the L. perennis feeding populations would go extinct. As L. perennis disappears from the northeast, so would L. perennis feeding frosted elfins, leaving a distinct regional difference in host plant use. Such a difference would provide the geogr aphic isolation necessary to further the host race speciation that is likely to already be occurring in C. irus in these areas (Mayr, 1942; Bush, 1969; Dres & Mallet, 2002). Efforts To Reduce The Decline Of C. Irus Habitat Management Well informed habitat management in areas that C. irus occupies will likely have the greatest positive effect on C. irus persistence. Research on this topic for C. irus has largely focused on restoring habitat it occupies, specifically managing for the plant community which inc ludes larval host plants and adult nectar sources, essentially returning it to a state prior to anthropogenic perturbation. Prescribed f ire A common, yet unavoidably controversial tactic to habitat restoration is the reintroduction of disturbance regimes, particularly fire. Fire is recognized as being the major source of disturbance in the pine barrens, oak savannahs, and upland pine sandhill (habitats that C. irus occupies). Fire suppression has a well known negative effect on the the persistence of these habitats (Vogl 1974; Nuzzo 1986; Borgerding et
34 al. 1995 ; Whelan, 1995; Albanese, et al. 2007; Swengel and Swengel, 2007; Pfitsch and Williams, 2009). However, legitimate concern has been raised on using fire as a restoration tactic, particularly when used on small, isolated, or fragmented habitats, as there is no refuge to escape the immediate mortality that fire can cause (Panzer, 2002 and references within there are about 10 in one section that I am referring to, as Panzer here summarizes this literatur Swengel 2007 < the fire refugia paper). In many areas, C. irus will certainly be negatively impacted by too aggressive of an approach to habitat restoration, and instead may need to be tempered somewhat. Fu rthermore, optimizing the use prescribed fire to have the best positive effect on the whole community is likely to have the greatest success. Singling out a particular species such as C. irus and managing specifically for it may not have great yields, as t he rest of the community or ecosystem may be ignored; increased knowledge of this system would help solve this potential problem. Some suggestions to improve fire as a habitat management tool for the persistence of species such as C. irus include designati ng portions of manag ed areas to be left unburned, using a long fire return interval, or the use of other types of management such as light grazing, mowing, or mechanical cutting (Swengel & Swengel, 2007; Vogel, et al, 2007) Tree r emoval One example of an alternate restoration technique to prescribed fire involved a C. irus colony in the Rome Sand Plains in New York. White pine, Pinus strobus L., had become established and was negatively impacting the lupine community through increased shade following years of urban expansion and fire suppression (Pfitsch and Williams, 2009). L. perennis growth and recruitment is poor in the shade, needing at
35 least a partially open canopy to grow well (Pavlovic and Grundel, 2009). After removal of a number of P. strobus tree s at this location, L. perennis performance increased, as measured by increased ground coverage and flowering rates. C. irus behavior was also shown to be less inhibited in areas of P. strobus removal, with males establishing mating territories within newl y cleared patches. Deer Population M anagement Reduction of O. virginianus populations would likely reduce their effects on host plants and of any C. irus eggs or larvae present In the absence of reduced O. virginianus populations, exclusion in sensitive areas could be feasible, as this was shown to increase both the chance of seeding and the number of L. perennis seeds produced (Frye, 2012). Flower stalk consumption and other behaviors of O. virginianus surely impact other organisms in the barrens and sa vannah habitats in which L. perennis and C. irus reside, so, a larger management plan for nuisance O. virginianus populations could then aid conservation or restoration of these species. Im plications Of C. irus Conservation L. melissa samuelis S pillover As L. melissa samuelis and C. irus occupy the same habitat s and utilize the same host plants and othe r resources, efforts to conserve or restore C. irus populations have the potential to aid both species, even other species that rely upon the same resources. Current conservation resources are largely focused on L. melissa samuelis so the current spillover effect is instead to C. irus The existing framework for butterfly monitoring could certainly be expanded to C. irus There may be conflict in protecting b oth species, and there are some differences in habits: C. irus emerges earlier in the year than L. melissa samuelis and feeds
36 primarily on flowers. In the light of any climate warming scenario, both species may shift towards earlier emergence, and if there is not a roughly equivalent shift in the phenology of L. perennis available plant tissues may change. Less mature plants would not have as much flower and seed pod tissues, possibly with a shift in C. irus to feeding on leaf tissue. This would possibly c hange the amount or quality of food for L. melissa samuelis assuming that they do not shift to feeding on flower and seed pods. Future study in this area is crucial for persistence of these communities, though this need is certainly apparent for most ecos ystems around globe. Community and Biodiversity P reservation C. irus is a representative of unique and disappearing habitats (Barnes, 2003). Protection and promotion of C. irus using ecologically and evolutionarily sound science could benefit other specie s. Such work would involve efforts improving features throughout the communities it is part of, and this has the side effect of an overall preservation of these habitats. Relatively high species diversity is present in the frequently disturbed barrens, sav annah, coastal plain, sandhill, and upland pine habitats, particularly if disturbance to these areas is heterogeneous Such a situation allows for many plant species of differing successional states to exist, and with a higher plant richness comes a wider range of animal species that rely upon these plants. This is in fact the situation predicted by the intermediate disturbance hypothesis, where species richness reaches a peak at levels of disturbance that are neither too rare nor too frequent (Grime, 1973; Connell, 1978) Lastly, as these habitats are disappearing and becoming increasingly fragmented, management of the small isolated remnants becomes more important now and in the future.
37 Scientific D iscovery A final benefit to conservation or restoration ef forts aimed at C. irus is that it could provide additional evidence to confirm hypotheses and support new emerging theories, and perhaps even reveal new knowledge. For example, a more in depth look at the distribution of C. irus populations or colonies und er the framework of metapopulation ecology or macroecology may reveal patterns that yield insight into aspects such as maintenance of genetic diversity of small popuations, genetic diversification or speciation resulting from isolation. Related to these f actors are characteristics such as host plant choice and specialization, and host plant or host race speciation. An example of scientific study is that initial studies on C. irus have revealed that it pupates in the leaf litter or soil, remaining there for most of the year. While residing as a pupa in the ground for extended periods is not entirely unique among insects, it is an extreme example within the butterfly clade Papilionidea, of which little is known about immature life stages for many species. Fur thermore, the implications of such a life history strategy are manifold, from an evolutionary perspective (avoidance of predators, environmental stress, or stochastic events) to the practical, surely influencing any management plans that include C. irus or similar species. Future Outlook In Florida C. irus is listed as critically imperiled, and has not been observed at or near several locations it was once known to inhabit. It was first discovered in 1990 in Clay County, but has not been seen at this locat ion for several years. In 2008 C. irus was discovered in several sites in the Apalachicola National Forest, near Tallahassee, but once again has not been observed at the Munson Hills site since 2010. It is possible that these populations in ANF make up a network of connected sub populations, as a
3 8 metapopulation, though genetic and dispersal studies would be necessary to confirm or deny this. NatureServe indicates C. irus is a metapopulation species, and there is evidence for such in New Jersey and Massachu setts, so it seems likely (Albanese, et al. 2007). Aside from this potential metapopulation in the ANF spanning multiple counties in the Florida panhandle, there is a significant lower density trend in the southern edge of the range of C. irus compared to populations in the north, as evidenced by larger clusters of counties having records of C. irus (Figure 1 7 ). Metapopulation theory predicts l imited or nonexistent connectivity as a result of this lower density in the southern range, leading to colony iso lation and then increased chance of extinction from stochastic or catastrophic events. Regarding phenology, one implication of earlier C. irus adult flights in Florida compared to northern locales is the potential of the northern populations to adapt to ea rlier emergence times and the earlier phenology of host plants they may encounter as a result of a warming climate. Larvae feed primarily on L. perennis leaf tissue in Florida, where the flight and growth seasons begin much earlier, a local adaptation to a warmer climate. Emerging earlier as a result of warmer temperatures could increase competition for the resources that were previously temporally or spatially partitioned between multiple species, as is the case for C. irus and the federally endangered L. melissa samuelis As a recap, b oth species feed on L. perennis but are active at slightly different times and feed on different plant tissues: C. irus emerge s earlier and feed s on flowers and seed pods, L. melissa samuelis emerg es about a month later and feed s on lea ves Assuming a shift towards earlier emergence, the flower and seed pod tissues would not be available, and C. irus might shift to feeding on leaf tissue. This would
39 reduce the amount of food for L. melissa samuelis larvae, assuming that they do not shift to feeding on flower and seed pods if they are still available Regardless, there is a potential for C. irus occupying a colder climate to adapt to warmer and earlier conditions, and an equal potential for this to have significant effects on other organisms within the same community.
40 Table 1 1 Regional overview of habitat affinities of extant populations of C. irus *Likely source for type specimen. State Habitat Florida Sandhill and upland pine forests Michigan Oak savan nah, oak pine barrens Massachusetts Coastal plain pitch pine scrub oak barrens New York Pine barrens of Long Island and Hudson Valley, including the Albany Pine Bush New Jersey Dry clearings and natural open areas such as savannahs or power line right of ways/roadsides Table 1 2. Conservation status of C. irus from NatureServe as of April 2013. Conservation status Location Description G3 Global Imperiled Across Entire Range SH IL, DC Presumed Extirpated SX CA ON, ME Extirpated S1 DE, FL, IN, KY, MD, NH, OH, OK, RI, WV, WI Critically Imperiled S1S2 NY Critically Imperiled Imperiled S2 NJ, NC, PA, VA Imperiled S2S3 CT, MA, MI, TN Imperiled Vulnerable S2S4 GA SU AL SNR AR, KS, SC, TX, VT Not ranked, not assessed
41 Table 1 3. Conservat ion status of L. perennis from USDA PLANTS database as of April 2013. Location Description Maine Possibly extirpated Maryland Threatened New Hampshire Threatened Pennsylvania Rare Rhode Island Special concern Vermont Endangered
42 Figure 1 1. Adults of Callophrys irus (Lepidoptera: Lycaenidae) April 2010 (A : Male ) and March 2011 (B : Female ) at Ralph E. Simmons Memorial State Forest, Nassau County, Florida Photos courtesy of M. Thom (A) and Brian Camposano (B ).
43 Figure 1 2 Placement of eggs of C irus on, L perennis at Blackwater River State Forest, Okaloosa County, Florida and Ralph E. Simmons Memorial State Forest, Nassau County, Florida, between 2010 and 2012. A) Developing flower stalk; B) Upper leaf sur face; C) At the base of new growth Photos courtesy of M. Thom (A) and B. Camposano (B and C).
44 Figure 1 3 Larvae feeding behavior, C. irus on L. perennis leaves (A C) and flowers (D) at Ralph E. Simmons Memorial State Forest, Nassau County, Florida between 2010 and 2012. In D), the arrow points to a hole bored into the flower from a C. irus larva. Photos courtesy of M. Thom.
45 Figure 1 4. Late instar larva of C. irus at Ralph E. Simmons Memorial State Forest, Nassau County, Florida. Ph otos courtesy of M. Thom.
46 Figure 1 5. Pupa of C. irus discovered at the forest floor June 2012, at Ralph E. Simmons Memorial State Forest, Nassau County, Florida. For B), major ruler marks are in centimeters. Photos courtesy of M. Thom.
47 Figure 1 6 State and province distribution of C irus Extirpation status is adapted from NatureServe distribution map on C. irus Figure 1 7 USA county records for C irus Records are from the Butterfly Occurrence Database from the National Atlas of the United States, USGS, and the Florida Natural Areas Inventory.
48 Figure 1 8. C. irus habitat series following a February controlled burn at Ralph E. Simmons Memorial State Forest, Nassau County, Florida, 2011. A) March 24; B) April 6; C) May 4; D) June 1. Photos courtesy of M. Thom.
49 Figure 1 9. U reversalis on L. perennis at Ralph E. Simmons Memorial State Forest, Nassau County, Florida. A) Arrow points to an egg mass; B D) larvae. Photos courtesy of M. Thom.
50 Figure 1 10. Lepidopteran larvae on L. perennis A) unidentified Geometridae larva and C. irus larva at Apalachicola National Forest; B) unidentified Geometridae larva at Ralph E. Simmons Memorial State Forest, Nassau County, Florida; C D) unidentified Arctiidae larvae at Ralph E. Simmons Memorial State Forest, Nassau County, Florida. Photos courtesy of M. Thom.
51 Figure 1 11. Blister beetle Epicauta sp. feeding on L. perennis May 5 th 2010, at Ralph E. Simmons Memorial State Forest, Nassau County, Florida. Photo courtesy of M. Th om.
52 Figure 1 12. Florida records for C. irus ANF = Apalachicola National Forest; BRSF = Blackwater River State Forest; RESMSF = Ralph E. Simmons Memorial State Forest, Nassau County; CLAY = Clay County. Records courtesy of John Calhoun of Butterflies a nd Moths of North America, Southeast Division; Dean Jue of the Florida Natural Areas Inventory.
53 Figure 1 1 3. Southeastern USA county records for C. irus Shaded counties represent records from the Butterfly Occurrence Database from the National Atlas of the United States, USGS, and the Florida Natural Areas Inventory Dark grey shaded county = Thomas County, Georgia.
54 Figure 1 14 State and province records for L perennis and B tinctoria in North America. (Records are from the USDA PLANTS database, USDA NRCS, 2013). Figure 1 15 Larval host plant use by C irus Confirmed plant use records are from: Butterflies and Moths of North America; Allen, 1997; Gatrelle, 1991; Butterflies of Wisconsin; Michigan Natural Features Inventory; Schweitzer, 1992; Bried, 2012; Frye, 2012.
55 Figure 1 16 Simultaneous feeding on L. perennis by L epidopteran larvae. A) Callophrys irus and a n unidentified larva (Geometridae) at Apalachicola National Forest, Leon County, Florida in 2010; B) to D) C. irus and Uresephit a reversalis (Pyralidae) at Ralph E. Simmons Memorial State Forest, Nassau County, Florida in 2012. Photos courtesy of M. Thom.
56 CHAPTER 2 PATTERNS OF MICROHAB ITAT AND LARVAL HOST PLANT USE BY CALLOPHRYS IRUS GODART AT RALPH E. S IMMONS MEMORIAL STAT E FOREST Introduction Habitat quality is fundamentally important to the persistence of a given species (Pulliam 1988). For herbivorous insects, habitat quality is a function of a highly specific combination of complementary resources that includes host plan ts or nectar sources, and physical features that serve useful functions (Baurefeind, et al., 2009). Examples of physical features include perch or shade structures, and can serve important roles for courtship, mating, or protection from adverse environment al conditions. Because several features of habitat may be relevant to population growth and persistence, identifying what features of the habitat are most important to species remains a focus for ecology and conservation. One method used to answer this is to sample the patterns of distribution or abundance of the organism within a particular area while simultaneously measuring habitat features, drawing conclusions from the correlations between abundance and habitat features Several recent studies focused o n the ecology of habitat and larval host plant specialist insects have done just this (Grundel and Pavlovic, 2007; Trager, et al., 2009). One species of butterfly in particular, the frosted elfin Calloph r y s irus Godart (Lepidoptera: Lycaenidae) has been s tudied in part of its range. As a general rule, adult C. irus are correlated to areas of high host plant density and intermediate amounts of tree cover or canopy openness. An isolated population in southeastern Massachusetts was found to be positively corr elated to the density of wild indigo, Baptisia tinctoria (L.) R. Br., the local larval host plant (Albanese, et al., 2007). Furthermore, patterns of
57 adults were associated with areas of at least some shade, and areas of low shrub undergrowth. Similarly, a population of adult C. irus in New York was found to be correlated to high densities of wild blue lupine, Lupinus perennis L., the larval host plant exclusively used in this population (Pfitsch and Williams, 2009). Areas of extensive cover of white pine, P inus strobus L., were negatively correlated to C. irus adults, indicating that open areas are important. Areas of heavy shade restrict C. irus behavior, as male mating territories are limited to open areas. L. perennis also performs poorly in heavy shade, with seedling survival and longevity lowest in full shade compared to open or partially shaded areas (Pavlovic and Grundel, 2009). Finally, populations of C. irus in Wisconsin were shown to be positively associated to higher densities of L. perennis and m ost commonly with estimated canopy cover between 10 50% (Swengel, 1996). Canopy cover has been previously shown to be important to the Karner blue butterfly, Lycaeides melissa samuelis Nabokov, for both quality of larval host plant L. perennis and for ovip osition by L. melissa samu e lis females (Grundel, et al., 1998a; Grundel, et al., 1998b). While it is clear from previous work on habitat affinity of C. irus that high host plant density and low to intermediate levels of shade are important, these studies f ocused entirely on patterns of adults, seen from transect walks or point observations. Immature stages such as larvae and pupae may be associated with different patterns, and finer scale differences may be of equal importance. A study from a similar region in southeastern Massachusetts followed female C. irus egg laying and survival to late instar larvae, while measuring similar microhabitat variables used for adults such as host plant density ( B. tinctoria ), canopy cover, and several variables related to w oody
58 vegetation (Albanese, et al., 2008). Presence of late instar larvae was correlated most strongly to larger host plants and shaded areas, however, female oviposition did not show any such preference. It was concluded that female C. irus indiscriminatel y laid eggs in a variety of microhabitats, suggesting that fine scale differences were not important features for oviposition preference. Microhabitat variables analyzed were r dead ground cover or whether plants exhibited signs of previous feeding damage by other organisms. Such factors are likely to be important for immature stages, as they influence plant quality and access, and are indicators of predation or competition. W hile considered rare and imperiled, C. irus is a wide ranging insect, and is found throughout the central and eastern US, historically as far north as Ontario, Canada, and as far south as northern Florida. No studies have focused on populations of C. irus south of the Midwest or Northeast, and so the patterns of habitat association may be different. Larval host plant species are also different throughout the range of C. irus and so there is the possibility of additional differences. Ignored in the above st udies is the fact that late instar C. irus travel to the ground to pupate in the leaf litter or soil, and so these conditions may also be important. Even if female C. irus are somewhat random in their oviposition, the fate of the larvae is not, and it is s pecifically tied to the microhabitat they are in (Albanese, et al., 2008) I seek to answer a series of questions related to C. irus and habitat affinity in an isolated population Florida: (1) Determine which L. perennis plants female C. irus are choosing /not choosing to lay their eggs on, (2) Characterize these plants and the
59 microhabitat surrounding them, and (3) describe the patterns of density of L. perennis occupied and unoccupied by C. irus immatures. Investigations of the above patterns are relevant for several reasons. The s trong relationship between female choice (preference) and offspring survival (performance) has been suggested previously, in that f emale insects will oviposit on resources that maximize offspring fitness and is likely to be most important to host plant specialists ( Dethier, 1959a, 1959b; Singer, 1972; Jaen ike, 1978; Craig and Itami, 2008). Given that a species like C. irus is a host plant specialist, one should assume that females are looking to maximize offspring fitness, though some evidence suggests otherwise (Albanese, et al. 2008). A look at the preference side of this relationship is the focus of this study, to reveal the pattern of female choice. Methods Study Site Ralph E. Simmons Memorial State Forest (RESMSF) in Nassau Co unty, Florida, contains an extant colony of C. irus and was the study site for this research RESMSF is a 1,472 hectare forest containing a diverse array of natural communities including sandhills, seepage slopes, low pinelands, and riparian habitats alon river which forms its northern border. The C. irus colony is restricted to an approximately 20 hectare section ( 30.797 N, 81.949 W) in a xeric sandhill pine oak upland, where the larval host plant sundial lupine, Lupinus perennis ssp. g racilis (Nutt.) Dunn, is located (Figure 2 1). Other non woody plants common to the colony area include wiregrass, Aristida stricta Michx. gopherweed, Baptisia lanceolata (Walter) Elliot, wooly pawpaw, Asimina incana (W. Bartram) Exell, and pinewoods milkweed, Asclepias humistrata Walter. The woody plant community is characterized by longleaf
60 pine Pinus palustris Mill., slash pine, Pinus el lioti Engelm., turkey oak, Quercus laevis Walter, b lack cherry, Prunus serotina Erhr., and various other Quercus species. Soils consist of moderately well drained fine sand. RESMSF is managed for multiple uses including timber, hunting, and other public us es, with an emphasis on ecosystem management and ecological restoration of native communities. The prominent management technique used is prescribed fire, with dormant and growing season fires conducted in subdivided sandhill units approximately every othe r year since at least 2008 using a variety of techniques including aerial incendiary ignition and hand ignition by ground crews. Study Organisms Callophrys irus Godart (1824) is a hairstreak butterfly in the Lycaenidae family of the insect order Lepidopt era, found in small locally rare colonies in oak pine barrens and savannah habitats in Eastern North America It is most often found closely associated with its two main larval host plants Lupinus perennis L. and Baptisia tinctoria (L.) R.Br. Adult C. irus emerge and become active in the early spring, which can be as early as February at RESMSF. After mating, female C. irus lay eggs on the new growth of the L. perennis plant, the join between two leaflets, the growing flower stalk, or on ope ning or mature f lowers (Figure 2 2 A B ) Eggs are a pale green upon first being laid, drying to almost white before hatching after approximately 5 days. Larvae appearance is typical for the family Lycaenidae, as they are onisciform (sow bug like), pubescent with many shor t setae, with prolegs and head hidden from view. The head becomes more visible in later instars when feeding, as the feeding mode changes from scraping of leaf or flower surfaces to consumption of whole tissues. Early instar larvae are yellow to pale green deepening to a light green in later instars (Figure 2 2 C D). Early instar L.
61 perennis feeding larvae typically can be found on the underside of leaves where they eat away the bottom and inner tissues leaving behind a series of opaque holes (Figure 2 2 E ) Feeding on leaves continues as the larvae mature, proceeding to complete consumption of leaf tissues. Flowers and early seed pods are sometimes eaten by larvae, b ut not developing flower heads. Lupinus perennis L. ssp. gracilis (Nutt.) D. Dunn is a pere nnial legume in the pea family Fabaceae that is found in oak and pine savannahs in eastern North America, from southern Canada to northern Florida. The gracilis subspecies exists in the southern states of the US, and is referred by its common name sundial lupine (Figure 2 3) It is more diminutive than the northern version, but otherwise is th e same: both are host plants for C. irus Other notable organisms found in close proximity to C. irus include another L. perenni s herbivore, and two species of plants. The larvae of the pyralid moth Uresiphita reversalis Guenee (Lepidopera: Pyralidae) are sometimes found feeding on L. perennis Multiple larvae of this moth are typically found on a single L. perennis genet, and hatch from an egg cluster deposited on a le af (Figure 2 4 A). Larvae of U. reversalis are highly distinctive, brightly patterned with yellow and black, and are reported to be aposematic, sequestering quinolizidine alkaloids that are present in consumed Lupinus tissues ( Bernays & Montllor, 1989 ) (Fi gure 2 4 B). Wiregrass, A. stricta and small sprouts of turkey oak, Q laevis are the two most commonly encountered species of understory vegetation in the immediate vicinity of C. irus at RESMSF. Along with larger Q. laevis trees, several pine species in cluding P. palustris P. ellioti contribute to form the forest canopy.
62 Focal Follow and Mark omg of C. irus In an effort to document female oviposition behavior and generate a meaure of population size, C. irus adults were observed and captured in 2012. To find adult C. irus transects were walked through the ~20 hectare L. perennis area, also in areas where C. irus had been seen earlier in the year and in previous years. Once an adult C. irus was discovered and positively identified, it was captured with a net, and size measurements were taken: forewing cord length and body length recorded to the nearest millimeter. The individual was photographed, and then carefully marked with a small dot on the dark colored region of the hind wing with a silver felt tip m arker (Sharpie Metallic Silver) unique to the individual Marking individuals made it possible to identify if the it had been captured previously. The location of the dot was recorded along with the focal follow and size measurements for reference of a rec apture. Following the workup, the now recorded and marked individual was released, and one researcher equipped with a video camera filmed and followed the individual. The other researcher, armed with binoculars, would describe behaviors of the focal C. iru s which are described in Table 2 1. Observers kept an approximately 3m distance from the target, to reduce the effect of observation on the behavior of the individual. The protocol was altered for the final C. irus individual found, and instead the focal follow occurred before the capture and measurements. This change was due to the fact that captured C. irus quickly flew away once they recovered from capture, and little to no behavior observations were able to be made.
63 Microhabitat Surveys Sampling surve ys of the C. irus and L. perennis microhabi tat at RESMSF were conducted in prings of 2010 and 2012. Methods were slightly different between years and so are split up in the following sections. 2010 Sampling was conducted from April 4 to June 9 th with spe cific sets of data collected on separate days (Table 2 2). The sampling was discontinued in a portion of the site following a prescri bed burn April 21 22 (Figure 2 5 A). Approximately 200 250 previously surveyed L. perennis genets were consumed by the fire of which at least 50% had C. irus eggs, larvae, or the presence of feeding damage. Plants were marked by the placement of a flag labeled with a unique number. Multiple samplings of the same plants were taken during the spring, with different types of da ta taken over this period representing snapshots through larval C. irus development. Plant information recorded included total number of stems and leaves for a genet, and flowering state ( no flower, stem but not yet in bloom, blooming, and seeding ) Each L perennis genet was observed for eggs, larvae, or feeding damage of C. irus Eggs presence, whether hatched or un hatched was recorded, and were most often visible as pale green to white <1mm spheres on the join of leaves, a developing leaflet, or flower stalk (Figure 2 2). The presence or absence of feeding damage of C. irus larvae is highly distinctive, and was recorded along with the numbers and instar stage of any C. irus larvae observed. Microhabitat information was collected on different days than wh en plant and larval presence was sampled, and included identification of the nearest non host plant vegetation, illuminance measurements, and ground vegetation cover estimates. The
64 nearest non host plant was either identified on site or photographed for la ter identification. Illuminance (log(lumens/m 2 ) was measured at breast height (~1.7m) once per focal plant sample, taken sequentially through the ~5 hour sampling period using a photographic light meter set to display the exposure value, to be converted in to lux = lumens/m 2 (Sekonic model L 508). N on lupine plant cover estimates were conducted by placing a 1 meter square PVC frame around the focal plant and total plant covered estimated separately by two observers. The total cover estimates were averaged to gether, and then expressed as proportions of herbaceous, graminoid, and woody plant cover. 2012 Flagging of the entire L. perennis population was conducted between mid March and continued through to mid May. Each flag was marked with a unique number, with a flag placed for each ramet, or individual emergent L. perennis stem as could be easily observed without digging into the soil. Microhabitat and C. irus sampling was conducted over an 11 day period from late March to early April, after the initial flaggin g revealed the spatial extent of the L. perennis distribution (Figure 2 5 B). Individual L. perennis ramets were chosen randomly from the available plant distribution marked on an Roth ignment of samples was conducted by dropping five to ten small oak leaves approximately 30cm above the map, with sampling locations marked where the leaf petiole landed. A 2:1 ratio of C. irus occupied and unoccupied plants was set, biased towards C. irus occupied plants to ensure the variation in C. irus immature numbers was accounted for. Sampling then was conducted by walking to the marked locations and finding the nearest plant fitting the occupied or unoccupied criteria. Each plant sampled had exactly
65 the same data gathered, and is summarized in Table 2 3. Flower state was recorded as no flower, flower stem, blooming flower, or seeding. Un hatched C. irus eggs appear pale green to yellow, and are completely colored; hatched eggs are white with a dark ce nter. C. irus larvae number and instar were recorded, with length measurements taken for each that were encountered. Ground cover visual estimates were conducted in the same manner as in 2010, but centered on the focal ramet (single stem), instead of the g enet (group of stems). The type, diameter, and distance away of the nearest non focal plant was sampled, and if field identification was not possible, it was photographed. Litter and duff measurements were taken at the base of the focal plant by insertion of a dowel to the top of the next layer, measured by ruler and recorded. Finally each sampled plant was assessed for the presence of other feeding damage, identified if possible. The presence of U. reversalis is distinctive with large aposematic larvae tha t lay a web of silk over and around the plant they feed on (Figure 2 4 B). If present, U. reversalis larvae were counted and their length measured to the nearest millimeter. L. perennis and C. irus Census Full color printed and laminated maps were created from 1:800 scale Roth photos of RESMSF described above. At this scale, individual evergreen trees ( Pinus sp. and some Quercus sp.) were able to be identified and correspondingly located on the ground. L. perennis was then mapped by placing a reference poi nt on the map, belonging to either a single genet or a collection of ramets. In the case of multiple ramets (single identifiable stem) or genets (collection of stems) for a reference point, the arrangement was diagrammed and measured off. The diagrams were always drawn in reference to magnetic north, and distances were measured to the nearest centimeter. A
66 clumped arrangement was common within a given reference point (many individual L. perennis stems), with the longest distance measured across a clump of i ndividually flagged L. perennis to the nearest centimeter. Each flagged plant in each diagram was marked with its individual plant number assigned by the numbered survey flags placed to mark the ramet. Additional information was gathered on the status of C irus presence or absence, as determined by eggs, larvae, or the highly distinctive pattern of feeding damage C. irus produces. Diagrammed L. perennis were then added as a point feature layer in Arc Map (ESRI Inc.) over the ortho photo image of RESMF used to create the ortho map laminates. Precise editing tools within this program allowed the distance measurements to be transcribed exactly as diagrammed in the field, restricting any error to the original diagramming during the field survey. Data Analysis 2010 microhabitat survey. Most variables in 2010 were not sampled simultaneously, for example, ground cover and illuminance (May 5 th ) were collected on plants 3 4 weeks after L. perennis and C. irus data were collected (April 13 18 th ). As a result of this a conservative approach to analysis was employed to avoid compromising the patterns seen at the different times. Instead, two subsets of the sampling period were used in two separate regression analyses, allowing initial conclusions to be made concerning patterns of C. irus and features of the microhabitat. The first regression conducted was to test the correlation to numbers of C. irus eggs as predicted by L. perennis genet leaf and stem number, and flowering state. A generalized linear model with a log link function assuming a negative binomial distribution was used in place of a Poisson distribution as the negative binomial should be used when the standard deviation of the response is
67 larger than the mean a characteristic of the C. irus egg count data The second regression was correlation of ground cover ocular estimates and illuminance to C. irus presence or absence, as observed from C. irus eggs, larvae, or larval feeding dama ge (using logistic regression). 2012 microhabitat survey Spearman rank corr elation coefficients were generated for continuous predictor variab les using the cor function in R Performing this step allowed for exclusion of highly correlated variables during regression modeling. Generalized linear regression was performed to identif y microhabitat variables correlated to C. irus egg numbers. A negative binomial distribution was used in place of a Poisson distribution as it is used when the standard deviation of the response is larger than the mean (overdispersion). Model selection was conducted using an information theoretic approach, and involved generating and comparing evidence ratios of a priori candidate models with the AICcmodavg package in R (Anderson, 2008). Hypotheses, components (sub hypotheses), and sampled variables are lis ted in Table 2 4. The first step in model selection was to choose a single term of a component of one of the two hypotheses explaining C. irus presence. Simplification of each component or sub hypothesis will prevent adding terms that provide little extra information to the model but penalize it due to the additional terms added For example, the three terms included in the focal plant sub hypothesis, number of L. perennis leaves, diameter of the L. perennis plant, or flowering state of the L. perennis plan t were compared, and the term with the lowest evidence ratio was chosen. Evidence ratios were calculated using the corrected Akaike information criterion (AICc) values that are
68 specific to each model (Anderson, 2008) Corrected AIC (AICc) penalizes models which have a large number of terms that are gen erated from small sample sizes. After simplification, conceptual models generated from the two hypotheses in Table 4 were constructed and analyzed in the same model selection framework as described above. The best model will have an evidence ratio equal to one, but models of higher values should also be considered possible. Where the evidence ratio becomes too high to consider the model is up to personal choice, but models with evidence ratios above around 10 b ecome increasingly less likely exp lanations. Pseudo R 2 values were generated for the likely models using the pR2 function in the pscl package in R, with Cr agg and Uhler's listed (Nagelkerke, 1991). Cragg r to the adjusted R 2 from ordinary least squares regression, with a range of 0 1, and 1 equaling perfect model fit, and 0 equal to a model without any predictors. Finally, parameter averaged estimates were also generated with AICcmodavg, and are included i n output summaries for the model chosen with the lowest evidence ratio. L. perennis census To document the patch size arrangement of of L. perennis and C. irus presence, density of L. perennis and C. irus presence was generated using the Density tool in Ar cMap. Kernel density was used for both analyses, as it spreads the value of a point outwards from the point location using a quadratic formula The highest value is the point itself, and decreases to zero at the maximum search radius distance, and the valu e for a given cell is calculated as the sum of intersections with other points. What results is a smooth contour surface, which in this application, essentially serves as the probability of finding the particular organism ( L. perennis plant or C. irus pres ence).
69 To assess for spatial structure causing non independence from the best fit regression model of the 2012 microhabitat dataset utilized to assess spatial autocorrelation of the residuals from the best model. A spline co rrelogram was constructed using the ncf function in R and the spatially referenced residuals extracted from ArcMap data entry from the L. perennis census. The spline residua ls, and includes a 95% confidence envelope from 10 replicate runs Attention is paid to the similarity index at short distances (<100m), as samples taken at these distances are where non independence may occur. A plot that shows high similarity at short di stances and decreases as distance increases indicates spatial autocorrelation. If instead the correlogram plot has a low similarity and is bound by zero at short distances, this suggests that spatial autocorrelation is not occurring, sample points are inde pendent, and regression modeling is not violated. Results Focal Follow and Mark of C. irus A total of seven C. irus adults were observed between March 13 15 th and 21 22 rd over 50 combined hours of searching. Behaviors observed included flying, no movement, abdomen probing, oviposition, and hind wing rubbing (Table 2 1). No adults were observed following this period while performing microhabitat surveys or the L. perennis census, implying that the adult flight had ended. Observation periods following capture marking, and release were quite brief, the longest lasting 6.5 minutes, but all others much less. Behaviors following release consisted of no movement for 1 5 minutes, followed by flying out of view. The final C. irus found in this study was observed fir st prior to capture. This adult was followed for over 10 minutes prior to
70 capture, and was observed to fly, land on and walk around on L. perennis plants, and touched the tip of its abdomen to the leaf surfaces. Due to a lack of replication of observations unaltered by capture and marking procedures, this information will not be included in this dissertation Future work specifically dedicated to observing C. irus adults prior to capture would help to illuminate individual behavior at RESMSF. A total of f our individual C. irus adults were captured, and the results of the capture are summarized in Table 2 5. Adult C. irus were observed or captured in the immediate vicinity of L. perennis and never before 1200 hours. Winds were calm or very light, with only a wind speed of ~8km/h recorded during a single capture. As evidenced by the exposure values, C. irus were observed or captured only during bright, mostly sunny conditions. Microhabitat Surveys 2010 Mean, range, counts, and standard deviation for microhab itat sa mpling are summarized in Table 2 6 The presence of C. irus was detected on over 53% of L. perennis plants surveyed between April 13 th and 18 th as measured by the presence of eggs, larvae, or feeding damage (n=287). Up to 18 eggs or 9 C. irus larvae were recorded on L. perennis genets over this sampling period. Regression revealed a correlation between C. irus eggs and L. perennis genet leaf number (Table 2 7). During the May 7 th non focal vegetation cover estimation sampling, C. irus presence was de tected on 54% of L. perennis plants surveyed. Total ground cover/m 2 quadrat centered on sampled L. perennis genets varied from 2 90%, with an average cover of 43.8%. Logistic regression of illuminance and averaged ocular estimates of non focal vegetation c over revealed a significant negative correlation between C. irus
71 presence and total estimated non focal plant vegetation cover but not illuminance (Table 2 8). This relationship was plotted with the logist ic regression line in Figure 2 6 2012 Mean, range counts, and standard deviation for microhabitat sa mpli ng are summarized in Table 2 9. Correlation analysis of the predictor variables revealed that total percentage averaged ocular estimate of non focal vegetation cover per m 2 was highly correlated to av eraged ocular estimate of non focal graminoid cover (r = 0.75). As such, total cover was removed from any regression analysis. No other predictor variables were found to be highly correlated (i.e. with a correlation coefficient of <|0.5|). Focal L. perenni s diameter was removed from the analysis, as L. perennis have significant three dimensional structure. Leaf number is a better measure of size as it accounts for both area and volume, scaling as the plant increases in leaves. Regression analyses of the com ponents (sub hypotheses) of the main hypotheses are displayed in Table 2 10. For the focal L. perennis component, number of L. perennis leaves resulted as the best model, with L. perennis flowering state clearly being a worse model with an evidence ratio o f ~41. In the proxy canopy cover component selection, duff was the best model, but litter had an evidence ratio of 2.56, which includes it in model building. This doubles the overall number of models for the final comparison, as duff and litter must be put in all the final models separately. Presence of other feeding damage was the best model out of the non C. irus feeding damage component, with presence of U. reversalis having an evidence ratio of ~11. Finally, the nearest non focal plant diameter and dist ance away were the best models out of this model selection, with ground cover ocular estimates being progressively poorer predictors, with evidence ratios of 3 8.
72 A total of nine models were compiled from the component selection detailed above. As there we re two likely predictors for both the canopy cover proxy and ground cover components, this doubled the total number each time. A final addition to the d to the rankings as C. irus egg counts could be entirely due to whether or not other organisms fed on L. perennis and in turn, did or could potentially consume C. irus eggs. The result of the model selection from the nine candidates is displayed in Table 2 11. The best model to describe the relationship of C. irus eggs per L. perennis L. perennis earest non focal pl focal plant wa s a large drop off in evidence ratio, with the three term model L. perennis earest non 14.28, which is a much less likely model. These results indicate that Hypothesi s 2 from Table 2 4 is clearly better at explaining the patterns of C. irus eggs in the spring of 2012 at RESMSF than Hypothesis 1, and is due to the greater information gained by the 2 for th e top three models indicate the same trend as in model selection, with the best Hypothesis 1 model having t he best measures of model fit, (Table 2 12). Model averaged estimates across these nine models are listed in Table 2 13.
73 Spatial autocorrelation anal ysis of the residuals from the best regression model, L. perennis earest non <100m. The spline correlogram plot of the residuals as a function of distance show the similarlity to be bound by zero over this interval (Figure 2 7 ) L. Perennis a nd C. i rus Census A total of 6,253 L. perennis ramets were flagged and surveyed in 2012, with 207 showing at least some s ign of C. irus presence, be it either egg, larvae, or the characteristic larval f eeding damage Kernel density for L. perennis was generated using the default search radius of 17.4m, which included 99% (6184/6253) of L. perennis ramets. Seven density class es were assigned by the natural breaks (jenks) classification feature in ArcMap, with the very lowest density class set as a null display value (Figure 2 8 A). Kernel density for C. irus was generated using a search radius of 30m as used by Albanese, et al ., and included every record (207/207) of C. irus presence. Density classes were assigned in the same manner as L. perennis density, with seven classes of natural breaks and with the lowest density class set as a null display value (Figure 2 8 B). Discussi on Overall, C. irus presence at RESMSF in Nassau County, Florida varied directly with larval host plant size, size of ground cover plants, duff depth near the larval host plant, and the presence of feeding damage from other organisms. C. irus presence was often observed, but adults were rarely encountered, and instead immature stages including eggs and larvae proved to be the best measure to compare to measures of the microhabitat. In 2010, larger L. perennis plants were found to have more C. irus
74 eggs. Are as of lower estimated ground cover were also more likely to have the presence of C. irus This first year of study confirmed that C. irus on L. perennis at RESMSF in Florida have similar patterns seen in the B. tinctoria utilizing population in SE Massachu setts. In the 2012 survey in the same location, larger L. perennis plants again had more C. irus eggs. The larger the diameter of non host plants nearest to L. perennis with C. irus eggs became, the fewer C. irus eggs were present. Numbers of C. irus of eg gs increased as duff at the base of focal L. perennis increased. Finally, presence of other feeding damage had a negative relationship with the number of C. irus eggs observed on a L. perennis plant.. This survey in 2012 also confirms the patterns seen in B. tinctoria in SE Massachusetts, and adds additional information in that presence of other feeding damage is also correlated to egg numbers. The presence of competitors such as U. reversalis may be important, and was the most common component of other fe eding damage observed (Table 2 9). Deer browse by O virginianus could also be important, as inadvertent predation is known to occur in other areas such as Maryland (Frye, 2012). The effect of deer browse could be a complicating factor with the measurement of other feeding damage, specifically, C. irus egg detection on plants that had feeding damage from deer may be compromised. While not directly quantified, oviposition by C. irus at RESMSF was most common on new leaf growth and larvae were commonly obser ved feed ing on these tissues; O. virgnianus typically consume L. perennis inflorescences, so this may not be as important of concern compared to other areas of the country where larvae tend to feed on inflorescences (Barnes, 2003). Indirectly, deer browse may have long term
75 consequences for L. perennis persistence due to poor seed set and seedling recruitment (Frye, 2012). L. perennis plants with C. irus present are arranged in clearly visible clumps at RESMSF (Figure 2 8 B). L. perennis also shows an aggr egated distribution, with areas of highest density not always the same as C. irus (Figure 2 8 A). Further work investigating the overlap of the larval host L. perennis and C. irus may reveal larger scale patterns at work related to seasonal patterns in can opy cover or patterns of use of L. perennis by other organisms. The density pattern observed was only for a single year, and is likely to vary in other years, so a larger time scale data set would be required to answer this question. The present study prov ides confirmation of the relationship of C. irus presence to the microhabitat variables of large host plant size, small size of ground cover vegetation, and the presence of at least some shade for a population in northeastern Florida. Additional evidence w as provided documenting a negative relationship between feeding damage from other organisms and C. irus egg numbers. This suggests some validation of the nave adaptationist hypoth esis for host plant specialists, in that egg presence, a relic of female ovi position, was correlated to features known to be linked to offspring performance in other populations. Furthermore, a snapshot of the patterns of L. perennis and C. irus immature stage presence were captured in an isolated population in the southernmost ex tent of the wide ranging C. irus butterfly. Information such as this exists for at least one other location, but is limited to adult C. irus and includes no information about host plants (Albanese, et al., 2007). General knowledge
76 of plant and insect distr ibutions is sure to be useful input for land management activities, and provides a record of the state of the organisms in a given year.
77 Table 2 1. Observed behaviors from Callophrys irus adult focal follow March 2012 at Ralph E. Simm ons Memorial State Forest, Nassau County, Florida. Behavior Observed Location Behavior Observed Flying Plant species/description Perching/sitting/resting/remaining still Location on the plant Abdomen probing Oviposition (egg laying) Hind wing rubbin g
78 Table 2 2. C. irus microhabitat sampling dates and features recorded during 2010 at Ralph E. Simmons Memorial State Forest, Nassau County, Florida. = Number of L. perennis leaves/genet not recorded. Date Sampled Feat ure 4 6 2010* 4 13 2010 4 18 2010 Number of L. perennis stems/genet Number of L. perennis leaves/genet Total number of C. irus egss Number of C. irus larvae Instar of each C. irus larvae Presence/absence of C. irus larvae feeding damage L. perennis flowe r state: none or bloom 4 11 2010 Exposure value at breast height above L. perennis Nearest non L. perennis vegetation Diameter of nearest non L. perennis vegetation Number of stems of nearest non L. perennis Diameter at breast height of nearest non L. pe rennis vegetation 4 21 10 Total number of C. irus egss Number of C. irus larvae Instar of each C. irus larvae Presence/absence of C. irus larvae feeding damage 4 27 10 C. irus egg presence/absence C. irus larvae presence/absence Instar of C. irus larvae present Presence/absence of C. irus larvae feeding damage 5 5 2010 Number of L. perennis stems/genet Number of L. perennis leaves/genet Total number of C. irus egss Number of C. irus larvae Instar of each C. irus larvae Presence/absence of C. irus larvae feeding damage L. perennis flower state: none, stalk, or bloom Litter depth (cm) Duff depth (mm) 5 7 2010 Presence/absence of C. irus larvae feeding damage Exposure value at breast height above L. perennis Ground cover estimates in 1m 2 around focal L. p erennis 6 9 2010 Number of L. perennis stems/genet
79 Table 2 3. C. irus microhabitat features sampled over an 11 day period March April 2012 at Ralph E. Simmons Memorial State Forest, Nassau County, Florida. General Date Time Focal plant number Photograph number taken at breast height 1 m 2 quadrat centered at the focal L. perennis plant Total percentage averaged ocular estimate of non focal vegetation cover Averaged ocular estimate of non focal herbaceous vegetation cover Averaged ocular es timate of non focal graminoid vegetation cover Averaged ocular estimate of non focal woody vegetation cover Non focal plant Nearest non focal plant species (photographed if unknown) Distance to nearest non focal plant from focal L. perennis (cm) Diam eter of nearest non focal plant (cm) Focal L. perennis plant Number of leaves Diameter (cm) Flowering state Litter depth at the base (cm) Duff depth at the base (mm) Number of hatched C. irus eggs Number of unhatched C. irus eggs Presence or absence of C. irus larval feeding damage Number and instar of C. irus larvae observed Length of C. irus larvae observed (mm) Presence or absence of other feeding damage Presence or absence of U. reversalis feeding damage Number of U. reversalis larv ae observed Length of U. reversalis larvae observed (mm)
80 Table 2 4. Hypotheses for model selection of number of eggs of C. irus per L. perennis plant as a function of elements of microhabitat sampled March 27 to April 6 2012 at Ralph E. Simmons Memorial State Forest, Nassau County, Florida. Hypothesis Components (Sub hypotheses) Model terms 1. Large larval host plant size, low amounts of ground cover, and intermediate amounts of canopy best predict C. irus : Components A, B, and C A. Focal L. pe rennis plant size and flowering state indicate future larval resources Leaf number Plant diameter Inflorescence state (none, stalk, blooming, seeding) 2. Large larval host plant size, low amounts of ground cover, intermediate amounts of canopy cover, and low presence of non C. irus feeding damage are the best predictors of C. irus : Components A, B, C, and D B. Ground cover influences apparency or access of potential host plants Total percentage averaged ocular estimate of non focal vegetation co ver per m 2 Average ocular estimate of non focal herbaceous vegetation cover per m 2 Averaged ocular estimate of non focal graminoid vegetation cover per m 2 Averaged ocular estimate of non focal woody vegetation cover per m 2 Nearest non focal plant species D istance to nearest non focal plant from focal L. perennis Diameter of nearest non focal plant C. Litter and duff amounts as a proxy for canopy cover Litter depth at the base (cm) Duff depth at the base (mm) D. Presence of non C. irus larvae feeding da mage negatively influences egg laying Presence or absence of other feeding damage Presence or absence of U. reversalis feeding damage
81 Table 2 5. Summary of individual size and environmental conditions at capture of C. irus between March 13 22 nd 2012 at Ralph E. Simmons Memorial State Forest, Nassau County, Florida. ID = individual identification number; FWC = forewing cord length (mm); EVSU = exposure value in the sun; EVSH = exposure value in nearest shade; WD = wind direction; WSL = wind speed at 0.2m; WSH = wind speed at 2.0m; TEMP = temperature in Celsius (Fahrenheit). Blank entries indicate data not taken. D ATE TIME ID FWC EVSU EVSH WD WSL WSH TEMP 3/13/2012 1223 1001 28.3 (83.0) 3/14/2012 1700 1002 13.6 11.0 11.0 27.3 (81.2) 3/15/2012 1228 1003 13.2 14.8 11.7 N 0 5 31.6 (88.8) 3/15/2012 1517 1004 15.3 14.7 11.7 0 0 30.8 (87.4)
82 Table 2 6. Summary statistics of the 2010 C. irus microhabitat survey at Ralph E. Simmons Memorial State Forest, Nassau County, Florida. = in the case of binomial or categorical variables, the three most common categories (if applicable) are given. Date Variable Mean* & Range Standard Deviation April 13 & 18 Number of leaves/genet 24.4 3 117 18.1 Number of stems/gene t 2.1 1 11 1.5 Total number of C. irus eggs 1.8 0 18 2.4 Presence or absence of C. irus larval feeding damage Yes (153) No (134) Total number of C. irus larvae 0.8 0 9 1.5 L. perennis flower stalk, bloom or none None (226) Stalk (28) Bloom (33) May 5th Burned on April 22nd Burned (93) Unburned (23) Litter depth at the base (cm) Unburned Burned 2.4 0 0 7 0 1.2 0 Duff depth at the base (mm) Unburned Burned 3.2 1.8 0 10 0 5 2.6 1.2 May 7th Illuminance (lux = lumens/m 2 ) 38868 415 9 100855 3064 Total percentage averaged ocular estimate of non focal vegetation cover 43.8 2 90 26.9 Averaged ocular estimate of non focal herbaceous vegetation cover 4.4 0 22 4.7 Averaged ocular estimate of non focal graminoid vegetation cover 28. 4 0 86 23.4 Averaged ocular estimate of non focal woody vegetation cover 11.0 0 66 16.6 Presence or absence of C. irus larval feeding damage Yes ( 49 ) No ( 41 )
83 Table 2 7. Output summary of the best generalized linear regression model of C. iru s eggs on L. perennis as a function of genet leaf number and flowering state sampled April 13 th and 18 th 2010 at Ralph E. Simmons Memorial State Forest, Nassau County, Florida. Model Estimate Std. Error t value Pr (>|z|) Intercept 0.1 6 0.1 3 1.16 0. 25 Number of leaves on L. perennis genet 0.0 3 0.0 04 6.26 < 0.0 001 Table 2 8. Output summary of the best logistic regression model of C. irus feeding damage on L. perennis as a function of illuminnance and ocular estimates of vegetation cover May 7 th 2010 a t Ralph E. Simmons Memorial State Forest, Nassau County, Florida. Model Estimate Std. Error t value Pr (>|z|) Intercept 1.06 0.43 2.45 0.01 Total percentage averaged ocular estimate of non focal vegetation cover 0.02 0.01 2.39 0.02
84 Table 2 9. Summary statistics of the C. irus microhabitat survey (n= 128) conducted between March 27 and April 6 2012 at Ralph E. Simmons Memorial State Forest, Nassau County, Florida. = in the case of binomial or categorical variables, the three most common categories (if applicable) are given. # = summarized for L. perennis plants where C. irus larval feeding damage, eggs, or larvae were present. Variable Mean* Range Standard Deviation Ground cover (1m 2 ) Total percentage averaged ocular est imate of non focal vegetation cover 27.1 4 85 18.4 Averaged ocular estimate of non focal herbaceous vegetation cover 4.9 0 24 4.0 Averaged ocular estimate of non focal graminoid vegetation cover 14.0 1 65 13.3 Averaged ocular estimate of non focal woo dy vegetation cover 8.3 0 60 11.8 Non focal plant Nearest non focal plant species L. perennis (35) Grass sp. (29) A. stricta (28) Distance to nearest non focal plant from focal L. perennis (cm) 7.0 0 28 5.1 Diameter of nearest non focal plant (cm) 8.0 1 50 7.0 Focal L. perennis plant Number of leaves 14.5 1 80 9.7 Diameter (cm) 13.0 4 25 4.1 Flowering state None (99) Stalk (17) Bloom (11) Litter depth at the base (cm) 1.7 0 5 1.0 Duff depth at the base (mm) 1.8 0 10 1.6 Num ber of hatched C. irus eggs # 0.96 0 7 1.3 Number of unhatched C. irus eggs # 0.02 0 2 0.2 Total number of C. irus eggs # 0.99 0 7 1.4 Presence or absence of C. irus larval feeding damage, eggs, or larvae No = 45 Yes = 83 Total number of C. irus larvae # 1.1 0 5 1.1 Presence or absence of other feeding damage No = 101 Yes = 27 Presence or absence of U. reversalis larvae feeding damage No = 122 Yes = 6 Number of U. reversalis larvae observed 0.04 0 2 0.2
85 Table 2 10. Output summary of generalized linear regression models of sub hypotheses of C. irus eggs on L. perennis as a function of microhabitat variables sampled between March 27 and April 6 2012 at Ralph E. Simmons Memorial State Forest, Nassau County, Florida. denotes terms that were selecte d for final model selection. Model K AICc Delta AICc Wi Cumulative Wi Evidence Ratio *Number of L. perennis leaves 2 274.24 0 0.98 0.98 1 L. perennis flowering state 4 281.66 7.42 0.02 1 40.79 *Duff 2 278.74 0 0.72 0.72 1 *Litter 2 280.62 1.88 0.28 1 2 .56 *Presence of other feeding damage 2 276.99 0 0.92 0.92 1 Presence of U. reversalis 2 281.82 4.83 0.08 1 11.18 *Nearest non focal plant diameter 2 279.02 0 0.37 0.37 1 *Nearest non focal plant distance away 2 279.55 0.54 0.28 0.65 1.31 Ocular estim ate of woody vegetation cover per m 2 2 281.32 2.30 0.12 0.76 3.16 Ocular estimate of herbaceous vegetation cover per m 2 2 281.53 2.52 0.10 0.87 3.52 Ocular estimate of graminoid vegetation cover per m 2 2 281.77 2.75 0.09 0.96 3.96 Nearest non focal plan t type 5 283.33 4.32 0.04 1 8.66
86 Table 2 11. Output summary of generalized linear regression final conceptual models of C. irus eggs on L. perennis as a function of microhabitat variables sampled between March 27 and April 6 2012 at Ralph E. Sim mons Memorial State Forest, Nassau County, Florida. (lfnum = Number of L. perennis leaves; dufmm = Duff depth; otrfeeddam = Presence of other feeding damage; nvdiacm = Nearest non focal plant diameter; nvdiscm = Nearest non focal plant distance away; litmm = Litter depth) denotes Hypothesis 2: host plant size, canopy cover, presence of other feeding damage, and ground cover. # denotes Hypothesis 1: host plant size, canopy cover, and ground cover Model K AICc Delta AICc Wi Cumulative Wi Evidence Ratio l e a fnum+duf f + ot herfeeddam+ dia mneare stnonhost 5 266.27 0 0.73 0.73 1 l eaf fnum+duf f + ot herfeeddam+ dis tancen earestnonhost 5 269.69 3.42 0.13 0.86 5.53 # l eaf fnum+duf f+ dia mnear estnonhost 4 271.59 5.32 0.05 0.91 14.28 lfnum+litter +ot herfeedda m+diamnearestnonho st 5 272.21 5.94 0.04 0.95 19.47 l eafnum+litter +ot herfeedd am+ dis tancenearestnonh ost 5 273.17 6.90 0.02 0.97 31.52 l ea fnum+duf f+ dis tanceton earestnonhost 4 274.17 7.90 0.01 0.98 52.00 leafnum+litter+ dia metern earestnonhost 4 275.23 8.96 0.01 0.99 88.32 l e a fnum+lit ter +dis tanceto nearestnonhost 4 276.14 9.87 0.01 1.00 139.19 ot he rfeeddam 2 276.99 10.72 0.00 1 212.87
87 Table 2 12. Model fit (pseudo R 2 ) of the most likely generalized regression models of C. irus eggs on L. perennis as a function of microhabitat variables sampled between March 27 and April 6 2012 at Ralph E. Simmons Memorial State Forest, Nassau County, Florida. (lfnum = Number of L. perennis leaves; dufmm = Duff depth; otrfeeddam = Presence of other feeding damage; nvdiacm = Nearest non focal plant diameter; nvdiscm = Nearest non focal plant distance away) Model Nagelkerke lfnum+dufmm+ otrfeeddam+nvdiacm 0.08 0.17 lfnum+dufmm+ otrfeeddam+nvdiscm 0.07 0.14 lfnum+dufmm+nvdiacm 0.05 0.10 Table 2 13. Model averaged estimates from AICc model selection of C. irus eggs on L. perennis as a function of microhabitat variables sampled between March 27 and April 6 2012 at Ralph E. Simmons Memorial State Forest, Nassau County, Florida. (lfnum = Number of L. per ennis leaves; dufmm = Duff depth; otrfeeddam = Presence of other feeding damage; nvdiacm = Nearest non focal plant diameter; nvdiscm = Nearest non focal plant distance away; litmm = Litter depth). Model Estimate Std. Error Lower CI Upper CI lfnum 0.04 0.0 1 0.02 0.07 dufmm 0.26 0.10 0.07 0.44 nvdiacm 0.06 0.03 0.12 0.00 otrfeeddamyes 1.37 0.53 2.41 0.33 otrfeeddamno 1.19 0.46 2.08 0.29 nvdiscm 0.03 0.03 0.03 0.09 litcm 0.21 0.17 0.13 0.55
88 Figure 2 1. Location of Ralph E. Simm ons Memorial State Forest and the Lupinus perennis and Callophrys irus in Nassau County, Florida. Ortho photo courtesy
89 Figure 2 2. Eggs and larvae of C. irus on L. perennis from 2010 2012 at Ralph E. Sim mons Memorial State Forest, Nassau County, Florida. A) Newly laid egg. B) Mature egg. C) Early instar larva. D) Late instar larva. E) C. irus feeding damage on L. perennis Photos courtesy of B. Camposano (A and B) and M. Thom (C, D, and E).
90 Figur e 2 3. L. perennis May 2011 at Ralph E. Simmons Memorial State Forest, Nassau County, Florida. Photo courtesy of M. Thom.
91 Figure 2 4. Most commonly encountered organisms in close proximity to L. perennis during 2010 2012 at Ralph E. Simmons Memorial State Forest, Nassau County, Florida. A) U. reversalis egg mass on L. perennis. B ) U. reversalis larvae on L. perennis C) Newly sprouted A. stricta following a February 2011 prescribed burn. D) Shrub like Q. laevis (arrows) among A. stricta Phot os courtesy of M. Thom.
92 Figure 2 5 Survey areas at Ralph E. Simmons Memorial State Forest, Nassau County, Florida. A) 2010. B) 2012. Ortho photos courtesy of Management District
93 Figure 2 6 Plot of logistic regression of presence of C. irus feeding damage on L. perennis as a function of total vegetation cover per m 2 May 7 th 2010 at Ralph E. Simmons Memorial State Forest, Nassau County, Florida.
94 Figure 2 7. Spline correlog ram of the 2012 best regression model ( L. perennis earest non similarity is bound by zero, suggesting no pattern of spat ial autocorrelation for the best fit model.
95 Figure 2 8 1m resolution 1:3549 color digital ortho image and density overlays from a census conducted March May, 2012 at Ralph E. Simmons Memorial State Forest, Nassau County, Florida. Density classes wer e generated using the natural breaks (jenks) option in ArcMap 10.0 (ESRI, 1999 2010)A) o verlaid is kernel density of L. perennis : Low = 160 501 L. perennis ramets/hectare, High = 3,505 5,804 L. perennis ramets/hectare; B) overlaid is kernel density of C. irus presence: Low = 6.0 19.6 C. irus occupied L. perennis ramets/hectare, High = 98.7 139.0 C. irus occupied L. perennis ramets/hectare. Ortho District.
96 CHAPTER 3 CONSEQUENCES FOR MO RTALITY BY FIRE: THE EFFECT OF PUPATION LOCATION OF THE FROS TED ELFIN, CALLOPHRYS IRUS GODART (LEPIDOPTERA: LYCAEN IDAE ) Introduction Fire is a ubiquitous yet stochastic force of disturbance in virtually all terrestrial ecosystems. The disturbance caused by fire is highly influ ential in shaping, promoting, and sustaining certain successional stages, and is a major contributor to the dynamic nature of living systems. The effects of fire are multifaceted in time, scale, and degree of severity. Fire may be destructive, by causing d irect mortality of plants and animals in its path, and by conversion of available nutrients into forms that are unavailable, such as nitrogen released into the atmosphere or runoff and leaching of nutrients post fire. Indirectly, the loss of food plants or other resources for animals at higher trophic levels post fire can also have a negative effect. However, fire can have many positive effects on an ecosystem, including releasing nutrients that were previously locked up in inaccessible tissues such as dead wood, litter, and duff, to live vegetation and animal matter. The structural changes as a result of fire are also very important in places in areas such as a forest floor or grassland; areas where flammable materials have been consumed are made available for species dependent on bare mineral soil for seed germination, and for ruderal species that quickly occupy released growing space. Because of the myriad effects fire on the structure, characteristics, and processes of the ecosystem in which it occurs, f ire to some degree has impacted the life history of countless organisms. There are many examples of plants with traits that confer a higher probability of survival in fire prone areas than plants without those traits, but in animals such traits are not as common or as immediately apparent. It is critical to
97 understand for both the economically important and the rare, imperiled, or endangered species, as human management of areas containing these species commonly involves the manipulation of fire (i.e. eithe r its suppression, or its prescription). One such example of a rare animal in a fire prone and fire managed ecosystem is the frosted elfin butterfly, Callophrys irus Godart (1824) (Lepidoptera: Lycaenidae). C. irus inhabits the oak pine barrens and savann ah habitats in Eastern North America, and is closely associated with its two main larval host plants Lupinus perennis L. and Baptisia tinctoria (L.) R.Br. Throughout its range, C. irus occurs in relatively small, localized, or fragmented colonies or metapo pulations. This butterfly has been noted to pupate in the leaf litter or soil, and while not common for a butterfly, this behavior is common throughout the Lepidoptera and in other insect groups. What makes the case of this butterfly special is that it is quite rare in both time and space, being univoltine and often in single isolated populations. In north Florida, the phenology of C. irus is strongly timed to coordinate with the phenology of new host resources provided by its sole host plant in this state, L perennis var. gracilis L., a somewhat ephemeral plant that is not available later in the year. Emergence of adult C. irus from overwintering pupae early in the spring coincides with new growth of L. perennis Shortly after emergence, C. irus adults mat e and females proceed to oviposit on L. perennis Larvae hatch after a short period of time, and feed exclusively on the lupine, completing their development in 4 6 weeks. At this point they enter a wandering stage, where they move off the host plant and i nto the leaf litter or soil to pupate, where they remain until the following spring, about 9 months later. North Florida is the southernmost extent of the range of C. irus, and fire is frequent (historically burning every 1 5 years) .T he butterflies require a plant
98 that is widely regarded as being fire adapted in the short and long term; lupine vigorously resprouts following fire with higher biomass and larger ground cover (Grigore and Tramer, 1996; Meyer, 2006; Pavlovic and Grundel, 2009). Due to these obse rvations, Florida populations of C. irus are extremely likely to be influenced by fire, and there are many possible strategies to explain the persistence or occurrence of this insect with fire. Evolutionarily speaking, one of the simplest strategies to cop ing with mortality from fire is avoidance: movement away in direct response to a fire, then returning to the area after the fire passes, or after the negative effects are over (no open flames, smoldering, excessive smoke, etc.) (Whelan, 1995). Return to th e area post fire could even be extended farther into the future, where movement back into the burned area occurs after food plants have resprouted or germinated, or even farther ahead in time to where the successional stage is back to the acceptable range of the organism. However, primarily sessile stages such as a larva or pupa do not have this ability to freely move about. In addition, the butterfly might not even survive in the area it flees to or survive long enough to return to the area that has recove red post fire The result is that the butterfly, if it is able to move (i.e. not an immature), would likely not be able to recolonize a burned area in its life time, and instead would rely upon fire free years and the constant dispersal of nearby populatio ns into previously burned and uninhabited areas. Applying this to C. irus would imply that the southernmost populations are an artifact of fire suppression: fire suppression by humans allowed C. irus to expand into areas where fire previously kept them out Furthermore, reintroduction of fire into these areas would cause their populations to decline or go extinct. The southernmost habitat
99 areas, where fire is now common, would act as an ecological sink: C. irus would continue to expand into this habitat, bu t would never perpetuate populations due to the ir negative response to fire. A second possible response from fire is colonization of a burned area post fire similar to that above The population would become locally extinct due to the immediate effects of fire, but individuals from unburned yet nearby areas would disperse and recolonize the burned area after the negative effects are over, similar to the strategy above. This would require that neighboring areas containing the butterflies remain unburned to act as a source for re colonization Such an area may even be as small as an unburned and mostly unaffected patch that would act as a refuge. As in the first strategy, these butterflies would need to survive long enough to recolonize, and new resources mus t be available in the burned area such as nectar or larval host plants. In addition, the distance from the occupied and unburned areas to the burned areas must be within the dispersal range of the recolonizing butterflies. This strategy implies that the po pulations be a network forming a metapopulation, where dispersal and connectivity must be present between each sub population. Butterfly phenology and dispersal behavior would also need to align with certain aspects of the fire regime such as seasonality, intensity, and spatial extent. The strategy of recolonization would not work if dispersal from outside populations into a previously extinct area coincided with the time of year when fire is most common. Evidence for this strategy for C. irus in Florida is limited but certainly possible for certain areas where it is found, such as in the Apalachicola National Forest Here a number of small populations are known and are in relative proximity to each other. Other locations such as at Ralph Simmons Memorial
100 St ate Forest are likely to be isolated, and so recolonization post fire is not a likely strategy. As a result, fire management in an area such as this will influence the future of this butterfly in Florida. An understanding of the interaction or impacts of f ire on C. irus will be important for developing appropriate management strategies for ensuring the persistence of the population locally or throughout the state. This understanding also has the potential to impact management of C. irus in other regions. En during or escaping the immediate effects of fire is a third possible strategy. The butterfly might be able to tolerate temperatures above the typical lethal range for other organisms due to physiological or behavioral adaptation. Burrowing into the ground or fleeing to refugia is common in many vertebrates such as small rodents, turtles, and amphibians (Babbitt and Babbitt, 1951; Long, 1974; Lin zey and Packard, 1977; Merritt, 1981; Smollen, 1 981; Folk and Bales,1982; Owen, 1984; Lackey et al., 1985 ; Ford, et al., 1999; Russell, et al., 1999 ). In the Lepidoptera, the order containing butterflies and moths, many different moth groups commonly pupate in the soil, such as saturniids, noctuids and sphingids (Fye and Carranza, 1973; Fye, 1978 ; Triplehorn and Johns on, 2005; Zheng, et al., 2011). Pupation in the soil is less well known for butterflies, but escaping into the soil would surely provide some protection from a fire. Fire temperatures are typically highest at the top of the fuel being consumed, and so resi ding below ground where ground (duff) fires are rare would likely result in lower temperatures (Whelan, 1995; Neary, et al., 2008; Busse, et al., 2010). To burrow, the substrate to be penetrated would need to be accessible, and within the ability of the or ganism to move into. From all this, it is likely that there is a trade off for enduring or escaping lethal temperatures. The upper thermal tolerance limit for animals
101 is around 50 C, which should be noted as the upper limit of the core temperature; lower temperatures can also be lethal depending on the duration of the heat pulse (Whelan, 1995; Busse, et al., 2010). There are many examples of temperatures encountered by insects that exceed this temperature, but typically involves only appendages and for short time periods (Christian and Morton, 1992; Chown and Nicolson, 2004). High temperatures are physically demanding to organisms, as proteins denature and chemical reactions slow du e to poorly functioning enzymes. There are biological pathways to resisting heat shock, but they can be costly, and would take away from energy that is spent on growth. Similarly, escape behaviors such as burrowing also must have a cost, and there may not be enough energy to devote to such behaviors after the growth threshold is reached (minimum growth needed to advance to the next life stage) leading to individual level differences in escape behavior as a result of differences in nutrition and other facto rs during development. The enduring or burrowing strategy implies that a physiological or behavioral trait would increase an pulse generated by a surface fire. Such a trait may initially have been selected for by a di fferent force such as avoidance of predation, parasitism, or desiccation, and was co opted for survival where wildfires were historically prevalent. In the case of a butterfly pupa in Florida, adaptation to deal with summer temperatures, which can exceed t he lethal limits, could also be advantageous to dealing with fire. Pupation in the soil or leaf litter to avoid predator/parasitoids could also provide protection from the heat pulse of a fire. An investigation of this third strategy is the focus of this study, and serves to set up a comparison of the other strategies: (1) Pupae r esiding in the soil at depth will result
102 in a higher chance of survival from the direct effects of a fire compared with being in the leaf litter (2) h eat from a fire causes mort ality, and as it increases, so does the probability of mortality (3) h eat and duration of heat pulse are negatively corr elated to the depth in the soil, (4) t emperatures likely to cause pupal death are not reached at the depths in soil where pupae are fou nd in frequent fir e characterized pine uplands, OR (5) t he threshold temperature defining probable pupal mortality is not reached at the typical soil depths at which pupae are found in frequently burne d pine uplands. To test these hypotheses, several stud ies were conducted including: (1) p upal depth measurements of the frosted elfin, Callophrys irus in situ at Ralph E. Simmons Memorial State Forest, Nassau County, Florida (2) h eat tolerance of butterfly pupae via laboratory experiments (3) p upal surviva l at different litter/soil depths following prescribed burning at the Ordway Swisher Biological Station (OSBS), Putnam County, Florida and (4) and the r elationship between depth in soil to heat pulse during a prescribed burn at the OSBS. Methods Study Sit es Ralph E. Simmons Memorial State Forest in Nassau County, Florida, contains an extant colony of frosted elfins, and was the study site for field observations of frosted elfin pupae location. RESMSF is a 1,472 hectare forest containing a diverse array of natural communites including sandhills, seepage slopes, low pinelands, and riparian The C. irus colony is restricted to an approximately 20 hectare section (lat 30.797 N, 81.949 W) in one of the sandhills, where the larval host plant sundial lupine, Lupinus perennis ssp. g racilis (Nutt.) Dunn, is also located. Other non woody plants common to the colony area
103 include wiregrass, Aristida stricta Michx., gopherweed, Baptisia lanceolata (Walte r) Elliot, wooly pawpaw, Asimina incana (W. Bartram) Exell, and pinewoods milkweed, Asclepias humistrata Walter. The woody plant community is characterized by longleaf pine Pinus palustris Mill., slash pine, Pinus ellioti Engelm., American persimmon, Diosp yros virginiana L., turkey oak, Quercus laevis Walter, and various other Quercus sp. Soils consisted of moderately well drained fine sand. RESMSF is managed for multiple uses including timber, hunting, and other public uses, with an emphasis on ecosystem m anagement and ecological restoration of native communities. The prominent management technique used is prescribed fire, with dormant and growing season fires conducted in subdivided sandhill units approximately every other year since at least 2008 using a variety of techniques including aerial incendiary ignition and hand ignition by ground crews. The Ordway Swisher Biological Station in Putnam County, Florida, served as the location for three experiments conducted on pupal survival following prescribed f ire during July of 2012. OSBS is a 3,755 hectare year round field station (lat 29.683 N, long 82 W) comprised of a mosaic of upland and wetland habitats, including sandhills, upland mixed forest, xeric hammocks, marshes, swamps, and lakes. All three experiments were conducted in frequently burned (fire return intervals ranging from 1 3 yea rs) sandhill pine and oak upland units, characterized by relatively open (fewer than 150 trees/ha) forests with grass and herbaceous dominants in the understory. Fires in this habitat are typically fast moving and carried by wiregrass, A stricta but vary ing amounts of leaf litter from Q laevis and pine needle litter from several pine species including P. palustris also contribute to the fuel load. The plant community composition
104 is similar to the community described for RESMSF. Soils are also similar t o the soils at the RESMSF frosted elfin colony, ranging from moderate to excessively drained sands. Historical land use in these units included both prescribed fire and wildfire, along with extensive grazing, during the 1900s. When the land was transferre d to the University of Florida in 1980, active prescribed fire management was initiated. The OSBS management plan for this area has emphasized growing season prescribed fires for more than a decade. Study O rganisms Organisms used in this study include the previously mentioned frosted elfin butterfly, C. irus and a captive reared colony of the atala hairstreak, Eumaeus atala Poey (1832). Due to the unavailability and considerable logistical challenges to using the rare, univoltine frosted elfin butterfly f or this experiment, E. atala was instead used as an experimental surrogate. Like C. irus E. atala is a hairstreak butterfly in the subfamily Theclininae of the Lepidoptera family Lycaenidae, the same group as other ground pupating sandhill pine and oak fo rest inhabiting butterflies such as the red banded hairstreak, Calycopis cecrops Fabricius (1793), and the white M hairstreak, Parrhasius m album Boisduval & LeConte (1833). In addition to phylogenetic similarity, E. atala is also similar in ecology and in size to C. irus E. atala is native to the pine rocklands of southern Florida and the Caribbean, a known fire and disturbance prone habitat ( Snyder et al. 1990 ; Possley, et al., 2009). Like C. irus E. atala is a larval host plant specialist, relying upon a single genus of cycad Zamia and often only one species Z. integrifol i a L., to complete larval development. Pupae of E. atala are similar in size to pupae of C irus and are assumed to have similar physiological reactions to experimental manipulations involving heat or prescribed fire. Finally, a major advantage
105 to using the E. atala as an experimental subject is that it is multivoltine, allowing a lab colony to always have large numbers of any stage of development, including different ages of pupae. E atala stock were obtained from wild populations in Miami Dade County in December, 2011, and from Broward, Palm Beach, and a different location in Miami Dade County in May of 2012. Each set of E. atala consisted of approximately 200 final instar larvae an d pupae gathered from occupied Z. integrifol i a plants at the above locations. General colony rearing of E. atala occurred in an indoor lab at the Department of Entomology and Nematology at the University of Florida, Gainesville. This indoor area was maint ained between 24 27 C and varied from 20 50% RH. Pupae of E. atala were placed in open topped clear plastic containers (20cm X 30cm X 5cm) lined with unbleached paper towels that were set inside Live Monarch cloth mesh cages (35cm X 35cm X 61 cm). Pupae we re lightly misted with distilled water 1 to 2 times daily until adult emergence. Newly emerged adults were carefully transferred to the sides of the cloth cages to encourage full extension of wings. At least 4 hours after emergence, adult E. atala were giv en individual marks on the dorsal hind wing surface using a silver marker (Sharpie Metallic Silver, felt tip), and emergence time, date, and sex recorded. Marked individuals were then placed into a larger Live Monarch cloth mesh cage (175cm cube) containin g distilled water and nectaring sources. Nectaring sources included cuttings of Bidens alba L. and a variety of potted plants including sweet almond bush, Aloysia virgata (Ruiz & Pav.) Pers. and scorpion tail Heliotropium angiospermum Murray In addition to nectar plants, small 15mL plastic centrifuge tube feeders were adapted from previous designs and placed in the adult cages (Hughes, et al., 1993;
106 Clayton, 2004). At least two feeders per cage were modified to hang from the cage ceiling, while another tw o were placed on the bottom of the cage. One feeder of each pair was with filled with distilled water, and the other filled with Gatorade, a convenient and effective substitute to a lab prepared sugar water mixture (J. Daniels, pers. comm.). 100W incandesc ent bulb lamps were placed directly above each cage and set to turn on for 30 minutes every hour during a 12 hour period A fan was also set up in the rearing area, placed on the same on/off cycle as the light, and set approximately 2 m away from the cages Si ngle fronds of new growth Z. integrifolia fitted with a cotton ball were hung from the ceiling of the cage to allow female E. atala to lay eggs. These oviposition fronds were removed daily, labeled with the date and number of eggs present, and placed in a separate cage to be observed for hatching. Z. integrifolia fronds were collected daily from greenhouse grown plants and nearby landscape plantings on the University of Florida campus. Fronds containing hatchling E. atala were placed into the same kind of clear plastic containers used for pupae, and set inside Live Monarch cages. The cages were checked twice daily and provisioned with fresh fronds of water pic mounted Z. integrifolia as needed to feed the growing E. atala larvae, with dry or leftover fronds removed. Larvae were also monitored for pupation, and the date of the first larvae pupating recorded. Pupal Depth M easurements of C. irus The first step to understanding the effects of heat transferred during a fire to frosted elfin butterfly pupae was to locate and document the position of these pupae in a natural setting. The C. irus colony at RESMSF was the location chosen for pupal depth excavation, a where frosted elfins are known to persist with frequent f ire. Two methods
107 were used to target likely areas where frosted elfin pupae might be found. The first involved simply searching the immediate area (1 m 2 ) around a plant that was host to late instar larvae during previous field observations, similar to foll owing the procession of larvae in pine processionary caterpillar, Thaumetopoea pityocampa Dinis & Schiffermuller (Battisti, et al., 2000). The other method involved placing small enclosures around plants containing late instar larvae, to restrict their lat eral movement during the larval wandering period, and aid in the actual location. Two types of enclosures were used, Live Monarch cloth mesh cages with the bottoms cut away and staked down around a focal plant, and 30 cm sections of 20 cm metal ducting ins erted into the ground around a focal plant, with the top opening covered with tulle. Temperature sensors (EasyLog USB) were also placed in and outside of the metal duct enclosures, as temperatures might be elevated inside a metal ducting. Temperature senso rs were also used in shaded and unshaded areas, as there were likely to be differences in the effect of the duct on temperature in areas of differing sun exposure. Pupae were searched for in the leaf litter and soil on August 31 st 2010, May 18 th to June 6 th 2011, and May 21st to June 5 th 2012. Enclosures were used in 2011 and 2012, and the search for pupae began approximately 2 weeks after late instar larvae were last seen, or May 5 th 2011, and May 7 th 2012, to allow ample time for the wandering larvae to pupate. Previous work investigating pupation depth of insects has focused on either laboratory or semi field conditions, involving careful preparation of different types of soil in selected sites (Fye and Carranza, 1973; Fye, 1978; Dimou, et al., 2003; Montoya, et al., 2008). Laboratory techniques involved placing late instar larvae at the soil surface of various apparatuses such as boxes or PVC pipes filled with
108 soil, then 1cm sections searched or removed and run through water (Fye and Carranza, 1973; Dimou, et al., 2003). Full field investigation has previously involved digging out sections of 3cm deep sections that were then searched for pupae (Battisti, et al., 2000). For the present study, excavation of pupae under natural conditions with the highes t degree of precision was the goal, to achieve an accurate understanding about the actual pupation habits of C. irus As a result of this goal, two techniques were utilized: digging out 1cm sections of soil around suspected pupation sites measured from the top of the soil surface as in Battisti, et al., and the use of a leveled datum, a technique used in geological and archaeological excavation (Hester, et al., 1997; Garrison, 2003). A surface datum essentially serves as an invisible ceiling to which consis tent measurements can be made in reference. One can ensure precise measurements by measuring and recording the distance from the datum level to specific points such as the litter, duff, soil surface, and depth of objects (or pupae in this case) that are en countered, always referring to the level datum that was established prior to excavation. The datum setup used in this part of the study included setting a 24cm by 24cm square leveled datum centered on the focal plant that was caged with the metal ducting a nd tulle described earlier. The datum was created using straight wooden dowels upon which lengths of string were attached. At 6 cm intervals two lengths of string were also leveled across, forming one half of a grid, with each string marked at 6 cm interva ls to form the fina l dimension, and making it a 16 point grid (Figure 3 1) Prior to searching and excavation of the litter and soil, the immediate area of the excavation site was described and photographed, including measuring litter and duff depths. Litt er depth was measured by taking a dowel with a flat and level tip and placing
109 it just in contact to the duff layer, marking off the depth on the dowel with a ruler or thumb. This spot was then measured to the nearest millimeter, and the whole process repea ted for at least 5 other spots in a 40 cm radius. In the excavations involving the use of a datum, a total of 16 spots were measured equally across a grid (Figure 1b). Duff was measured in a similar way, by inserting a dowel into the duff until it reached the bare mineral soil surface. The litter was then carefully searched piece by piece for pupae, and set aside, as C. irus is known to pupate in either the litter or soil (Cook, 1906; Schweitzer, et al., 2011). If no pupae were found, then the duff was sear ched in a manner identical to that of the litter, and also set aside. Finally, if no pupae were found, the soil was carefully excavated and searched to a depth of 10cm or until a pupa was found. If a pupa was found, the depth was measured from the top of t he soil surface to the pupa (n= 4), or using the nearest grid point to the level datum (n=8). Pupae were replaced in the approximately same position, location, and depth they were found, and carefully covered with the corresponding amount of soil and leaf litter present before their excavation, to minimize any further disturbance. Heat Tolerance of Butterfly P upae In order to link heat exposure from fire to mortality in butterfly pupae buried in an area that was burned, a controlled lab experiment was condu cted to test the tolerance of butterfly pupae to a range of combinations of increased temperature and duration. The range of possible temperatures and durations were determined from a search of published heat pulses in similar soil types such as sand or sa ndy loam (Whelan 1995, Busse, et al. 2010). General consensus of lethal temperature for animals is at about 50 0 C so the range chosen included this plus an additional amount above to give a firm end point to the data. Lower temperatures were also of intere st, as high durations of
110 lower temperatures might also induce mortality. What resulted was a temperature range from 30 0 C to 65 0 C, and a duration range of 1 min to 55 min. Longer durations are certainly possible in some long smoldering ground fires, but we chose 55 min as the maximum duration beacuse the historically frequent fires in sandhill pine and oak forest were surface fires with relatively short residence times. E. atala pupae were assigned a random temperature and duration combination from the liste d ranges. Pupae themselves varied in age since pupation, a possible factor for successful survival to eclosion after exposure to a heat pulse. Pupae ages ranged from 5 to 15 days since pupation occurred, which was set as the point when the pupal molt was a pparent and the larval head capsule was shed. Since all possible combinations of pupa age, duration, and temperature would result in an unmanageable experiment, pupae were lumped into general categories of early (5 7 days, n=76), mid (8 11 days, n=63) to l ate (12 15 days, n=37) and each range was assigned a similar random assortment of temperature and duration combinations. This allowed for analysis of pupa age as a factor in survival to eclosion when exposed to a temperature and heat combination, as pupa m aturity may lead to different survival probabilities. Experimental setup included placing individual pupae into glass test tubes that were loosely capped with a cotton ball, and each tube was labeled with the assigned temperature and duration combination. The lead of a T type thermocouple was placed in contact with the pupa cuticle, and a data logger recorded the temperature reached over the course of each temperature and duration combination. Pupae were then set into a rack and placed into the appropriate temperature water bath for the assigned duration, carefully avoiding letting any water get into the tube. At the end of the
111 assigned duration, the pupae were promptly removed from the test tube and placed into a clear plastic cup with a lid, labeled with the age, temperature, and duration combination and set aside to be observed for eclosion. A small hole was opened in the lid, allowing the pupae to be misted with distilled water twice daily prior to eclosion. Also placed inside the plastic cup was a woode n stirring stick for the emerging butterfly to climb, aiding their movement to the underside of the cup lid. Viable adults were scored as those that were fully formed after eclosion from the pupa molt, and were capable of vertical flight as measured by rel ease at a 1.5m height. Those that flew weakly to the ground or not at all were considered unviable, and were euthanized by freezing. All adults were recorded (emergence date, sex), with viable adults given a unique number and letter mark with a silver felt tip paint pen ( Sharpie Metallic Silver, felt tip ) on the dorsal surface of both hind wings, and placed in adult rearing cages provisioned and treated identically but separately from the main adult colony. The rearing cages were checked daily for dead indi viduals, who were removed and their day of death recorded. Pupae were allowed 30 days to emerge, and after that were considered dead. A follow up to this experiment was conducted to confirm that this experiment and its outcomes were repeatable. Taking the results from the first experiment, the temperature at which mortality first occurred was identified. The duration just short of causing mortality was selected as a treatment, as was the duration where morality occurred. This selection process was repeated for the highest temperature of any duration where some survival occurred. Two durations were chosen in the same manner as previously, just short of causing mortality and the duration where mortality first occurred. Finally two temperatures equally distrib uted between the high and low
112 temperatures previously selected was chosen, again with a pair of durations on either side of the survival threshold. The result of the selection process was a total of eight unique temperature and duration treatments, to whic h a set of 6 differently aged E. atala pupae were each assigned (Table 3 1 ). A set of six pupae assigned as untreated controls were also handled and processed in the same manner as the heat treated pupae, only in a room temperature water bath (27 C). The e xperiment and follow up handling of pupae and adults was conducted in exactly the same ma nner as in the first experiment Pupal Survival Following Prescribed B urning In July of 2012, experiments were conducted at the Ordway Swisher Biological Station to te st the ability of E. atala pupae to become viable adults when exposed to prescribed burns Two experiments on consecutive days (July 5 th and 6 th ) involved placement of live E. atala pupae in a management unit that was subsequently burned. A final experimen t conducted on July 26 th was conducted in the same manner, minus placement of E. atala pupae, in order to increase the amount of temperature and treatment condition data that could be gathered. Within the management unit to be burned, a characteristic area in the interior was subjectively chosen to be the site where E. atala pupae were buried. Individual treatment sites were set up in such a way to account for small scale (1 3 m radius) heterogeneity in both high and low surface (litter and wiregrass) fuel loads, with sampling occurring within the range of fuel load variation present. Ocular estimates were used to quickly identify areas where litter and wiregrass cover was either patchy or continuous, as evidenced by the degree of exposed bare mineral soil, discontinuous wiregrass clumps, and sparse pine needle and oak leaf litter.
113 After a specific area was chosen, the depth of burial was randomly assigned, ranging from the soil surface (0 cm) up to 5 cm depth, at 1 cm intervals. Litter depth was measured pri or to placement of E. atala pupae, as was general information on the litter and fuel types in the immediate area (~1 m radius) surrounding the treatment plot. If assigned the soil surface treatment, litter was temporarily displaced to arrange pupae at the soil surface, and then replaced in as close to the same arrangement as prior to removal. The temperature for surface treatments was recorded using a non contact infrared thermometer (Raytek ST 20 Pro) having an operating range of 23 C to 510 C +/ 1% or 1 C. The thermometer was aimed at a fixed point near each surface treatment in sequence every 30 seconds. Temperatures were recorded 15 minutes before the fire front entered the experiment area, during the passing of the fire front, a nd for approximately 15 minutes after passing of the flame front. The researcher taking the measurements was standing in the bed of a fire engine to ensure a clear line of sight to the measurement points. Ignition in each burn was conducted by ground crews using a drip torch, with the final burn also involving the use of two horseback mounted crew members carrying drip torches. Spotted strip head firing techniques were used to establish head fires (burning in the direction of the wind) so that fire would sp read at a steady rate by the time the flaming front reached the research plots. Research plots were marked using surveying flags, to be visible by ground crews, and all research materials were removed post fire. Fire weather conditions were for the most part similar between experiments, though some differences were present (Table 3 2). The July 5 th fire was a wind driven head fire contrasting to the July 6 th fire that was a slower moving backing fire. Fire on
114 July 26 th exhibited elements of both head and backing fires. Both types of fire and the shift between each during a single burn are typical of prescribed burn behavior in this habitat, and so are representative of the conditions that would be experienced outside this expe riment. For pupae that were b uried, the litter was treated in a manner similar to the surface treatments. T hen a small round hole was dug into the soil as close to the assigned depth as possible either by hand or using a garden bulb planter Two pupae of the same age since pupation w ere placed at each cardinal direction, calibrated using a compass. In the center of this arrangement an iButton Thermocron Temperature Logger (DS1922L F5) was placed, which was programmed to begin logging approximately one hour prior to the passing of the fire front (Figure 3 2). The depth from the top of the iButton to the surface of the soil was measured and recorded, with the thickness of the iButton later added to the depth to get the actual depth. Controls were set up in the exact same manner in a dire ctly adjacent area that was not burned. A total of 8 pupae of 4 different ages were included in each depth treatment, with 6 treatments on July 5 th 9 treatments on July 6 th (Table 3 3 ). As mentioned before, no pupae were included in the July 26 th experim ent in order to get as much heat pulse data as possible in the limited time frame available for experimental setup prior to the pres cribed burn conducted that day. Fo r additional clarity, Table 3 4 displays treatment depths and the corresponding with pupae ages for the July 5 th and 6 th field experiements. Approximately 20 min after the fire front passed, surface temperatures returned to similar values prior to the fire passing through the treatment area. Each experimental unit was then carefully excavated, with pupae individually removed and placed into clear
115 plastic containers labeled with the pupal age, date, and treatment number. Pupae were transported back to the rearing lab, and monitored for eclosion in the following weeks. The designation of successfu l eclosion was performed in the same manner as in the controlled laboratory heat tolerance experiments (no malformations, positive vertical flight test). Viable adults were processed and set up in the exact same manner as described in the main lab colony a nd the laboratory experiment, but kept separate in a different cage. The iButton temperature sensors were also taken back to the lab for the uploading of temperature and time data. Data A nalysis Quantifying the temperature/time curves as a number comparabl e across the lab and field experiments was the first step in determining correlation to pupal survival. Using the Expedata software (Sable Systems, Inc.), the temperature versus time curves were imported and the integral was determined. For the temperature data from the lab study, the duration for the particular temperature experiment was highlighted and integrated. The integral value generated for all of the temperature data is essentially the cumulative heat that the pupae experienced, and will be referre The baseline was set as the room temperature during the trials, and was 26 +/ 0.1 0 C for one thermocouple and logger, and 26.5 +/ 0.1 0 C for another that was used. Other possible factors such as peak temperature, time to peak tem perature, temperature change, and the rate of temperature change were also extracted from the logged temperature file, and were used in regression analysis. For the field burn experiments, finding the integral was a slightly more complex because the baseli ne, the ambient temperature, was changing due to normal diurnal heating. To determine what actually was the pulse of heat that from the fire, beginning
116 and end points were established. The beginning point was set as the moment right before a significant in crease in temperature occurred over a short period of time, attributable to the passing of the flame front from the fire. Because this time point was slightly different for each treatment, it was determined from the rate of temperature change calculated b y dividing the current temperature for a reading by the previous reading, and subtracting one: (Equation 3 1) where T n is the temperature at time n and T n 1 is the temperature at the previous time interva l. T he beginning point was set where the rate was positive in 3 consecutive measurements. N o such increase was observed in the unburned controls and rate point also lined up perfectly with a visual estimate of the beginning of heat pulse from the fire. Th e end points were set as the lowest temp erature post fire, right before heating from the diurnal flux begins again. This point was obtained by finding the halfway point between where cooling from the heat pulse levels off and diurnal heating begins again. This is the halfway point between the last negative rate change value and the first positive rate value in a series of 3 changes in a row that were positive. This also is lines up exactly with visual estimates of the end of the heat pulse. To generate the baseline, the two point drift correction feature in the Expedata software was used Beginning and end points were selected in the program at the chosen time interval, as close to or slightly below the temperature recorded at that time. This was to ensure t hat when the line was connected it would not intersect or go above those points. What resulted was a baseline that sets the boundary between the heat
117 pulse generated by the fire and the background diurnal temperature change. The integral was then taken ove r the interval of the heat pulse, with the newly generated baseline acting as zero Statistical analysis was conducted on sets of data collected from the lab experiment and the field experiments, and consisted of several steps: regression analysis of survi val as a function of heat separately in both lab and field experiments, and then regression of the factors shared in both the lab and field experiments to burial and litter depth measured in the field experiments (heat, peak temperature, time to peak tempe rature) To determine what factors of the heat pulse were significantly correlated to the success or failure of viable adults to eclose, logistic regression was conducted in R using the separate data sets generated for lab and field experiments. Interactio ns between these factors were modeled, and the analysis also included pupal age as a predictive variable, and pupa weight for the lab experiment. A correlation matrix of the predictors was created for both experiments, and heat, peak temperature, and buria l depth (in the field experiment) were shown to be strongly correlated, indicating some possible difficulty including these terms in the same regression model. This was confirmed in both the lab and field experiments, as quasi or complete separation of the data occurred, resulting in perfect prediction when the variables of heat, peak temperature, and burial depth were modeled together. Due to this issue, regression analysis could not proceed with these variables present at the same time, so they were pulle d out and modeled without the other two present in the model. Model selection began by eliminating the most complex interaction terms one at a time, comparing it to the model with all the other similar level interaction terms. After they were determined to
118 be singly non significant, they were sequentially added to the model without that level of interaction terms. This process was repeated for all interaction levels and continued with removal of single terms. Models with all terms significant were compared to simpler models comparing the Akaike Information Criterion ( AIC ) and corrected AIC (AICc) values AICc was used as it takes into account model complexity and sample size, penalizing for large numbers of parameters in comparison to sample size. As a rule of thumb, a simpler model ( = lesser number of terms) was chosen if the AIC/AICc did not decrease or increase by more than 2 when additional terms were added. The second part of the statistical analysis involved l inear regres sion analysis to determine if se nsor depth (a proxy for depth of pupation) and litter depth in the immediate vicinity were correlated to the factors that significantly predicted survival to eclosion of E. atala pupae A pooled set of data was used for this analysis involving the July 5 th and 6 th experiments containing E. atala pupae and the July 26 th experiment where pupae were not buried, increasing sample size for the regression analysis. Interaction and quadratic terms were included during model selection. Data were checked for normali ty by residual and Q Q plots, with removal of points conducted to improve normality and to reduce the influence of single points. Model selection was conducted in the methods previously described, with systematic removal of non significant terms and compar isons of models using the anova function with Chi squared tests using R. Results Field Observations of Depth of P upae of C. irus From 2010 to 2012, a total of 39 excavations were undertaken to find frosted elfin pupae at Ralph E. Simmons Memorial State For est, Nassau County, Florida In general,
119 leaf litter was composed primarily of turkey oak, sand oak, and persimmon leaves, along with slash and longleaf pine needles. A lesser amount of grass blades contributed, with wiregrass being the most prominent. The litter also contained some small twigs or branches, of varying length. The average litter depth across all excavations and years ranged from 5 to 50 mm, resulting in a total mean depth of 18 mm. Average duff depth per excavation site ranged from trace amo unts to 7 mm, with almost one third of duff measurements being trace amounts. Trace amounts of duff were those that were at levels below the ability to measure, at or less than 1mm. In summary, low amounts of litter were present, and there was little to no duff. Litter and duff depths recorded were likely due to the presence frequent burns and the subsequent consumption of fine fuels in the area, as prescribed fires were conducted roughly every other year from at least the period of 2008 to 2012. A total o f 12 pupae were found from the 39 excavations. Of those 12, 8 of the pupae were located at the surface of the soil, in the duff and below the leaf litter. The following 4 pupae were found in the soil: 1 at 0.5cm, 2 at 2.0cm, and 1 at 3.0cm deep (Figure 3 3 ). The relatively low recovery rate of pupae could be attributed to a negative effect from the movement restriction devices. In a comparison of mean temperature in and outside the movement restricting, only 1 out of 4 comparisions showed significantly high er temperatures inside the metal restricting device: mean temperature inside = 27.9 0 C, outside = 19.5 0 C ( t = 7.7214, df = 320.27 p < 0.0001). Graphing this shows quite the disparity in temperature inside and outside the device (Figure 3 4 A). Graphing the 3 other devices indicates the temperatures inside the metal restrictor device tracks well with the temperature outside the device (Figure 3 4 B D).
120 Heat Pulse in Soil of Prescribed Burns in S andhill at Ordway Swisher Biological Station, Putnam County, Fl orida, July, 2012 Soil surface temperatures as measured by infrared thermometer for the July 5 th and 6 th burns exceeded 500 C ( Figure 3 5 ), with lower temperature peaks recorded throughout the treatment plot (Figure 3 5 A B) Peak soil surface temperatures recorded for the July 26 th burn were approximately 375 C, once again with considerable variation in peak temperature recor ded in the treatment area (Figure 3 5 C ). An unburned position was also measured during the July 26 th burn, with no sharp rise in temperature recorded (Figure 3 5 C orange line). No E. atala pupae survived to adult eclosion in any of the soil surface trea tments Soil depth temperatures as measured by iButton sensors buried with E. atala pupae during the July 5 th burn peaked at 43 C in the two shallowest depths, 1.5cm and 1.9cm (Figure 3 6 ). During the July 6 th burn, soil depth temperatures as recorded by iButton sensors buried with E. atala pupae peaked at approximately 51 C for the shallowest depth of 1.3cm (Figure 3 7 ). Sensors buried in unburned sections recorded some heating, with the 2.8cm treatment recording displaying the most heating. Finally, soil depth temperatures during the July 26 th burn recorded the highest peak temperatures in the shallowest treatments, peaking at 53 C at 1.9cm and 50 C at 1.4cm (Figure 3 8 ). Sensors placed in the unburned area recorded little to no increase in temperature during the experiment, as did a sensor placed in 7.1cm deep in the burned section (Figure 3 8 light brown line. Heat T olerance o f E. atala P upae F ollowing L aboratory W ater B ath T reatment A plot of survival of E. atala pupae to successful adult eclosion following the initial water bath heating experiment is displayed in Figure 3 9. Survival varied by
121 temperature and by duration, with a threshold between success and failure to successfully eclose that is roughly linear in nature. Failure to successfully eclose began at about 40 C with a 50 min duration, and can be seen to roughly decrease linearly to 51 C with a 3 min duration A plot of the survival of E. atala pupae from the follow up water bath heating experiment is displayed in Figure 3 10 Survival to successful adult eclos ion remained high for the 40 C treatment of both 10 and 50 min For the 44 C and 47 C treatments, some survival occurred in the long duration treatments, but much less so than for the short duration treatments: at 44 C, 6/6 survived the 7 min treatment vs 2/6 surviving the 30 min treatment; at 47 C 5/6 survived the 6 min treatment vs 2/6 for the 10 min treatment. Pupal survival of the 51 C treatment was even more markedly different between durations: 5/6 for the 3 min treatment vs. 0/6 for the 17 min treatm ent. Survival of B uried and U nburied E. atala P upae to S uccessful E closion F ollowing P rescribed B urning Successful eclosion of E. atala pupae following the burial, prescribed burning, and removal is displayed in Figure 3 11 No adults successfully emerged for all three treatments on the soil surface, and for those placed 1.3 centimeters below the surface. Successful eclosion was mixed for depths between 1.5 and 2.5cm, with the proportion successfully eclosing ranging from 25% to 88%. For burial depths of 2 .8 to 4.1cm, 100% of E. atala successfully eclosed, but there was a brief drop of ecolsion success for the final two depths of 4.9cm each (75 88%).
122 Factors C orrelated to S urvival of E. atala P upae to S uccessful A dult E mergence F ollowing W ater B ath H eating Perfect prediction from the logistic regression was occurring when the variables two terms were analyzed in separate models (Table 3 5 ). Stepwise regression of the model in survival to successful adult ecolosion (Table 3 6 ). An output summary of this model is give n in Table 3 7 and a graph of each factor is plotted against pupa survival to successful adult eclosion (Figures 3 12 and 3 13 ). in the best fitting and most parsimonious mode eclosion (Table 3 8 ). The output summary of this model is given in Table 3 9 and is graphed and displayed in Figure 3 1 4 Factors C orrela ted to S urvival of E. atala P upae to S uccessful A dult E mergence F ollowing P rescribed B urning ted for this and are displayed in Table 3 5 Stepwise regression of the model containing being a single term 3 10 ). An output summary of thi s model is given in Table 3 11 and is graphed and displayed in Figure 3 15 best fitting and most parsimonious model being
123 3 12 ). The output summary of this model is given in Table 3 13 and a graph of each factor is plotted against pupa survival to successful adult eclosion (Fi gures 3 16 and 3 17 ). fitting and most parsimonious model being significant predictor of survival of pupae to successful adult e closion (Table 3 14 ). The output summary of this model is given in Table 3 15 and is graphed and displayed in Figure 3 18 Factors C orrelated to the H eat P ulse D uring P rescribed B urning Diagnostic plots of the linear regression analysis revealed a single data poin t that was having a highly influential effect, so this point was removed. The resulting residual vs. fitted and Normal Q Q plots were much more reasonable as a result of this removal, and analysis resumed. Linear regression of heat as a function o f burial depth and litter depth resulted in a significant negative correlation between the heat and burial depth (p = 0.0257, R 2 =0.1654 ) Specifically, heat transferred by the fire decreased as depth in the soil increased. The amount of litter in the immed iate vicinity of the individual treatment was not found to be a significant predictor of the heat pulse at any depth. Discussion Heat and peak temperature were consistent predictors of the survival of E. atala pupae to successful eclos ion into viable adul ts. In both the laboratory water bath experiments and in the prescribed burn field experiments, logistic regression of the survival of E. atala pupae to successful adult eclosion was significantly negatively correlated to heat ( C seconds) and peak tempera ture. These results are consistent
124 with what is known about survival of organisms at high temperatures, as there are upper thermal limits in both maximum temperature and the amount of heat energy that can be endured (Whelan, 1995) The time taken to reach to peak temperature was also found to be a significant predictor in both of the heat only models of the lab and field experiments, but the nature of this relationship was different between experiments. Time to peak temperature in the lab experiment was sho wn to have a positive correlation to survival of pupae to successful adult eclosion; as the time to peak temperature increases, survival increases. This is consistent with general knowledge of organisms, that increasing the time taken to reach a critical t emperature results in a higher probability of survival, and is due to such mechanisms as the heat shock response and upr egulation of chaperone proteins. In the field experiment heat only model, time to peak temperature was shown to have a negative relation ship with survival of a pupa to successfully eclose into an adult, where survival was high at low to mid values of time to reach peak temperature before dropping rapidly at high values. A logical explanation for this result is not clear, and due to the inc onsistency in this result from the lab and field experiments indicates that this may be compounded by another unknown factor. That inclusion of the time to peak temperature term in both heat only regressions resulted in the best model indicates the importa nce of this term, but it is curious that it was also not part of the peak temperature only models. Additionally, survival probability was at its lowest (80%) at the shortest time to peak temperature, and rapidly increased to 100%; this contrasts to surviva l dropping to 0% at high er heat and peak temperatures. The lack of consistency in both the models and biological
125 explanation for the time to reach peak temperature term indicate a need to explore the relationship further in future work. Survival of E. atal a pupae to successful adult eclosion was also shown to be positively correlated to burial depth: survival probability increased as burial depth increased (Figure 3 18 ). The model predicts survival to reach 50% at 1.75cm deep in the soil, and to steadily in crease up to 80% at about 3cm deep. Regression analysis of burial depth, heat, and peak temperature reveals a significant correlation between depth and peak temperature, but interestingly, heat ( C seconds) drops out of the model (Table 3 16 ). This could be from high surface temperatures recorded that mask or overwhelm the heat measured from the prescribed fires below the soil surface, or perhaps some other factor. Heat may still be important, and when graphed along with burial depth against survival, it is apparent that it changes at a similar rate to burial depth and survival (Figure 3 19 ). Plotting in the same manner, but replacing heat with peak temperature, also demonstrates a strong similarit y between the change in burial depth and survival, and from regression analysis this is a significant interaction (Figure 3 20 ; Table 3 17 ). In summary, as heat or peak temperature increases, survival to successful eclosion decreases. The results from the prescribed burning experiments indicate that there is 50% survival of E. atala in the soil at depth confers a benefit from the immediate effects of fire. The present study is unique in that it measured both the temperature at increasing depths during prescribed burning while simultaneously testing the survival of animals. Data on
126 immediate survival to fire by fossorial animals is limited, so this study provides novel data and directly links certain fact ors such as heat and depth to survival. The majority of organisms that reside at depths greater than 1.75 cm would be able to escape or endure the immediate effects of fire if their tolerance is similar to that of E. atala Of C. irus pupae that were found during excavation at Ralph E. Simmons Memorial State Forest, 25% were found in the soil, resulting in roughly 25% of pupae being able to survive a fire event. Additional support of the fire endurance strategy is that the phenology of C. irus fits well wit h the typical fire timing: larvae pupate mid to late spring, and fires in Florida historically occur during late spring to early summer. Typical fire frequency of 3 10 years also would allow some recovery of the population to pre fire levels, even at the s hortest fire return interval at ~3 years. Furthermore, the relationship between C. irus and its host plant, L. perennis developed in the context of a long term frequent fire regime. Sundial lupine readily sprouts post fire, has seeds that germinate best wh en the seed coat is partially broken, and is a rapid colonizer due to its ability to acquire nutrients in poor soils due to bacterial symbionts in the root nodules (Grigore and Tramer, 1996; Pavlovic and Grundel, 2009). Continuing on with relating the exp erimental results to C. irus survival of an individual population would be dependent on the reproductive response of the surviving individuals the following spring and any particular demographic problems that the loss of a large proportion of the populati on could cause, such as a genetic bottleneck. Data on these two factors is beyond the scope of this current study, but would further give evidence for the fire survival/endurance strategy. Furthermore, a thorough investigation of the soil pupation trait it self would be in order, including genetic basis and cost benefit
127 analysis. Other Lepidoptera groups such as the Noctuiodea and Sphingidae are accomplished soil pupating insects, and so the likelihood of this trait having a genetic basis is likely. There ma y not even be any genetic variability or heritability for the burrowing trait ; d ifferences seen in pupation depth may be an artifact of a time limit during the wandering larva stage and the finding of a suitable location for pupation, or of the tradeoffs o f growth and pupation location. Extra energy to be spent on wandering or burrowing may be different for individuals, and some may lack the resources to burrow beyond the leaf litter. In the case of a butterfly pupa in Florida, adaptation to dealing with su mmer temperatures, which can exceed the lethal limits, could also be advantageous to dealing with fire. Maximum temperatures at the soil surface at RESMSF exceeded 50 C for portions of the day even in the spring months of April and May, well beyond the threshold for survival as observed and modeled for E. atala It is quite likely C. irus or other insects with ground dwelling pupae have a similar threshold mortality to heat, and those that were insulated by leaf litter or soil would be would be more likely to survive. The other strategies to dealing with fire by C. irus are not as likely in light of the results of this study. As was mentioned earlier, immediate movement away in response to fire would be very much dependent on the timing of the fire with the phenology of the particular organism. Frosted elfin adults are active in Florida from mid February to late March, and a fire during this time could potentially be avo ided by an individual by flying to a nearby unburned area. To perpetuate the population, this new area would need to have resources such as adult nectar sources and larval host plants, or else the host plants would need to quickly re sprout in the burned a rea within the lifespan of the adult
128 butterflies. Sundial lupine has been observed to re sprout approximately 15 days post fire at RESMSF, so it is possible that these plants would be able to support larvae if an adult could survive this long to lay eggs a gain. A major constraint with the adult movement strategy is that beyond the short adult flight period of 2 5 weeks, C. irus is either a larva or pupa, which is either mostly sessile or completely dormant and unable to move far enough away. This strategy o f immediate avoidance would not be possible for the majority of the year, and the result would be a butterfly dependent on fire free immediate effects of such a fire. Fu rthermore, fire historically is most common in this habitat in late spring and early summer, a time when there are no adults present, and makes it more likely that immediate survival by pupation in the soil is the more likely strategy. Recolonization of th e burned area from outside areas remains a possible strategy, and many insects are thought to employ it. Insects typically have a high reproductive rate and the population ca n quickly recover, b ut not all insects are as fecund as others. Univoltine species like C. irus have a lower reproductive rate than a species with multiple generations per year, assuming reproductive ability of an individual of each species is similar. Population recovery post fire by recolonizing frosted elfins would be slower, and the population may not return to the pre fire level before another fire or other disturbance occurred. To support recolonization, m etapopulation structure would need to be present, which is not consistent in Florida (some casual evidence for Apalachicola NF, no evidence for Simmons SF) There is more support for this strategy than for simple movement away in response to fire, and it
129 may be employed in addition to a limited amount of local survival to a fire event. Fires are often patchy, so some pupae residing on the surface of the soil would not be burnt, and a limited amount survival in a burned area is also possible. In fact, there may be a combination of strategies: limited local survival by enduring the fire in the soil, survival by the sheer chance of res iding in a patch that is not burnt, and by recolonization of burnt areas post fire or in fire free years. What is clear is that the relationship and response of organisms to stochastic disturbance processes such as fire is quite complex, and that there are a multitude of possible strategies that organisms employ to coping with the stresses of living in such dynamic areas. For butterflies, given the evidence in this study and the overall prevalence of soil pupation in insects, it is likely that pupation in t he soil is a common behavior that does increase survival probability in fire prone areas. An immediate implication of this study is that it aids in better understanding the impact of prescribed fire on litter and soil dwelling Lepidopteran pupae. Survival of pupae of the experimental surrogate E. atala was found to be directly related to the amount of heat experienced in controlled laboratory experiments and in controlled burn situations. It is therefore likely that individuals in a populat ion of a species such as C. irus a proportion of which were demonstrated to pupate in the soil at RESMSF, would survive fire similar to that conducted during the experiments at OSBS, if conducted at the time of the year when the butterfly exists as a pupa. While not direc tly tested, mortality of egg and larvae residing on host plants would be extremely high in burned areas, as evidenced by temperatures exceeding 350 C at the soil surface in the three prescribed burn experiments, and the complete mortality of all pupae placed at the soil
130 surface (Figure 3 5 ). Survival of fire by C. irus or other sandhill inhabiting organisms would therefore solely rest in their abilit y to avoid such temperatures, which for C. irus could only occur as a pupa that was buried in the soil to at least a couple of centimeters, corresponding to those times of the year when C. irus are pupae, approximately mid May through early February the fo llowing year in Florida. Additionally, only a quarter of C. irus pupae were found to be buried in the soil, leaving the remaining proportion subject to heat induced mortality from a prescribed fire. So along with the proper seasonal timing of a prescribed fire, spatial extent is also highly important, as a fire which completely burns the entire area where C. irus are pupae would result in the death of 75% of the entire population. This suggests that the best way to manage an area containing a C. irus colony would be to only burn a portion of it each year, rotating burns in such a way that would not reduce the population too low, yet allowing for the positive benefits of fire to occur in the habitat, in order to improve conditions for host plants. This conclu sion, that spatial extent of prescribed fire is an important factor in the management of C. irus populations in sandhill pine and turkey oak forests, falls in line with earlier work done in the Midwest in prairies and oak barrens communities. Specifically, non burned areas are important as refuge or source for butterfly populations in small isolated habitat remnants (Swengel & Swengel, 2007). However, it has also been shown that there is higher abundance and species richness of insects in burned versus unbu rned habitat remnants, emphasizing a need for the use of prescribed burning as a habitat management technique (Panzer & Schwartz, 2000; Panzer, 2002; Waltz & Covington, 2004). Further supporting this need are two
131 explanations relevant to the current and la tter work, the fire attrition and intermediate disturbance hypotheses. The f ire attrition hypothesis states that burning at an inappropriate frequency results in a reduction in population sizes by not allowing sufficient recovery time, that too frequent of burns would reduce populations to become extinct (Panzer & Schwartz, 2000). This is certainly a relevant concern for C. irus pupae at RESMSF, of which only 25% could potentially survive a prescribed fire. On the other hand, species such as C. irus that ar e dependent on habitat and host plants maintained by relatively frequent disturbance also indirectly rely upon disturbance at some frequency. The intermediate disturbance hypothesis states that species richness is highest when disturbance is neither freque nt nor too rare, and serves to fit with the ecology and life history of C. irus (Grime, 1973; Connell, 1978) Other species with narrow habitat requirements would also be prone to this apparent dichotomy: reliance upon disturbance, yet an increased risk of extinction because of it. There is surely a delicate balance with disturbance for many species, with certain aspects of their habitat or life history that seem to run counter to others. This complexity can unfortunately confound conservation, management, and restoration practices aimed at improving these species populations, but past and future studies will hopefully serve to untangle what to even trained eyes appears to be in disarray.
132 Table 3 1. Experimental treatment setup for the follow up labora tory water bath heat tolerance experiment using E. atala pupae. Each temperature and duration combination included a total of 6 pupae representing 6 ages since pupation: 2, 6, 7, 8, 11, and 15 days since pupation. Duration pairs for each temperature are fo r previously observed survival (low duration) and survival (high duration). Temperature ( C) Duration (minutes) 27 (room temperature) 53 40 10 40 44 7 30 47 6 10 51 3 17 Table 3 2. Fire weather data for experimental prescribed fires conducted dur ing July 2012 at the Ordway Swisher Biological Station, Putnam County, Florida. Fire date 7/5/2012 7/6/2012 7/26/2012 Air Temperature Range ( C) at 2 meters Low 33.7 31.4 32.9 High 34 .0 32 .0 33.3 Average 33.9 31.7 33.1 Average Windspeed and di rection (km/hour) Low W 3.5 W 3.8 SW 2.6 High W 3.7 W 4.5 SW 2.7 Average W 3.6 W 4.2 SW 2.7 Relative Humidity Range (%RH) at 2 meters Low 52.84 64 .0 63.6 High 50.8 61 .0 60.9 Average 51.8 62.5 62.2 Fire Rate of Spread (meters/minute) 10 .0 2.5 2.5 Estimated Maximum Flame Length (meters) 1 2 1 1
133 Table 3 3. Experimental treatment numbers for controlled burn experiments conducted at the Ordway Swisher Biological Station, Putnam County, Florida, in July, 2012. Two Eumaeus atala pu pae of four separate ages were buried in each treatment on the July 5 th and July 6 th experiments, for a total of eight pupae in each treatment. All experimental treatments included a single iButton temperature sensor buried at assigned treatment depths. E xperiment Date Control (Unburned) Treatments Experimental Treatments Pupae Ages (Days Since Pupation) Total Pupae 7/5/2012 1 6 1 9 11 16 56 7/6/2012 2 9 2 10 12 17 88 7/26/2012 5 20 0 Table 3 4 Fire experiment treatment and replication for p rescribed fire heat tolerance experiments of E. atala conducted at the Ordway Swisher Biological Station, Putnam County, Florida. For each treatment depth, 2 pupae of each age were included. See Figure 3 2 for general orientation of buried pupae. Experime nt Date Pupa age (days since pupal molt) Treatment Depth (cm) 7/5/2012 1, 9, 11, 16 0 1.6 1.9 2.8 3.8 4.9 7/6/ 2012 2, 10, 12, 17 0 0 1.3 1.5 2.5 2.9 4.1 4.5 4.9 Table 3 5. Regression model setup for laboratory and prescribed burning experiments (A and B, respectively) on the survival of E. atala pupae to successful adult ecolosion. Regression Models Laboratory heat tolerance experiment Successful eclosion of E. atala ~ Heat*Time to peak temperature*Pupa age at experiment*Pupa weight Successful e closion of E. atala ~ Peak temperature*Time to peak temperature*Pupa age at experiment*Pupa weight Prescribed burning experiment Successful eclosion of E. atala ~ Heat*Time to peak temperature*Pupa age at experiment Successful eclosion of E. atala ~ Pea k temperature*Time to peak temperature *Pupa age at experiment Successful eclosion of E. atala ~ Burial depth*Depth of litter at burial location Burial depth ~ Heat*Peak temperature*Time to peak temperature
134 Table 3 6. Logistic regression model selec tion of survival of E. atala pupae to successful adult eclosion as a function of heat, time to reach peak temperature, pupa age, and pupa weight following laboratory water bath experiment, February 2012. Model AIC AICc deltaAIC deltaAICc survival~ 216.600 216.625 69.080 69.031 survival~ heat 147.520 147.594 2.580 2.497 survival~ heat + time to peak temperature 144.940 145.097 Table 3 7. Output summary of the best logistic regression model of survival of E. atala pupa to successful adult eclosion a s a function of heat, time to reach peak temperature, pupa age, and pupa weight following laboratory water bath experiment, February, 2012. Estimate Std. Error z value Pr (>|z|) Intercept 1.132e+00 4.234e 01 2.675 0.00748 heat 1.078e 04 2.051e 05 5.2 56 1.47E 07 t ime to peak temperature 1.007e 01 4.817e 02 2.091 0.03648 Table 3 8. Logistic regression model selection of survival of E. atala pupae to successful adult eclosion as a function of peak temperature, time to reach peak temperature, pupa age and pupa weight following laboratory water bath experiment, February 2012. Model AIC AICc deltaAIC deltaAICc survival~ 216.6 216.625 149.790 149.714 survival~ peak temperature 66.835 66.911 Table 3 9. Output summary of the best logistic regression model of survival of E. atala pupa to successful adult eclosion as a function of peak temperature, time to reach peak temperature, pupa age, and pupa weight following laboratory water bath experiment, February, 2012. Estimate Std. Error z value Pr (>|z|) Intercept 23.00705 4.39265 5.2381 1.63e 07 peak temperature 0.52620 0.09904 5.313 1.63e 07 Table 3 10. Logistic regression model selection of survival of E. atala pupae to successful adult eclosion as a function of heat, time to reach peak temperat ure, and pupa age following prescribed burning July 5 th 6th, 2012, at the Ordway Swisher Biological Station, Melrose, Florida. Model AIC AICc deltaAIC deltaAICc survival~ 114.14 114.187 25.643 25.554 survival~heat 88.497 88.626
135 Table 3 11. Output su mmary of the best logistic regression model of survival of E. atala pupa to successful adult eclosion as a function of heat, time to reach peak temperature, and pupa age following prescribed burning July 5 th 6th, 2012, at the Ordway Swisher Biological Stat ion, Melrose, Florida. Estimate Std. Error z value Pr (>|z|) Intercept 3.202 6.065e 01 5.279 1.30e 07 heat 3.092e 04 7.526e 05 4.109 3.97e 05 Table 3 12. Logistic regression model selection of survival of E. atala pupae to successful adult eclosio n as a function of peak temperature, time to reach peak temperature, and pupa age following prescribed burning July 5 th 6th, 2012, at the Ordway Swisher Biological Station, Melrose, Florida. Model AIC AICc deltaAIC deltaAICc survival~ 114.14 114.187 36.77 1 36.692 survival~ peak temperature 77.369 77.498 4.412 4.101 survival~peak temperature + time to peak temperature + peak temperature :time to peak temperature 72.957 73.397 Table 3 13. Output summary of the best logistic regression model of survival of E. atala pupa to successful adult eclosion as a function of heat, time to reach peak temperature, and pupa age following prescribed burning July 5 th 6th, 2012, at the Ordway Swisher Biological Station, Melrose, Florida. Estimate Std. Error z value Pr (>|z|) Intercept 40.95401 11.02081 3.716 2.02e 04 peak temperature 0.98463 0.26329 3.740 1.84e 04 time to peak temperature 2.69600 0.96861 2.783 5.38e 03 peak temperature :time to peak temperature 0.06673 0.02511 2.658 7.868e 03 Table 3 14. Logi stic regression model selection of survival of E. atala pupae to successful adult eclosion as a function of burial depth, litter depth, and pupa age following prescribed burning July 5 th 6th, 2012, at the Ordway Swisher Biological Station, Melrose, Florida Model AIC AICc deltaAIC deltaAICc survival~ 114.14 114.187 21.949 21.867 survival~burial depth 92.191 92.320 1.599 1.289 survival~ burial depth+litter depth+burial depth:litter depth 90.592 91.031
136 Table 3 15. Output summary of the best logistic r egression model of survival of E. atala pupa to successful adult eclosion as a function of burial depth, litter depth, and pupa age following prescribed burning July 5 th 6th, 2012, at the Ordway Swisher Biological Station, Melrose, Florida. Estimate Std. Error z value Pr (>|z|) Intercept 1.875 0.676 2.774 5.54e 03 burial depth 1.073 0.267 4.017 5.89e 05 Table 3 16. Linear regression model selection of burial depth as a function of heat, peak temperature, and time to peak temperature following presc ribed burning July 5, 6th,and 26 th 2012, at the Ordway Swisher Biological Station, Melrose, Florida. The Pr(>Chi) value results in the rejection of the null model in favor of the single term peak temperature model Multiple R squared: 0.434, Adjusted R sq uared: 0.4138 Model Reds.Df RSS Df Sum of Sq Pr(>Chi) burial depth ~ 29 51.034 1 22.148 3.596e 06 burial depth~peak temperature 28 28.886 Table 3 17. Output summary of the best linear regression model of burial depth as a function of heat, peak t emperature, and time to peak temperature following prescribed burning July 5, 6th,and 26 th 2012, at the Ordway Swisher Biological Station, Melrose, Florida. Model Estimate Std. Error t value Pr (>|z|) Intercept 9.619 1.424 6.754 2.47e 07 peak temp eratu re 0.167 0.036 4.633 7.55e 05
137 Figure 3 1. Datum setup and reference point diagram for field excavation of C irus pupae at Ralph E. Simmons Memorial State Forest, Nassau County, Florida, May June of 2012. A) Datum. B) Reference poin t diagram. Photo courtesy of M. Thom.
138 Figure 3 2. Diagram of pupae locations within prescribed fire treatment units for burn experiments at Ordway Swisher Biological Station in July, 2012. (1 4) represent paired E atala pupae of four different ages fo r each treatment; (5) is the position of the iButton thermochron. Arrangment was identical for surface treatment minus the iButton thermochron.
139 Figure 3 3. Depth from top of soil of C irus pupae excavated from Ralph E. Simmons Memorial State For est from August 2010 to June 2012. Photo courtesy of M. Thom.
140 Figure 3 4 Temperature at ground surface as measured by EasyLog USB sensors, April May 2012 at Ralph E. Simmons Memorial State Forest. A and B = open canopy ; C, D = shaded canopy
141 Fi gure 3 4. Continued.
142 Figure 3 5. Soil surface temperatures as estimated using a non contact infrared thermometer during controlled burning at the Ordway Swisher Biological Station, Putnam County, Florida, July, 2012. A) July 5 th 2012; B) July 6 th 2012; C) July 26 th 2012. Flat orange line in c) is an unburned location monitored during the controlled burn.
143 Figure 3 5. Continued. Figure 3 6. Soil temperatures as measured with iButton thermocrons during a controlled burn at the Ordway Swisher B iological Station, Putnam County, Florida, July 5 th 2012.
1 44 Figure 3 7. Soil temperatures as measured with iButton thermocrons during a controlled burn at the Ordway Swisher Biological Station, Putnam County, Florida, July 6 th 2012.
145 Figure 3 8. Soil temperatures as measured with iButton thermocrons during a controlled burn at the Ordway Swisher Biological Station, Putnam County, Florida, July 26 th 2012.
146 Figure 3 9. Plot of eclosion of E. atala eclosin failure of adult to successfully eclose. Line represents a linear threshold between success and failure of E. atala eclose into an adult following heated water bath treatment. Relat ive symbol darkness is related to age of pupae at the experiment.
147 Figure 3 1 0. Proportion of E. atala pupae successfully eclosing following heated water bath treatment. For each temperature and duration combination, 1 pupa of 6 different ages were included: 2, 6, 7, 8, 11, and 15 days since pupation.
148 Figure 3 11. Proportion of E. atala pupae successfully eclosing following prescribed burning on July 5 th 6 th at the Ordway Swisher Biological Station, Putnam County, Florida For each burial depth, n=8 pupae.
149 Figure 3 12 Plot of survival of E. atala pupae to successful adult eclosion as a function of heat from the laboratory water bath experiment, February, 2012. Figure 3 13 Plot of survival of E. atala pupae to successful adult eclosion as a function of time to peak temperature from the laboratory water bath experiment, February, 2012.
150 Figure 3 14 Plot of survival of E. atal a pupae to successful adult eclosion as a function of peak temperature from the laboratory water bath experiment, February, 2012. Figure 3 15 Plot of survival of E. atala pupae to successful adult eclosion as a function of heat from prescribed burning July 5 6 th 2012 at the Ordway Swisher Biological Station, Melrose, Florida.
151 Figure 3 16 Plot of survival of E. atala pupae to successful adult eclosion as a function of peak temperature from prescribed burning July 5 6 th 2012 at the Ordway Swisher Biological Station, Melrose, Florida. Figure 3 17 Plot of survival of E. atala pupae to successful adult eclosion as a function of time to peak temperature from pre scribed burning July 5 6 th 2012 at the Ordway Swisher Biological Station, Melrose, Florida.
152 Figure 3 18 Plot of survival of E. atala pupae to successful adult eclosion as a function of burial depth from prescribed burning July 5 6 th 2012 at the Ordway Swisher Biological Station, Melrose, Florida. Figure 3 19 Plot of heat and burial depth as related to survival of E. atala pupae to eclosion following prescribed burning July 5 th and 6 th 2012, at the Ordway Swisher Bio logical Station, Melrose, Florida.
153 Figure 3 20 Plot of peak temperature and burial depth as related to survival of E. atala pupae to eclosion following prescribed burning July 5 th and 6 th 2012, at the Ordway Swisher Biological Station, Melrose, Florid a.
154 CHAPTER 4 STATUS OF THE FROSTED ELFIN, CALLOPHRYS IRUS GODART (LEPIDO PTERA: LYCAENIDAE) IN FLORIDA Introduction Once present in pine and oak savannahs and barrens throughout eastern North America, the frosted elfin C allophrys i rus Godart, is under a marked decline. It has disappeared from several areas including Illinois, Maine, Ontario, and the District of Columbia, and is considered critically imperiled or imperiled in the majority of remaining locations it inhabits ( NatureSer ve, 201 3). C. irus is tied to habitats maintained by a steady degree of disturbance from forces such as fire, and faces extinction in areas where these disturbances are altered or where there is loss of habitat due to development. For these reasons, there is a need to document the locations that C. irus inhabits, to provide an updated picture. C. irus was last reviewed in 2008, and remains listed as vulnerable (Natural Heritage Inventory rank G3) since assigned that designation in 1999 (NatureServe, 2013). Work directly regarding C. irus has focused largely on the patterns and process of single or scattered populations in Wisconsin, Massachusetts, and New York, including fragmentation and the effects of fire suppression and fire management (Swengel, 1996; Sw engel, 1998; Albanese, et al., 2007; Albanese, et al., 2008; Pfitsch and Williams, 2009). However, C. irus is widespread south of these areas, and includes the discovery of a population in northeastern Florida in 1990 (Gatrelle, 1991). Further search has r evealed a distribution of colonies throughout northern Florida and the Panhandle representing the southernmost range of this species.
155 In Florida, C. irus is limited to the sandhill pine and oak upland and savannah ecosystems that contain its larval host p lant, Lupinus perennis L., which includes both federal and state managed forests, as well as the margins of roads and power line cuts. State and federal forests are maintained for a wide range of activities including conservation of natural resources, visi tor recreation, and timber harvest, and often involve the use of herbicides or prescribed fire. Roadsides and power line right of ways are managed for their various functions, including practices such as mowing and prescribed fire. From this then, it is cl ear that some degree of disturbance is present in Florida C. irus habitats, and the disturbance or lack of disturbance will have an effect on the populations of C. irus and L. perennis that reside within them. Study of C. irus in Florida can provide genera l information on a species at the boundaries of its range, where the community and the forces that shape it are liable to be different at least at some scale. Monitoring of these populations serves the particular role of providing a basic understanding of presence in a given time frame. The goals of this study were to provide a clear picture of where L. perennis and C. irus are and where they might be by (1) documenting the historical distribution of C. irus and its larval host plant L. perennis in Florida, (2) identifying locations it currently resides in Florida, and (3) using a combination of both records to identify areas where C. irus might be found in Florida by using ecological niche modeling. The results of (1) establish a baseline from which to proc eed to the second objective of locating current populations of C. irus This second objective provides a snaps hot of records, updated to the current point in time. Finally, combining the results of (1) and (2) allow s for r a larger number of records to par ameterize the predictive model (3) that can be used for
156 identifying possible locations of C. irus or L. perennis populations, as well as providing an initial look into the larger scale patterns of distribution of these species in Florida. Methods Study Organisms Callophrys irus Godart is a hairstreak butterfly in the Lycaenidae family of the insect order Lepidoptera (Figure 4 1A B). It is most often found closely associated with its larval host plants, which in Florida is the sundial lupine variant or su b species of L upinus perennis, L. perennis gracilis (Nutt.) D. Dunn., a more diminutive form that is found in sandhill pine and oak uplands of northern Florida and the Panhandle (Figure 4 1B C). C. irus spends the majority of the year as a pupa, emerging a s an adult in early spring, which can be as early as mid to late February in northern Florida. Adults are most typically active in the middle of March to mid April, mating and laying eggs on L. perennis plants. C. irus larvae hatch and develop over a perio d of approximately 5 6 weeks, then retreat to the leaf litter or soil to pupate where they remain until the following spring. Species Record C ollection To establish the baseline distribution for C. irus and L. perennis in Florida, records were collected an d input into ArcMap for reference ( ESRI, Inc., Redlands, CA, USA ). Records for C. irus come from a number of sources databased in Butterflies and Moths of North America prior to 2010 (John Calhoun, pers. comm.). These include original discovery of C. irus in Florida, as well as results of butterfly surveys by the Florida Natural Areas Inventory (FNAI) and the members of the North American Butterfly Association (NABA). All C. irus records contained GPS coordinates and short
157 descriptions of the collection. L. perennis records come from detailed descriptions on vouchered specimens housed at three herbaria: the University of Florida Herbarium University of South Florida Herbarium (USF) Most records lacked GPS coordinates, and so were assigned coordinates (geo referenced) using a combination of Google Earth and paper maps. Only records with locality detail sufficient to assign a specific location were used. Records containing informatio 5.6 mi. E of Madison Co. line on FLA 6 Sandy Hill in Live Oak, Fla. C. irus and L. perennis records were then added to a base county map of Florida (Florida Geographic Data Library, GeoPlan Center, U niversity of Florida). C. irus M onitoring S urveys To assess the current status of C. irus in Florida, C. irus populations were monitored for presence or absence of adults or larvae over 2010 2012 through surveys by multiple parties. Surveys were conducted approximately every two weeks starting at the beginning March, continuing through early the beginning of May, or until C. irus was observed at the site. Locations at each site where new L. perennis sprouting and blooming occurred were targeted, including a reas of full sun to partial shade. 2010 monitoring included NABA member surveys at Blackwater River State Forest (BRSF) in Okaloosa County, a new location discovered at Munson Hills in Apalachicola National Forest (ANF), and surveys by the author at Ralph E. Simmons State Forest (RESMSF) in Nassau County, northwest of Jacksonville. 2011 monitoring again included NABA member surveys in BWRSF and ANF, an additional discovery in also in the ANF ~30km west of Munson Hills, while surveys by the author expanded R ESMSF to include the original dis covery location near Middleburg, Clay County. The final year of monitoring in
158 2012 included surveys by NABA members in BRSF, the two locations in ANF from 2011, and surveys by the author at RESMSF, and Middleburg. Niche M od eling In order to identify where C. irus might be for future surveys, ecological niche modeling was used in an attempt to describe locations where the probability of finding C. irus would be highest given a small set of environmental conditions. Niche mode ling included separate models of both C. irus and of the larval host plant, L. perennis C. irus has not been recorded utilizing any other host plant in Florida, and so C. irus should be effectively restricted to areas that contain L. perennis Niche model ing was conducted in MaxEnt (version 3.3.3k, Phillips, et al., 2006; Phillips and Dudik, 2008). This program uses the maximum entropy method, which for species distribution modeling, defines a probability distribution over a study area using a combination of geo referenced species occurrence records and features such as climate, land cover, vegetation, or many combinations of other such data. This method is advantageous for several reasons, the most obvious in that it uses presence only records, which are e asily obtained from museum collections and field surveys, and performs well with presence only data (Elith et al., 2006) Furthermore, the output is derived from clear mathematical me thods, allowing for straightforward analysis. Output predictions are continuous and are easily displayed and interpreted on a map of the study area; categorical or binary cutoff points can be made at desired intervals, further simplifying interpret ation (P hillips, et al., 2006). Predictions can serve many functions in a griculture, conservation, to species management, including using predicted future environmental conditions from global climate change estimates to both predict species distribution and potent ial invasiveness (Lentz, et al. 2008 ; DeMatteo & Loiselle, 2008; Mukherjee, et al., 2012). However,
159 some caution should be used when modeling with this and other methods, as there is considerable risk of error when predicting beyond the environmental extr emes currently recorded for the modeled organism, and there is considerable variation from the model procedure itself (Hijmans and Graham, 2006). Biotic factors are also highly important determining the distribution of any species, including elements of th e community such as food or competitors, as well as limitations to dispersal ( Davis et al., 1998; Kea rney & Porter, 2004; Guisan & Thuiller, 2005 ; Hijmans and Graham, 2006 ) Including these features into the model or during the interpretation of results is paramount, as they take the model output to a more realistic state. Location record data sources used in C. irus and L. perennis niche modeling come from the BAMOA database and the vouchered herbarium records that were geo referenced, as outlined above. F eature data used for L. perennis modeling include a mix of climate, geological, and geographical variables obtained from public sources: temperature, precipitat ion and altitude from WorldClim in the form of layers at 30 arc s econd resolution, and a geology layer of the dominant rock present in the soil from USGS ( Hijmans et al., 2005 ) These l ayers were clipped at the Florida administrative borders in Ar cMap as part of the ArcGIS 10.0 package (ESRI, Inc., Redlands, CA, USA). The accuracy of the model was evaluated using 20% of test data to calculate the area under the curve (AUC) ( Fiedli ng and Bell, 1997 ) A fair model is estimated to score an AUC value above 0. 8, and a good model greater than 0.9 (Swets, 1988; Thuiller, et al., 2006) All models were ru n 100 times and the average AUC w as trusted to measure accuracy.
160 Several models were f irst run for L. perennis using different combinations of Bioclim layers based on correlation and biolo gical importance (Table 4 1). Scenarios set up were: Model1: mean temperature of wettest quarter (BIO8 ) and precipitation of wettest month (BIO13); Model 2: BIO8 and precipitation of wettest quarter (BIO16); Model 3: BIO 13 and precipitation of driest month ( B IO 14 ); Model 4: precipitation of warmest quarter (BIO 18 ) and precipitation of coldest quarter (BIO19 ) These scenarios were all ran with the altitude a nd geology layers. In addition, climate variables were chosen by the results of correlation analysis of all 19, for which the combination of representing the maximum number of variables in the Bioclim dataset. Three were selected from this analysis and included max temperature of warmest month (BIO5), precipitation of warmest quarter (BIO18), and precipitation of coldest quarter (BIO19). Each of these climatological te rms make intuitive sense as predictors for a plant such as L. perennis which prefers well drained soil and is susceptible to root rot (Anderson, 2003). Each of these models were then run with the altitude and geology layers. Feature layers used for C. iru s modeling also included a mix of climate, geological and geographical variables. Climate variables were chosen by comparing sets that contained no more than four variables and that were hypothesized to have a large effect: Model A: BIO5 (max temperature o f warmest month), BIO9 ( mea n temperature of driest quarter) BIO14, and BIO18; Model B: BIO6 (minimum temp erature of coldest month), BIO11 (mean temperature of coldest quarter) and BIO19 ; Model C: BIO5 BIO 6, B IO14, and B IO15 (precipitation seasonality); M odel D: and B IO5 B IO 6, B IO 18 and B IO 19 These models were all ran with the altitude and
161 geology layers, and with and without the best L. perennis model. (Table 4 2). This method was similar to that used with L. perennis above: a priori knowledge about the environmental limits of C. irus were applied. U pper temperature limits may be reached in the long, hot, and periodically dry summers in Florida. Extreme cold is not likely to be as important to C. irus in Florida, as it extends its range far north into ar eas such as Wisconsin, Massachusetts, and New York, areas with much more pronounced of winters. It may also be possible that some period of cold temperatures is required in order to break diapause, and these temperatures are uncommon the farther south in F lorida. Results Species Record C ollection 134 vouchered L. perennis specimen records were gathered from the FLAS, FSU, and USF herbaria in July and September 2010 (Table 4 3). A total of 95 unique records were refined from this total, which included remov ing repeat locality records present at and between herbaria (Figure 4 2). Vouchered records include specimens listed from 18 counties in Florida including: Bay, Calhoun, Clay, Escambia, Franklin Hamilton, Jackson, L afayette, Leon, Liberty Madison, Okaloo sa, Santa Rosa, Suwannee, Wakulla, Walton, Washington Upon geo referencing, two records from Lafayette County were confirmed to actually be in Gilchrist County, and the Madison county records did not contain enough information to properly assign GPS coord inates. The refined list of counties with vouchered records of L. perennis is then: Bay, Calhoun, Clay, Escambia, Franklin Gilchrist, Hamilton, Jackson, L afayette, Leon, Liberty, Okaloosa, Santa Rosa, Suwannee, Wakulla, Walton, and Washington
162 Seven known records of C. irus were recorded by the Butterflies and Moths Database prior to 2010 (J. Calhoun, pers. comm; Table 4 4; Figure 4 3). These include locations in Clay, Franklin, Leon, Liberty, Nassau, and Okaloosa Counties. Locality records from Leon, Fran klin, and Liberty County are within the Apalachicola National Forest. The Nassau County record comes from inside Ralph E. Simmons Memorial State Forest. The locality record for Okaloosa County refers to a roadside that runs through the Blackwater River Sta te Forest. Finally, Clay County records include a power line cut near Middleburg, and a location near tree farms outside of Penney Farms. C. irus Monitoring S urveys Summarized results of monitoring of a selection of C. irus localities from 2010 2012 are di splayed in Table 4 5. C. irus was present at BRSF, Munson Hills (ANF), and RESMSF in 2010. 2011 surveys included the discovery of a new locality in ANF, approximately 25km west of the Munson Hills site. C. irus was present at BRSF and RESMSF in 2011, but n one were seen during the surveys at Munson Hills or at the Middleburg site in Clay County. The final year of surveys in 2012 was similar to 2011, with C. irus present at BRSF, RESMSF, and the new ANF location, but none observed at either Munson Hills or th e Middleburg locality. MaxEnt Niche M odel ing A table summarizing t he AUC results from the L. perennis modeling is presented in Table 4 6 With an AUC of 0.882, model 5 ( BIO5, BIO18, BIO19, ELEV, ROCK ) had the highest AUC of the entire set of models. This model was parameterized by the combination of Bioclim layers picked by the coefficient of >0.7 and is considered the most accurate model out of the set. A visual projection is
163 displayed in Figure 4 4; yellow to red area s indicate areas of high prediction, while blue colors represent areas of low prediction. AUC results for the C. irus modeling are displayed in Table 4 7, and are separated between whether or not the best L. perennis prediction layer (Model 5: BIO5, BIO18, BIO19, ELEV, ROCK ; AUC = 0.882) was included. For the models that included the L. perennis layer, Model B (BIO6, BIO1, BIO19, ELEV, ROCK) was the best, with an AUC of 0.716 (Figure 4 5). For models without the inclusion of the best L. perennis layer, Mode l A (BIO5, BIO9, BIO17, BIO18, ELEV, ROCK) was the best, with an AUC of 0.861 (Figure 4 6). Interpretation of the two C. irus predictive projections are the same as with the L. perennis model, with yellow to red indicating areas of high likelihood, and blu e areas low likelihood. Discussion There are very few locality records for C. irus in Florida. Combining records from 1990 2009 with the results of monitoring from 2010 2012 results in only a maximum of nine records. It remains unknown if C. irus is limite d to these localities. Considering the general nature of C. irus to be a locally rare species, this might be true. It is also important to note that there may be some bias in the records for C. irus As they are known from museum collections and as part of other butterfly surveys limited in scope, some areas are more favored for collection over others due to a multitude of reasons (ease of access, chance of encounter, lack of identification of potential areas). The amount of time that C. irus is easily obse rvable is quite limited as it is univoltine and has a relatively quick development of immature stages. This is compounded by several factors that make it difficult to detect such as drab adults and cryptic immature life stages. The time when different life stages were present was also variable across
164 Florida, with the timing of life stages at RESMSF sometimes a week or two later than those in ANF, for example. Variable spring weather also is likely to affect observation, as late cold periods in spring disru pt the adult flight season into a reduced and bimodal response (J. Daniels, M. Friedman, pers. comm.). Based purely on herbarium records, L. perennis is much more plentiful than C. irus From herbarium records, L. perennis appears more or less con tinuous throughout the Florida P anhandle. Records drop off east of Leon and Wakulla Counties, before appearing in Suwannee, Lafayette, Gilchrist, Hamilton, and Columbia Counties. Continuing eastward, county records occur again only in Clay County and Nassau Count y. The same potential issues of bias with records of C. irus mentioned above are also likely to apply to L. perennis particularly the lack of survey of potential areas and ease of access. On the other hand, L. perennis detection is much easier than C. iru s particularly when the plant in is bloom, and the fact that it is sessile. Even beyond herbarium records, L. perennis is quite plentiful, from clusters of plants to even an almost carpet like appearance in locations like Munson Hills (Figure 4 7). The u se of Maxent to produce species distribution models of L. perennis and C. irus had a moderate amount of success. The best L. perennis model had an AUC of 0.881, which should be considered at least a fair model (Thuiller, et al., 2006). Similarly, the best C. irus distribution model had an AUC of 0.861, also a fair model. However, this model did not contain the L. perennis model as a predictor. When this parameter was added, the best model instead had an AUC of 0.716, a value indicating a poor model. While C irus will be tied to areas where L. perennis is present, the output of the L.
165 perennis model does not improve the C. irus predictions; no information is gained by the addition of this term. Based on the best niche model tested, the highest probability o f C. irus to be present is approximately 0.75 (Figure 4 6). For most of the Florida panhandle, the predicted probability C. irus is over 0.5. Also, the prediction for L. perennis is high through this range, providing additional support that C. irus may be found in these areas. In contrast, the large geographic range that is projected to have >0.5 prediction also suggests that the model is not very discriminatory, that other factors have a role. This is a firm possibility, considering that C. irus is very pa tchy throughout its range outside the Northeast, even at a coarse resolution (Figure 4 8). Perhaps then the possibility is high for C. irus to be in any given area, but other factors serve to limit the actual presence. The general understanding of an ecolo gical niche model is that the prediction acts as the realized niche, the actualized area where the species will occur after taking into account the effects of predation or competition (Hutchinson, 1957; Guisan and Thullier, 2005). Two other factors outside of the realized niche might serve to explain the observed patterns of C. irus and include source sink dynamics and dispersal limitations (Pulliam, 2000). A habitat might be suitable, but the population growth rate is < 1, resulting in extinction of the p opulation if not constantly replaced by immigration, which may also be hindered by barriers to dispersal. There may already be evidence for these factors in the current observed occurrences of C. irus manifesting in one particular area that is striking in the difference in prediction between L. perennis and C. irus In the northeastern part of Florida, L. perennis has a band of high prediction that is not matched by the C. irus model. There are C. irus collection records in this
166 area, as this includes Clay and Nassau County. However, the Middleburg site that was monitored during 2010 2012 did not have C. irus so the records may represent a past sink dynamic and that is now limited by dispersal. Nassau County remains one of the best locations to regularly o bserve C. irus to this area. Based on the evidence of the Maxent models, Middleburg and the area around it represents a current disjunction of C. irus and L. perennis in contrast to the past records of C. irus presence. Several future directions are apparent from the current study, the first of which includes a thorough census of all recorded and currently known areas where C. irus is found in Florida. This could then be compared against the model and the hist orical records provided in this study to develop hypotheses for any changes in presence, the how and the why. This census effort could be enhanced by simple monitoring efforts of forest biologists and naturalist groups such as NABA, even to just alert of o bservations of C. irus throughout the state or the onset of spring. The present monitoring effort was sampling of convenience, and so the results must be taken with a certain amount of caution, as for example, several areas in ANF were not monitored, inclu ding the Franklin and Liberty County records. The same amount of caution should be applied to the distribution modeling, as there are surely a multitude of other factors that influence the presence of an insect such as C. irus besides some simple climate and geological data. Certainly for L. perennis an improvement on s imple elevation would be useful; the range of elevation it can be found is quite variable, 10m at Munson Hills in ANF, ~20m at RESMSF in Nassau County, and ~75m at BWRSF in Okaloosa. In fac t, relating the elevation to the
167 surrounding area in a manner such as Topographic Position Index (Jenness Enterprises) would be ideal, as L. perennis is limited to the well drained sandy soil characteristic of Florida sandhills, compared to lower lying or drainage areas of surrounding areas. An improved C. irus distribution model would aid in identifying areas where new populations could be found or candidate locations for captive rearing and release if ever the need may arise. Despite the limitations that this modelling may have, it remains a worthy tool to indentify where undiscovered populations of C. irus may be. Discovery of such populations could then be used to provide additional location records to refine the current model. Even if no new records are found, this helps identify any features that may be restricting the presence of C. irus and provide additional knowledge about its broad scale distribution. Furthermore, if C. irus is not recorded in a highly predicted area, this area may remain a viable area for mass rearing and release projects. These areas may be in high quality sites, but are limited due to dispersal or other factors outside of the ability of C. irus colonize.
168 Table 4 1. Climate variable conceptual models for maximum entropy species distribution modeling of L. perennis in Florida. BIO5 = Max Temperature of Warmest Month; BIO6 = Min Temperature of Coldest Month; BIO9 = Mean Temperature of Driest Quarter; BIO11 = Mean Temperature of Coldest Quarter; BIO13 = Precipitation of Wett est Month; BIO14 = Precipitation of Driest Month; BIO18 = Precipitation of Warmest Quarter; BIO19 = Precipitation of Coldest Quarter. Hypothesis Terms Temperatures and precipitation of the warmest and wettest part of the year limit species distribution Model 1: Monthly precipitation BIO8, BIO13 Model 2: Quarterly precipitation BIO8, BIO16 Precipitation in the driest and wettest parts of the year limit species distribution Model 3: Monthly precipitation BIO13, BIO14 Model 4: Quarter ly precipitation BIO18, BIO19 Table 4 2. Climate variable conceptual models for maximum entropy species distribution modeling of C. irus in Florida. BIO5 = Max Temperature of Warmest Month; BIO6 = Min Temperature of Coldest Month; BIO9 = Mean Temperatur e of Driest Quarter; BIO11 = Mean Temperature of Coldest Quarter; BIO13 = Precipitation of Wettest Month; BIO14 = Precipitation of Driest Month; BIO18 = Precipitation of Warmest Quarter; BIO19 = Precipitation of Coldest Quarter. Hypothesis Terms Temperatu res and precipitation of the warmest part of the year limit species distribution: heat and desiccation stress Mode l A nm nm nm nm nm nm nm n BIO5, BIO6, BIO17, BIO18 Temperatures and precipitation in the coolest part of the year limit species distribut ion: breaking of diapause requires some cold Model B nm nm nm nm nm nm nm n BIO6, BIO11, BIO19 Temperatures and precipitation of the warmest and coolest parts of the year limit species distribution: heat and cold important Model C: Seasonality precip itation BIO5, BIO6, BIO14, BIO15 Model D: Quarterly precipitation nm BIO5, BIO6, BIO18, BIO19
169 Table 4 3. Vouchered Florida L. perennis specimen records collected from FLAS, FSU, and USF herbaria in 2010 separated by county. The number of geo referen ced records is followed by the total number of records present. Duplicate localities were only included once in geo referenced totals. The total record omits repeat specimen records present in between FLAS and FSU. *Record listed as Lafayette County, but d etermined to be Gilchrist County during geo referencing. **Incomplete locality information failed to confirm this record. Herbarium Number Counties FLAS 36 (69) Bay, Calhoun, Clay, Escambia, Gilchrist*, Hamilton, Jackson, Lafayette, Leon, Liberty, Madiso n**, Okaloosa, Santa Rosa, Suwannee, Wakulla, Walton, Washington FSU 35 (37) Bay, Calhoun, Clay, Franklin, Gilchrist*, Hamilton, Jackson, Lafayette, Leon, Liberty, Santa Rosa, Suwannee, Wakulla, Walton, Washington USF 26 (28) Bay, Calhoun, Clay, Escambia Hamilton, Lafayette, Liberty, Nassau, Okaloosa, Santa Rosa, Suwannee, Wakulla, Walton Total = 95 (134)
170 Table 4 4. Florida records of C. irus prior to 2010 from Butterflies and Moths of America Database. A brief description is included, but exact locations withheld. County Description Clay Original discovery of C. irus ; near Middleburg Clay Near Penney Farms Franklin Apalachicola National Forest Leon Apalachicola National Forest Liberty Apalachicola National Forest Nassau Ralph E. Simmons Memorial State Forest Okaloosa Blackwater River State Forest Table 4 5. Florida C. irus surveys 2010 2012 by NABA and M. Thom. A brief location description is included, but exact locations withheld.*New record in 2010. **No adults or larvae se en in 2011 or 2012 following initial discovery in 2010. ***New record discovered in March 2011. Year Location Status 2010 BRSF, Okaloosa County *Munson Hills (ANF), Leon County RESMSF, Nassau County Detected Detected Detected 2011 BRSF, Okaloosa County *Munson Hills (ANF), Leon County ***ANF, Leon County RESMSF, Nassau County Middleburg, Clay County Detected Not observed Detected Detected Not observed 2012 BRSF, Okaloosa County **Munson Hills (ANF), Leon County ***ANF, Leon County RESMSF, Nassau County Middleburg, Clay County Detected Not observed Detected Detected Not observed
171 Table 4 6. MaxEnt model output for L. perennis using Bioclim, elevation, and major rock type layers. AUC = area under the curve, a measure of model fit. Model 5 Bioclim va riables selected by Model Terms AUC 1 BIO8, BIO13, ELEV, ROCK 0.840 2 BIO8, BIO16, ELEV, ROCK 0.850 3 BIO13, BIO14, ELEV, ROCK 0.825 4 BIO18, BIO19 0.877 5 BIO5, BIO18, BIO19, ELEV,ROCK 0.882 Table 4 7. MaxEnt model out put for C. irus using Bioclim, elevation, major rock type, and L. perennis model 5 species distribution layer. AUC = area under the curve, a measure of model fit. Model Terms AUC With L. perennis layer A BIO5, BIO6, BIO17, BIO18, ELEV, ROCK 0.694 B BI O6, BIO11, BIO19, ELEV, ROCK 0.716 C BIO5, BIO6, BIO14, BIO15, ELEV, ROCK 0.698 D BIO5, BIO6, BIO18, BIO19 0.698 Without L. perennis layer A BIO5, BIO9, BIO17, BIO18, ELEV, ROCK 0.861 B BIO6, BIO11, BIO19, ELEV, ROCK 0.842 C BIO5, BIO6, BIO14, BIO1 5, ELEV, ROCK 0.849 D BIO5, BIO6, BIO18, BIO19, ELEV, ROCK 0.845
172 Figure 4 1. C. rus (A, B) and L perennis var. gracilis (C) at Ralph E. Simmons State Forest, Nassau County, Florida. Photos courtesy of M. Thom.
173 Figure 4 2 Geo referenced records of Florida L. perennis from vouchered specimens gathered from FLAS, FSU, and USF herbaria in July and September 2010.
174 Figure 4 3. Records of C. irus in Florida prior to 2010, courtesy of Butterflies and Moths of Nort h America.
175 Figure 4 4. Ecological niche model of L. perennis using the maximum entropy method (Maxent version 3.3.3k), projected across Florida. Model terms are BIO5, BIO18, BIO19, elevation, and rock. Yellow to red represent high probability of L. per ennis presence given the modeled terms. AUC = 0.882.
176 Figure 4 5. Ecological niche model of C. irus with the L. perennis layer using the maximum entropy method (Maxent version 3.3.3k), projected across Florida. Model terms are BIO5, BIO9, BIO 17, BIO18, elevation, and major rock type. Yellow to red represent high prediction of C. irus given the modeled terms. AUC = 0.787.
177 Figure 4 6. Ecological niche model of C. irus without the L. perennis layer using the maximum entropy method (Maxent ver sion 3.3.3k), projected across Florida. Model terms are BIO5, BIO9, BIO17, BIO18, elevation, and major rock type. Yellow to red represent high prediction of C. irus given the modeled terms. AUC = 0.861
178 Figure 4 7. L. perennis at Munson Hills Apalachicola National Forest, Leon County, Florida. A) 2010. B) 2011 Photos courtesy of M. Thom.
179 Figure 4 8. USA county records of C. irus Records are from the Butterfly Occurrence Database from the National Atlas of the United States, USGS, and the Florida Natural Areas Inventory
180 CHAPTER 5 C ONCLUSIONS AND SYNTH ESIS General Importance One of the main reasons of this research, like all such work in science, is to fill in the gaps. This is of special importance due to the nature of t he organisms invol ved: C. irus is an imperiled species and is a representative of a habitat and community in decline C. irus then provides a g reater understanding of biodiversity and the wealth that it can provide specifically the exceptional relationship to anthropogenic disturbance from fire. The particular story investigated this relationship: the phenomena and consequences of living in a fire prone habitat, and what that represents. T he comp ilation and review of information rangewide for C. irus was conducted as an introductory chapter to this disseratioin. While done previously in such formats as NatureServe and in Schweitzer et al. 2012, the current work represents a sing ular focus on a sin gle taxa, including detailed study of C. irus south of the Midwest and Northeast US. Figures displaying the statewide and countywide distribution of C. irus were generated, as well as larval host plant distributions and use of larval host plants rangewide. Study of Florida populations of C. irus also helped to describe new ecological patterns A population of C. irus was shown to vary according to several features of the microhabitat identified in other studies, but also to the presence of other L. perennis herbivores and browsers. A major piece of this dissertation linked C. irus pupation behavior to the fire prone habitat it resides in, specifically investigating the mortality fire
181 can induce on C. irus pupae C. irus has evolved in fire prone habitats, pe rsisting despite the challenges of such an environment. It was shown that C. irus mortality from fire is significant, and thus fire plays a major role in population persistence. This is of particular importance because of the ubiquitous alteration of ecosys tems by humans specifically habitat loss and fragmentation Current C. irus populations are often small and isolated as a result and are particularly vulnerable to changes to population genetics or male to female ratio. Fire represents a selection event which has a major effect on a small population, and so changes to the fire regime such as season of burn, intensity, and frequency by humans is changing the relationship C. irus has with fire. Need for Communication The e xpanded knowledge of C. irus ecolog y aids conservation and management for the species and the habitat and community it lives within. Communicating the new information learned from this research in an accessible format is critical to it being used effectively. One such use that several chapt ers aid in is the development of a degree day model The model could be used to predict when C. irus will be present and be employed by amateurs, researchers, or to land managers with conservation objectives Communication of this work also acts as a snaps hot of the current state of knowledge and realtity of a declining species and habitat, providing evidence to be used in future inquiry. An additional need for communicating this work is that it involved an i nnovative method for determining effects of fire on insects of which not much is known. A series of field and lab experiments were conducted to describe the conditions experienced during a common scenario (prescribed fire in upland pine/oak sandhill) and it has built the basis for further work on fire, heat, and survival of ground or soil dwelling organisms
182 Influence of Current T hinking The entire dissertation has an u nderlying theme of investigation into the effects of anthropogenic d isturbance on organisms. Work similar to this theme include s study o f biodiversity patterns, the effect of global climate change on organisms and the environment, and the role of species conservation and habitat restoration. The research detailed in this dissertation documents the decline of a species and habit at, factors related to its distribution, and the effects on mortality from prescribed fire Though C. irus evolved with fire present ( the habitat and larval host plant persistence is maintained by frequent fire), current fire regimes are different and will result in d ifferent outcomes, such as extinction of the small populations that C. irus now commonly found in. Management of areas where C. irus inhabits must take into account the fac t that small populati o n s are more vulnerable to extinction, and that techniques for maintaining the habitat must be balanced with reducing the negative impact those activities may have on C. irus Lastly, the pupation of C. irus in the soil is an interesting feature of life history which was shown to increase the probability of survival from fire. This behavior may have evolved for a variety of different reasons such as predator or parasitoid avoidance or an escape from environmental extremes, but in the human altered landscape of the present, it may also confer a survival benefit from mo re frequent fire. This provides additional evidence that traits can be successfully co opted for multiple uses, and that species may have an ability to adapt to even rapid environmental change. Such ability will become paramount for species as anthropogeni c influence on the natural world continues to accelerate.
183 LIST OF REFERENCES Albanese, G., P. D. Vickery, and P. R. Sievert. 2007. Habitat characteristics of adult frosted elfins (Callophrys irus) in sandplain communities of southeastern Massachusetts, USA Biological Conservation 136 :53 64. Albanese, G., P. D. Vickery, and P. R. Sievert. 2008. Microhabitat use by larvae and females of a rare barrens butterfly, frosted elfin ( Callophrys irus ). Journal of Insect Conservation 12 :603 615. Allen, Thomas. 1997. The Butterflies of West Virginia and Their Caterpillars. University of Pittsburg Press, Pennsylvania, USA. Anderson M. K. 2003. Lupinus perennis L. plant guide. USDA, NRCS, National Plant Data Center c/o Department of Plant Sciences, University of Califor nia, Davis, California http://plants.usda.gov/plantguide/pdf/cs_lupe3.pdf Anderson, D. R. 2008. Model based inference in the life sciences: a primer on evidence. Springer Science + Busines s Media, LLC. Barnes, J.K. 2003. Natural History of the Albany Pine Bush. New York State Education Department, New York, USA. Battisti, A., M. Bernardi, and C. Ghiraldo. 2000. Predation by the hoopoe (Upupa epops) on pupae of Thaumetopoea pityocampa and th e likely influence on other natural enemies. Biocontrol 45:311 323. Bauerfeind, S. S., A. Theisen, and K. Fischer. 2009. Patch occupancy in the endangered butterfly Lycaena helle in a fragmented landscape: effects of habitat quality, patch size and isolati on. Journal of Insect Conservation 13 :271 277 Bernays, E. A. and C. B. Montllor. 1989. Aposematism of Uresiphita reversalis larvae (Pyralidae). Journal of the Lepidopterists' Society 43 :261 273. Borgerding E A G. A. Bartelt and W. M. McCown. 1995. The future of pine barrens in northwest Wisconsin: a workshop summary. PUBL RS 913 94. Wisconsin Department of N atural Resources, Madison Bried, J.T., J.E. Murtaugh, and A.M. Dillon. Local distribution factors and sampling effort guidelines for the rare frost ed elfin butterfly. Northeastern Naturalist 19 :673 684. Bush, G. L. 1969. Sympatric host race formation and speciation in frugivorous flies of genus Rhagoletis (Diptera : Tephritidae). Evolution 23 :237 251. Busse, M. D., C. J. Shestak, K. R. Hubbert, and E. E. Knapp. 2010. Soil Physical Properties Regulate Lethal Heating during Burning of Woody Residues. Soil Science Society of America Journal 74 :947 955.
184 Calhoun, J. V. 2004. Histoire Generale et Iconographie des Lepidopteres et des Chenilles de l'Amerique S eptentrionale by Boisduval & Le Conte (1829 (1837)): Original drawings used for the engraved plates and the true identities of four figured taxa. Journal of the Lepidopterists' Society 58 Chown, S.L. and S.W. Nicolson. Insect physiological ecology. 2004. Oxford University Press Inc., New York. Christian, K. A. and S. R. Morton. 1992. Extreme thermophilia in a central Australian ant, Melophorus bagoti. Physiological Zoology 65:885 905. Connell, J. H. 1978. Diversity in tropical rain forests and coral reefs. Science 199 :1302 1310 Cook, J. H. 1906. S tudies in the genus Incisalia The Canadian Entomologist 38 :141 144. Craig T.P, and J.K. Itami. 2008. Evolution of Preference and Performance tionary ey Jean Tilmon. University of California Press, Berkley, California. Davis, A. J., L. S. Jenkinson, J. H. Lawton, B. Shorrocks, and S. Wood. 1998. Making mistakes when predicting shifts in species range in r esponse to global warming. Nature 391 :783 786. DeMatteo, K. E. and B. A. Loiselle. 2008. New data on the status and distribution of the bush dog (Speothos venaticus): Evaluating its quality of protection and directing research efforts. Biological Conservat ion 141 :2494 2505. Dethier, V.G. 1959a. Foodplant distributions and larval dispersal as factors affecting insect populations. Canadian Entomologist. 91 :581 596. Dethier, V.G. 1959b. Egg laying habits of Lepidoptera in ration to available food. Canadian Ent omologist. 91 :554 561. Dres, M. and J. Mallet. 2002. Host races in plant feeding insects and their importance in sympatric speciation. Philosophical Transactions of the Royal Society B Biological Sciences 357 :471 492. Dimou, I., C. Koutsikopoulos, A. P. Ec onomopoulos, and J. Lykakis. 2003. Depth of pupation of the wild olive fruit fly, Bactrocera (Dacus) oleae (Gmel.) (Dipt., Tephritidae), as affected by soil abiotic factors. Journal of Applied Entomology 127 :12 17. Elith, J., C. H. Graham, R. P. Anderson, M. Dudik, S. Ferrier, A. Guisan, R. J. Hijmans, F. Huettmann, J. R. Leathwick, A. Lehmann, J. Li, L. G. Lohmann, B. A. Loiselle, G. Manion, C. Moritz, M. Nakamura, Y. Nakazawa, J. M. Overton, A. T. Peterson,
185 S. J. Phillips, K. Richardson, R. Scachetti Pere ira, R. E. Schapire, J. Soberon, S. Williams, M. S. Wisz, and N. E. Zimmermann. 2006. Novel methods improve prediction of species' distributions from occurrence data. Ecography 29 :129 151. Fielding, A. H. and J. F. Bell. 1997. A review of methods for the a ssessment of prediction errors in conservation presence/absence models. Environmental Conservation 24 :38 49. Ford, W. M., M. A. Menzel, D. W. McGill, J. Laerm, and T. S. McCay. 1999. Effects of a community restoration fire on small mammals and herpetofauna in the southern Appalachians. Forest Ecology and Management 114 :233 243. Frye, J. A. 2012. The effect of deer browse on sundial lupine: implications for frosted elfins. Northeastern Naturalist 19 :421 430. Fye, R. E. and R. L. Carranza. 1973. Cotton pests: overwintering of three lepidopterous species in Arizona. Journal Econ Ent 66:657 659. Fye, R. E. 1978. Pupation preferences of bollworms, tobacco budworms, and beet armyworms and impact on mortality resulting from cultivation of irrigated cotton. Journal of Economic Entomology 71:570 572. Garrison, E.G. 2003. Techniques in Archaeological Geology. Springer Verlag, Berlin. Gatrelle, R.N. 1991. The taxonomic implications of the discovery of Incisalia irus in 33 :57 58. Gatrelle, R. R. 1999. An evolutionary subspecific assessment of Deciduphagus henrici (Lycaenidae) based on its utilization of Ilex and non Ilex hosts: description of a third Ilex associated subspecies. Designation of a neotype and type locality for Dec iduphagus irus. Taxonomic Report of the International Lepidoptera Survey 1 :1 10, 11A 14A. Grigore, M. T. and E. J. Tramer. 1996. The short term effect of fire on Lupinus perennis (L). Natural Areas Journal 16 :41 48. Godart, J.P. 1824. Encylopedie Methoidiq ue. Histoire Naturelle. 9 : 674. Golden, D. M., and L. Pettigrew. 2005. Frosted Elfin Management Plan. New Jersey Department of Environmental Protection: Division of Fish and Wildlife, Trenton, New Jersey. Grime, J. P. 1973. Competitive exclusion in herbace ous vegetation. Nature 242 : 344 347. Grundel, R., N. B. Pavlovic, and C. L. Sulzman. 1998. The effect of canopy cover and seasonal change on host plant quality for the endangered Karner blue butterfly (Lycaeides melissa samuelis). Oecologia 114 :243 250.
186 Gr undel, R., N. B. Pavlovic, and C. L. Sulzman. 1998. Habitat use by the endangered Karner blue butterfly in oak woodlands: the influence of canopy cover. Biological Conservation 85 :47 53. Guisan, A. and W. Thuiller. 2005. Predicting species distribution: of fering more than simple habitat models. Ecology Letters 8 :993 1009. Hester, T.R, H.J. Shafer, and K.L. Feder. 1997. Field Methods in Archaeology. Mayfield Pusblishing Company, California. Hijmans, R. J., S. E. Cameron, J. L. Parra, P. G. Jones, and A. Jarv is. 2005. Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology 25 :1965 1978. Hijmans, R. J. and C. H. Graham. 2006. The ability of climate envelope models to predict the effect of climate change on species distributions. Global Change Biology 12 :2272 2281. Hutchinson, G. E. 1959. Homage to Santa Rosalia, or why are there so many kinds of animals? American Naturalist. 93 :145 159. Jaenike, J. 1978. On optimal oviposition behavior in phytophagous insec st. Theoretical Population Biology 14 :350 356. Kearney, M. and W. P. Porter. 2004. Mapping the fundamental niche: Physiology, climate, and the distribution of a nocturnal lizard. Ecology 85 :3119 3131. Klots, A. B. 1951. A field guide to the butterflies of North America, east of the Great Plains. Houghton Mifflin, Boston. Lentz, D. L., R. Bye, and V. Sanchez Cordero. 2008. Ecological niche modeling and distribution of wild sunflower (Helianthus annuus L.) in Mexico. International Journal of Plant Sciences 16 9 :541 549. Mayr, E. 1942 Systematics and origin of species. New York: Columbia University Press. Meyer, Rachelle. 2006. Lupinus perennis. In: Fire Effects Information System, [Online]. U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station, Fire Sciences Laboratory (Producer). Available: http://www.fs.fed.us/database/feis/ [2013, February 19]. Montoya, P., S. Flores, and J. Toledo. 2008. Effect Of Rainfall And Soil Moisture On Surv ival Of Adults And Immature Stages Of Anastrepha Ludens And A Obliqua (Diptera: Tephritidae) Under Semi Field Conditions. Florida Entomologist 91 :643 650. Mukherjee, A., R. Diaz, M. Thom, W. A. Overholt, and J. P. Cuda. 2012. Niche based prediction of esta blishment of biocontrol agents: an example with Gratiana
187 boliviana and tropical soda apple. Biocontrol Science and Technology 22 :447 461. Nagelkerke, N. J. D. 1991. A note on a general definition of the coefficient of determination. Biometrika 78:691 692. NatureServe. 2012. NatureServe Explorer: An online encyclopedia of life [web application]. Version 7.1. NatureServe, Arlington, Virginia. Available http://www.natureserve.org/explorer (Accessed: December 12, 2012 ). Neary, Daniel G.; Ryan, Kevin C.; DeBano, Leonard F., eds. 2005. ( revised 2008). Wildland fire in ecosystems: effects of fire on soils and water. Gen. Tech. Rep. RMRS GTR 42 vol.4. Ogden, UT: U.S. Department of Agriculture, Forest Service, Roc ky Mountain Research Station. 250 p. Nelson, M.W., 2002. Element abstract: Callophrys irus. Massachusetts Natural Heritage Program and Endangered Species Program, Division of Fisheries and Wildlife. Westborough, MA. Nuzzo VA (1986) Extent and status of midwest oak savanna: presettlement and 1985. Nat ural Areas J ournal 6 :6 36 Pagel, M. D., R. M. May, and A. R. Collie. 1991. Ecological aspects of the geographical distribution and diversity of mammalian species. American Naturalist 137 :791 815. Panzer, R. and M. Schwartz. 2000. Effects of management burning on prairie insect species richness within a system of small, highly fragmented reserves. Biological Conservation 96: 363 369. Panzer, R. 2002. Compatibility of prescribed burning with the conse rvation of insects in small, isolated prairie reserves. Conservation Biology 16 :1296 1307. Pavlovic, N. B. and R. Grundel. 2009. Reintroduction of Wild Lupine (Lupinus perennis L.) Depends on Variation in Canopy, Vegetation, and Litter Cover. Restoration E cology 17 :807 817. Pelham, J.P. 2008. A catalogue of the butterflies of the United States and Canada. The Joural of Research on the Lepidoptera. 40 :222. Pfitsch, W. A. and E. H. Williams. 2009. Habitat restoration for lupine and specialist butterflies. Res toration Ecology 17 :226 233. Phillips, S. J., R. P. Anderson, and R. E. Schapire. 2006. Maximum entropy modeling of species geographic distributions. Ecological Modelling 190 :231 259. Phillips, S. J. and M. Dudik. 2008. Modeling of species distributions wi th Maxent: new extensions and a comprehensive evaluation. Ecography 31 :161 175.
188 Possley, J., J. Maschinski, C. Rodriguez, and J. G. Dozier. 2009. Alternatives for Reintroducing a Rare Ecotone Species: Manually Thinned Forest Edge versus Restored Habitat Re mnant. Restoration Ecology 17 :668 677. Pulliam, H. R. 1988. Sources, sinks, and population regulation. American Naturalist 132 :652 661. Pulliam, H. R. 2000. On the relationship between niche and distribution. Ecology Letters 3 :349 361. Rabinowitz, D., S. C airns, and T. Dillon. 1986. Seven forms of rarity. Pages 182 204 in M. E. Soule, ed. Conservation biology. Sinauer, Sunderland, Mass. Russell, K. R., D. H. Van Lear, and D. C. Guynn, Jr. 1999. Prescribed fire effects on herpetofauna: review and management implications. Wildlife Society Bulletin 27 :374 384. Schweitzer, D.F. 1992. Incisalia irus revisited: a response to Reverend Ronald Gatrelle. 34 :69 70. Schweitzer, D.F., Minno, M.C., and D.L. Wagner. 2011. Rare, declining and poorly known butterflies and moths (Lepidoptera) of forests and woodlands in the eastern United States. U.S. Forest Service, Forest Health Technology Enterprise Team, FHTET 2011 01. Scudder, S.H. 1872. A Systematic Reivew of American Butterflies. Rep ort of the Peabody Academy of Science 32 Singer, M.C. 1972. Complex components of habitat suitability within a butterfly colony. Science 176 : 75 77. Swengel, A. B. and S. R. Swengel. 1996. Factors affecting abundance of adult Karner blues (Lycaeides melis sa samuelis) (Lepidoptera: Lycaenidae) in Wisconsin surveys 1987 95. Great Lakes Entomologist 29 :93 105. Smith, Jane Kapler, ed. 2000. Wildland fire in ecosystems: effects of fire on fauna. Gen. Tech. Rep. RMRS GTR 42 vol. 1. Ogden, UT: U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station. 83 p. Snyder, J. R., A. Herndon, and W. B. Robertson. 1990. South Florida Rockland. Pages 230 277 in R. L. Myers and J. J. Ewel, editors. Ecosystems of Florida. University of Central Florida Pre ss, Orlando. Swengel, A. B. 1996. Observations of Incisalia irus (Lepidoptera: Lycaenidae) in central Wisconsin 1988 95. Great Lakes Entomologist 29 :47 62. Swengel, A. B. and S. R. Swengel. 2007. Benefit of permanent non fire refugia for Lepidoptera conser vation in fire managed sites. Journal of Insect Conservation 11 :263 279.
189 Swengel, A. B. and S. R. Swengel. 2007. Benefit of permanent non fire refugia for Lepidoptera conservation in fire managed sites. Journal of Insect Conservation 11 :263 279. Swets, J. A. 1988. Measuring the accuracy of diagnostic systems Science 240 :1285 1293. Thuiller, W., O. Broennimann, G. Hughes, J. R. M. Alkemade, G. F. Midgley, and F. Corsi. 2006. Vulnerability of African mammals to anthropogenic climate change under conservative land transformation assumptions. Global Change Biology 12 :424 440. Trager, M. D., B. M. Boyd, J. C. Daniels, and J. A. Pence. 2009. Host Plant Selection, Larval Survival, and Reproductive Phenology in Megathymus yuccae (Lepidoptera: Hesperiidae). Environm ental Entomology 38 :1211 1218. Study of Insects. Brooks/Cole, Belmont, California. USDA, NRCS. 2013. The PLANTS Database ( http://plants .usda.gov 18 January 2013). National Plant Data Team, Greensboro, NC 27401 4901 USA. USFWS. 2012. Karner Blue Butterfly ( Lycaeides melissa samuelis ) 5 Year Review: Summary and Evaluation. United States Fish and Wildlife Service, Ecological Services Field Office, New Franken, Wisconsin. Vogel, J. A., D. M. Debinski, R. R. Koford, and J. R. Miller. 2007. Butterflal, responses to prairie restoration through fire and grazing. Biological Conservation 140 :78 90. Vogl, R.J. 1974. Effect of fire o n grasslands. I n: Kozlowski TT, Ahlgren CE (eds) Fire and ecosystems. Academic Press, New York Waltz, A. E. M. and W. W. Covington. 2004. Ecological restoration treatments increase butterfly richness and abundance: Mechanisms of response. Restoration Ecology 12 :85 96. W helan, R. J. 1995. The ecology of fire. Cambridge University Press, Cambridge ; New York. WI DNR. 2010. Wisconsin Statewide Karner Blue Habitat Conservation Plan ( http://www .fws.gov/midwest/endangered/permits/hcp/kbb_wi/index.html 3 May 2013). PUBL SS 947 2010 Rev. Zheng, X. L., X. P. Cong, X. P. Wang, and C. L. Lei. 2011. Pupation behaviour, depth, and site of Spodoptera exigua. Bulletin of Insectology 64 :209 214.
190 BIOGRA PHICAL SKETCH Matt received his PhD. in entomology with a minor in wildlife ecology and c onservation at the University of Florida in the summer of 2013. Prior to enrolling at UF, he worked as a teaching assistant in several middle schools, specializing in helping students with a wide range of developmental or behavioral issues Matt received his Bachelor of Science in b iology at the University of Oregon in the s ummer of 2006 He conducted undergra duate research on social spider ecology in Amazonian Ecuador and physiological ecology of porcelain crabs at the Oregon Institute of Marine Biology in Charleston, Oregon He is interested in researching and communicating invertebrate conservation and restoration ecology, biodiversity, and life history evolution.