1 ROAD NETWORK DEVELOPMENT AND LANDSCAPE DYNAMICS IN THE SANTA FE RIVER WATERSHED, NORTH -CENTRAL FLORIDA, 1975 TO 2005 By ALISA W. COFFIN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009
2 2009 Alisa W. Coffin
3 To my husband Oswaldo and my parents
4 ACKNOWLEDGMENTS The road that I have traveled to complete this dissertation was long and certainly not conventional Like the road s of the Santa Fe River watershed, my journey has involved several blind turns and dead ends. However, years ago, when I became determined to study road networks and their ecological effects, a few people agreed that this was a very important subject and they sanctioned my wayfaring. I am very grateful to them Among them, my supervisory committee C hair, Michael Binford, has provided me with insightful and timely guidance especially during the development of these manuscripts I ts members, Timothy Fik, Joann Mossa and Francis Putz support ed me with enthusiasm and confidence over the long years of this dissertation research. A few individuals perhaps unknowingly, pl ayed key roles at various stages through their thoughtful suggestions and insights Among them were Richard Forman who pointed me down the network analysis path several ye ars ago, and Jochen Jaeger and Evan Girvetz, who patiently responded to my questions about effective mesh size methodology. My progress and research were supported by many organizations throughout my career as a PhD student and I am sincerely grateful to them I thank the Suwannee River Water Management District, the Florida State Parks, the Alachua Conservation Trust Osceola National Forest and UFs Ordway Swisher Biological Reserve for giving me access to their land for field data collection In partic ular, I thank the staff at the Itchetucknee Springs State Park and OLeno State Park for their assistance. Among the institutions that supported this research with awards of financial aid at the University of Florida were: the Graduate School and the Depar tment of Geography through the J. Wayne Reitz Presidential Fellowship; the College of Liberal Arts and Sciences through the O. Ruth McQuown S cholarship; and the Graduate School Office of Minority Affairs through a Supplemental Retention S cholarship. The Association of American
5 Geographers provided me with a dissertation grant; which allowed me to purchase much needed field equipment. In addition to these institutional forms of assistance I was very fortunate to benefit from the good will of my employers, who, geographers themselves, sanctioned my use of their labs and computing systems for my dissertation work during off -hours In particular, I am very grateful to Ann Foster, of the U.S. Geological Survey (USGS) for allowing me to use the GIS lab at USGS during the four and a half years that I worked there. More recently, Robert Swett and Charles Sidman of Florida Sea Grant were very generous, providing me with work space and access to computing resources while employed there. More importantly, they allowe d me to borrow time from Corina Guevara a Sea Grant GIS technician and graduate student, to assist with field and technical work Corina and I spent many hot, humid days making our way through dense vegetation of north central Florida both amazed and h umored by the things we discovered here in our own backyard. Who knew that a jungle experience awaited us behind the WalMart in Starke? Thanks to Corinas experience as a field biologist in the Petn, we never missed a beat, and I am confident that our fie ld data collection was first rate. Corinas competence extended to the lab where she patiently assisted with digitizing and editing the road network datasets. In addition to my colleagues at work, I thank the friends and fellow grad students who pro vided m e with moral support, good advice and help on this journey. I am grateful to Matt Marsi k for preprocessing several of the satellite images used in my analysis, and helping me understand the steps involved. Katerie Gladdys accompanied me on several field tr ips and kept the conversation very interesting as we searched for the intersection s of science and art amongst the turkey oaks and loblolly bays Kristen Conway, who preceded me in the Department of
6 Geography, was one of the first graduate students I met a t UF and is a good friend who always provided a listening ear when juggling family, work and a PhD got to be too much. I am also thankful for the support and love of our parents. Mafalda Suarez Eg ez, was the best abuelita a daughter in law could want. She was a calming presence in our home, and helped over many months so that I could continue to advance my research My parents, Bea triz and Laurence Coffin, have provided me with unending confidence and love over the years The field work acc omplished for this dissertation is dedicated to them, as it was their keen interest in nature and their work in landscape architecture that inspired me to pursue an environmental education. My forays into the field were made possible thanks to their encour agement and material assistance. Finally, I am deeply grateful to my best friend, Oswaldo Rodriguez, who is also my spiritual and life partner, for his help, patience and love through all these years of hard work. During the course of this research, we welcomed Beni and Noah into our lives, who joined their brothers Jorge and Jasper in our happy and boisterous Familia Rodriguez. Throughout a ll, Oswaldo has been a constant compan ion, an excellent father, and filled my li fe with joy. The challenges of a PhD can be formidable, and taking it on while raising children might be insane, if it werent for the fact that nurturing a child in a loving home is one of the sanest things that a human being can do My sons have, through it all, kept me from wandering astray in back roads of academic frivolity and kept my mind focused on the goal of finishing. I thank them for their love, smiles and sacrifices as they accompanied me on this fascin ating journey.
7 TABLE OF CONTENTS P age ACKNOWLEDGMENTS .................................................................................................................... 4 LIST OF TABLES ................................................................................................................................ 9 LIST OF FIGURES ............................................................................................................................ 11 LIST OF OBJECTS ............................................................................................................................ 13 ABSTRACT ........................................................................................................................................ 14 CHAPTER 1 INTRODUCTION ....................................................................................................................... 16 Study Area ................................................................................................................................... 19 Patterns of Road Networks and Landscape Dynamics in the Santa Fe River Watershed ...... 22 2 FROM ROADKILL TO ROAD ECOLOGY: A REVIEW OF THE ECOLOGICAL EFFECTS OF ROADS ............................................................................................................... 29 The Effects of Roads on Abiotic Components of Ecosystems ................................................. 32 Changes to Hydrology and Wat er Quality ......................................................................... 33 Erosion and sediment transport ................................................................................... 33 Introduction of chemical pollutants ............................................................................ 34 Noise and Other Atmospheric Effects ................................................................................ 36 The Effects of Roads on Biotic Components of Ecosystems ................................................... 37 Roads as Sources of Mortality and Barriers to Animal Movement .................................. 38 Roads as Habitat, Corridor and Conduit ............................................................................ 41 Ecological Effects of Road Networks ........................................................................................ 43 Landscape Change and Fragmentation ............................................................................... 44 Road Edges and the Road -Effect Zone ........................................................................... 46 Ecological Road Network Theory ...................................................................................... 47 Conclusion ................................................................................................................................... 49 3 A NETWORK ANALYSIS OF ROADS IN THE SANTA FE RIVER WATERSHED FROM 1975 THROUGH 2005 .................................................................................................. 50 Introduction ................................................................................................................................. 50 Road Network Development and Analysis ............................................................................... 51 Map Based Measures of Road Networks ........................................................................... 53 Graph Theoretic Methods for Road Network Analysis .................................................... 53 Measures of the entire network ................................................................................... 55 Measures of network elements .................................................................................... 57
8 Methods ....................................................................................................................................... 59 Study Area ............................................................................................................................ 59 Data Development ............................................................................................................... 61 Data Analysis ....................................................................................................................... 63 Results .......................................................................................................................................... 66 Extent .................................................................................................................................... 66 Conn ectivity ......................................................................................................................... 68 Accessibility ......................................................................................................................... 69 Discussion .................................................................................................................................... 71 Connectivity and Funct ional Capacity of the Network ..................................................... 73 Accessibility of the Landscape vs. Accessibility of the Road Network ........................... 75 General Remarks and Further Research ............................................................................. 76 Conclusions ................................................................................................................................. 77 4 LANDSCAPE DYNAMICS IN THE SANTA FE RIVER WATERSHED FROM 1975 THROUGH 2005 ........................................................................................................................ 97 Introduction ................................................................................................................................. 97 Study Area ................................................................................................................................. 102 Methods ..................................................................................................................................... 104 Results ........................................................................................................................................ 111 Discussion and Conclusio ns ..................................................................................................... 116 5 SYNTHESIS .............................................................................................................................. 137 APPENDIX A DETAILED DESCRIPTIONS OF ROAD NETWORK METRICS ..................................... 146 Detailed Descriptions of Map -Based Road Network Metrics ................................................ 146 Detailed Descriptions of Graph Theoretic Based Road Network Metrics ............................ 148 B TRAINING SAMPLE DATA SHEET .................................................................................... 152 C FIELD NOTES DATABASE ................................................................................................... 154 D EFFECTIVE MESH SIZES, IN HECTARES, FOR SUB BASINS OF THE SFRW ......... 155 LIST OF REFERENCES ................................................................................................................. 158 BIOGRAPHICAL SKETCH ........................................................................................................... 174
9 LIST OF TABLES Table P age 3 1 M easures and indices of road networks. ............................................................................... 88 3 2 Basic measures of graphs commonly used to describe transportation networks (Garrison and Marble, 1962; Kansky, 1963; Cliff e t al., 1979). ......................................... 89 3 3 Alpha and gamma index ranges for basic network configurations (Taaffe and Gauthier, 1973). ...................................................................................................................... 89 3 4 1975 to 2005 population estimates for the top five counties (by area) in the SFRW based on 1970 to 2000 census data (U. S. Census Bureau, 2008). ..................................... 89 3 5 Average annual population growth rate (percent) for the top five counties (by area) in the SFRW based on 1970 to 2007 estimates. ....................................................................... 90 3 7 Topological rules from ArcGIS used to verify and correct the SFRW road networks. ..... 91 3 8 Measures of road network extent in the SFRW. .................................................................. 91 3 9 Percent change matrix for measures of road network extent: Length (L), vertices ( v ) and edges ( e ). .......................................................................................................................... 92 3 10 Percent change matrix for measures of road network extent: diameter ( ), average path length (APL), diameter path length ( d ) and shape index ( ). ...................................... 92 3 11 Percent change matrix for measures of road network extent: network weight ( w ), proportion of first -order vertices ( vo=1:vo>1), and mean vertex order (MVO) .................... 93 3 12 Measures of road network connectivity and structure. ........................................................ 93 3 13 Percent change matrix for measures of road network connectivity and structure. ............ 94 3 14 Measures of landscape accessibility. .................................................................................... 94 3 15 Measures of road network accessibility. ............................................................................... 95 3 16 Percent change matrix for measures of landscape accessibility: road density ( L/A ), roadless volume ( RV ), mean distance to road ( Mean DTR ), and maximum distance to road (Max DTR). ................................................................................................................ 95 3 17 Percent change matrix for measures of shortest path accessibility in the road network: mean shortest -path accessibility ( Mean Ai), and minimum shortest -path accessibility (Min Ai). ............................................................................................................ 96 3 18 Percent change matrix for measures of mean vertex betweenness ( MVB ), mean edge betweenness ( MEB ), and mean closeness ( CLO ). ............................................................... 96
10 4 1 Satellite, sensor, acquisition dates and pixel size of remotely sensed images (WRS2, Path 17, Row 39) used for land cover cl assification. ......................................................... 133 4 2 Area of landcover type in SFRW (in hectares). ................................................................ 133 4 3 Percent change matrix for change in area of land -cover type for the SFRW. .................. 133 4 4 Effective mesh size (ha) and patch number, size (ha) and area information reported by fragmentation geometry (FG) and year. ........................................................................ 134 4 5 Percent change matrix of effective mesh size for the SFRW basin for the two fragmentation geometries (FG). .......................................................................................... 134 4 6 Contingency table summarizing relative changes in FG1 and FG2 in sub-basins of the SFRW, categorized by percent change in effective mesh size. ................................... 135 4 7 4 6. ........................................................................................................................................ 136
11 LIST OF FIGURES Figure P age 1 1 The study area: the Santa Fe River watershed in the southern reaches of the Suwannee River basin. ........................................................................................................... 25 1 2 Artesian spring located in Itchetucknee Springs State Park. ............................................... 26 1 3 Longleaf pine forest and diverse ground cover in Olustee Battlefield State Park. ............ 26 1 4 Bald cypress forest along the Itchetucknee Springs run. ..................................................... 27 1 5 Remnant cat -face scars from sap collection on a pine tree stump at the Lake Alto North conservation land. ........................................................................................................ 27 1 6 Hay fields in San Felasco Hammock State Preserve. .......................................................... 28 3 1 The study area: the Santa Fe River watershed in the southern reaches of the Suwannee River basin. ........................................................................................................... 80 3 2 A planar graph with five vertices and eight edges. .............................................................. 81 3 3 Roads in the SFRW study area.. ............................................................................................ 81 3 4 The General Highway Map o f Bradford County, Florida.. ................................................. 82 3 5 Alachua County 1973 Highway Map superimposed over the mosaic of aerial photography used as a base map for georeferencing. .......................................................... 82 3 6 The 1975 digitized road network for the SFRW. ................................................................. 83 3 7 SFRW road network showing diameter path(s) in red and basin outline in blue. ............. 84 3 8 Distance to road values.. ....................................................................................................... 85 3 9 Map showing a small area of the SFRW road network in 1975 (black) with accumulated changes in 2005 (red). ...................................................................................... 86 3 10 Idealized process of network development with decrease in connectivity resulting from addition of first -order vertices. ..................................................................................... 87 4 1 The study area: the Santa Fe River watershed in the southern reaches of the Suwa nnee River basin. ......................................................................................................... 123 4 2 Field data collection in June and July 2007. ....................................................................... 124 4 3 Sub -basins of the SFRW. ..................................................................................................... 125
12 4 4 Landsat TM images of the SFRW and corresponding land cover classification in forest (green) and non -forest (gold). ................................................................................... 126 4 5 Effectiv e mesh sizes in hectares for SFRW sub basins. .................................................... 128 4 6 Change in effective mesh size of SFRW sub basins ordered by greatest increase to greatest decrease in FG2meff. ............................................................................................... 130 4 7 Percent chan ge in effective mesh size in SFRW sub basins from 1975 to 2005. ............ 132
13 LIST OF OBJECTS Object P age C1 Field notes database ............................................................................................................. 154
14 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 ROAD NETWORK DEVELOPMENT AND LANDSCAPE DYNAMICS IN THE SANTA FE RIVER WATERSHED, NORTH -CENTRAL FLORIDA, 1975 TO 2005 By Alisa W. Coffin May 2009 Chair: Michael W. Binford Major: Geography Roads are known to have numerous ecological effects. It is postulated that networks of roads have cumulative ecological effects that are unaccounted for in the analysis of single road effects. Common ways researchers account f or road networks include the calculation of road density or distance to road metrics. Transportation network analysis provides an alternative, quantitative measure of road networks that also provides information about the structure and function of the road network. Road networks in the Santa Fe River Watershed of north central Florida were studied using network analysis to understand how the network has changed from 1975 to 2005. This investigation was coupled with a parallel analysis of landscape fragmenta tion in the same region. As low -order rural roads were added to the network, the network expanded but the overall network connectivity decreased. While accessibility of the landscape increased over time, accessibility of the road network declined. The landscape became more fragmented as effective mesh size, a connectivity -based measure of fragmentation, decreased over time. Fragmentation rates were greater during the final decade of the study, 19952005, corresponding to a dramatic increase in road network extent. The changes in the watershed signal a shift from
15 agricultural expansion, predominant in 19751985, to suburbanization. Watershed sub-basins located along highways linking regional urban centers experienced the greatest increases in fragmentation. T he addition of dead-end roads was an important change in the road network structure, particularly during the second decade(19851995). This class of roads has specific characteristics and potential ecological effects which may not be fully accounted for in the effective mesh size metric. The use of network analysis can provide additional information about how roads fragment landscapes and the dynamics of landscape change caused by road networks.
16 CHAPTER 1 INTRODUCTION For several decades now, some sectors of society have called into question the paradigm of unfettered economic growth, pointing out that there are costs or externalities associated with continual growth and that these costs are paid in ecological and social capital (Meadows et al. 1974; World Commission on Environment and Development 1987) Ecological capital includes the natural resource s and environmental services upon which all life depends. Clean water, biological diversity, nutrient cycling an d climatic systems are among these critical resources and services and are vulnerable when land cover and land use change spurred on by econom ic growth (Wilson 1988; Karl and Trenberth 2003) In the pursuit of valuable commodities, human beings have transformed landscapes, affecting most of the ice -free parts of the planet (Vitousek et al. 1997) The ramifications of such extensive influence are now being observed in the loss of biological diversity and changes in regional and global climate s w ith serious implications for humanitys future (Ehrlich 1988; Milleniu m Ecosystem Assessment 2005) Humanitys ability to so comprehensively affect planetary systems is a testament to the reach of human activity, which is dependent in one way or another on our capacity to move and access resources. The movement of people, goods and ideas is the fundamental idea of transportation (Lowe and Moryadas 1975) Transportation is a critical element in economic development. Without the existence of transportation systems that can move goods and people regions do not develop economically (Black 2003; Puu 2003) Until the 20th century, ideas and information also depended on physical transportation systems. With the advent of electronic and then wireless communications technology, this rela tionship is less direct, although communications infrastructure is oftentimes co -located with transportation routes and still dependent on the public spaces associated with transportation infrastructure (Gorman 1999) In short, economic growth and political power
17 remain tightly linked to the developm ent of safe and secure transportation routes. For this reason, road development projects have, throughout history, garnered a large share of attention, resources and spending from central governments, from the prehistoric Inca and Romans to modern nations (Jackson 1984; Denevan 1991) Wherever road building becomes difficult, in steep terrain or across water for example, more resources, engineer ing and greater political capital are needed to create the road There are numerous examples of roads built in very unlikely places at great cost, as well as roads left unfinished, often due to changing economic and political conditions that suddenly reve aled them as exorbitant (Hayasaki 2008) Road networks are a proximate cause of land cover change (Geist and Lambin 2001) driving the transformation of land by lowering the cost of basic economic activities, particularly for agriculture, logging and hunting. In more sophisticated economies road networks facilitate the transformation of land for industrial u ses by lowering the cost of manufacturing and distributing mass produced goods for human consumption (Lowe and Moryadas 1975) In a spiraling feedback system land -cover changes with land use intensification, placing more demand on the transportation network. The system in turn develops, inducing further intensification and more demand (Vuchic 1999) While road networks are the physical manifestation of human social connections, they are also persisten t features in the landscape with substantial ecological ramifications (Forman and Alexander 1998) Their design and regulation are brought about through engineering and political organization, but their development both affects and depends on the physical context where they are located. Road networks are known to have a variety of ecological effects that change as the road network grows (Forman et al. 2003)
18 Numerous reviews of the ecological effects of roads have been published (Andrews 1990; Forman and Alexander 1998; Jones et al. 2000; Carr et al. 2002; Evink 2002; Forman et al. 2003; Coffin 2007) Briefly, roads consume land and facilitate land cover change. They affect living systems by fragmenting habitat and populations of plants and animals, thereby under mining population dynamics. They affect landscape processes such as hydrologic flows by channeling water, and they can act as fire breaks. Roads provide habitat for many species as well. Some of these are exotic invasive species, which supplant native spec ies and disrupt ecological relationships. Road networks can also provide for the transport of pathogens, implying a role as facilitators of disease spread. Roads and their traffic affect ecosystems by introducing chemical pollutants into the environment. R oadway traffic is a significant source of mortality and a cause of behavioral modification in many species of animals. Perhaps most notably, road networks provide access to the landscape for people to do what they will. Although landscape access for recre ation and social reasons are important facets of human culture, most often people access the landscape in pursuit of tangible resource s These resources may be bush -meat, timber, agricultural products, or minerals. Landscape access may also be sought for s peculative r easons, such as prime location s for industrial or urban land use s In their pursuit to add value to the extractable goods or the land itself, people further develop roads by paving, widening, realigning, etc., providing better accessibility, ov ercoming conditions that would otherwise constrain transportation (Lowe and Moryadas 1975). The concomitant modification or loss of habitat is often dismissed as necessary in the interest of progress. By virtue of their structure and function, road networks are critical spaces where the effects of human activities are registered profoundly in ecosystems. As such, their careful planning and management can provide opportunities for fundamentally affecting how human s interact with
19 biological systems. For this reason, road networks are important subjects for ecological and socio -economic resear ch. The ecological interaction of road networks with the landscapes they penetrate is the main focus of road ecology. Road ecology is a relatively new interdisciplinary endeavor that seeks to bring together scientists from both transportation and ecology t o explore basic questions about the ecological effects of roads (Forman et al. 2003) This dissertation research contributes to road ecology by deepening our understanding about how road networks interact with the environment. For this research, I drew from lands cape ecology and transportation geography theory to explore questions about the development of a rural road network, the dynamics of land-cover change and road related landsca pe fragmentation within a north -central Florida watershed. The research program involved several phases the results of which have become the chapters of this dissertation. The first step was to review the literature of transportation geography, landscape ecology and road ecology. Next, I developed research questions and chose a stud y area for the project. The research questions focused on how the road network and the associated landscape changed over time. The research involved using methods from transportation network analysis to analyze changes to the road networks from 1975 to 2005. On a complementary track, I applied methods and techniques in land -cover change science and landscape ecology to study landscape dynamics in the watershed. Study A rea The study focused on the Santa Fe River watershed in north central Florida (SFRW ; Fig ure 1 1 ). I compared landscape and road network dynamics over a thirty year period (19752005. The Santa Fe River is a tributary of the Suwannee River, and the SFRW is within the southern reaches of the Suwannee River Basin which extends north into Georgia The watershed area is
20 approximately 3584 km2 and covers almost all of Bradford and Union Counties and major portions of Alachua, Gilchrist and Columbia Counties The study region is underlain by limestone, giving this buriedkarst landscape the unusual characteristic of having rapid and substantial sur face -groundwater interactions. One of the more notable features of the Santa Fe River is its disappearance below ground fo r several kilometers in OLeno State Park at the boundary Alachua and Columbi a Counties The river flows west from central Florida to disc harge into the Suwannee River. The Santa Fe and Suwannee River corridors comprise a famously unique North Florida landscape of forest clearings with first magnitude artesian springs that boil to the surface and then run to the rivers (Figure 1 2; Carr 1994) The unusu ally high quality of the water and the constant temperature of the springs make them unique ecosystems and valuable resources to a variety of plant and animal species (Nordlie 1990) The extensive network of underground caverns and si nkholes play host to rare endemic arthropods and fish, and are considered world-class cave -diving sites. They constitute rare and fragile ecosystems embedded in the landscape, and are vulnerable to threats of pollution brought about by urbanization and lan d use intensification in the drainage basin. In fact, nitrate levels in the Suwannee River Basin have been increasing over the last two decades (Upchurch et al. 2007) In the SFRW nitrate levels have been increasing due in part to both instensive agricultural practices (Sabesan 2005) and increasing urbanization. Prior to agricultural intensification, this region was dominated by a mosaic of forested land -cover types (Ware et al. 1993) Subtle variations in topography, soils, hydrology and fire regime create d a mosaic of vegetation types ranging from relatively open pine and oak woodlands to dense r broadleaf wooded swamps T he upland native vegetation of this area was predominantly flatw oods consisting of pine savanna s These patterns are evident today in
21 protected areas, where la nd has reverted to less intensive uses. T hese fire -adapted ecosystems are dominated by more -or less open canopies of longleaf pine ( Pinus palustris ), slash pine ( P. elliotti) and pond pine (P. serotina ), depending on hydroperiod and fire regimen (Abrahamson and Hartnett 1990) These pine forests are interspersed with hammocks of oak s and hickories (Platt and Schwartz 1990) The open canopy of the pine and oak savanna s allows light to reach the ground, where a highly diverse assemblage of shrubs, grasses and forbs thrive (Figure 1 3) In low lying areas and palustrine environments protected from fire the vegetation is dominated by mesic hardwood hammocks and swamps. The vegetative characteristics of the swamps depend on hydroperiod, fire frequency, organic matter accumulation and the source of the water (Ewel 1990b) Bald cypress ( Taxodium distichum ) trees cover the riverbanks along the Santa Fe River (Figure 1 4). The organic leachates, or tannins, fro m the swamps of North Florida give the water a dark coloration throughout the entire Suwannee River Basin. Hardwoods and other coniferous trees such as pond cypress ( Taxodium ascendens ), swamp tupelo ( Nyssa bicolor ) and loblolly bay (Gordonia lasianthus ) a re also found in forested wetlands within the upland matrix where shallow groundwater is expressed on the surface for long periods The study area, which is a short distance from the Spanish colonial city of Saint Augustine, has experienced continuous huma n occupation for millennia: first by nomadic Paleoindians and subsequent aboriginal cultures (Milanich and Borremans 1990; Milanich 1990) ; and most recently by the descendants of European immigrants and African slaves (Gannon 2003) One of the first transit systems to cross Florida ran through the SFRW: the Florida Atlantic and Gulf Central Railroad; and alongside it, the Lake City to Jacksonville highway, now U.S. Route 90. This railway was the objective of the Civil War Battle of Olustee, on February 20, 1864, while the highway was the thoroughfare for the troops (Gannon 2003) With European settlement, the
22 region was used extensively for timber and naval stores production (Stout and Marion 1993) Vestiges of the turpentine business can occasionally be seen in todays natural forests as the characteristic cat -fa ce scars on the resin -soaked stumps of pine trees whose tops have long since decayed (Figure 1 5). Today, land use in the SFRW is still dedicated mostly to forestry and agriculture with timber companies owning large tracts of land in the region. Pine fore sts and agriculture are the most common land cover types, and hay is the most commonly produced agricultural crop (Figure 1 6; Pearlstine 2002) Transportation syst ems in the area consist of several now lightl y used railroads, interstate highways I 75 and I 10, and a netw ork of state and county roads connecting the regional urban centers of Gainesville, Alachua, Starke and Lake City. A fine mesh of minor roads p erva des most of the landscape. Most of the road network in this watershed is unpaved. Apart from the extensive commercial forestry operations, industrial operations includ e power plants, a ce ment plant, and several state and county prisons. Perhaps the most en vironmentally important industry is a titanium mine operated by Dupont de Nemours to the east of Starke. Patterns of Road Networks and Landscape Dynamics in the Santa Fe River Watershed The SFRW provided the location to study the characteristics of a rura l road network and the landscape it penetrates. In preparation, I carried out an extensive review of the literature in both transportation geography and road ecology. The road ecology section of the review became chapter two of this dissertation, and was published in the Journal of Transport Geography in 2007 (Coffin 2007) This chapter reviews the ecological effects of roads beginning with effects on the abiotic components of ecosys tems. Subsequent sections address the effects of roads on the biological components of ecosystems. Finally, the chapter points to possible contributions that transportation geographic theory can make to develop our knowledge about road networks.
23 Following this review, chapter three delves into a network analysis of roads in the SFRW from 1975 to 2005. This chapter characterizes how the road network changed over time. It begins with a brief overview of road network analysis describing the theory and metrics used in the analysis (Kansky 1963; Haggett and Chorley 1969; Taaffe and G authier 1973) This section is supplemented with a more detailed description in Appendix A. I analyzed the road network dataset as it changed over time. The results and discussion sections describe the outcome of the detailed network analysis and show t hat the network did not strictly follow expectations described by theories of transportation geography. In a complementary analysis, chapter four describes the landscape dynamics and landscape fragmentation in the SFRW, focusing on how the dynamics of roa d network development and land covers affected patterns of landscape fragmentation. In this study, I examined how roadrelated fragmentation patterns differed between forested and agricultural land -cover types in the SFRW. Using a nested hierarchical scale I explored how fragmentation varied across space and time, and how these two landcover types interacted as the road network evolved. Supporting field work yielded an extensive database of vegetation information and imagery from 265 field locations. The field data collection sheet and database are available in Appendices B and C of this dissertation, and the imagery is available from the author until a suitable web service can be found to host these images. Landcover classifications were created from fou r satellite images collected throughout the study period. Using this information, I applied current methods in road ecology (Jaeger 2000; Girvetz et al. 2008) to describe fragmentation of the land covers for each time period. By including two spatial scales and two temporal scales to measure fragmentation in two groups of land-cover types, I was able to develop a detailed picture of land-cover dynamics in the watershed, and how the road network has affected landscape connectivity there.
24 The final chapt er of this dissertation attempts to bring together these two different studies in a synthetic analysis of their implications. It examines the conclusions provided from the separate studies and presents conclusions about the how the road network affected la ndscape ecological processes in the in light of these two analyses. Finally, I highlight important gaps in road ecology research and remark on a recent exchange of views published in Science in 2007 (Girvetz et al. 2007; Watts et al. 2007a; Watts et al. 2007b) offering another perspective on the issues raised
25 Figure 1 1. The study area: the Santa Fe River watershed in the southern reaches of the Suwannee River basin.
26 Figure 1 2. Artesian spring located in Itchetucknee Springs State Park. Figure 1 3. Longleaf pine woodland with a hyper -diverse ground cover in Olustee Battlefield State Park.
27 Figure 1 4. Bald cypress forest along the Itchetucknee Springs run. Figure 1 5 Remnant cat -face scars from sap collection on a pine tree stump at the Lake Alto North conservation land.
28 Figure 1 6. Hay fields in San Felasco Hammock State Preserve.
29 CHAPTER 2 FROM ROADKILL TO ROAD ECOLOGY: A REVIEW OF THE ECOLOGICAL EFFECTS OF ROADS The idea that roads interact with the surrounding landscape is fundamental in Geography. Von Thnen's 1826 theory described how land use is a function, at least in part, of the cost of transport to markets (Wartenberg 1966) Ullman (1956) pointed out the importance of transportation in modifying the surface of the earth Transportation geographers showed the correlation between transportation network expansion and economic development of regions (Kansky 1963; Taaffe et al. 1963; Haggett 1965) Roads, in particular, are physical manifestations of social connections and the economic and politi cal decisions that lead to land use change. The debate over whether landscape transformation is a cause or an effect of road network development belies the complexity of the interactions between the social and biological realms which ultimately produce these networks. Their existence depends on social structures, and their physical characteristics depend partly on landscape structure. The environmental effects of transportation systems are of interest to transportation geographers, but are relegated to the margins of the field, leaving room for exploration. Transportation g eographers of the mid to late 20th century who examined the structure of transportation systems focused on their network properties, and their effects on land use, allocation, and competition between producers, manufacturers, distributors and consumers (Garrison and Marble 1962; Beckman 1967; Taaffe and Gauthier 1973; Lowe and Moryadas 1975) They produced a wealth of knowledge about the structure of transpo rtation networks and derived a number of quantitative tools for their study (Haggett and Chorley 1969) In this body of work, roadway systems were considered part of the required infrastructure for increasing productivity in a region, their physical structure a benign necessity in the promotion of progress. More recently, environmental topics have gar nered more attention from transportation
30 geographers, however, most of the discussion is focused on issues of sustainable transport and the quality of human life (Black 1989; Gordon 1991; Black 1996; Hunter et al. 1998) Little attention has been given by geographers to the unintended consequences of road networks, or how their expansion affects the landscapes that they bisect. Far from being anonymously inert features, roads and their traffic introduce pollutants and exotic elements, fragment populations of plants and animals, kill animals and cause behavioral changes (Forman et al. 2003) As early as the 1970s, wildlife biologists began publishing research on the effects of roads on wildlife populations as barriers to movement (Oxley et al. 1974; Wilkins 1982; Mader 1984; Mech et al. 1988; Brody and Pelton 1989; Mader et al. 1990; Develey and Stouffer 2001; Bhattacharya et al. 2003) sources of mortality (Bellis and Graves 1971; Wilkins and Schmidly 1980; Davies et al. 1987; Cristoffer 1991; Groot Bruinderink and Hazebroek 1996; Mumme et al. 2000; Main and Allen 2002; Smith and Dodd 2003; Dodd et al. 2004) and the cause of behavior modification (Rost and Bailey 1979; Van Dyke et al. 1986; Brody and Pelton 1989; Norling et al. 1992; Kerley et al. 2002; Tigas et al. 2002) Coincident with the development of landscape ecology and landscape scale analyses, attention has turned to the broader scale effects of landscape connectivity and habitat fragmentation, and, specifically, the effects of roads in fragmenting the landscape a nd interacting with landscape processes (Andrews 1990; Reed et al. 1996; Canters et al. 1997; Strittholt and DellaSala 2001; Heilman Jr. et al. 2002; Saunders et al. 2002; Bhattacharya et al. 2003; Hawbaker and Radeloff 2004) In recent years, interest in the ecological effects of roads on ecosystems and landscapes has increased, evidenced by a number of review papers published in scientific journals and edited volumes (Andrews 1990; Bennett 1991; Forman and Alexander 1998; Spellerberg 1998; Trombulak and Frissell 2000; Carr et al. 2002; Havlick 2002) With the clamor to review and
31 consolidate information about the ecological effects of roads, research into this field is surging forward at t he impetus of landscape ecologists and conservation biologists. One result of this attention has been to underscore the weaknesses of a landscape fragmentation paradigm which generally ignores anthropogenic causes such as land use intensification and urb anization (Laurance and Cochrane 2001) The term road ecology was coined by landscape ecologist Richard T.T. Forman in 1998 (Forman 1998) It refers to an emerging subject of ecological inv estigation building on the mounting evidence that roads are having dramatic effects on ecosystem components, processes and structures, and that the causes of these effects are as much related to engineering as to land use planning and transportation policy Road ecology is rooted in ecology, geography, engineering and planning. While research on the ecological effects of roads and transportation has been occurring in North America, Europe and Australia for some decades, the publication of the book Road Ecol ogy (Forman et al. 2003) heralded the consolidation of this endeavor at a new conceptual scale, under the auspices of an interdisciplinary scientific umbrella. The purpose of this paper is to present an overview of the literature describing the various ecological effects of roads and the development of road ecology as a body of scientific inquiry. The paper first gives a synopsis of the ecological effects of roads on the abiotic components of ecosystems including interactions with hydrologic systems, sediment eros ion and deposition dynamics, environmental chemistry, and ambient noise. The next section of the paper deals with the effects of roads on biota, from the direct effects of road related mortality, i.e. roadkill, to population fragmentation and road avoida nce behavior. This section also considers the fact that roads create habitat for many plants and animals. The final section of the paper attempts to address the more complex aspects of the cumulative ecological effects of roads on landscapes.
32 The most obvi ous of these is the fragmentation of landscapes as roads bisect large patches of a contiguous land cover. In addition to the fragmentation of the landscape caused by roads, however, are the cumulative ecological effects of roads when considered as networke d systems. Ecological road network theory suggests that these cumulative effects may be influenced by the design and function of the network structure. In this, transportation planners have an important role to play in being able to analyze and predict the potential ecological effects of alternative transport scenarios by using the tools that have already been developed by transportation geographers. Applying transportation geography theories and methods to research in road ecology could advance our underst anding about the dynamic between road systems and landscapes and help lessen the negative ecological effects of roads on the environment. The E ffects of Roads on A biotic Components of E cosystems Roads affect the abiotic components of landscapes including the hydrology, the mechanics of sediment and debris transport, water and air chemistry, microclimate and levels of noise, wind, and light adjacent to roadsides The extent and intensity of the effects vary with the position of the road relative to pattern s of slope, prevailing wi nds and surrounding land cover (Forman and Alexander 1998) Roads can increase the energy of stream systems, causing channel erosion and scouring on one hand; on the other hand, cut banks of roads near streams can cause sedimentation to occur. Either way, the presence of roads and related infrastructure has measurable effects on the morphology of stream and river channels which in turn affects the biota. Air and water pollution is one of the most often recognized environmental effects of roads. Toxic chemical s associated with air and water -borne particulates cause diseases and increased mortality in humans, and indeed, this aspect of transportation has been the focus of intense scrutiny by government researchers, regulators and lawmakers for several decades. However, the broader ecological effects of chemical pollution due to roadrelated transportation has been less
33 well -studied, although it is clear that toxins enter and persist in the environment and interact with biota. Changes to Hydrology and Water Q uality The nature of the interaction of roads with aquatic systems depends on their location relative to the drainage network and the slope Roads act as a source of water where water runs off the surface of the road They can serve as sinks for water, where water accumulates on roads (this is far less significant) Roads can act as barriers to water flowing downhill, but can also speed the removal of water (Jones et al. 2000) A t the landscape scale, road networks interact with stream networks, increasing the stream drainage density, the overall peak flow in the stream drainage, and the incidence of debris flows in the drainage basin (Jones et al. 2000) Roads extend the drainage network of the stream network when drainage swales along roads directly connect to stream networks (Forman and Alexander 1998) Faster moving water enters the stream channels increasing the energy of the strea m system, eroding channel banks, scouring the channel and can increase the likelihood of flooding downstream (Dunne and Leopold 1978) Erosion and sediment transport Roads are often associated with land uses that can, in tandem, cause changes to erosion and deposition rates of sediments in stream channels. Logging roads and logging are notable because forestry is commonly the first broad -scale land use causing the whole -sale anthropogenic removal of vegetation and exposure of soil in a watershed. The likelihood of mass movement of earth is higher following logging and floodplains experience overbank deposition following logging events in watersheds (Johnson et al. 2002) The sediment pulses throughout the stream basin and results in changes to the morphology of streams, depositing in channels and creating shallower pools. The shallowness of the pools, combined with increased turbidity of the water and less vegetated ban ks, raises the temperature of the water in the streams. This stresses fish
34 species that require colder temperatures, and favors other species that do not. Such a case was discovered in the Navarro watershed where logging practices and the associated roads created favorable conditions for the reproduction and growth of the common fish species California roach ( Lavinia symmetricus ), while stressing the steelhead trout ( Oncorhynchus mykiss ) and coho salmon ( Oncorhynchus kisutch). In the North Fork basin of the Navarro watershed, 100% of the sediment eroded from cut banks along a highway in close proximity to the main stream channel was delivered to the channel network (Johnson et al. 2002) Introduction of chemical pollutants Sources of chemical pollutants along roadsides include the vehicles that use the road as well as the roads and bridges themselves, and the maintenance activities associated with the roadway. Chemical spills along roads are also an important source of chemi cal pollutants ( U.S. Environmental Protection Agency 1996; U.S. Environmental Protection Agency 2001; Grant et al. 2003) Some chemicals affect only the areas nearest the road itself, while other chemicals are transported, via water or wind, greater distances from the road (Forman et al. 2003) Toxic contaminants from roads enter the broader landscape most importantly via stormwater runoff. The contaminants in runoff vary greatly in size, over six orders of magnitude, and include hydrated ions, dissolved, colloidal and gravitoidal particles, and suspended matter. This makes the research and assessment of their ecological effects diff icult, as a variety of tests must be used to analyze the different fractions of contaminants. Heavy metals and organic compounds are often adsorbed onto particles such as clay, silt and sand, associated with the road and roadbed. Many best management pract ices (BMPs) aimed at mitigating for chemical contaminants at the roadside are geared toward reducing the influx of particles into the surrounding landscape (Grant et al. 2003) The toxicity of contaminants depends on the way particulates affect organisms, such as altering the level of exposure to the toxi n. The
35 effectiveness of mitigation for chemical toxicity associated with roadway runoff depends on the extent to which contaminants associate themselves with particles that are removed by BMPs and the effectiveness of the BMPs (Grant et al. 2003) A complex and wide array of contaminants associated with v ehicles are introduced to the landscape via roadway runoff. Among them are hydrocarbons, asbestos, lead (Pb), cadmium (Cd), and copper (Cu). In addition, chemicals associated with the road itself or its maintenance, including pesticides, insecticides and deicing salts combine with runoff and make their way into storm water drainage systems ( U.S. Environmental Protection Agency 2001; U.S. Environmental Protection Agency 1996; Trombulak and Frissell 2000; Grant et al. 2003) Volatile chemicals associated with roads are introduced to the envir onment from vehicle emissions. These include carbon monoxide (CO), nitrogen oxides (NOX), volatile organic compounds, sulfur dioxide (SO2), particulates from exhaust and road dust, lead (Pb), methane (CH4), and toxics including benzene, butadiene and form aldehyde. In addition to these primary emissions, some chemicals react to form secondary pollutants in the air. Chief among these is ozone, which is produced when nitrogen oxides combine with volatile organic compounds in the air. In the United States the emissions of chemicals increased rapidly until, in the 1970s and 1980s, pollution controls helped to reduce some emissions from vehicles, with lead emissions seeing the most dramatic declines in recent decades. Despite this decline, the estimated premature death in 1991 due to respiratory ailments caused by motor vehicle air pollution was equivalent to the number of deaths from motor vehicle accidents, approximately 40,000 (U.S. Environmental Protection Agency 1996) For this reason, air pollution is widely considered to be the most significant dire ct environmental effect of road related transportation. Air pollutants also enter aquatic systems and compound effects of stormwater runoff, with substantial inputs of
36 nitrogen, metals and hydrocarbons to water bodies from atmospheric sources ( U.S. Environmental Protection Agency 2001) Noise and Other Atmospheric Effects Increased noise levels are one of the most important environmental effects of highways, and are considered a nuisance to human populations in urban and suburban areas. In the United States, the Federal Highway Act of 1970 mandated the development of standards for noise and noise abatement relative to land use. Noise abatement studies are mandatory elements of the environmental impact assessm ent of highway construction projects. Mitigation for noise is a substantial part of the budget of any highway construction project and often results in the design and construction of specific noise abatement structures along highways ( U.S. Department of Transportation and Federal Highway Administration 2000) Despite a several decades long concern over the impact of highways on ambient noise levels, the effects of n oise on populations of wildlife have not been as extensively researched. The Road Ecology Center at the University of California, Davis, recently sponsored a series of lectures in Winter 2005, to focus attention on the subject of the effects of noise on wi ldlife populations (2005) and a session of the 2005 International Conference on Ecology and Transportation (ICOET) also focused on this theme (West 2006; Dooling 2006) Road noise has a variable effect on animals. The most significantly impacted by road noise are those species that incorporate sound into their basic behavior, such as birds. Much of the effect depends on the fr equency to which the species in question is attuned. The effects of roads will disproportionately affect those species for whom the frequency of the road noise interferes with the frequency of their calls. For example, great tits ( Parus major) in the city of Leiden, the Netherlands, were found to sing at higher pitches in noisier environments to overcome the problem of masking caused by low -frequency noises of the urban din (Slabbekoorn and Peet
37 2003) In addition, the patterns of noise produced by traffic fluctuate in time. There may be a varying effect of road noise on animals as determined by time of day or season of the year, depending on the daily and life cycle patterns of that animal. Aside from road -related noise problems, other atmospheric effects are produced by the physical structure of roads. Roads affect patterns of wind direction and speed, temperature, rel ative humidity and insolation. Generally, roadsides are windier and more turbulent, hotter, dryer and sunnier (Forman et al. 2003) In addition, the air is dustier near roads, particularly near unpaved roads. Road dust affects vegetation by covering surfaces and affecting photosynthesis, respiration and transpiration thereby resulting in injury and decreased productivity (Farmer 1993) Dust provides adsorption surfaces for volatile contaminants that are subsequently deposited either by dry or wet deposition, causing phytotoxic pollutants to enter plant tissues, and causing respiratory ailments in animals and humans. These microclimatic changes can affect areas great distances from the road, changing the vegetation composition for some distance away from roads (Farmer 1993; Forman and Deblinger 2000) The E ffects of R oads on B iotic Com ponents of E cosystems Roads are agents of change that have both primary, or direct effects, as well as secondary, or indirect effects on the biota (Bennett 1991) Roads affect animal and plant populations directly by entirely obliterating the ecosystems in their path. While for some species, the destruction of a small area for a roadbed may not be significant, for other species, particularly small animals wit h high levels of site fidelity, it can be ruinous. Populations of slow -moving animals and those which regularly cross roads suffer in particular from the negative effects of increased mortality due to vehicle collisions. Roads also act as conduits introduc ing and facilitating the spread of exotic species. The indirect effects of roads include changes or impacts that result from increased con tact with humans and human landuse activities.
38 Road s as Sources of Mortality and Barriers to Animal Movement In the United States, roadkill has surpassed hunting in its effect on vertebrate mortality (Forman and Alexander 1998) While some species with high roadkill rates ( e.g., house sparrow, Passer domest icus ) seem unaffected by the high rate of mortality associated with roads, others are much more affected, such as the Florida panther ( Felis concolor coryir ), which had an annual roadkill mortality rate of 10% of its population before 1991 (Forman and Alexander 1998) In Florida, road related mortality is the primary source of mortality for panther, black bear ( Ursus americanus ), key deer ( Odocoileus virginianus clavium ), and American crocodile ( Corocodylus acutus ) (Harris and Scheck 1991) This dubious distinction extends to the marine environment as well, with boat collisions being the largest known human related cause of mort ality for the W est Indian manatee ( Trichechus manatus latirostris ) in Floridas coastal waters (O'Shea et al. 1985; U.S. Fish and Wildlife Service 2001) While much of the focus on anim al mortality has been on large mammals, herpetofauna are also significantly affected by roadkill. One of the most important mitigation projects in this regard is the ecopassage to alleviate roadkill rates on U.S. Highway 441 through Paynes Prairie State Preserve near Gainesville, Florida (Smith and Dodd 2003; Dodd et al. 2004; Smith et al. 2005) Harris and Scheck (1991) identified several reasons why roads and traffic are such significant sources of mortalit y for wildlife: migration routes and home ranges or territories are bisected by roads; animals intermingle with traffic as they move along open road corridors; new food resources, such as carrion and forage, are available in road corridors; and roadside en vironment is attractive and serves as an ecological trap or habitat for some species. These observations suggest that the reasons that animals are killed by vehicles are driven mostly by the spatial arrangement of resources. Animals die when they are str uck while trying to reach resources (food, water, den sites, etc.). Smith (1999; 2003) carried out an extensive spatial
39 analysis of roadkill in Florida and suggested where planning and design efforts could mitigate vehicle -wildlife collisions taking into account the existing locations of roadkill, landscape patterns, animal distribution and movement patterns and questions of land and road ownership. Many researchers have also discovered a temporal pattern to roadkill that depends on varying resources, such as standing water during wet cycles, and life history, such as dispersal, hibernation, or foraging patterns (Davies et al. 1987; Main and Allen 2002; Saeki and Macdonald 2004) The direct mortality of animals due to vehicle collisions is a primary and obvious effect that reduces animal populations. In less populated areas, such as tropical forests where traffic counts are low, animals may not risk being struck and killed by a vehicle. In these areas, increased human access corresponds with reduced densities of animals due in part to increased hunting pressure a secondary effect of roads (Bennett 1991; Robinson and Bodmer 1999) This phenomenon is well understood in Amazonia where loggers access remote forests by paths, trails and rudimentary logging roads and, while logging in an area, trap and hunt game for local consumption or trade (Peres and Lake 2003) This is one of the many secondary effects of logg ing (and logging roads) which also include land transformation as well as a number of socioeconomic responses. It is thought that these cumulative secondary effects of logging may be more detrimental to the overall long-term health of tropical forests tha n the actual logging itself (Laurance 2001) In the wake of loggers, poachers and local settlers also venture into the local forests and preserves to hunt for subsistenc e and to augment their incomes (Barnes et al. 1995; Altrichter and Boaglio 2004) Thus the secondary effects of roads on wildlife mor tality extend far beyond the road corridor per se.
40 With the pressure of increased mortality due to roadkill and hunting, many species have been observed to alter their behavior near roads or in areas where there are higher densities of roads. It has already been noted that birds are known to change their calls in response to the disturbance of road related noise. It also is apparent that some species learn how to avoid vehicle collisions with age. At 3 years old, the mortality rate for Florida scrubjays ( A phelocoma coerulescens ) living near roads is the same as birds living in nonroaded environments (Mumme et al. 2000) One possible explanation for this is that surviving jays learn to avoid automobiles. Other types of modification include adjusting behavior to avoid, spatially and temporally, hum an activities. Such is the case for bobcats and coyotes observed in urban environments (Tigas et al. 2002) For some species, the stress accompanied with road relat ed disturbance and behavioral changes can affect overall survivability. Amur tigers ( Panthera tigris altaica ) living in roadless areas stayed longer at kill sites, ate more meat, and ultimately survived longer than tigers living in areas with roads (Kerley et al. 2002) That roads act as barriers which hinder the movement of animals and fragment breeding populations is one of the most often noted effects of roads. Every published review of the ecological effects of roads notes the importance of the barrier effect of roads. The extent of the effect is determined by the characteristics and behaviors of the species in question, the phys ical qualities of the road and road related infrastructure, the characteristics of the road traffic, and the spatial configuration of the road relative to adjacent landscape. The phenomena of population fragmentation has been noted across taxa, and affects small mammals (Oxley et al. 1974) large mammals (Nellemann et al. 2001) understory birds (Develey and Stouffer 2001) insects (Bhattacharya et al. 2003) and herpetofauna (Smith et al. 2005) It arises when populations of animals are subdivided into smaller groups and genetic exchange between the groups ceases to
41 occur because the road is impassable. The overall effect is to make local extinctions more likely as sources of immigrants are disconnected (Johnson and Collinge 2004) Roads as H abitat, Corridor and Conduit Roads and road verges do provide habitat for some animals, particularly smal l mammals and insects (Oxley et al. 1974; Getz et al. 1978; Vermeulen 1994; Brock and Kelt 2004) and provide a source of food for carrion -feeders (Bennett 1988) The use of roadsides by animals depends greatly on design and management of the ve rge (Forman and Alexander 1998) Differences in mowing regimes or planting designs can vary the effect of roads on bird, insect and mammal populations. In some cases, where the surrounding l and cover has been extensively transformed, as in parts of Australia and The Netherlands, roadsides, or verges, are the only remnants of native vegetation remaining, and are important sources of biodiversity in the landscape (Hussey 1999; Deckers et al. 2005) In these cases, the contribution of railway corridors is also very significant. Railway rights -of -way are often important reserves of remnant native vegetation in landscapes that are predominantly agricultural. The benefits of remnant vegetation to animals depend on the width of the verge as well as the design characteristics of the roads. In the wheat belt of Western Australia, where the road verge is a considerable portion of the remaining native vegetation, as the width of the verge increased the number of species in the verge increased (Arnold and Weeldenburg 1990) The design characteristics of the road, including the width of the road, the height of the road above grade and the surface of the road determines the habitat characteristics for species that might use it thus. For example, kangaroo rats (Dipodomys stephensi ) were more active in using dirt roads than gravel roads (Brock and Kelt 2004) For some small mammals, road verges constitute a long, ribbonlike habitat along which they can move and disperse (Vermeulen 1994)
42 Some large animals are known to use roads and the space above roads to move more easily through the landscape. Observations have discovered t hat these are wide ranging animals, using lightly traveled roads and tracks. They include red fox ( Vulpes vulpes ), dingo ( Canis familiaris dingo), wolf ( C. lupus ), cheetah ( Acinomix jubatus ), and lion ( Panthera leo). Bats are known to use these spaces muc h as they would a gap in the forest. Bennet (1991, p. 100) identified four types of movement patterns that utilize roadside habitat: local foraging movements; dispersal between separated populations; long distance migratory movements; and local or geographical range expansion. To the extent that roads can serve as conduits for movement, it is important to recognize t hat not all species can take advantage of the road space to forage, disperse or colonize. Many species, it has been noted, experience the road as an inhospitable environment and a barrier to movement. The plants and animals that are facilitated by the roa d as a conduit for movement are often generalist species. These species are able to exploit highly variable ecological conditions, such as those found in roadside environments (Forman and Alexa nder 1998) They are often very successful at using road verge to facilitate their persistence and spread across the landscape. For this reason, roads are often cited as major causal factors in the successful invasion of exotic flora and fauna (Gelbard and Belnap 2003) Many exotic plants, which fit into the generalist category, exist disproportionately in roadside c orridors (Tyser and Worley 1992; Watkins et al. 2003; Paucha rd and Alaback 2004) In addition to the non -native plants that are frequently used in roadside landscapes, non native seeds and propagules are dispersed by vehicles and encounter environments in road verges where they can thrive (Schmidt 1989; Lonsdale and Lane 1994) Roadsides have abundant light, littl e competition for runoff water from established shrubs and trees, and are flushed with nutrients
43 periodically as land is cleared adjacent to roads. The easy availability of limiting factors (i.e. light, water, nutrients), combined with aggressive dispersal mechanisms, repeated human introductions, and the high contiguity of roadsides that extend for hundreds of miles uninterrupted, make roadsides highly invasible spaces (Davis et al. 2000; Parendes and Jones 2000; With 2003) This introduction and establishment process is illustrated well in the case of Cogon grass (Imperata cylindri ca ), which was first introduced to Florida in the 1940s and 1950s for purposes of forage and erosion control. While Cogon grass can disperse by seed, the seeds do not travel great distances. Dispersal by rhizomes is far more significant, and is particularl y problematic when spread by road construction equipment, or by contaminated roadway fill. Following its intentional introduction in the 1960s, extensive road construction occurred in the regions where it was introduced. By the mid1980s Cogon grass was considered a noxious weed along highway rights -of -way, with severe infestations in north -c entral Florida, near Gainesville, which was one of the points of introduction (Dean et al. 1 989; Willard et al. 1990) Red imported fire ants ( Solenopsis invicta) followed a similar trajectory. Although it was accidentally introduced to Mobile, Alabama, in the 1930s this species spread quickly throughout the southeastern United States. It has s ince spread to the Florida Keys where it is threatening many rare and endangered endemic species. While fire ants are found in all habitat types, they were most often encountered within 150 meters of a road or development (Forys et al. 2002) Ecological Effects of Road Networks While roads have many direct ecological effects on adjacent aquatic and terrestrial systems, as network structures, they also have far reaching, cumulative effects on landscapes which have been less well -studied (Riitters and Wickh am 2003) Some major effects to landscapes that directly relate to roads include the loss of habitat through the transformation of
44 existing land covers to roads and road -induced land use and land cover change (Angelsen and Kaimowitz 1999) ; and reduced habitat quality by fragmentation and the loss of connectivity (Theobald et al. 1997; Carr et al. 2002) Together they point to the larger issue of the synergistic effects of roads and road networks on ecosystems at broader scales (Forman et al. 2003) Landscape Change and Fragmentation In tropical forested areas, economet ric models of land use and land-cover change have revealed important relationships between biophysical and economic variables relative to roads. Not surprisingly, in rural areas, particularly in developing countries, the presence of roads has been most strongl y correlated with processes of land cover change by facilitating deforestation (Chomitz and Gray 1996; Angelsen and Kaimowitz 1999; Mertens and Lambin 1997; Lambin et al. 2001) The impact of roads, however, is not uniform. The spatial model developed by Chomitz and Gray (1996) examining the effects of road building on deforestation in Belize shows a sensitivity to soil quality, land tenure regulations and market access. They concluded that these three factors h ave strong interactive effects on the likelihood and type of cultivation (Chomitz and Gray 1996, p. 501) In an area such as Belize with a low -d ensity population, farmers will not simply follow logging roads to establish farms on poor quality land. The conversion of forest to agriculture is more likely to occur on lands that have higher quality soils and greater access to markets, and are not prot ected State Reserves. Spatial models that take into account the effects of roads on deforestation processes have become much more common in recent years (Mertens and Lambin 1997; Stone 1998; Munroe et al. 2002) A recent study examined the process of deforestation by crossing spatial analysis studies with livestock economic studies to understand the processes of land cover change in the Brazilian Amazon. In this case, road construction unambiguously increases the incentives to convert forests to other uses (Mertens et al. 2002, p. 291) However, the effect of road
45 construction varies and depends on the type of road, the stage of economic development in the region, and variably affects land owners according to their production status. The main roads that formed the lines of penetration conceived of by Taaffe et al (1963), were more important in the early colonization process, where the recent forest clearing is more closely associated with the development of feeder lines. In the Brazilian Amazonian case, the roads were more important in providing market access to small scale producers rather than large ca ttle ranchers, for whom the presence or absence of roads is less of a constraint. Therefore, the development of roads was more important in explaining deforestation processes in the case of colonization than in the case of large -scale cattle ranching. As elicitors of landscape fragmentation, roads appear to have the upper hand over other anthropogenic causes. To the extent that the road network is extended and connected, the landscape become s more fragmented and less well -connected. Reed et al (1996) found in the Rocky Mountains that roads c reated more forest fragmentation than clearcut logging as roads divided large forest patches. In nu merous studies, densities of species are correlated either with road density (negatively) or with distance from road (positively) (Mech et al. 1988; Barnes et al. 1995; Canaday 1996; Huijser and Bergers 2000; Develey and Stouffer 2001) These tend to be species that require interior forest conditions, requ ire extensive home ranges and are shy, or are hunted. Whats more, there is a time lag between the time of road construction and the effect of species decline, which varies by taxa (Findlay and Bourdages 2000) Conversely, the quality of roadlessness may be a significant determinant of survivability for some species. In a regional analysis of Wyoming, Montana and Idaho, the overall regional habitat connectivity increases, as measured by four landscape metrics of area, configuration, isolation and co ntagion when roadless areas were included along with conservation areas (Crist
46 and Wilmer 2002) Another study in Alaska gave similar results, with roadless areas contributing significantly to regional measures of habitat connectivity (Strittholt and DellaSala 2001) Road Edges and the Road Effect Zone The fragmentation caused by roads is alternatively measured by the amount of edge created by them. In many studies devoted to effects of roads on landscapes, references to the fragmentation of landscapes and the creation of edges are often coupled and used complementarily to describe the processes of change (Laurance and Wi lliamson 2001; Hawbaker and Radeloff 2004) Ecological edges occur naturally throughout the landscape. They create the spatial patterns that result from environmental heterogeneity and the interactions between and among organisms (Turner et al. 2001) Ecological transitions, or ecotones, can result from changing resource availability, such as the transition from a marsh to a forest, corresponding with changing hydrologic conditions. Alternatively, edges are created when an ecological disturbance, such as a fire, creates a localized opening, like a gap in a forest (Forman 1995) In this case, increased light and nutrients coupled with decreased competition from other plants allows for the colonization and establishment of other plant species, creating a distinctive edge. Road edges, however, apart from being entirely human induced, are peculiar in their linear shapes, and are affected by the dynamics of transport economics, distinguishing them from other types of edges. Reed et al. (1996) found that the amount of edge created by roads was 1.54 to 1.98 times that created by clearcuts. While it is true that road edges provide resources for some species (section 3.2), they are most often noted by ecologists for their far reaching negative consequences to ecosystem structures and flows. The edge ef fect of roads variably extends several meters into the adjacent landscape and is responsible for rendering vast areas as uninhabitable to many others (Forman et al. 2003) The microclimatic changes produced by even narrow roads affect the leaf litter and vegetatio n composition, soil macroinvertebrates, interior -
47 dwelling forest birds, herpetiles, mammals and overall species richness (Willard and Marr 1971; Haskell 2000; Godefroid and Koedam 2004) In the Amazon, overall pos itive feedbacks have been noted between increasing fires and drought conditions, i.e. regional climate change, and the amount of forest fragmentation and deforestation, directly related to the construction of roads (Laurance and Williamson 2001) The strength of the ecological effects of roads on adjacent nonroad areas is a variable phenomenon, changing both in space and time. This road effect zone is determined as the zone adjacent to roads where one or more direct ecological ef fects of the road can be discerned. The convoluted shape of the zone can extend laterally to cover areas many times greater than the dimension of the road and its verges, depending on the ecological process and sensitivity of the species in question (Forman et al. 1997) While the significance of this concept cannot be underestimated, measuring the road-effect zone is as yet complex and imprecise, and the development o f tools to assess the landscape -scale effects of road edges is not yet well developed (Ries et al. 2004) Despite the difficulty of precisely locating the road effect zone, it is c lear that the area of land ecologically affected be roads is vast, by virtue of proximity alone. On a continental scale, Riitters and Wickham (2003) estimated that approximately 83% of the land area in the c onterminous United States to be within slightly more than one kilometer of any road, and only 3% of the area slightly more than 5 kilometers away. Ecological Road Network Theory Cumulatively, road-effects interact with each other when roads are considered as systems. Ecological road network theory, which is comprised of basic principles of land use, transportation, network theory and ecology, provides a framework to interpret the ecological effects of road networks. With this theory as a basis, an analysis of the effects of a road network in a terrestrial ecosystem suggests that they extend over large areas of the landscape, that long -
48 distance effects can saturate a landscape even in moderately roaded areas, and that isolated patches of habitat are created by road -effect patterns (Forman et al. 2003) At this early stage in the development of road network theory, measuring the cumulative ecological effects of roads is far from well -developed. Indices that have been proposed include measuring road density, road locat ion (Forman et al. 1997) and, using empirical methods, measuring road -effect zones on wildlife populations (Carr et al. 2002) Landscape ecological researchers are also considering the use of various modeling approaches to examine the problems (Carr et al. 2002; Forman et al. 2003) It is clear that developing robust quantitative methods which model, explain and predict the interactions between road network structures and landscapes are important for future research. As a start, Jaeger et al. (2005; 2006) have used simulation mod eling to predict the effects of road configuration on animal population persistence. They concluded that the effect of a gridded vs. parallel road network configuration depends on the target species behavior, i.e. to what degree that species avoids crossi ng roads, and the probability of it being killed if it does. Bundling traffic (locating roads in close proximity to each other) is beneficial to population persistence, and core habitat areas that are unfragmented should be protected from road constructi on. They also conclude that while their modeling studies are an important step in developing ecological road network theory, empirical analysis comparing the effects of different road network configurations should be done. Clearly there is potential in th is area for a collaborative effort between road ecologist and transportation geographers. Although transportation geographers were prolific in developing quantitative indices of transportation network structures in the 1960s, this work is only beginning to find its way into road ecology literature (Forman and Alexander 1998; Forman et al. 2003) and the indices have yet to be tested in modeling landscape/road-network interactions.
49 Conclusion Wildlife biologists, observing the effects of tra ffic on animal mortality were the first to point out ecological effects of roads, and the concern over their effects has grown such that Forman and Alexander (1998) refer to ro ad ecology as the sleeping giant of conservation ecology. Roads affect both the biotic and the abiotic components of landscapes by changing the dynamics of populations of plants and animals, altering flows of materials in the landscape, introducing exoti c elements, and changing levels of available resources, such as water, light and nutrients. Historically, the field of transportation geography has concerned itself mostly with the economic and structural aspects of transportation, and implications for land use change. However, transportation geographers are in a unique position to contribute to the emergence of the science of road ecology, as the wealth of knowledge already developed can provide both theoretical and analytic tools to study the landscape scale effects of road networks.
50 CHAPTER 3 A NETWORK ANALYSIS OF ROADS IN THE SANTA FE RIVER WATERSHED FROM 1975 THROUGH 2005 Introduction Transportation networks are a defining feature of human social space s Of these networks, roads are the most ancien t and pervasive. Roads provide conveyance for people, ideas and resources to be introduced, extracted, transported, and processed for exchange. Road networks are a global imprint of human influence persisting long after agricultural fields and settlement s have vanished (Jackson 1984; Trombold 1991) Transportation geography theory generally assumes that with sustained human population growth and economic development, road networks become more extensive and interconnected (Taaffe and Gauthier 1973) The capacity of the network increases and accessibility improves with these changes. The entire region becomes more accessible as roads are built or upgraded. The cause and effect question about whether economic development causes road network development or vice -versa depends on a variety of controlling factors operating from local to international scales: among them financial markets, economic and social policies, and land use and tenure laws. An important chara cteristic of transportation systems is that the decision rules governing their development change over time, so the functional relationships between explanatory variables and the rules implied by them change as the network evolves, creating a rather diffic ult research problem (Garrison and Marble 1962) Resolving the question of this dynamic was not a central focus of this study. As road networks become denser and better connected, human accessibility improves, intensifying land use while diminishing and fragmenti ng the landscape for many species (Andrews 1990) Our knowledge about the structure and function of road networks is largely based on the analysis of regional and national transportation systems. But the ecological effects
51 of roads are experienced at a finer scale, and the cumulative ecological effects of road networks are important to the long-term survival of numerous animal and plant populations (Forman and Alexander 1998) The dynamics of local road networks in rural areas where people most often encounter wildlife are poorly understood. This study explores questions about the development of a road network in a predominantly rural, increasingly populated region of north Florida. Among the questions are, how does the road network change over time, and what are the nature of those changes? Based on principles of transportation geographic theory, I hypothesized that the network characteristics of extent, connectivity and accessibility would all increase over time. I used GIS and network analysis to examine the development of the local road network from 1975 to 2005 in the ar ea of the Santa Fe River watershed (SFRW ; Figure 3 1). The road networks were characterized according to twenty three distinct metrics and are described later. The objective of this research was to describe road network dynamics at the scale of a watershed (103 104 km2), including local roads in the analysis. This chapter describes the changing structure of a rural road network in the SFRW from 1975 through 2005. Section two is a review of research on road network analysis, describing the various ways that road networks have been modeled. Section three describes the study area and the methods that have been used to analyze the road networks Section four is a straightforward accounting of the results, and section five provides interpretation of those results, suggestions for further research and offers some conclusions. Road N etwork D evelopment and A nalysis An ideal typical sequence of transportation development was proposed by Taaffe, Morrill and Gou ld (1963) based on studies examining how transportation lines developed in colonial Ghana and Nigeria (Gould 1960) The process, comprising six phases, begins with scattered ports along a coast line (phase a). Penetration lines, or roads into the hinterland, develop from
52 some of the ports (phase b), thereby reducing costs for certain market points, and increasing their dominance at the expense of the less accessible ports. These penetration lines develop feeder lines which result in the development of nodes, or small towns at the intersections of these lines (phase c). As the lines continue t o expand and interconnect, the size and dominance of the best connected nodes increases (phase d) until the system is completely interconnected (phase e) Finally, the development of major trunk lines, or highway systems, enhance s opportunities for certain towns, bypassing others (phase f). The authors caution against seeing this as a historical sequence, pointing out that all six phases could be occurring simultaneously in different regions of a country. Although they do not discuss issues of transportation mode, presumably the process could be thought of as a conceptual model capable of recurring as the technologies and economics of transportation adjust over time. Alternative models of transportation development i n stages were proposed by other researcher s including a mid -continental model (Lachene 1965) and a model based on a theoretical Lschian landscape (Haggett 1965) All of these models share common ideas of expansion and centralization of the networks, with certain urban centers rising in prominence and accessibility, while others were bypassed. These theoretical sequences were informed by empirical studies of transportation networks as part of a broader effort by geographers in the mid 20th century to quantify spatial interactions (Burton 1968) They recognized that r oad networks and their traffic lent themselves to the quantitative analysi s of their spatial characteristics Several mod els have been used to describe the properties and structure of road networks so as to observe their interaction with environmental and economic variables. As physical features embedded in a landscape, roads have spatial qualities of length, surface type, d imensions, location and density Using the mathematics of graph theory, road networks can also be described as structures of interconnected
53 points or vertices, and lines or edges, with characteristics of connectivity, flow and accessibility When the r esearch objective was to describe accessibility of the landscape for resource consumption, road density, distances to roads, and other factors were used to model effects of roads (for example, Heilman Jr. et al. 2002) When the objective was to study network routing and configuration topological characteristics of the network itself were used (for example, Black 1993) A full list of the metrics used in this study is given in Table 3 1, with more detailed descriptions provided in Appendix A. Map -B ased M easures of Road N etworks A number of basic measures of road networks are based on their mapped characteristics. The most obvious of these is a direct measure of their total length As roads are linear features, a direct measure of their total length is a simple, straightforward descriptor of their extent which depends som ewhat on the scale and resolution of the map (Haggett and Chorley 1969) O ther map -based measures of road networks include: area -related measures, such as road density; distanceto road (DTR) measures, which are often used in spatial and ecological analyses; and other measures related to the flow of traffic or the capacity of the network. Graph -The oretic M ethods for R oad N etwork A nalysis Optimal methods for simultaneously considering characteristics of the physical structure of road networks and their traffic flows were developed from the principles of graph theory and network analysis. These method s were developed by Garrison and Marble (1962) elaborated upon by Kansky (1963) and are well described in textbooks on the subject published in the1970s (e.g., Taaffe and Gauthier 1973; Lowe and Moryadas 1975) In their analys es, road networks were abstracted as graphs so that places, cities or intersections, here referred to also as junctions, were abstracted as vertices or nodes; the roadway connections between vertices here referred to also as road segments, were abstracte d as edges or linkages.
54 Graph -theory based network measures and indices summarize the topological properties of road network s thus enabling comparison of structural and functional characteristics of different networks and the study of their evolution. The fundamentals of graph theory provide that any given network can be described as a graph, G consisting of three elements: E, or the edge set of a graph; V vertex set; and the relation between edges and vertices, where the edges of E are defined by two vertices in V that can be identical (West 2001) The vertices are represented as points and the edges as lines. Figure 32 shows a graph with five vertices and eight edges. The indices used to describe graphs of road networks consider them to be planar graphs: that is, they exist in two -dimensional space. The implication is that a vertex exists wherever two edges cross. This assumption limits the number of possible connections for each ver tex. Although strictly speaking, road networks are not necessarily planar there are overpasses and underpasses over most of the area of the network, they are planar with a few rare exceptions (Garrison and Marble 1962) The numerous indices used in road network a nalysis are derived from a few basic properties (Table 3 2; Cliff et al. 1979) The simplest of these are the number of edges, or road segments, ( e ) and the number of vertices, or junctions, ( v ) in the graph. Subgraphs ( p ) are separate subcomponents of a graph. In this study, the entire road network was considered to be connected, so p = 1. The order (o ) of a vertex refers to the number of edges connecting at that vertex, i.e. the number of road segments intersecting at a junction. In graph theory order is also sometimes referred to as the valence or degree of the vertex. For the graph in Figure 3 2, the vertex in the center has o = 4, while the vertices on the edges have o = 3. Finally, the shortest topological route between two vertices, j and k is the geodesic, gjk. Geodesic related indices help
55 describe the relative importance of individual el ements within the overall network structure and are considered measures of centrality or accessibility. Measures of the entire network Graph -theoretic indices used to describe the structure of national transportation networks were correlated with a technological scale derived from national economic indicators by Berry (1960) and certain spatial characteristics of each country. While variability existed between each of these indices and the predictor variables, the relationship was consistently strong between certain indices and t he economic factors. That is to say, a s countries developed economically, their highway systems and railroads became more extensive and inter connected The number of higher -order junctions and road segments increased as functional capacity improved. Conseq uently, circuitry, connectivity and functional capacity of the network increased with a nations economic development. This same relationship held for state and county level transportation systems, although relationship was not as strong at this scale (Garrison and Marble 1962; Kansky 1963) The measures and indices of an entire network are derived from the basic elements of edges and vertices Measures of the aggregate network explored here include the cyclomat ic number ( ), a basic measure of circuitry, and diameter ( ), the geodesic between the two most distant points of the graph As a network develops, the number of circuits and connections are expected to increase resulting in an increase for and a decrea se in T he cumulative characteristics of graphs can be measured with indices of connectivity and circuitry including the alpha ( ), g and beta ( ) indices. Alpha ( ), a measure of the circuitry of a graph is a ratio of the number of actual circuits in the graph, expressed as e (v -1) and the maximum number of circuits possible in a planar graph, given as 2v -5 The theoretical upper limit of is one, and its lower limit is zero (a graph
56 with no circuits) A s such, when multiplied by 100, it is sometimes presented as a percentage of connectivity Alpha ( ) increases a s a network becomes more connected The g amma index of actual edges, e and the maximum number of edges possible in a planar graph, e(v -2) As a network Taaffe and Gauthier (1973) associated the network configurations based on limits of their values The se classes used by engineers to describe planar networks include spinal, grid and delta structures, where a spinal structure is a minimally connected graph with no circuits, a delta structure was a maximally c onnected graph in which most nodes have three linkages, and a grid was an intermediate graph structure, considered as a transitional stage between spinal and delta configurations G raph types and their index limits are given in Table 3 3. Beta ( ), a connectivity index, is the ratio of edges to vertices in the graph. For simple tree graphs and disconnected graphs < 1 For a graph with only one circuit, = 1 Higher values of up to a maximum value of three, are produced when network structures become more complex and the number of edges relative to vertices increase (Kansky 1963; Rodrigue et al. 2006) In addition to these, Kansky (1963) and Garrison and Marble (1962) created a set of indices to gauge characteristics of graphs that are unique to transportation networks. These included the shape index ( ), as well as the average edge length ( ) and the structure index ( ). This group of indices has the shared trait of comparing network length with one or another network feature (diameter, edges and network weight respectively) to measure structure and f unctional capacity. As a road network develops, the value increases, so highly developed networks have very high values, while undeveloped networks have values approaching one
57 (Haggett 1965) Conversely, values for and are negatively correlated with economic development. Decreasing and are associated with increasing functional capacity of the network. Measures of network elements Graph theory also provides for measurements of properties of vertices and edges the individual elements of graphs These methods are derived from analysis of binary connectivity matrices in which connections between individual elements are represented by a one (connected) or a zero (not connected). Through matri x multiplication one can determine the level of accessibility of each vertex, and discover geodesics in the network. Vertices can be ranked to recognize important (or vulnerable) infrastructure in any networked system, identifying, for example, critical no des and linkages in computer (McIntee 2004) or highway systems (Demsar et al. 2008) The Shortest -path accessibility index used by transportation geographers (Taaffe and Gauthier 1973) and the closeness and betweenness indices, developed by social scientists (Wasserman and Faust 1994) all make use of these concepts. S hortest -path accessibility ( Ai) uses t he shortest path algorithm (Shimble 1953; Dijkstra 1959) to eliminate redundancies in alternate paths finding the geodesics in the network. Where each row in the shortest path matrix pertains to an individual vertex i t he row summaries give Ai for each vertex. The vertex with minimum Ai ( Min Ai) is the most accessible in the graph. In applying this concept to transportation networks, it is important to keep in mind that the shortest topological route may not be the fastest or the shortest distance between an origin and a destination Taaffe and Gauthier (1973) demonstrate a method to measure Ai using cost values as opposed to simple binary connections The values may be related, for example, to the capacity, allowable s peed or length of the road segment. In an analysis of the security of the U.S. Internet backbone, McIntee (2004) used these principles to derive an unweighted relative connectivity
58 index (URCI) a weighted relative connectivity index (WRCI) and a network connectivity index (NCI) to examine the relative connectivity and vulnerability of major nodes in t he network. In a social network, closen ess measures how close an actor, or vertex is to all the other actors or vertices C loseness is the inverse of Ai and measures distances in geodesics. As the sum of distances decreases (i.e. the actor becomes more connected and accessible), the value of closeness increases, approaching a limit of one. For this study, summary values of vertex closeness were considered a s measure s of the accessibility of the road junctions. Increasing values fo r mean closeness overt time indicate that distances across the overall network are decreasing, and that, on average, road junctions are becoming closer to one another Similar to closeness, the idea of betweenness describes the degree of importance for par ticular vertices or edges The vertex betweenness index is the sum of the probabilities that a vertex i is a member of the geodesic gjk between two vertices, j and k The more central an actor is in the network, the higher the level of betweenness (Wasserman and Faust 1994) Betweenness should increase as the element becomes more central relative to the network as a whole D emsar et al. (2008) creatively applied these social network analysis algorithms to identify vulnerable points in the transportation infrastructure of the Helsinki metropolitan area Combining a GIS based representation of the road network with grap h theory measures of centrality, the authors were able to distinguish spatially critical points in the infrastructure that could then be prioritized as important security concerns for transportation network administrators and planners An increase in mean and maximum values of betweenness in a network could indicate that a centralization process is occurring, either through the expansion of the network or through interconnection.
59 Methods Study A rea The SFRW is within the southern reaches of the Suwannee River Basin which extends northward into Georgia (Figure 3 1). The watershed area is approximately 3584 km2, centered at 30.0 N, 82.5 W. It covers almost all of Bradford and Union Counties and major portions of Alachua, Gilchrist and Columbia Counties which comprise 94% of the SFRW. The edges of the watershed stretch into Suwannee, Baker and Clay Counties with the most extreme southeastern corner reaching into Putnam County. The limits of this study inclu ded the entire area of the SFRW that intersected with the system of 7.5 quadrangles defined by the U.S. Geological Survey (USGS) Each quadrangle, or quad, is divided into four sections known as quarter quads. The outer boundary of the intersecting quarte r quads was used as the boundary of the study area, giving a total area of approximately 4974 km2, and comprises 119 quarter quads. This geographic delineation provided a system for defining the study area that was independent of the geometry of the road n etwork. At the same time, it provided an unbiased system for partitioning the study area into a grid of 119 nearly equal area regions that were used to systematically manage data entry and editing, and will be used in forthcoming analyses on road network s tructure variability The p opulation in the five counties comprising most of the study area has increas ed over the study period (Table 3 4 ), with growth rates for most counties close to that of the S tate of Florida (Table 3 5 ). Recent increases in populat ion growth rates in Union and Bradford Counties are concomitant with economic development in the Jacksonville Gainesville corridor as small urban centers like Starke grow to accommodate new businesses. Land cover in the SFRW shows small areas classified as urban and residential, while agriculture is the second most common land cover type (Pearlstine 2002) The region is also an
60 important timber production area. Land use is dedicated mostly to forestry and agriculture with a few urban centers Starke, Lake City and Alachua being the largest While the land cover classification makes no distinction between pine plantations and native pine forests, pine forest is the most common land cover type. H ay production ranks higher in area than any other crop in the study area counties The SFRW is located in a buried karst region, characterized by rapid surface -groundwater interactions. Perhaps the most important natural feature s of this region are the scores of high quality, artesian springs (Nordlie 1990) Itchetucknee State Park is a protected area in the SFRW that sits astride a spring run, a river produced by the outflow of several of these springs. The protection of water q uality and biota of these unique ecosystems is seen as an important conservation priority in the region, and increasingly figures in discussions and plans for future growth and land use (Suwannee River Partnership 2007) Transportation systems in the area consist of several l ightly used railroads, U.S. Interstate highways I 75 and I 10, and a network of state and county roads (Figure s 3 -3A and B). A fine mesh of minor roads pervades most of the landsca pe (Figures 3 3C and D). Most of the road network in this watershed is unpaved. Among the industries in the watershed are power plants, a cement plant, and three prisons. Perhaps the most environmentally important industry within the watershed area is a titanium mine operated by Dupont de Nemours to the east of Starke in Bradford County. All of the above -mentioned facilities are concentrated in the eastern half of the watershed with the exception of a cement plant near the Ichetucknee River. In general, the SFRW is a rural region experiencing moderate to rapid population growth. Wh ile little of the land is urbanized, several major thoroughfares pass through it and there is little
61 to suggest that population growth or economic development will not continue on its current trajectory. Data D evelopment Maps of roads in the SFRW are avail able from many sources, including the USGS topographic maps and the TIGER/Line GIS data files of the U.S. Census Bureau. For local roads, these data sources are often incomplete or geographically inaccurate. The Florida Department of Transportation (FDOT) has maintained accurate maps of local roads in the region since the mid 1930s, and I chose to use them for this study. FDOT maps are published for each county and are based on both aerial photographic interpretation and field checking. The maps undergo fr equent revisions ; new maps were redrawn every ten to twenty years from the 1950s through the 1990s. Digital copies of the maps were retrieved for the 1970s to 1990s from the Publication of Archival and Museum Materials (PALMM) project of the State University System of Florida ( http://palmm.fcla.edu ; Figure 3 4). Although the publication and revision dates for individual county maps were not uniform, the latest revisions were within four years of the middle of each decade, with one exception (Table 3 6 ). For this study, the composite data layer titled 1975, for example, was based on FDOT county road maps with publication dates ranging from 1971 1975 with revisions occurring as late as 1977. To georeference the raster ma p images, a base map of mosaicked aerial photograph was created using orthophoto aerial photography from 2004. These images are available from the Land Boundary Information System (LABINS), a digital repository of online survey information and maps for the State of Florida ( http://data.labins.org/2003/) operated by the Bureau of Surveying and Mapping in the Florida Department of Environmental Protection (FDEP). This base map was created using Erdas Imagine (ver. 9 .0) software which was also used for georeferencing (Figure 3 5).
62 After georeferencing the county highway maps, the roads were converted to vector data structures in a GIS using the ArcInfo level of ArcMap (ver. 9.1), ArcGIS software by ESRI. While all roads on the principal map were digitized, only the main roads connecting through urban areas were represented. Inset maps were not incorporated, thus, the local urban roads within municipal boundaries, as defined in the FDOT maps, were excluded from the s tudy. The first composite map (Figure 3 6) was the 1975 data layer and served as a template for the subsequent 1985 and 1995 data layers. The fourth time period in the study, represented by the 005 data layer also originated from the FDOT. In contra st to earlier data these were vector data layers, and, while methods for including them differed slightly, the same level of detail was maintained. Current county highway maps from the FDOT are available from their website ( www.dot.state.fl.us/surveyingandmapping /). Prior to the late 1980s FDOT field personnel would include local, private roads in their surveys. Subsequently questions over property rights and liability, and concerns for the safety of field personnel led to a decision to stop verifying these roads and, thus cartographers stopped showing them on published maps (A. Shopmyer, personal communication 9 April 2007). Nonetheless, private roads shown on earlier drawn maps in most cases still exist. Field technicians working for decades within the FDOT became very familiar with back roads in the regions where they work and were able to indicate where they existed. In such cases, the FDOT maintained the roads in their data files, but they were n ot shown on the publicly available maps. To solve this problem for later dates I obtained the most recent data from the FDOT Surveying and Mapping Office that included private roads where known to exist. Once digitized, the roads data were edited for tempo ral and topological consistency. The logic was that if a local road appeared in early data layers as well as the 2005 layer and the 2004
63 aerial photograph, then it probably existed in the intervening years (i.e. 1985 and/or 1995). Roads are highly persist ent landscape features, and I considered it unlikely if not implausible that a road would exist in the 1970s or 1980s, disappear and then reappear one or two decades later. In these cases, the missing roads were copied and inclu ded in the intermediate dat es. Topology rules were established to eliminate topological inconsistencies and to verify the topology of each feature class (Table 3 7 ). Once the topology was verified, individual errors were corrected by editing the feature class. Further network -based analysis of the data required their conversion into network dataset data types which interpret line features as a network data structures within ArcGIS. In the SFRW road network datasets, the topology rules dictated that roads connected at any junction. The network dataset consisted of the original digitized roads and a point data layer of junctions that was automatically created when the network topology was built in ArcGIS. The junction points were then included as a separate layer of point features in the geodatabase. These were, in turn, edited to remove junctions that occurred at the few overpasses in the study area, the majority of which were on the I 75 corridor. Data A nalysis To characterize the road networks, I calculated twenty three descript ive metrics for each dataset For six metrics describing elements of the region or network, I calculated summary statistics. I also examined the direction and rate of change of these metrics over the study period. The total length of the road network was calculated by summing the values for all the lines in each dataset. The road density was determined by dividing the total length of roads by the area delimited by the study boundary line. DTR measures were calculate d for a 30 m x 30 m grid of the study area, with the distance calculated as the Euclidian distance from the center of the grid
64 cell to the nearest road. Summary statistics for DTR were observed for each data layer. RV was calculated according to methods de scribe by Watts, et al. (2007b) For this analysis, the value of e was calculated as half the number of edges shown in the network dataset properties. The value for v was simply the number of junctions given by the network dataset properties. These basic values were used to derive the graphtheoretic indices such as , and I identified the number of vertices along the boundary of the study area ( vBOUND) and subtracted this from v to give an adjusted number. I used this figure in calculations of the proportion of first -order vertices ( vo=1:vo>1) and w. Further study of the road networks and their elements, including diameter and other shortest path measures, required the use of softw are designed for network analysis. However, the network analysis functions available in ArcGIS do not include a number of indices developed by transportation geographers. For this, I used igraph (ver. 0.5; Csrdi and Nepusz 2006) a graph analysis module for use with the statistical language and environment R (ver. 2.6.7 and 2.7.0; R Development Core Team 2008) I transformed the network datasets into graphs that could be analyzed using these applicatio ns. Using igraph I calculated the of each of the road networks along with average path length s and average length s of geodesics. In addition, the number of alternate paths was found for each date, and their paths were described as a list of adjacent vertices. Each was mapped to the set of junctions in the GIS data layers (Figure 37), and their distances were calculated. The value of d for each date was derived from t he mean of the diameter distances. With this value, was calculated for all four dates. In igraph, the networks were imported as directed graphs, wherein each road segment had two unique directions, i.e., from vertex j to vertex k and vice -versa. In this situation, a vertex
65 with o = 3, was reported by igraph with a degree value of six ( D = 6). To calculate w I summed degree values ( n vv D1) ( ), and subtracted the number of vertices with D = 2 ( 2 ) ( v Dv ), or first order ( o1) verti ces, and vBOUND (Equation 3 1). This eliminated boundary vertices, and gave a value of one to all o1 vertices. ) ( ) (1 2 ) ( n v BOUND v Dv v v D w (3 1) The proportion of o1 vertices ( vo=1:vo> 1) gives the ratio the number of o1 vertices to all vertices in the network. First-order ( o1) vertices correspond to the end points of roads, so their presence indicates the existence of roads that are joined at one end. In road network terms, these are dead -end roads, such as one would e ncounter in a cul -de -sac. More typically, in this study area, these are local roads that lead into forest plantations ( e.g., logging roads), farms or small clusters of residential or commercial buildings These small roads are more than simply driveways encountered in a typical urban environment. They are often times graded roads that provide vehicular access into iso lated areas distant from higher -order improved roads. For each date, I calculated vo=1:vo> 1 for the entire network, excluding vBOUND. Multiplied by 100, this could also be interpreted as the percentage of o1 vertices in the network. The final group of metrics used was a series of graph related measures of individual elements of the networks. For vertices, I calculated values for order, b etweenness, closeness and shortest path accessibility ( Ai). For edges, I calculated values for betweenness. Vertex order ( o ) was determined by halving the vertex D values reported in igraph and removing vBOUND from the set before calculating summary stati stics. Vertex closeness and vertex and edge betweenness were calculated using igraphs automated algorithms. Calculating Ai in igraph was a process that involved several steps: 1), ensuring that the graph was a single, connected graph; 2), collapsing it
66 in to an undirected graph; 3), creating a matrix of shortest path values between each vertex; and 4), summing the rows of the matrix. This column of sums was itself summarized, and the mean, minimum, maximum and standard deviation values were found. Results I calculated the metrics described above for each of the four datasets, and observed how they changed from one date to the next. Results for the twenty-three metrics were grouped for interpretation into three categories of extent, connectivity and circuitr y, and accessibility. Extent Measures of extent included basic network properties, topological network characteristics and certain summa ry values of vertices (Table 3 8 ). Network properties of extent were measured by L v and e Basic topological properti es of the network were measured by the average path length ( APL ), d and Important information about the network extent and structure was gleaned from summaries of vertex attributes. These included w, the proportion of o1 vertices, and the summary values of o All of the measures characterizing the road network extent increased during the thirty-year period of the study. In 1975, L was 5887 km and increased by 23% to 7304 km in 2005. Length increased during each ten year i nterval, but the greatest increase occurred in 19952005 (Table 3 9 ). The increase in length was accompanied by an increase in network elements. The number of road segments ( e ) increased by 55%, and the number of junctions ( v) increased 69% during t he thir ty -year period As with length, e and v increased most in 19952005. Network extension was also measured by increases in shortest -paths. Geodesics are the shortest topological path between any pair of vertices. Over thirty years, the number of geodesics in the SFRW road network nearly tripled from 13,017,753 in 1975 to 37,087578 in 2005. Diameter ( ) and APL both increased during each ten -year increment of the study. The changes
67 in these two metrics were very similar, ranging from 3.6 to 9.8% f or any given interval (Table 3 10). As with other measures of extent, the greatest change occurred in 1995-2005. Increase in extent was also calculated in the related values of d and the ratio between the L and d Of all the metrics, d and were the only two that did not increase or decrease continually during all three decadal intervals. The number of alternate pathways was found for each dataset: in 1975, there were eight paths; in 1985, there were two; in 1995, there were four; and in 20 05, there was only one path. The average of their individual lengths gave the value for d which grew from about 124 km in 1975 to 152 km in 2005. As a network becomes more intricate, d will increase. As a network becomes more connected, values for d will decrease. In the first decade d increased slightly (Table 3 10), but decreased by nearly 4% in the second decade and increased sharply in the final decade. Altogether, d experienced a total increase of 24% from 1975 to 2005. The shape index ( ) resp onded to changes in d and L by increasing during the first two decades and then decreasing to 47.98 km in 2005, a value only slightly greate r than the 1975 value (Table 3 10). In 19851995 the slight decrease in d (<4%) was accompanied by an increase in L causing to increase by nearly 10%. However by the third decade, d was substantially greater than all previous decades (Figure 3 7). This change caused a drastic reduction in In other words, according to the shape of the network was very similar in 2005 and in 1975. Vertices of the network provided important clues about the networks structure and potential function. Certain summaries of attributes having to do with o documented the characteristics of network expansion from 1975 to 2005. In the SFR W, w increased from 1975 to 2005 as v increased. As with the increase in v the greatest change in w at 28%, came in 19952005 ( Table 3 11). This increase was accompanied by a decline in summary values for o
68 Summary statistics indicated a slight decrease in mean o for the road networks, from 2.6 in 1975 to 2.5 in 2005. The proportion of o1 vertices (vo=1:vo> 1) increased from about 22% in 1975 to nearly 28% in 2005, reflecting a drastic increase (+160%) in the number of o1vertices in the network. Unlike other metrics of road network extent, the greatest increase in vo=1:vo> 1 came in the second decade, 19851995. Connectivity Connectivity of the road network, which was measured with and was expected to increase over time. Also in this category were complex indices that relate length of the network to network elements, with implications about functional capacity, including and (Table 3 12). These values were expected to decrease as the network developed over t ime. In the SFRW road network circuitry increased overall, but relative connectivity declined. The cyclomatic number ( ), t he minimum number of road segments that must be removed to make the network free of circuits increased as the extent of the road net work increased, suggesting increasing connectivity. However, the ratio of the actual number of circuits to the maximum number possible, decreased over time. In 1975, was 15.2%, but this had declined to 12.4% by 2005, a decrease of more than 18%. Relat ive road network circuitry ( ) decreased most during 19851995 (Table 31 3 ). As with indices of road network connectivity and also slightly decreased in the SFRW. The beta index ( ) declined from 1.304 to 1.248, and decreased from 0.435 to 0.416, at a nearly identical rate for all intervals. Although the magnitude of change for and was under 5% in all decades, over the entire thirty years, their values indicated a continual negative trend. Changes to network stru cture, connectivity and functional capacity were measured using and which related the length of the network to various network elements and properties Values for generally decreased as expected from 1975 to 2005. Average edge length ( ) decreased b y
69 200 m per edge, from 0.87 km to 0.67 km, a drop of 23% from 1975 to 2005. Iota ( ) relates length of the network to its function, as given by w, the weighted sum of the network vertices. Due to the very high number of vertices in the road network, the va lue for was very small to begin with, at 0.23 km in 1975. This value dropped steadily by 50 m to a value of 0.18 m in 2005 a decline of about 22%. Accessibility Accessibility was described in two ways: accessibility of the landscape and accessibility of the road network. While these are two rather different ideas, they are related through the physicality of the road network which fundamentally defines them. In this study, road density, DTR, and RV measured accessibility of the landscape (Tabl e 3 14). Net work accessibility was measured using unweighted shortest -path methods including Ai, closen ess and betweenness (Table 3 15). While closeness and betweenness are considered measures of centrality, they also describe a relative level of accessibility at the individual vertex or edge scale, and are related to Ai in the use of geodesics. Numerous studies have shown that, as regions develop over time, accessibility increases. Not surprisingly, road density in the SFRW, as measured by the ratio of the length of the network to the footprint of the study area, increased over time. Road density was well above 1 km/km2 at the beginning of the study period. This increased to about 1.5 km/km2 by 2005. The rate of change also increased throughout the study period, from 3% in the first decade to 14% in the third decade. After thirty years, road density in the SFRW had incr eased by nearly 24% (Table 3 16). As road density increased, values for DTR decreased. Summary statistics for DTR, which included mean, median, maximum and standard deviation all decreased during the period of study. From 1975 to 2005, t he average distance to a road decreased by 83 m. The median was
70 consistently less than the mean value for DTR; however the gap between these two values decreased from 111 m to 74 m over thirty years. Maximum DTR decreased modestly in the first two ten -year intervals, but in the third decade, the decrease was dramatic. In 1975 the farthest point from a road was over 3 km. But in 2005, this distance had decreased to about 2.5 km. This point, located in the Santa Fe Swamp, a protected area managed by the Suwannee River Water Management District (Figure 3 8), remained relatively fixed as surrounding roads encroached upon it over time. Roadless volume (RV) relates the mean DTR t o the area of the footprint. By measuring the volume of unroaded space, RV is useful as an index of the degree of landscape penetration by roads. Proportionate to the changes in DTR, RV also decreased substantially. In the SFRW, RV declined by 312 km3 during the study period, a decrease of 18% from 1975 to 2005 (Table 3 1 4 ). As a road region develops, accessibility of the road network is expected, sometimes even assumed to increase. In the SFRW, summary values of Ai, including the mean, minimum, maximu m and standard deviation all increased by nearly 100% during the study period. Growth in the minimum Ai increased from around 16% in the first decade of the study to 43% in the final decade (Table 3 17). All other Ai summary measures followed suit, increas ing substantially in 19952005. The mean Ai doubled over time, and the minimum value, Min Ai, representing the most accessible intersection, followed suit. Vertex betweenness and edge betweenness are measures of centrality that indicate the relative importance of an individual element in the overall structure of a network graph. Betweenness increases as a vertex or edge becomes more central to a network. Summary statistics were calculated for each of these values to represent trends across the network. In t he SFRW road network, average values for the betweenness of junctions, or mean vertex
71 betweenness (MVB), increased from 1975 to 2005. The increase in the MVB was dramatic; more than doubling over thirty years (Table 3 18). Maximum vertex betweenness also i ncreased. The average value for the betweenness of road segments, or mean edge betweenness (MEB), echoed MVB by more than doubling during the study period. Maximum values for edge betweenness also increased, corresponding to a steady increase in variabilit y of edge betweenness as measured by standard deviation. Vertex closeness measured the relative topological closeness of a junction to the rest of the network. As with betweenness, closeness increases as an element becomes more central to the rest of the network. Summary statistics of closeness showed a general decrease. The largest decrease in mean values for closeness occurred in 19851995. Minimum and maximum values for closeness decreased in the same fashion, along with a corresponding decrease in standard deviation of closeness values. Discussion Transportation geography theory postulates that as economies develop, road networks expand and interconnect. While the road network extent and connectivity are expected to increase, new connections between es tablished roads reduce topological distances decreasing the path length between pairs of junctions, including network diameter. This in turn leads to greater functional capacity of the network and greater accessibility of its junctions and road edges. Gene rally, as road networks expand and develop, road densities increase (Gould 1960) Conversely, distance to road and roadless volume decreases over time with road network expansion (Watts et al. 2007b) Based on theories of transportation network development, I hypothesized that the road network in the SFRW would become more extensive and connected from 1975 to 2005. Increased connectivity in the road network should lead to an overall increase in network accessibility, road density and landscape accessibility.
72 Extent and network development The extent of the SFRW road network increased from 1975 to 2005. All measures of extent clearly indicated that the networ k expanded considerably, especially in the last decade, 19952005. Overall length ( L ) and the number of network elements, including e and v increased as expected. Conversely, measures of topological distance did not decline as expected. Values related to the diameter of the network increased over thirty years, mostly from 1995 to 2005, including average path length, and d The increase in the topological breadth of the network meant that the two most distant points in the network were further apart in 2005 than in 1975. That is to say, as more junctions and road segments were added, the distance between these two points increased. Summary measures of vertices als o provided insights about structural changes in the network during the expansion process. Network weight ( w ), a weighted value of all vertices in the network, normally increases as a network expands and adds new vertices. Vertex order (o ) is expected to in crease as the network becomes more interconnected, so the relative presence of o1 vertices should diminish and the ratio vo=1:vo> 1 should decline. Network weight ( w ) did, indeed, increase as expected. As it increased, however, mean o declined and the numbe r of o1 vertices more than doubled. The proportion of first order vertices ( vo=1:vo> 1) moreover increased, especially during the second decade. The increase in this class of vertices, and, more importantly, the increase in the proportion of these vertices relative to higher -order vertices, i.e. road intersections, suggests the active development of dead-end roads. A major cause of the SFRW road network expansion therefore, particularly in 19851995, was the creation of new dead-end roads penetrating the lan dscape. This process is clearly illustrated in a map that compares the 1975 road network with the 2005 road network (Figure 3 9).
73 Network diameter is related to the overall shape and intricacy of the road network, summarized by Large values for indica te an elaborate transportation network, corresponding with more developed regions as in the SFRW. In this case, increased for the first two decades and decreased in the third. However, in view of the increases in L, geodesics and network elements, the SFRW road network was not less intricate in 2005 than in 1995. This apparent contradiction highlights the importance of considering alternative evidence when interpreting such indices. An increase in can be an indicator of road network extension (increas ing L ) and/or interconnection (decreasing d ), whereas a decrease in could signify contraction of the network (decreasing L ) and/or elaboration (increasing d ). The value of was high to begin with in 1975. In this case, the increase in during the first two decades was due to greater relative increases in L The reduction of in the final decade was due to the greater relative increase in d All the while, d and L were simultaneously increasing, indicating a process of extension and ela boration of the road network. Connectivity and Functional Capacity of the Network Garrison and Marble (1962) showed that countries that were more economically developed had higher levels of highway connectivity. It is often presumed that road network conne ctivity will increase as a region develops, and that this will be reflected in higher values for connectivity metrics of and In a road network, increases in and are associated with increasing circuitry. Increases in indicate an increase in the number of road segments per junction. Values of approaching one are associated with road networks approaching the maximum number of road segments possible, which is 3( v 2) (assuming a planar graph). As the SFRW network expanded increased as expected. With increased there were more alternative routes through the network in 2005 than in 1975. The increase in circuits suggested greater connectivity in the network. However, a decrease in other measures implies
74 that connectivity a nd circuitry declined, particularly from 1975 to 1995. The actual number of road segments and circuits relative to the maximum number possible, according to and declined over thirty years. The connectivity index also decreased in all three decades. While the network expanded, the expansion was not characterized by interconnection, and so, the changes registered as decreases in connectivity measures, apparently contradicting theorized increases in connectivity. In the classification system of basic network configurations (Table 3 3), the index values indicate that the road network in the SFRW was within the range of a spinal configuration, while index values placed it in the range of a grid configuration (Taaffe and Gauthier 1973). Although the indices classified the networks differently, they agreed that the changes caused by the addition of many small low order roads were pushing the road networks closer to a spinal configuration. Although and other measures of extent showed that the road network was becoming more elaborate, on the whole, the addition of so many dead-end roads was driving the network toward a more tree -like topological configuration. Changes in the complex indices of networks such as and gauge the relative changes in network structural characteristics. Eta ( ), the average edge length, and will increase with expansion of the road network (increasing L ) and/or a decline in the number road segments (for ) or network weight (for w ). These indices will decrease as the n etwork contracts (decreasing L ) and/or and increase in the number road segments (for ) or network weight (for w). The negative trends in from 1975 to 2005 indicate d a shortening of the average length of road segments as the numbers of road segments and junctions increased faster than the overall length of the network. The decrease in from 1975 to 2005, suggests that the capacity of the network to accommodate traffic flow increased over the period of study reinforcing the results of These
75 findings ar e consistent with correlations published by Kansky (1963) and Garrison and Marble (1962), who found that as economic development increased, values for these indices would decrease. While these indices changed as expected, in the SFRW, the decreases were li kely due to the overwhelming increases in dead -end roads and not interconnection or an increase in the number of higher -order junctions. This highlights the importance, as in the case of of viewing the results in light of alternative evidence about chan ges in the road network. Accessibility of the Landscape vs. Accessibility of the Road Network Accessibility within the study area, as described by measures of road density and distance to road s changed as expected, increas ing during the study period. Road density has always been well above 0.6 km/ km2a threshold for certain large mammals above which populations are affected (Forman et al. 1997) The continued increase in road density suggests that the landscape became less habitable to wildlife, although the effects of the increases of the 1980s may only now be observable due to a time lag (Findlay and Bourdages 2000) P rocesses of road network expansion and development that caused increases in road density also caused decreases in DTR and RV measures Changes to RV in the SFRW followed a similar trend discovered in a sixty year analysis of RV change in Colorado (Watts et al. 2007b) Clearly the landscape became mo re accessible to people using roads The in between spaces diminished as roads penetrated more of the landscape and fragmented previously contiguous unroaded areas. However, these map -based metrics told only part of the story. While road density and DTR implied that the landscape was more accessible by road, network analysis pointed toward declining accessibility in the road network itself, again, apparently contradicting the presumption that transportation systems become more concentrated and centralize d over time (Haggett and Chorley 1969) Accessibility of the road network was determined using me trics that describe the role of vertices and edges within the geodesics of the
76 network. These included Ai, as well as closeness and betweenness, both measures of centrality. Individual values were summarized to represent network trends. For all metrics, th e summary measures indicated that, on average, the accessibility of junctions and road segments diminished substantially over time with the sharpest declines occurring in 19952005. The decrease in accessibility can be attributed to a global increase in th e number and length of geodesics in the road network. On the whole there were more shortest -paths between any given pair of junctions in 2005 than there were in 1975. With the addition of junctions resulting from dead-end roads intersecting with connected, through roads, shortest paths between junctions became lengthier. As topological distances between the junctions in the road network increased, the value of Ai increased, and its reciprocal, closeness, decreased, affecting mean values for the entire netw ork. Likewise, with more geodesics the probability of an individual road segment or junction being a member of a geodesic increased, thereby raising the mean values for vertex and edge betweenness. This illustrate s an interesting and important point While accessibility to the SFRW landscape increased over time, the road network itself became less accessible This is because expansion of the network was mainly caused by dead -end roads penetrating previously inaccessible areas demonstrated by the decrease in RV. Notably both groups of accessibility measures experienced the greatest decadal change in the period from 19952005 (Table 313), corresponding with the very high rates of road network expansion. General Remarks and Further Research The SFRW is a rural environment with a growing population where extractive activities such as logging and agric ulture, are the primary land uses. Where comparable economic and land use conditions exist, similar patterns in the topology of rural road networks might be found Such a dynamic might be expected w herever there is an expanding rural frontier associated with
77 resource extractio n and land use intensification, for example, in Amazonia (Nepstad et al. 2001) Fur ther research could indicate whether or not this is the case. An idealized process of how a network could expand, with length increasing while connectivity decreases is shown in Figure 3 10. Over time, the network expands with the addition of edges at exis ting junctions and by splitting an edge (Figure 3 10A). By the end of the process numerous small lines have developed along the periphery of the network (Figure 310C). The length of the network has increased, along with its density. But these changes have occurred through a process of adding lines of penetration to the existing network (Taaffe et al. 1963) and not by a process of interconnection. The result of addin g semi connected edges is a lowering of connectivity indices. As software capabilities improve, network analysis of extensive and large spatial datasets will become more tractable. Suggested improvements to the analysis described in this study include the use of road attributes to create weighted analyses. For example, it is possible to use a classification system of road segment capacity or length to compute Ai and betweenness. The results of such an analysis could profoundly affect notions of accessibilit y in this scenario. Finally, the next step in this research is to relate network analysis results to the spatial characteristics of the data, tying the network analysis to spatial analysis via a GIS. This would provide opportunities for a range of geograph ic analyses that could combine information about network structure with information about the structure of the physical landscape, thereby helping to inform theories of social and ecological interactions. Conclusions The analysis of roads in the SFRW from 1975 to 2005 revealed the nature of changes in the structure of a rural road network. The study included both mapbased analyses of the road network and graphtheoretic based analyses of the network structure. The network included all
78 classes of roads, fr om small rural roads to interstate highways, but excluded local, urban streets in metropolitan areas. Based on these analyses, changes in the rural road network apparently contradict predictions from theoretical transportation models that road networks bec ome more connected and accessible as regions develop. The results of this study indicate that in the SFRW, processes of road network development have favored expansion rather than interconnection. Road network development from 1975 to 2005 included the add ition of new roads that penetrated into formerly unroaded areas presumably facilitating resource extraction and land use intensification. Accessibility to areas between the roads increased with increases in road network density and decreases in average di stances to roads. However, w hile accessibility to the landscape increased over thirty years the road network itself became less accessible The cumulative effect of adding many small local roads was to increase the overall extent of the network. S ince the roads were mainly dead end s they were only connected to the network at one end. Each time one of these small roads connected to the wider network it created a junction, thereby splitting the road in to two segments where there had been only one The cumulative topological effects of this repeated action overwhelmed the effects that any new connector roads might have had, resulting in a relatively less connected, less accessible and more elaborate road network with a greater capability of accessin g unroaded rural areas. Research by transportation geographers in the mid 20th century provided important models of transportation network evolution and regional development. The conclusions reported here were at variance with the se earlier findings One possible explanation is that previous analyses con centrated on higher -order systems at regional or national scale s excluding local roads The results reported here provide important new insights regarding road network dynamics at scale s
79 where many of the ecological effects of roads are experienced. It was shown that in the SFRW penetration of new roads into unroaded areas was so extensive as to affect overall road network connectivity and accessibility. The numerous ecological implications of these developments include land use intensification, increased landscape fragmentation, and increased roadrelated effects on the biota and natural processes of the watershed (Coffin 2007) Fur thermore, as persistent features in the landscape, these new inroads provide ready accessibility, thus hastening the potential for urbanization.
80 Figure 3 1. The study area: the Santa Fe River watershed in the southern reaches of the Suwannee R iver basin.
81 Figure 3 2. A planar graph with five vertices and eight edges. A B C D Figure 3 3. Roads in the SFRW study area. A) northbound I 75 in Alachua County, B) southbound SR 441 in Alachua County C) typical improved loc al rural road D) typical local unimproved or private road. vertex edge
82 Figure 3 4. The General Highway Map of Bradford County, Florida. An example of FDOT maps used for the study available online via the PALMM project (http://p almm.fcla.edu ). Figure 3 5. Alachua County 1973 Highway Map superimposed over the mosaic of aerial photography used as a base map for georeferencing.
83 Figure 3 6. The 1975 digitized road network for the SFRW.
84 A B C D Figure 3 7 SFRW road network showing diameter path(s) in red and basin outline in blue. A) 1975, B) 1985, C) 1995, D) 2005.
85 Figure 3 8. Distance -to -road values A) for 1975, B) for 2005. Darker green values are farther from the road. Red triangle indicates farthest point from any road. A B
86 Figure 3 9. Map showing a small area of the SFRW road network in 1975 (black) with accumulated changes in 2005 (red).
87 A B C Figure 3 10. Idealized process of network development with decrease in connectivity resulting from addition of first -order vertices Subfigures A, B, and C represent development of the road network over time. e = 25 v = 20 = 0.17 = 1.25 = 0.46 e = 20 v = 15 = 0.24 = 1.33 = 0.51 e = 15 v = 10 = 0.4 = 1.5 = 0.625
88 Table 3 1 Measures and indices of road networks. Class Index Description or equation (assuming planar graphs) Map -based indices Total length (L) Total length of the road network in distance units (meters, km, etc.) Total area (A) Total area of the network in area units (m 2 km 2 etc.) Road density (L/A) Density of roads per unit area or per capita. Distance to road (DTR) Euclidian distance to the nearest road. Value is usually calculated for a given point or a grid of points within the space of the network area (A). Roadless Volume (RV) A ( A is the area of the footprint area being analyzed and is the mean DTR of the study area (Watts et al. 2007b )) TNA Cyclomatic number ( ) v + p Diameter ( ) and diameter distance ( d ) x max y d (x,x) d(x,x) (units) : topological length in distance units (meters, km, etc.) Connectivity or gamma ( ) ) 2 ( 3 v e Alpha ( ) 5 2 v Beta ( ) v e Shape index or Pi ( ) d L Eta ( ) e L (in distance units ) Weight ( w ) 2 v ; ;1 > o 2 1 = o 1 2 1w v w w w w Structure index or Iota ( ) w L (in distance units ) Shortest path accessibility (Ai) n jj i d i A1) (
89 Table 3 1. Continued Class Index Description or equation (assuming planar graphs) SNA Closeness g j j i ij Cn n d n C1 1) ( ) ( Betweenness jk k j i jk ij Bg n g n C / ) ( ) ( Note: These indices were derived from mapped characteristics and graph theoretic indices of transportation network analysis (TNA; Garrison and Marble 1962; Kansky 1963; Taaffe and Gauthier 1973; Rodrigue et al. 2006) and social ne twork analysis (SNA; Wasserman and Faust 1994; Csrdi and Nepusz 2006) In this table e is the number of edges (road segments), v is the number of vertices (junctions), o is the vertex order, gjk is the length of the geodesic between vertices j and k and p is the number of separated subgraphs (for this study p = 1). Table 3 2 Basic measures of graphs commonly used to describe transportation networks (Garrison and Marble 1962; Kansky 1963; Cliff et al. 1979) Graph measure Description Edges ( e ) number of edges Vertices ( v ) number of vertices Subgraphs ( p ) number of subgraphs Order ( o ) vertex order Table 3 3. Alpha and gamma index ranges for basic network configurations (Taaffe and Gauthier 1973) Network configuration lower limit upper limit lower limit upper limit Spinal 0 0 Grid > 0.50 < 0.66 > 0 < 0.50 Delta Table 3 4 1975 to 2005 population estimates for the top five counties (by area) in the SFRW based on 1970 to 2000 census data (U. S. Census Bureau, 2008) County 1975 Pop. est. 1985 Pop. est. 1995 Pop. est. 2005 Pop. est. Alachua 132,600 169,515 195,535 231,864 Bradford 17,300 22,629 23,854 28,023 Columbia 30,600 39,312 49,436 63,939 Gilchrist 4,900 7,724 12,289 16,230 Union 9,900 10,610 12,180 14,652 Top 5 Total 195,300 249,790 293,294 354,708
90 Table 3 5. Average annual population growth rate (percent) for the top five counties (by area) in the SFRW based on 1970 to 2007 estimates County 1975 1985 1985 1995 1995 2005 1970 2007 Alachua 2.36 1.53 1.70 2.29 Bradford 2.99 0.54 1.43 1.86 Columbia 2.68 2.29 2.67 2.72 Gilchrist 4.65 4.73 2.85 4.35 Union 1.49 1.31 1.63 1.73 Top 5 Total 2.48 1.67 1.88 2.36 Florida 2.87 2.31 2.21 2.72 Table 3 6 List of source FDOT maps and their publication dates. Composite data layer County Publication Date / Last revisions 1975 Alachua 1972 / 1973 Baker 1974 / 1975 Bradford 1973 / 1974 Clay 1971 / 1977 Columbia 1973 / 1974 Gilchrist 1974 / 1974 Suwannee 1975 Union 1973 / 1974 1985 Alachua 1976 / 1983 Baker 1974 / 1981 Bradford 1973 / 1983 Clay 1981 / 1988 Columbia 1979 / 1983 Gilchrist 1983 / 1984 Suwannee 1975 / 1984 Union 1988 / 1989 1995 Alachua 1989 / 1993 Baker 1996 / 1999 Bradford 1998 / 1999 Clay 1995 / 1999 Columbia 1989 / 1990 Gilchrist 1998 / 1998 Suwannee 1991 / 1992 Union 1997 / 1999
91 Table 3 7 Topological rules from ArcGIS used to verify and correct the SFRW road networks. Rule Description 1 Must be single part The line vector must be one single part, not a multi part element. Multipart lines should be merged into one element or edited to remove extra parts. 2 Must not self overlap The line vector must not overlap itself. Overlapping lines should be removed. 3 Must not have pseudo Nodes must occur at the beginning and end of a line. Nodes occurring in the middle of the line are pseudo-nodes and should be removed 4 Must not have dangles Lines should be connected to other lines at their ends. Exceptions to this rule are dead-end roads or roads at the edges of the study area. L ines that undershoot an intersection should be extended, and lines that overshoot an intersection should be clipped. Table 3 8 Measures of road network extent in the SFRW. Note: m easures of extent include length ( L ), the number of vertices ( v ), the number of boundary vertices ( vBOUND), the number of edges ( e ), the diameter of the graph ( ), the average path length (APL) the average diameter distance ( d ), the network weight ( w), the proportion of first -order vertices (vo=1:vo>1) and summary statistics for vertex order ( o ). For all datasets minimum vertex order was one Measure 1975 1985 1995 2005 increase + decrease L (km) 588 7 607 8 643 1 730 4 + v 5162 5765 6679 870 5 + v BOUND 227 226 230 283 + e 6734 7479 8480 10865 + 125 133 139 152 + APL 43.81 45.39 47.60 52.24 + d ( km ) 123.93 126.48 121.75 152.21 + 47.5 0 48.05 52.8 2 47.98 +/ w 25,373 28,145 31,695 40,576 + v o=1 :v o>1 0.22 3 0.238 0.26 3 0.275 + o mean max SD 2.61 5 1.011 2.59 6 1.018 7 2.5 4 6 1.030 2. 50 6 1.02 1 + +
92 Table 3 9 Percent change matrix for measures of road network extent : Length (L ), vertices (v ) and edges ( e ). 1975 1985 1995 1985 L 3.24 v 11.68 e 11.06 1995 L 9.24 5.49 v 29.39 15.85 e 25.93 11.80 2005 L 23.31 19.07 13.58 v 68.64 50.98 30.33 e 55.23 39.93 28.13 Note: b old type indicates the greatest change for any ten or twenty year interval. Table 3 10. Percent change matrix for measures of road network extent: diameter ( ), average path length ( APL ), diameter path length ( d ) and shape index ( ). 1975 1985 1995 1985 6.40 APL 3.61 d 2.06 1.16 1995 11.20 4.32 APL 8.65 4.64 d 1.76 3.74 11.20 9.93 2005 20.30 13.67 9.35 APL 18.57 14.39 9.75 d 23.82 20.34 25.02 1.01 0.15 9.16 Note: b old type indicates the greatest change for any ten or twenty year interval.
93 Table 3 11. Percent change matrix for measures of road network extent: network weight ( w ) proportion of first -order vertices ( vo=1:vo>1), and mean vertex order (MVO) 1975 1985 1995 1985 w 10.92 v o=1 :v o>1 6.93 MVO 0.77 1995 w 24.92 12.61 v o=1 :v o>1 18.09 10.50 MVO 2.68 1.93 2005 w 59.92 44.17 28.02 v o=1 :v o>1 23.32 15.55 4. 5 6 MVO 4.22 3.48 1.58 Note: b old type indicates the greatest change for any ten or twenty year interval. Table 3 12. Measur es of road network connectivity and structure Note: i ndices for circuitry and connectivity of the road networks includ e cyclomatic number ( ), alpha ( ), beta ( ), and gamma ( ). Complex indices of road network structure and function includ e eta ( ) and iota ( ). Index 1975 1985 1995 2005 increase + decrease 1573 1715 1802 2161 + 0.152 0. 149 0. 135 0. 124 1.304 1.297 1.270 1.248 0.435 0.433 0.423 0.416 (km) 0.87 0.81 0.7 6 0.6 7 (km) 0.23 0.21 0.20 0.18
94 Table 3 13. Percent change matri x for measures of road network connectivity and structure 1975 1985 1995 1985 9.03 2.36 0. 5 4 0.46 6.90 8.70 1995 14.56 5.07 11.45 9.31 2.61 2.13 2.76 2.31 14.47 6.17 13.04 4.76 2005 37.38 26.01 19.92 18.50 16.53 7.97 4.29 3.78 1.73 4.37 3.93 1.66 22.99 17.28 11. 84 21.74 14.29 10.00 Note: b old type indicates the greatest change for any ten or twenty year interval. Table 3 14. Measures of landscape accessibilit y. Note: i ndices include map-based measures of road density (L/A), roadless volume (RV), and summary statistics for distance to road (DTR). Measure 1975 1985 1995 2005 increase + decrease L/A 1.18 1.22 1.29 1.47 + RV(km 3 ) 1724.16 1683.51 1592.28 1412.17 DTR (m) Mean Median Max SD 367 256 3118 321.4 339 247 3092 319.6 320 240 3090 304.2 284 210 2580 273.4
95 Table 3 15. Measures of road network accessibility. Vertex and edge measures 1975 1985 1995 2005 increase + decrease A i Mean Min Max SD 223,553 154,968 417,481 43,670 258,425 179,734 498,730 51,237 313,771 222,189 597,181 61,509 449,948 309,633 829,757 87,457 + + + + Vertex betweenness Mean Max SD 215,957.4 5,566,774 488,813.1 249,621.9 5,498,836 563,757 303,226.5 9,472,276 745,164.5 436,673.4 14,868,844 1,191,422 + + + C loseness Mean Min Max SD 0.00973 0.00019 0.01123 0.00127 0.00858 0.00017 0.00979 0.00112 0.00746 0.00015 0.00838 0.00097 0.00692 0.00012 0.00784 0.00084 Edge b etweenness Mean Max SD. 84,705.3 2,679,801 199,899.9 98,374.8 2,708,032 229,336.7 121,976.0 4,571,242 307,206.5 178,344.3 6,504,066 498,157.2 + + + Note: i ndices include graph measures of shortest path accessibility ( Ai), vertex betweenness, closeness, and edge betweenness. For all datasets minimum betweenness of vertices and edges was zero. Table 3 1 6 Percent change matrix for measures of landscape accessibility : road density ( L/A ), roadless volume (RV ) mean distance to road ( Mean DTR ), and maximum distance to road ( Max DTR ). 1975 1985 1995 1985 L/A 3.39 RV 2.36 Mean DTR 7.6 3 Max DTR 0.83 1995 L/A 9.32 5. 74 RV 7.65 5. 42 Mean DTR 12.81 5.73 Max DTR 0.90 0.0 7 2005 L/A 2 4 58 20.49 13.95 RV 18 .10 16.12 11.31 Mean DTR 22.62 17.04 11. 25 Max DTR 17. 26 16.57 16.5 1 Note: b old type indicates the greatest change for any ten or twenty year interval.
96 Table 3 1 7. Percent change matrix for measures of shortest -path accessibility in the road network: mean shortest -path accessibility ( Mean Ai), and minimum shortest -path accessibility ( Min Ai). 1975 1985 1995 1985 Mean A i 15.60 Min A i 15.98 1995 Mean A i 40.36 21.42 Min A i 43.38 23.62 2005 Mean A i 101.27 74.11 43.40 Min A i 99.81 72.27 39.36 Note: b old type indicates the greatest change for any ten or twenty year interval. Table 3 1 8. Percent change matrix for measures of mean vertex betweenness ( MVB ), mean edge betweenness ( MEB ) and mean closeness ( CLO ). 1975 1985 1995 1985 MVB 15.59 MEB 16.14 CLO 11.82 1995 MVB 40.41 21.47 MEB 44.00 23.99 CLO 23.33 13.05 2005 MVB 102.20 74.93 44.01 MEB 110.55 81.29 46.21 CLO 28.88 19.35 7.2 4
97 CHAPTER 4: LANDSCAPE DYNAMICS IN THE SANTA FE RIVER WATERSHED FROM 1975 THROUGH 2005 Introduction Transportation systems provide essential conduits for movement. Transportation geography conceptualizes movement primarily as the transfer of people, resources, ideas, information and goods between locations (Lowe and Moryadas 1975) Transportation systems make resources accessible and therefore usable by people. By reducing the cost of travel between locations, transportation systems are a requirement for economic s pecialization and development. Human land use, including all forms of primary production such as agriculture and forestry, are fundamentally affected by the presence or absence of accessible transportation (Lowe and Moryadas 1975) Typically, transportation in rural areas is via road networks which are easily discerned as relatively permanent landscape signals of human activity. R oad networks facilitate land use and land-cover change by providing access to larger functional region s (Chomitz and Gray 1996; Geist and Lambin 2001) While road networks are essential to the economic development of regions and provide critical societal functions, they are physical structures that participate in landscape ecological dynamics (Bennett 1991; For man and Alexander 1998) This characteristic of road networks, albeit important, has largely gone unstudied by transportation geographers. Roads can act as ecological barriers, conduits, sources and sinks. R oads and their traffic cause ecological chang es to the surrounding landscape, affecting ecosystem function and structure at various scales (Forman et al. 2003) There may also be cumulative effect s of road networks that ha ve gone unaccounted for in studies that examine the discrete effects of individual road s Studies of landscape fragmentation at national and state levels support the conclusion that highway networks fragment the unroaded landscape into ever -smaller patches (Heilman Jr. et al. 2002;
98 Jaeger et al. 2007) Agriculture was also found to be an important factor in forest fragmentation in rural watersheds where agricu lture is a key economic activity and forest is the natural / historical land cover (Girvetz et al. 2008) The research described here explores how the dynamics of road network development and la nd cover affected patterns of landscape fragmentation. Specifically, how did road-related fragmentation patterns differ between forested and agricultural land -cover types? Did fragmentation vary across space and time? How did these two land-cover types int eract as the road network evolved? The growth of human populations and economic development are significant contributors of land use intensification which results in habitat modification and loss (Lambin et al. 2001) These, in turn, reduce habitat connectivity and increase isolation of plant and animal populations Increased isola tion is particularly problematic for species that require contiguous preferred habitat for dispersal and movement (Tischendorf and Fahrig 2000b) L oss of connectivity and population isolation can cause problems such as genetic drift, inbreeding, and local extinctions which increase the likelihood of extinction of entire species (Schonewald-Cox et al. 1983; Ehrlich 1988; Fahrig 2001) Consequently, biodiversity can be negatively affected (Fahrig 2003) potentially reducing the resilience of entire ecosystems (Folke et al. 1996) with serious repercussions for global evolutionary processes (Myers 1988) Reducing the isolation of species in a landscape is a critical s trategy for maintaining the viability of wild animal and plant populations (Forman 1995) The characterization and modeling of landscape connectivity dynamics are essential features of predictive habitat models and spatial simulations (Gilpin 1996; Tischendorf and Fahrig 2000b; Tewksbury et al. 2002; Tischendorf et al. 2005)
99 Landscape connectivity implies a relationship between the struc ture of the landscape and its ability to impede or facilitate movement, hence, as a concept, it is species and landscape specific (Tischendorf a nd Fahrig 2000b, p. 8) Where a river is a barrier to one species, it may be a movement corridor for another species. Habitat fragmentation, which is the breaking apart of habitat (Fahrig 2003, p. 509) also implies a functional -structural relationship. If connectivity and fragmentation are both landscape and species -specific concepts, then they can be inversely related. In such a paradigm, landscape fragmentation occurs when barriers in the landscape impede the move ment of species, increasing isolation and reducing connectivity. A number of remote sensing and geographic information system (GIS) based techniques have been developed recently to measure landscape structure, in an effort to model the connectivity of habitats (Theobald et al. 1997; T ischendorf and Fahrig 2000a; Urban and Keitt 2001; Opdam et al. 2003) Land -cover information derived from satellite imagery is often used to represent landscape elements in a GIS. There are numerous metrics that describe the characteristics of landscap e structure (O'Neill et al. 1988; Gustafson 1998; McGarigal 2002) and can be used to measure the geometry of landscape elements, such as patches and corridors. However, landscape structural indices can be inconsistent in their cor relation with functional measures of connectivity such as dispersal success (Tischendorf 2001) O nly a few link landscape structure with ecological function (Tischendorf and Fahrig 2000a; Tischendorf and Fahrig 2000b) Fr agmentation studies that consider the functional aspects of fragmenting elements can potentially provide for more ecologically meaningful analyses. For example, if roads exacerbate the isolation of particular species, then their inclusion in an analysis of landscape fragmentation will be meaningful for those species that are sensitive to roads, but not necessarily for others.
100 T est ing the link between measures of structural and function al connectivity often requires the identification of focal populations, s pecies or guilds (Bunn et al. 2000; Bruinderink et al. 2003; Chardon et al. 2003) and knowledge about behavioral responses to landsc ape features Then again, r ecent reviews of habitat connectivity research recog nize the difficulty of understanding behavioral responses in movement patterns for many species (Brooks 2003) and suggest avenues for further research into the relationship between structural connectivity and functional aspects of landscape networks (Tischendorf and Fahrig 2000b; Olff and Ritchi e 2002; Goodwin 2003) without relying on species -specific behavioral information The concept of effective mesh size ( meff), developed by Jaeger (2000) explicitly uses information about barriers in the landscape to characterize landscape fragmentation (Jaeger et al. 2006) The meff metric is an expression of the probability of two points chosen randomly in a region being connected (Jaeger et al. 2008, p. 742) which was revised from its original form to account for cross boundary connecti ons, i.e. the CBC procedure (Moser et al. 2007) Values are calculated for planning units which are the areas of interest. These could be watersheds, counties, physiographic regions, or any meaningful area. In this approach, barrier elements penetrate and fragment the landscape, preventing the connection of two locations. Fragmenting elements can include any barriers that prevent accessibility such as roads, large rivers, mountains or particular land cover types (Jaeger et al. 2008) The selection of fragmenting elements can vary, depending on the choice of organism or ecolog ical process of interest. The arrangement of remaining landscape spaces is the fragmentation geometry. Alternative arrangements, or geometries, are used to compare the relative effects of different fragmenting elements (Girvetz et al. 2008; Jaeger et al. 2008)
101 One way to interpret meff is in the case of animal movement and biodiversity, which is the approa ch chosen for this study. In this case, meff is the average area within a planning unit that an animal can move without encountering a barrier. The implications of movement are important for dispersal and the resilience of populations. For many animals, ro ads act as barriers in the landscape, and are sources of mortality (Forman et al. 2003) In Florida, road related mortality is the leading cause of death for several large animal species (Harris and Scheck 1991) For s ome species, open areas with no forest cover are barriers to movement, inhibiting dispersal and increasing isolation (Canaday 1996; Lowery and Perkins 2002; Deckers et al. 2005) The transformation of the forest into agricultural fields where no traces of forest structure or organisms remain is particularly problematic for biodiversity (Millenium Ecosystem Assessment 2005) In the southeastern U.S., vast forested areas have been transformed due to the extensive logging supplanting the original forest, and the expansion of agriculture (Ware et al. 1993) This study characterized changes to landscape fragmentation, as measured by meff, caused by the developing road network, urban lands and agriculture, in a semi rural watershed of northcentral Florida, from 1975 through 2005. I studied landscape fragmentation at two spatial and temporal scales for two fragmentation geometries. Given evidence fro m other road fragmentation studies, I hypothesized that fragmentation in the SFRW increase d over the thirtyyear study period, measured by a decrease in meff. I also expected differential rates of fragmenta tion between the two geometries. I also observed t he variability in fragmentation rates across the watershed and from one decade to the next. The planning unit for this study was a watershed chosen in reference to a regional ecological system that was unrelated to county or state level road -network pla nning and
102 development It was both large enough to observe spatial variation in the landscape, and small enough to warrant inclusion of local roads in the analysis. Watershed subbasins provided a nested hierarchical system for spatial differentiation of l andscape processes. The study period coincides with the most recent decades of the last 150 years, when biodiversity has been most affected by habitat transformation (Millenium Ecosystem Assessment 2005) particularly in Florida, as human population growth rates increased (Ewel 1990a; Gannon 2003) The beginning date of the study refle cts the availability of consistent land -cover information derived from Landsat satellite imagery and coincident with systematic, detailed mapping of local roads by the Florida Department of Transportation (FDOT). Landscape dynamics were analyzed for the entire thirty year period as well as in three ten -year increments, providing decadal variation. Two fragmentation geometries that distinguished agricultural and forested lands were chosen to reflect important factors for biodiversity conservation in the region. A comparison of changes in the two fragmentation geometries across space and through time revealed interactions between road networks and land cover in the watershed. Given the nature of changes in the SFRW land cover and road network, described la ter, I expected that most of the increase in fragmentation would occur in the final decade of 19952005, and that fragmentation rates for forest s would be greater than for agricultural lands. Study A rea The location for this study was the Santa Fe River wa tershed (SFRW) in north central Florida (Figure 4 1). The SFRW was an accessible study area where both road network and land cover information could be ascertained. The SFRW is within the southern reaches of the Suwannee River basin, which extends north into Georgia. The boundaries of the SFRW watershed were designated by the U.S. Environmental Protection Agency (EPA; U.S.
103 Environmental Protection Agency 1998) The Suwannee River Management District, which has responsibility for managing waters within all tributaries of the Suwannee River, delineated the 115 subbasins within the SFRW used in this study. The entire basin area is approximately 3583 km2 and covers almost all of Bradford and Union Counties and major portions of Alachua, Gilchrist and Columbia Counties Prior to agricultural intensification, the uplands of north Florida hosted a rich mosaic of fire adapted forests dominated by longleaf pine ( Pinus palustris ) (Abrahamson and Hartnett 1990) These ranged from xeric, relatively open pine and oak savannahs, to hydric, dense, bay swamps and cypress stands (Ewel 1990b; Platt and Schwartz 1990) The range of forest types makes drawi ng sharp distinctions between vegetation associations rather difficult at times. Forest composition of varies according to complexities of geology, topography, soils, frequency of flooding and fire, and plant and animal interactions. With the construction of intentional fire breaks, drainage ditches and roads, the structure of the landscape changed, driving changes in natural fire and hydrologic regimes. Over time, mixed hardwood forests dominated by laurel oak (Quercus hemisphaerica ) developed in areas wh ere the original pine forests were logged and fire was excluded (Ware et al. 1993) Land use in the SFRW is dedicated mostly to commercial forestry and agriculture with a few urban centers. The onceextensive longleaf pine forests of t his region were an important source of timber products for ship building and naval stores (Stout and Marion 1993) Today, large areas are now intensively managed for pine sawtimber and pulpwood production. These plantations dominated by slash pine ( Pinus elliotti ) (Ware et al. 1993) are in various stages of development and under differing management regimes. In this region, it is common for highly productive pine pla ntation s to have cutting/ regrowth cycle s of 18 to 20 years (Binford et al.
104 2006) According to Florida Gap Project analyses, the mesic -hydric pine forest compositional group, which includes pine plantations, and agriculture are the m ost common land cover classes and hay is the most commonly produced agricultural crop (Pearlstine 2002) Recent trends in population growth, land-cover conversion and road network development, are changing the SFRW landscape. Population of the five -county area of the SFRW increased from 195,300 in 1975 to 354,700 i n 2005. Columbia and Gilchrist Counties which had the greatest population increases in the study area, experienced average annual gr owth of 2.72% and 4.35% respectively from 1970 to 2000, equaling or outpacing the Florida state average during this period ( U.S. Census Bureau and Population Division 2008) Sabesan (2005) discovered an overall shift toward more intensive agricultural land uses from 1990 to 2003, away from rangeland and pine plantations. In chapter 3 of this dissertation, it was noted that road networks in the SFRW expanded by 23% in the three decades from 1975 to 2005, that most of this occurred in the last decade, and that the extension of dead -e nd roads built from established roadways was an important aspect of road network expansion. Taken together, these studies indicated that, while the landscape became more populated, land use in the SFRW was intensifying as forests and pasturelands gave way to agricultural and urban land uses. Methods GIS data about road networks, were de rived from maps created by FDOT In contrast to earlier studies that measure meff, this study included all local roads in all levels of analysis, including those that are not connecting roads, or dead end s. Railroads were also provided from FDOT digital data, but did not vary over time. Power transmission lines were obtained from a data layer created by Thinkburst Media, Inc and modified by the Florida Geographic Data Lib rary (FGDL ; www.geoplan.ufl.edu ). Digital orthophoto aerial photographs (DOQQs) were available for both 1995 and 2004 from the Land Boundary Information System (LABINS ;
105 http://da ta.labins.org). Mosaics of the relevant DOQQs were made for both times to be used as ancillary reference s Major rivers were extracted from a GIS data layer of Florida Rivers created by the EPA and the National Oceanic and Atmospheric Administration. Thes e datasets along with others used for reference in the analysis were downloaded from the FGDL. R iver data were referenced to the mosaic of 2004 aerial photography. Rivers in the study area were extracted and, where necessary, edited so that their alignment s coincided with the photography. Urban land areas were digitized from the FDOT road maps and combined with a visual interpretation of satellite imagery and aerial photography. These data layers were rasterized and combined into one image for each date cor responding to road, river and urban -related fragmenting elements that were later masked from the thematic land cover. Thematic land cover classes including open water, agricultural/open land s and forested land s were derived from the classified satellite imagery of all four time periods (Table 4 1). Forested lands included non urban areas where >5 m tree canopy covered more than 40% of the ground. This class included a range of wooded land covers from intensively managed pine plantation fiber farms, to natural forests The agricultural/open land s class was similarly diverse including cropped lands and recent clearcuts as well as fallow fields and marshes Image dates were constrained to winter months. Image processing and classification followed met hods described by Lillesand et al. (2004) All image processing was done using Leica Erdas Imagine (ver. 9.2) software, while GIS analysis was done using the ArcInfo level of ArcMap (ver. 9.2), ArcGIS software by ESRI. Preprocessing the Landsat imagery included radiometric correction, subsetting and georectification (Lillesand et al. 2004) The study area is i n the center of one Landsat image
106 (WRSII, Path 17, Row 39). S ubset s of the area corresponding to the intersecting USGS quarter quadrangles was selected from each Landsat scene. These were further georectified to digital aerial photography and road maps. RM S report errors ranged from 2 to 17 m for the four images. Land cover reference information was collected from specified locations in the watershed (Figure 4 2). A protocol and data collection sheet, based on one developed by the CIPEC group at Indiana Uni versity (Kauneckis et al. 1998) was used for collecting informatio n about vegetation and canopy density at each sample point (Appendix B). Permission was obtained to enter and collect field data on public and some private lands in the study area. Using the Isodata algorithm from Imagines automated classification routi nes, an unsupervised classification of the 2005 image provided 20 spectrally distinct information classes. A stratified random sampling approach was used to locate sample points in each information class within accessible lands. Efforts were made to gather data from at least ten representative samples from each class scattered across the SFRW. Two hundred sixty -five field locations were visited between January and December 2007, using a handheld Garmin GPSmap 76CSx to navigate. The average position dilution of precision (PDOP), a measure of GPS positional accuracy, was 5. Field notes were recorded and either a video or photographic record was made of each data collection point. Some classes were not well represented in the accessible sites, with few or no po ints occurring within them. These tended to correspond to agricultural lands. For these classes, points were located near road rights of -way such that the sample points were easily seen from the road without physically accessing them. This field work yielded a detailed database of reference information (Appendix C) and imagery about land cover in the SFRW. T wo of the four images (1985 and 1995) had active fires with smoke obscuring parts of the scene for certain spectral bands. To reduce the effect of the s moke and classify the images more
107 accurately, classification was done using a combination of bands 4, 5, and 7, three of the spectral bands available in the TM and ETM+ sensors and corresponding to near and mid-infrared radiation. All bands were used in th e classification of the MSS (1975) scene. An unsupervised classification of the imagery yielded 15 spectral classes. The first class corresponded to open water and changed little from one time step to the next. The pixels in this class were then removed fr om the image. Finally, each spectral class was assigned a thematic class of non-forest or forest. This process was done first with the 2005 image and carried backward. Following classification, thematic maps were checked for accuracy. For the 2005 image, field sample points provided reference data for the classification. For the remaining images, a minimum of 80 points per land cover class were randomly located in the image and their classification accuracy was checked by comparing them with aerial photogr aphy and a posteriori knowledge of the SFRW landscape. Accuracy of image classification was assessed using methods described by Lillesand et al. (2004) Overall, producers and users accuracy values were noted, as well as overall and conditional Kappa values. The image was considered sufficiently accurate if overall Kappa > 0.80. Methods used here for determining meff (Equation 4 1) are similar to those followed in Jaeger et al. (2008) and Girvetz et al. (2008) : cmpl 1 total CBC1i n i i effA A A m (4 1) where n = the number of patches corresponding to the land cover being analyzed, Atotal is the size of the study area ( e.g., county, watershed, etc.), Ai is the area of patch i within the boundary of the study area ( i = 1, 2, 3, n ), and Ai cmpl is the complete area of patch i as extended beyond the
108 study area boundaries. So, if Ai is compl etely within the study area boundary, then it is equal to Ai cmpl. Two fragmentation geometries (FG) were created for the analysis that pertained to the two most dominant land -cover groups in the region: wooded and agricultural lands The first fragmentati on geometry (FG1) was created by combining both land -cover classes and masking out th e date -specific roads and urban-related infrastructure. The resulting data layer consisted of the spectrum of forested land covers along with clearcut areas, rangelands, croplands and any other non urban open land essentially any areas that were not water, urban land or part of the road and railway network. In FG1, fragmenting elements consisted of open water, which had been removed in the first step of classification, l arge rivers, and transportation and urban infrastructure. The remaining landscape spaces were the combined forested, agricultural and non urban open spaces in the watershed. The second fragmentation geometry (FG2) was created by selecting only forested la nd cover types and masking roads related infrastructure. In FG2, open and agricultural landcover classes were added to the fragmenting elements used in FG1. These included clearcuts, pastures and croplands as well as any other nonurban open land cover t ype. As FG2 excludes all but forested land cover s the analysis of this second level of fragmentation gives the effective mesh size of all combined forested land covers in the study area. Eight thematic land cover maps, two for each date, were produced by this method. These eight raster maps were converted to vector polygon data layers for meff analysis in a GIS. Effective mesh size was calculated for the entire watershed and its sub -basins for all four dates. One sub -basin, the drainage along the Santa Fe River itself, stretched across the entire watershed. For ease of interpretation this sub -basin was subdivided into 4 areas, at points where
109 the distance across the boundary was < 500 m, increasing the number of sub-basins from 115 to 118 (Figure 4 3). Effective mesh size ( meff) was calculated at using the CBC algorithm provided by Moser at al. (2007) in an automated ArcGIS script (J. Jaeger and E. Girvetz, personal communication, 23 Se ptember 2008) In addition to effective mesh size, basic measures of landscape structure were recorded at the watershed scale for each date. The minimum patch size was the grain of analysis, measured at 576 m2. Metrics included: the number of patches > 576 m2; the number, total area ( km2) and percent of the total area of patches > 2000 ha; and the number, total area and percent of the total area of patches > 1000 ha. Patches included in this analysis were those whose centroids fell within the watershed bounda ry. In addition to these measures, the area of land cover was calculated for five land cover types: forested, agricultural/open, urban, open water and road networks calculated as the area within a buffer 15 m on either side of the roadway center line Ro ad network land cover areas also included the unchanging fragmentation elements of large rivers, power transmission lines and railroads. Change s in land cover were calculated as percent change for each land cover type. To observe how effective mesh size c hanged both over time and across the area of the watershed change and percent change in effective mesh size for FG1 ( FG 1 meff) and FG2 (FG2 meff) were computed. Values were calculated for the entire watershed and for each sub basin of the SFRW; at decadal i ncrements and for the entire study period. Changes in FG 1 meff were compared with changes in FG2 meff in each sub -basin to elucidate the dynamics between road networks and urban areas, agricultural/open lands and forested lands. A contingency table was created to summarize the dynamics of change for FG 1 meff and FG2meff. Percent changes in FG 1 meff and FG2meff were categorized into an ordinal scale, as
110 decreasing ( 10%), showing no change (> 10% and < 10%) or increasing ( The 10% threshold s were chosen, after inspecting the distribution of the data, as reasonable limits to both accommodate variability across all eight datasets, and to provide a uniform system of categorization. In this 3 x 3 matrix, there were nine possible outcomes, numbered by row and column (1.1, 1.2, 1.3, 3.3). The number of cases in each outcome was reported ( nij), and the joint probability ( ij) of that outcome was calculated as the proportion of cases in each outcome (ij = nij/n ) for the entire gr oup of sub basins (n = 118; Agresti 2002) Changes in meff were summarized for the three decades and for the entire thirty years. Thirty -year percent changes in FG 1 meff and FG2 meff were mapped to show spatial variation in rates of change across the watershed from 19752005. Gamma is a measure of association between variables in a contingency table (Agresti and Finlay 1997) that compares the proportion of concordant and discordant pairs in the table (Equation 4 2). Values for gamma fall between 1 and +1and the sign of the value indicates whether the association is positive or negative. As the value of gamma increases, the strength of D C D C (4 2) where C is the number of concordant pairs and D is the number of discordant pairs in a contingency table. A pair of subbasins is concordant if the sub-basin that is higher on the ordinal scale for percent change in FG 1 meff is also higher on the scale for percent change in FG2 meff. The pair is discordant if the sub -bas in that is higher on the ordinal scale for one variable is lower on the scale for the other variable. Gamma values were calculated and interpreted to explain the strength of the association between the percent changes in FG 1 meff and FG2 meff and the
1 11 dynamic s of the association as it changed over time. The numbers and proportions of concordant and discordant pairs were tabulated for each decade and the thirty -year period. Results Land cover classification produced thematic maps of forested and non-forested l and cover for the four time periods (Figure 4 4). Overall accuracy for all four classifications never fell below 94%. Producers and users accuracy rates for each land cover in each time period were all similarly high, with the lowest value at 91% for 1995 forest land cover. The lowest overall K appa value of the four land cover classifications was for the 1995 image at 0.8875. The lowest conditional K appa values related to the forest cover in 1995 at 0.8353. These results were higher than the overall Kappa limit of 0.80 set prior to analysis and so all four classifications were considered valid. A reas of landcover type changed substantially in some classes and less in others (Tables 4 2 and 4 3). In the 358,335 ha study area, f orested land -cover area declined from 219,923 ha in 1975 to 206,503 ha in 2005. However, most of this decline came in the first decade when it dropped to 206,438 ha. Forested land covers continued to decline in the second decade slightly, but then increased slightly in the third decade. Agricultural/open lands, in contrast increased in the first decade from 101,982 in 1975 to 112,846 in 1985. But then, over the next twenty years this decreased substantially, and by 2005 constituted only 89,093 ha. Both urb an and road network land covers increased continually over thirty years. Urban areas increased from 4854 ha in 1975 to 9430 ha in 2005. The area occupied by road networks increased from 26,832 ha in 1975 to 47,762 ha in 2005. The area classified as open wa ter fluctuated over the image dates around an average value of 5382 ha. In 1975, 4744 ha were classified as open water. This increased until 1995, and then decreased to 5547 ha in 2005.
112 For the entire watershed, meff decreased in the SFRW from 1975 to 2005, as the number of patches increased, the size of the largest patch decreased, and the number and area covered by large patches generally declined (Table 4 4). This trend held true for both fragmentation geometries. The decline in meff was more pronounced in the third decade, 19952005, and most evident in FG1. The first two decades (19751985) saw a steady decline in FG1 meff of about 6% per decade while the final decade saw a drastic reduction from 1030 ha to 775 ha, a 21% loss (Table 4 4). Overall the v alue for FG1 meff in 2005 was nearly 31% less than in 1975, indicating a substantial increase in fragmentation due to expansion of the road network and urban areas Values for FG2 meff declined by nearly 29% from 19752005. As with FG1, most of this change o ccurred in the final decade of the study (19952005) when it dropped by 25%. In fact, up until that period, the decline in FG2 meff changed little, never exceeding -3% per decade The total area covered by large forested patches (FG2 patches > 2000 ha) shra nk by nearly half in the thirty year period, but this reduction occurred mostly in the last decade of the study. Effective mesh size values for sub -basins of the SFRW are listed in detail in Appendix D. Values varied spatially within the watershed (Figure 4 5), with northwestern regions consistently more fragmented than the rest of the watershed. In the northcentral and northeastern regions, the proportion of forested to non -forested land was consistently higher. Over time, the distribution of FG2 meff valu es became more skewed with a few sub basins maintaining rather high values while the rest decreased. Summary statistics of meff for sub -basins show a declining trend similar to that seen for the watershed as a whole (Figure 4 5). Median meff values of the sub basins were consistently lower than values for the watershed and declined steadily throughout the study period, although the
113 decline was sharper for median FG1meff than for median FG2meff. Maximum FG1meff increased by 8% over the thirty year interv al. However, there was a 19% increase between 1975 and 1985, after which maximum FG1meff declined steadily. In contrast, maximum FG2meff decreased by 14% from 1975 to 2005, with a substantial decrease of 25% in the first two decades but a slight increase i n the third. Minimum FG1meff showed the most dramatic change by dropping 39% over the thirty year period. Most of this change occurred in the final decade, when the minimum meff value dropped from 161 ha in 1995 to 104 ha in 2005. Minimum FG2meff values we re consistently very low, hovering between 10 and 20 ha throughout the study period. While the summary statistics gave a clear indication for overall trends, meff values did not decrease uniformly across the watershed. Sub-basins with the greatest decreases in FG1meff were in the east (Figure 4 9) northwest of Gainesville and along the corridor between Waldo and Lawtey The top three sub -basins with the greatest decreases in FG1meff over thirty years included: 115, Blues Creek; 90, Southeast 15th St.; and 15, Bradford Olustee Creek (Figure 4 3). Th e western half of the watershed in the sub-basins of Columbia County had comparatively low meff values. These regions also showed the greatest decreases in FG2meff. However, as with FG1, two of the top three sub basins experiencing the greatest decreases in FG2meff were in Bradford County. They included: 90, Southeast 15th St.; 44, New River Drain; and 15, Bradford Olustee Creek. Change in meff also varied from one decade to the next in the watershed (Figure 4 6A -D). Median values for change in both FG 1 meff and FG2meff were lower in the first and third decades than in the second, suggesting greater fragmentation during those periods The range of decadal changes in FG2meff and FG1meff also narrowed as variability in change values decreased during the second decade. The standard deviation for change in FG1meff decreased from about 266 in
114 the first decade to about 13 in the second decade, but increased to 335 in the third. These values suggest that rates of change were slower and less variable in the second decade (1985 1995). In the three ten -year intervals, changes in FG2mef f in some sub -basins did not always concord with changes in FG 1 meff. In most cases, especially in the final decade, the two values correspondingly decreased. However, for each decade there were alternative outcomes: where one or both meff values increased; where one meff value increased and the other decreased, or where one or both values for meff did not change. The percent change contingency table (Table 4 6) shows both the joint probabilities (in table cells) and marginal probabilities (in row and colum n totals) for the relative changes in FG 1 meff and FG 2 meff. The marginal probability for FG 1 meff and FG 2 meff over the three decade period, 19752005 (D for 1+ and +1), showed that most sub-basins experienced declines in either FG 1 meff or FG 2 mef f. The joint probability for this outcome ( 1.1 ) was also the highest at .695 with 82 sub-basins experiencing declines in both FG 1 meff and FG2meff over the entire period. Decadal increments provided more details about the relative changes in these variables. The marginal distributions given by row totals, indicated that 78% of sub basins in 19751985, and 81% of sub -basins in 19851995 experienced little or no change for FG 1 meff (outcome 2.+ ) during these two decades. This pattern changed dr amatically in 19952005, when nearly 76% of sub -basins experienced decreases in FG 1 meff (outcome 1 .+). In contrast, decreases in FG 2 mef f were a consistent pattern of change for sub -basins in the watershed, although the proportion of sub -basins in this category ( +1) decreased from 59% in the first decade to 48% in the third decade. While decreasing FG 2 mef f was the most probable outcome for all three decades, in 1985 2005 less than 50% of the sub-basins fell into this category, and the combined number of sub-
115 basins e xperiencing increases or no change in FG 2 mef f (n+2 + n+3) was 63 in 19851995 ( +2 + +3 = .534), and 62 in 19952005 ( +2 +3 = .526). Of these, most experienced no change in FG 2 mef f in 19851995 ( n+2 = 39), and increases in 19952005 ( n+3 = 33). Throug hout the study period, a few sub basins showed increases in FG2meff (outcome +.3 ). The proportion of cases in this category increased from 8% in 19751985 to 28% in 19952005. The specific outcomes, or combinations of cases, were described by the joint probabilities of the individual cells in the contingency tabl e. During the first two decades (19751995) outcome 2.1 which was little or no change to FG 1 meff and declining FG 2 meff valu es, was the most common outcome: 52 sub-basins in the first decade ( 21 = .441); and 46 in the second ( 21 = .390). In the third decade there was an overwhelming shift toward outcome 1.1 declines in both FG 1 meff and FG 2 meff values, with the number of cases rising sharply from 8 in 19851995 (11 = .068), to 52 in 19952005 (11 = .441). There was always a small proportion of sub-basins showing increases in both values (outcome 3.3 ) during any given decade and during the entire study period (<5%) However, by the third decade, and for the study period as a whole, no cases fell in outcomes 3.1 or 3.2 where a decline or no change in FG2meff was coupled with an increase in FG1meff The row and column totals of the contingency table were mapped to sh ow the spatial outcome of the thirty-year changes of meff in the watershed (Figure 4 7). Decreasing values were graphically differentiated in 10% increments to visualize which areas had changed most. The greatest declines in FG 1 meff were associated with su b basins ne ar or adjacent to urban centers (e.g. Lake City, Gainesville, Starke, etc .) and a reas with intensive agriculture ( e.g. the area between Fort White and Lake City). The 15 subbasins that experienced no change were grouped in the center of the watershed. The 4 sub-basins where FG 1 meff increased were in the southern
116 half. Changes to FG 2 meff followed a spatial pattern similar to the changes in FG 1 meff, but there was greater variation in the magnitude of change. The gamma analysis provided additional information about the association of fragmentation dynamics in the watershed (Table 4 7). A positive gamma value indicated a positive association between percent changes in FG1 meff and FG2 meff. There was a predominance of concordant pairs, where decreases in FG1 meff were associated with decreases in FG 2meff, clearly the most common situation in the SFRW. Since FG2 is a subset of FG1, it seems obvious that gamma should be positive, but the tempora l shifts provide for a more detailed understanding. Gamma increased from 0.213 in the first decade to 0.575 in the third indicating a strengthening positive association over time. While the positive association was present in 1975, it was not strong, and the proportion of discordant pairs, where decreases in one value were associated with increases in another value, was about 0.4. By the end of the study period the positive association between changes in FG1 meff and changes in FG2 meff strengthened considera bly. The proportion of discordant pairs had decreased to 0.21 with a corresponding increase in the proportion of concordant pairs. Discussion and Conclusions Urban land area in the SFRW increased by 94% over the thirty year period, and most of that occurre d in the second decade, although decadal rates of increase were never below 20% for any ten -year period. Likewise, the road network also increased in extent and density with most of the increase occurring in 19952005 (see chapter 3) when there was a 62% increase in land cover area occupied by roads. Most of this increase was at the expense of agricultural/open land which declined 18%, nearly 20,000 ha in the third decade. Forested land covers experienced the biggest decline in the first decade, as agricul tural/open lands expanded. In 19751985, a 6% decline of 13,485 ha in forested land, was met with an 11% increase of 10,864 ha in
117 agricultural/open land, along with increases in urban land and roads. However, this increase was short lived as agricultural/o pen land made way for urban areas and roads during the last two decades. As expected, landscape fragmentation in the SFRW increased in 19752005 for both the forested land covers and the combined forested with agricultural/open land covers. This trend was substantiated by the simultaneous decrease in meff for both FG1 and FG2 in 19752005 for the entire basin and in 44% of the subbasins of the watershed, outcome 1.1 of Table 4 6. Based on other effective mesh size studies (Jaeger et al. 2007; Girvetz et al. 2008) this was the expected outcome of change over time, as road networks expanded and increased fragmentation in the remaining land scape. Other commonly used metrics of landscape fragmentation also pointed to increasing landscape fragmentation. These included an increasing number of small landscape patches, a decrease in the size of the largest landscape patch, and decreases in the nu mber and area of large landscape patches (Table 4 2). Thirty -year trends in meff varied across the watershed. Declines in meff were generally associated with sub -basins intersected by major highways. The greatest decreases in meff occurred in sub -basins of Alachua and Bradford Counties near the Waldo -to Lawtey corridor. However, substantial decreases were also noted in sub basins northwest of Gainesville and near Lake Butler in Union County. The northwestern sub-basins in Columbia County registered large decreases in percent change in meff because small changes in these already low values made them more sensitive to change. Exceptions to the trend in decreasing meff were the sub basins near the cities of Alachua and La Crosse, wher e values increased. Sub -basins where meff increased, did not change, or slightly decreased were often adjacent to waterways.
118 Decadal changes in meff indicated that during the first two decades (19751995) fragmentation rates increased at a steady rate for both FG1 and FG2 across the watershed. There were also declines in median FG1meff and FG 2 meff values for the sub-basins of the watershed. During these two decades, the variability of change in the sub -basins was greater in the first decade than in the sec ond decade. From 1975 to 1995, the most common change in a sub-basin was increased fragmentation of FG2 with little or no change in FG1 (outcome 2.1 Table 4 6). During this period, the majority of sub-basins experienced little or no change in the combined forest and agricultural/open land covers (A 2+ = .780; B 2+ = .813). The dynamic from 1975 to 1995 implied that land cover was actively shifting between open and forested land cover possibly signaling the cuttingregrowth cycles common in plantation fore stry. Nevertheless the increase in maximum FG1meff combined with decreasing FG2 meff during 19751985 suggests that changes were oriented toward the expansion of open areas and agriculture at the expense of forests. In 8 sub -basins, FG1meff increased in 19 751985, further suggesting a decrease in roadrelated fragmentation of agricultural fields, possibly associated with enlargement and consolidation of croplands. This is reflective of a contemporaneous push in the mid to late 20th Century toward the expans ion of agriculture in the U.S. powered by cheap fossil fuels (Jackson 1980) D uring the third decade of the study (19952005) fragmentation rose sharply for FG1 and FG2 at the watershed scale. The very strong positive association between changes in FG1 meff and changes in FG2 meff, which increased over time, provides evidence for the assertion that that forest fragmentation and agricultural fragmentation increased jointly in response to road network and urban expansion. During this decade, 76% of subbasins experienced increases in the fragmentation of FG1, the combined forest/agricultural/open land s while nearly 48% of sub-
119 basins experienced increases in fragmentation of FG2, or forested lands The median change in meff was always negative for both FG1 and FG2, and in 19952005 the trend continued. M edian change in FG1meff decreased from 12 in 19851995, to 116 in 19952005, but median change in FG2meff increased slightl y from 15 to 11 per decade respectively Although the change in FG2meff still maintained a negative trend, this differential response was indicative of the changing dynamics in the watershed in 1995 2005. While for est fragmentation increased in nearly ha lf of the watershed sub -basins, forest land cover in the watershed actually increased slightly in 19952005. The increase in the standard deviation of meff change during this decade indicated that there was much more variability in the change dynamic in th e third decade than the previous two. For slightly more than half of the sub basins, forest fragmentation did not increase during 19952005, and in 28% of the cases, it declined. Maximum FG2meff also rose during this decade. There was a simultaneous decrease in FG1meff and increase in FG2meff in 21 sub-basins, indicating an increase agricultural fragmentation combined with a decrease in forest fragmentation. These factors combined with a simulta neous expansion in the road network, described a fundamental change in la nd cover dynamics. They provide evidence that during the last decade of the study, there was a shift in l andscape fragmentation and land use dynamics away from agricultural expansion and toward reforestation. This pattern seems counter -intuitive: as road networks and urban areas expand, forests should become more fragmented. However, in suburban areas where land may be more valued as real estate rather than productive farmland or timbe r stands, land management shifts to less intensive practices and forests regrow in abandoned pastures, while road networks expand In the Santa Fe River watershed, road s increasingly fragmented both the forested lands and the agricultural/open lands from 1975 to 2005, as hypothesized. Over thirty years, as the road
120 network expanded and population increased, meff declined by nearly 30%. As expected, most of the increase in fragmentation occurred in the final decade (19952005). In addition fragmentation wa s not uniform between forested and agricultural land cover types or from one decade to the next. Based on reports of an increase in agricultural intensification in the SFRW from 1990 to 2003 (Sabesan 2005) I expected higher fragmentation rates for forested than for agricultural lands. In the first decade (1 9751985), the fragmentation of forested lands (FG2) increased as expected, indicating defo restation. In the second decade (19851995) this trend slowed, and the fragmentation of forested land s in most sub -basins either did not change or decreased. Taken together, in the first two decades deforestation was slightly more common than reforestation, while agric ulture tended to expand. Agricultural expansion at the expense of forests was also substantiated by increases in agricultural/open land covers and decreases in forest land covers during this time. However, in the third decade, fragmentation increased more dramatically for combined forest and agricultural lands than forested lands alone, contrary to my expectations. Possible reasons for this unexpected result include the added consideration of comprehensive, accurate local road network data (see chapter 3) The inclusion of this information, along with data about land cover from earlier and later dates, and details about fragmentation dynamics, lead me to an alternate conclusion. Roads were a major fragmenting factor in many sub -basins, especially over the final ten years. However, forests have also been fragmented by agricultural expansion, particularly during 19751985. There was comparatively less change in 19851995. V ariability of changes in fr agmentation was the lowest during this period, although fragmentation tended to increase for both forests and the combined forested/agricultural/open landscapes. During 19952005, the
121 dynamics of land cover and fragmentation in the watershed underwent a fundamental shift. This decade saw a sharp increase in the extent of the road network and urban land area, with substantial increases in fragmentation for both land cover types. In some areas, fragmentation of agricultural/open lands was accompanied with for est expansion, as might be expected in suburban expansion. From these salient points, I deduced that in the first decade (19751985) fragmentation and land-cover change patterns expressed the dynamic of agricultural expansion and ongoing intensive commerc ial forestry. In the final decade (19952005) this dynamic changed, and several sub -basins in the watershed shifted toward a dynamic of urbanization and suburbanization. The sub -basins where this pattern occurred were concentrated along highway corridors and near urban areas. The comparatively low variability in meff changes in the middle decade (19851995) suggest ed that this was a period of transition in land use dynamics. Whether or not land use in the watershed was shifting from agricultural expansion to suburbanization might be demonstrated with further analysis of economic and productivity indicators from this period. This process might be detected in variables related to, for example, the value and ownership of land ( e.g., subdivision and sale of agr icultural lands), and the productivity and earnings from agricultural and timber products. The SFRW is part of a larger region in the southeastern US with which it shares an ecological and cultural history (Delcourt et al. 1993) The SFRW participates in the reg ional economy that for the last several decades, has experienced rapid growth as human populations continue to grow mostly through migration. Changes in land cover and fragmentation in this north -central Florida watershed are not unique to this particular site, and could be extrapolated to the context of the southeaster n US. Natural biodiversity in the southeastern US is among the highest in the temperate world (Martin and Boyce 1993) Ongoing destruc tion of the once -
122 extensive longleaf pine forests of the southeast, and their failure to reestablish themselves was an ecological event of spectacular proportions, the impact of which may never be properly assessed (Ewel 1990a; Ware et al. 1993) Existing forests and areas of native vegetation are important remaining sources of natu ral biodiversity in the region R oad network development and suburban expansion continue to fragment the landscape potentially affecting biodiversity in the region by restricting the movement of species.
123 Figure 4 1. The study area: the Santa Fe River watershed in the southern reaches of the Suwannee River basin.
124 Figure 4 2. Field data collection in June and July 2007. A) Taking field notes in OrdwaySwisher Biological Station, B) measuring canopy closure in Olustee Battlefield State Park. A B
125 1 New River 2 West Olustee Drain 3 Alligator Lake 4 Price Creek 5 Blue Lake 6 Cannon Creek 7 Bradford Turkey Cr. 8 Birley Road Slough 9 Hwy 47/Hwy 441 Sl. 10. Union Swift Creek 11. Columbia Rose Creek 12. Clay Hole Creek 13. Grannybay Drain 14. West Lulu 15. Bradford Olustee Cr. 16. Swift Creek Pond 17. Cypress Lake Road 18. Center Bay Drain 19. Swift Creek Drain #4 20. Richard Creek 21. Ichetucknee Trace 22. Ebenezer Cemetery 23. Pounds Hammock 24. Lawtey Alligator Cr. 25. Swift Creek Drain #2 26. Piney Bay Drain 27. Olustee Creek 28. Swift Creek Drain #1 29. Swift Ck Swamp Drn 30. Washington Road 31. High Falls Road 32. New River Drain #2 33. Cliftonville 34. Mason Slough 35. Rogers Road 36. Lake Butler Outlet 37. Water Oak Creek 38. Cedar Hammock Drn 39. Old Ichetucknee Road 40. Swift Creek Drain #3 41. Bradford Gum Creek 42. Fivemile Creek 43. Mckinney Branch 44. New River Drain #1 45. Bethany Cemetery 46. Fern Pond Drain 47. Browns Still Run 48. Hopewell Church 49. 25th Street Basin 50. Cypress Run 51. Lake Crosby Outlet 52. Ichetucknee River 53. New River City 54. Burlington Church 55. Starke Alligator Cr. 56. New River Church 57. Harmony Church 58. Hammock Branch 59. Sampson River 60. New River Drain #5 61. Mined Area 62. Santa Fe River 63. Little Springs Church 64. New River Drain #4 65. Clayno 66. Markey Cemetery 67. Fort White 68. Wilson Spring 69. Near Graham 70. New River Drain #3 71. Old Bellamy Road 72. West Brooker 73. Santa Fe Ranch Airpo 74. Prevatt Creek 75. Robinson Sinks 76. West Hasan 77. Pareners Branch 78. Hol der Branch 79. Double Run Creek 80. Rum Island 81. Hampton Ditch 82. Braggs Branch 83. Brooker 84. Hampton Lake Outlet 85. East Hasan 86. Theressa Slough 87. Spring Hill 88. Alachua Rocky Creek 89. Townsend Branch 90. Southeast 15th St. 91. Mill Creek Sink 92. West Lacrosse 93. Sunshine Lake 94. Gilchrist Cow Creek 95. North Alachua Drain 96. Hornsby Spring Run 97. Monteocha Creek 98. Trout Pond Outlet 99. West Cow Creek Drn 100 East Louise 101 East Lacrosse 102 High Springs 103 Little Monteocha Cr. 104 Hainesworth Branch 105 West Flats Drain 106 University Creek 107 East Waccasassa Flats 108 Rocky Creek Sink 109 Burnett Lake Drain 110 Alachua Slough 111 Dry Basin 112 Hague Branch 113 Alachua Turkey Cr. 114 Sanchez Prairie 115 Blues Creek 116 Santa Fe River 117 Santa Fe River 118 Santa Fe River Figure 4 3. Subbasins of the SFRW (Suwannee River Water Management District 2008) Bold type are sub basins with important increases in fragmentation in 1975 2005.
126 Figure 4 4. Landsat TM images of the SFRW and corresponding land cover classification in forest ed (green ; including plantations ) and nonforest ed (gold). Santa Fe River watershed boundary is shown in white, and the Santa Fe River is shown in blue on the classified images. A) Landsat MSS image acquired on 01 January 1975; B) corresponding 1975 land cover classification; C) Landsat ETM+ image acquired on 20 March 2005; and D) corresponding 2005 land cover classification. A B
127 Figure 4 4. Continued C D
128 Figure 4 5. Effective mesh sizes in hectares for SFRW sub basins. A) 1975; B) 1985; C) 1995; D) 2005. Subbasin values are plotted in decreasing order for FG2meff (Y axis). Summary statistics are provided in adjacent table. FG1 comprises combined nonurban foreste d (including plantations), a gricultural and open land covers; FG2 comprises only nonurban forested land covers. A B
129 Figure 4 5. Continued C D
130 Figure 4 6. Change in effective mesh size of SFRW subbasins ordered by greatest increase to greatest decrease in FG2meff: A) 1975 1985, B) 1985 1995, C) 1995 2005, D) 1975 2005. A 1975 1985 B 19851995 2000 1500 1000 500 0 500 1000 meff (ha)19751985 effeffMean 31.21 56.23 Median 5.14 34.81 Std. dev. 265.65 128.34 25th perc. 26.75 86.94 75th perc. 4.98 0.74 2000 1500 1000 500 0 500 1000 meff (ha) FG1 FG2 19851995 effeffMean 44.99 26.04 Median 12.06 15.37 Std. dev. 12.71 8.85 25th perc. 36.05 48.50 75th perc. 3.37 9.27
131 Figure 4 6. Continued C 1995 2005 D 19752005 2000 1500 1000 500 0 500 1000 meff (ha)19952005 effeffMean 210.77 58.56 Median 115.88 11.39 Std. dev. 335.05 234.52 25th perc. 225.73 66.57 75th perc. 52.18 16.06 2500 2000 1500 1000 500 0 500 1000 meff (ha) FG1 FG2 19752005 effeffMean 286.36 140.84 Median 154.39 78.43 Std. dev. 427.88 251.36 25th perc. 319.28 185.60 75th perc. 74.00 19.23
132 Figure 4 7. Percent change in effective mesh size in SFRW sub basins from 1975 to 2005. A) FG1 (forested, agricultural/open lands) with lakes and rivers in blue, B) FG2 (forested lands) with roads and towns. Red subbasins showed a decrease in effective mesh size. Gre en subbasins showed an increase in effective mesh size. A B
133 Table 4 1. Satellite, sensor, acquisition dates and pixel size of remotely sensed images (WRS2, Path 17, Row 39) used for land cover classification. Decade Satellite, sensor Acquisition date Pixel size (m) 1975 Landsat 1, MSS 01 January 1975 60 x 60 1985 Landsat 5, TM 16 January 1985 30 x 30 1995 Landsat 5, TM 12 January 1995 30 x 30 2005 Landsat 7, ETM+ 20 March 2005* 30 x 30 Level 1G SLC -off gap -filled products include a scan gap mask for each band. Gap acquisition dates for this image are: 2005 0304, 20050405, and 200504 21. Table 4 2. Area of land-cover type in SFRW (in hectares). Date Forest Agriculture/Open Urban Open water Roads* 1975 219923 101982 4854 4744 26832 1985 206438 112846 5843 5401 27807 1995 206360 109088 7627 5837 29423 2005 206503 89093 9430 5547 47762 Roads include 3542 ha. in power transmission lines (2408 ha), railroads (867 ha) and large rivers (267 ha) which did not change from one date to the next. Total basin area was 358,335 ha. Table 4 3. Percent change matrix for change in area of land -cover type for the SFRW. 1975 1985 1995 1985 FOR 6.13 AG 10.65 URB 20.38 OW 13.85 R 3.63 1995 FOR 6.17 .03 AG 6.97 3.33 URB 57.13 30.53 OW 23.04 8.07 R 9.66 5.81 2005 FOR 6.10 .03 .07 AG 12.64 21.05 18.33 URB 94.27 61.39 23.64 OW 16.93 2.70 4.97 R 78.00 71.76 62.33 Note: Land cover types are denoted as follows: FOR forestry; AG agriculture/open; URB urban; OW open water; R roads* (* see Table 4 2 ). Bold type indicates greatest change in a ten or twenty -year interval.
134 Table 4 4. Effective mesh size (ha) and patch number, size (ha) and area information reported by fragmentation geometry (FG) and year. Fragment. geometry Year m eff (ha) Number of patches ( 2) Size of largest patch (ha) Patches > 2000ha (total area/% area) Patches > 1000 ha (total area/% area) FG1 1975 1084.5 1547 6563 16 (538 km 2 /15%) 58 (1101.7 km 2 /30.7%) 1985 1059.9 1668 6380 18 (602.3 km 2 /16.8%) 59 (1149.4 km 2 /32.1%) 1995 1030.9 1769 6130 18 (602.2 km 2 /16.8%) 54 (1091.8 km 2 /30.5%) 2005 775.0 1814 5480 12 (390 km 2 /10.9%) 49 (874.7 km 2 /24.4%) FG2 1975 494.9 8138 4384 8 (253.2 km 2 /7.1%) 27 (516.8 km 2 /14.4%) 1985 464.2 12103 4510 7 (236.4 km 2 /6.6%) 25 (494.8 km 2 /13.8%) 1995 435.5 21808 4512 8 (252.8 km 2 /7.1%) 23 (466.4 km 2 /13%) 2005 342.8 21788 4093 4 (129 km 2 /3.6%) 20 (359.4 km 2 /10%) Note: FG1 includes forested land covers combined with agricultural/open land covers. FG2 includes forested land covers. Values reported for patches whose centroids fall within the boundaries of the study area. Table 4 5. Percent change matrix of effective mesh size for the SFRW basin for the two fragmentation geometries (FG). 1975 1985 1995 1985 FG1 6.2 0 FG2 2.27 1995 FG1 12.02 6.18 FG2 4.94 2.74 2005 FG1 30.73 26.15 21.29 FG2 28.54 26.88 24.82 Note: FG1 includes forested land covers combined with agricultural/open land covers. FG2 includes forested land covers. Bold type indicates greatest change in a ten or twenty -year interval.
135 Table 4 6. Contingency table summarizing relative changes in FG1 and FG2 in sub basin s of the SFRW categorized by percent change in effective mesh size Percent chang e in effective mesh size for FG2 (% FG2 meff) D ecrease (% FG 2meff 10%) N o change (10% > % FG 2meff > 10%) I ncrease (% FG 2meff Total Percent chang e in effective mesh size for FG1 (% FG1 meff) Decrease (% FG1 meff 10%) 1.1 FG1 contracts FG2 contracts -----------------A : n11= 13; 11= .110 B: n11= 8; 11= .068 C : n11= 52; 11= .441 D : n 11 = 82; 11 = 695 1.2 FG1 contracts FG2 no change -----------------A : n12= 4; 12= .034 B : n12= 4; 12= .03 4 C: n12= 16; 12= .136 D: n 12 = 9; 12 = .076 1.3 FG1 contracts FG2 expands -----------------A: n13= 1; 13= .008 B : n13= 0; 13= 0 C: n13= 21; 13= .178 D: n 13 = 8; 13 = .068 1.+ FG1 contracts -----------------A: n1+= 18; 1+= .152 B : n1+= 12 ; 1+= .102 C: n1+= 89; 1+= .755 D: n 1+ = 99; 1+ = .839 N o change (10% > % FG1 meff > 10%) 2.1 FG1 no change FG2 contracts -----------------A : n21= 52; 21= .441 B : n21= 46; 21= .390 C : n21= 4; 21= .03 4 D : n21= 5; 21= .042 2.2 FG1 no change FG2 no change -----------------A : n22= 33; 22= .280 B: n22= 35; 22= .297 C : n22= 13; 22= .11 0 D : n22= 10; 22= .085 2.3 FG1 no change FG2 expands -----------------A: n23= 7; 23= .059 B: n23= 22; 23= .186 C : n23= 8; 23= .068 D : n23= 0; 23= 0 2.+ FG1 no change -----------------A: n2+= 92; 2+= .780 B: n2+= 103; 2+= .813 C : n2+= 25; 2+= .212 D : n2+= 15; 2+= .127 Increase (% FG1 meff 3.1 FG1 expands FG2 contracts -----------------A : n31= 5; 31= .042 B: n31= 1; 31= .008 C : n31= 0; 31= 0 D : n 31 = 0; 31 = 0 3.2 FG1 expands FG2 no change -----------------A: n32= 1; 32= .008 B : n32= 0; 32= 0 C : n32= 0; 32= 0 D : n 32 = 0; 32 = 0 3.3 FG1 expands FG2 expands -----------------A: n33= 2; 33= .017 B: n33= 2; 33= .017 C: n33= 4; 33= .034 D : n 33 = 4; 33 = .034 3.+ FG1 expands -----------------A: n3+= 8; 3+= .067 B: n3+= 3; 3+= .025 C: n3+= 4; 3+= .034 D : n 3 + = 4; 3 + = .03 4 Total +.1 FG2 contracts -----------------A: n+1= 70; +1= .593 B: n+1= 55; +1= .466 C: n+1= 56; +1= .475 D: n +1 = 87; +1 = .737 +.2 FG2 no change -----------------A: n+ 2= 38; + 2= .322 B : n+ 2= 39 ; + 2= .331 C : n+ 2= 29 ; + 2= .246 D : n+ 2= 19; + 2= .161 +.3 FG2 expands -----------------A: n+ 3= 10; + 3= .084 B: n+ 3= 24; + 3= .203 C: n+ 3= 33; + 3= .280 D : n+ 3= 12; + 3= 102 Note: c olumns represent categories of percent change in FG2meff (% FG2meff) Rows represent categories of percent change in FG1meff (% FG1meff) For each outcome, labeled as i.j landcover dynamics are described, and changes in meff are illustrated. FG1 includes forested and agricultural/open land covers combined; FG2 includes f orested land covers. Date intervals are as follows: A) 1975 1985; B) 19851995; C) 19952005; D) 1975 2005. Results are reported by date intervals as the number of cases for each outcome ( nij), and the joint probability of the outcomes ( ij; n = 118). Tota l row and columns give the marginal distributions, denoted by i+ for the row and +j for the column. Bold type indicates greatest proportion of cases for that particular date interval.
136 Table 4 7. iscordant pairs in Table 4 6. Time interval C PropC D PropD 1975 1985 817 0.60 531 0.40 0.213 1985 1995 730 0.75 245 0.25 0.497 1995 2005 1560 0.78 421 0.21 0.575 1975 2005 1244 0.88 165 0.12 0.766 Note: C is the number of concordant pairs. D is the number of discordant pairs. Proportions of concordant and discordant pairs are denoted as PropC and PropD
137 CHAPTER 5 SYNTHESIS The purpose of my dissertation research was to investigate the relationship between road network development and landscape dynamics, including fragmentation. The research questions focused on how a changing road network interacted with land cover and fragmentation dynamics in a semi -ru ral watershed over time. An important objective was to study how theories and methods developed in transportation geography might help explain roadrelated landscape fragmentation dynamics. The fact that roads fragment landscapes is well known and axiomati c in studies that examine the ecological effects of roads (e.g., Hawbaker and Radeloff 2004) In light of the loss of biological diversity attributed t o habitat loss and fragmentation (Ehrlich 1988) roads have come under increased scrutiny by ecologists as humanbuilt structures that have negative consequences for many species (Harris and Scheck 1991; Forman and Alexander 1998) As a concept, fragmentation, the breaking apart of habitat (Fahrig 2003) implies a link between changes of landscape structure and c hanges in ecological functions. For example, a more fragmented landscape is prone to local plant or animal extinction due to the isolating effects of the fragmenting elements (Forman 1995) Many metrics used to describe fragmentation rely on purely formal dimensions, such as patch size, area to -perimeter ratio, etc. (O'Neill et al. 1988) but a few measures of fragmentation take into account functional considerations, for example, information about the potential for movement in the landscape (e.g., Jaeger 2000) Road networks are landscape structures created by humans. They are meaningful social sp aces that have critical economic and political functions (Jackson 1984) Like ecological systems, road networks are imbued with characteristics of structure and function. Road network structure is readily discerned from aerial photography and information about road networks is
138 maintained in road maps, which vary in scale, accuracy and consistency. Road netwo rk function is less discernable, but relatively easy to measure and can be represented by traffic counts, freight tonnage and network capacity. In studies that consider the fragmenting effects of road networks, conclusions about the dynamics of landscape patterns depend on information in maps, which describe the structural characteristics of road location, length, and density (e.g. Heilman Jr. et al. 2002) In some cases, information about the function of a road network is related to functional aspects of fragmentation, for example, studies that relate a nimal mortality with road traffic (e.g., Saeki and Macdonald 2004) In the 1960s and 1970s transportation geographers in the U.S. A. developed methods based on graph theory and network analysis to describe structural and functional properties of transportation systems (e.g., Garrison and Marble 1962; Kansky 1963) These methods, however, have not yet been integrated into roadrelated fragmentation studies. This dissertation research did not delve into questions of how the structure of the road netwo rk was affected or determined by biophysical variables. Nor did it examine local fragmentation effects of local changes of road network structure. For the most part, transportation geographers have failed to consider biophysical variables in models of tr an sportation network development. Historically, research in transportation route development has generally abstracted the landscape unrealistically as an isometric plane, avoiding the question of landscape variability almost entirely, or relegating it to an engineering problem to be dealt with at the local scale in the actual route design and construction Exceptions include Kanskys (1963) use of gene ral shape and topographic variables in predictive models of transportation network structure, and Blacks (1993) use of a topographic cost factor in modeling profitability of transport route locations. This study, which documents in de tail the physical location and
139 structure of the SFRW road network, provides excellent datasets for the exploration of both how the structure landscape has affected road network structure, and how local changes in network structure might have affected local fragmentation dynamics. For example, it stands to reason that, in sub -basins with large rivers, or large bodies of water, there will be more dead-end roads and fewer through roads. These are interesting areas of research which I will, hopefully, pursue mo re fully in future studies. This dissertation draws from transportation geography, using techniques of network analysis to describe the structure and function of road networks, and landscape ecology, using methods for describing landscape fragmentation. A road network in the Santa Fe River watershed (SFRW), a semi rural region of north central Florida, was chosen for the analysis. Its structural and functional properties were observed as they changed from 1975 to 2005. A complementary analysis of landscape fragmentation was conducted for the same area using effective mesh size ( meff), a metric that takes into account both structural and functional aspects of fragmentation (Jaeger 2000; Moser et al. 2007) The expectation was that the road network would become more extensive over time, and that its connectivity and accessibility would increase as it developed. At the same time, it was expected that road related landscape fragmentation would increase as the road network developed. The SFRW road network expanded greatly from 1975 to 2005, especially from 1995 to 2005. Proc esses of road network development favored expansion rather than interconnection. Road network expansion was documented by increases in total length, road density, and network diameter values. Accessibility to areas between roads increased with increases in road network density and decreases in average distances to roads. However, w hile accessibility to the landscape increased over thirty years the road network itself became less accessible Road
140 network development included the addition of new roads that p enetrated into the formerly unroaded landscape, presumably facilitating resource extraction and land use intensification. S ince the roads were mainly dead end s they were only connected to the network at one end. Each time a new dead -end road connected to the wider network it created a junction thereby splitting the road in to two segments where there had been only one The cumulative topological effects of this repeated action overwhelmed the effects of new connector roads, resulting in a relatively less c onnected, less accessible and more elaborate road network with a greater capability of accessing formerly unroaded rural areas. Based on these findings, the changes in the SFRW road network did not strictly follow the model of in creasing connectivity and accessibility as the road network developed. Fragmentation of forested and agricultural /open lands in the SFRW increased from 1975 to 2005, a s the road network developed. Increased fragmentation was measured by declining meff. For forested lands (FG2) meff decreased by 29% over thirty years; for combined forested and agricultural/open lands (FG1) meff dropped by 31%. Most of the increase in fragmentation occurred in the final decade of the study ( 19952005). Fragmentation rates varied between watershed sub basins, land cover types and from one decade to the next. In the first decade of the study (19751985), fragmentation of fores ted lands increased as expected During this period, expansion of agricultural/open lands was an important fragmenting dynamic f or forested lands. The second decade of the study (19851995), was a period of comparatively less change. Variability of changes in fragmentation was the least during this ten-year period and the magnitude of changes was smaller. Although fragmentation te nded to increase in the aggregate for both FG1 and FG2 most sub-basins experienced little or no fragmentation in the combined
141 forested and agricultural/open land covers, and more than half experienced either decreases or no change in fragmentation of fore sted land covers. Taken together, deforestation was slightly more common than reforestation in the first two decades, as FG1 tended to expan d while FG2 fragmented This pattern of change was also substantiated by increases in the area of agricultural/open land covers and corresponding decreases in the area of forest ed land covers during this period In contrast, during the third decade of the study (1995 2005), there was a rapid expansion in the extent of the road network and urban land area, coupled with substantial increases in fragmentation for both FG1 and FG2 in general. F ragmentation increased more dramatically for combined forest and agricultural lands than forested lands alone. Moreover, i n some sub -basins, increased fragmentation of FG1 was accomp anied by a decline in fragmentation of FG2, indicating increasing connectivity of forested areas as agricultural/open areas fragmented From these salient points, I deduced that fragmentation and land cover dynamics in the SFRW experienced a fundamental s hift over the thirty -year study period. I n 19751985, fragmentation and land-cover change patterns expressed agricultural expansion and ongoing intensive commercial forestry. In 19952005, however, several sub -basins in the watershed shifted toward urbaniz ation and suburbanization. These sub-basins were concentrated along highway corridors and near urban areas. The comparatively low variability in meff changes in 19851995 suggests that this was a period of transition in landuse dynamics. Whether or not la nd use dynamics in the watershed were shifting might be further demonstrated with an analysis of economic and productivity indicators from this period. This pattern might be detected in variables related to, for example, the value and ownership of land ( e. g., the subdivision and sale of agricultural lands), and the productivity and earnings from agricultural and timber products.
142 The road network analysis clarified how fragmentation patterns in the SFRW corresponded with the development of roads In general a s the road network expanded, the landscape became more accessible, and fragmentation of the landscape increased throughout the entire thirtyyear study period. The decadal changes to the road network corresponded with decadal changes in the fragmentatio n rates at the scale of the entire watershed. Variations in this trend were observed, however, in sub-basins of the watershed. Steady increases in the road network extent in 19751995 corresponded with steady increases in the fragmentation rates in general In 19851995, the road network continued to expand and landscape accessibility increased. While road expansion corresponded with increased overall fragmentation in the aggregate, the effects were mostly experienced in a few sub -basins; most sub -basins ex perienced little or no fragmentation during this period In 19952005, the greatest decadal increases in road network extent also coincided with the greatest general increases in fragmentation rates. In some sub basins, however agricultural/open land cove rs were more severely fragmented than forested l and covers and fragmentation of forested land covers decreased (i.e., forested lands became more connected) These facts suggest that during this period road network development was less a factor in the frag mentation dynamics for forested lands than for agricultural/open land covers. One clear characteristic of change in the SFRW road network was development of deadend roads throughout the thirty -year study period. This process was particularly common in 1985 1995. However, these changes did not register as changes in landscape fragmentation during that time because the meff metric changes little or not at all, with the addition of deadend roads. In the conceptualization of meff, dead -end roads do not cause fragmentation. This phenomenon was documented in an exchange of views published in Science in 2007 (Girvetz et al. 2007; Watts et al. 2007b; Watts et al. 2007a) In this discourse, Givertz et al. (2007) compared
143 two idealized landscapes of the same road density but different configurations of roads. They showed how, although landscape accessibility decreased with increasing roadless volume, functionally, the connectivity of the landscape declined substantially, measured by decreasing meff. The chief difference between the two idealized road networks was the inclusion of dead -end roads. Watts et al. (2007a) countered that the two metrics were related yet distinct (p. 1240c) and should not be used to validate, or invalidate, one another. This debate was interesting because it documented a gap in road ecology research by overlooking an important point. The comparison failed to consider the drastic difference in the structure and implied function of the two road networks an omission that is symptomatic of the ways roads are considered studies of road re lated fragmentation. In this example, while road length and density remained the same for each of these scenarios, the number of junctions and the degree (or order) of the junctions in the networks changed. The different measures of road network connectivi ty suggest disparate functional cap acities, implying distinct land use and land -cover change possibilities for each scenario. Thus, the potential for road -related effects could be quite different. Since meff measures the probability of connection in lands cape patches, and dead -end roads do not completely bifurcate unroaded areas, meff does not in itself account for the effects of deadend roads unless they are built into the characterization of the fragmenting geometries. Forman and Deblinger (2000) explained a variable road -effect zone and showed how to map it. Presumably the differential effects of road endings could be taken into account in such a mode ling scenario, although the detailed mapping of road -effect zones is complex and cumbersome. Addressing this issue, Jaeger et al. (2007) incl uded variable buffers in modeling the effects of road traffic noise. Watts et al. (2007a) suggest ed using a reciprocal of DTR to model landscape vulnerability to road effects. These are workable approaches, yet neither
144 satisfactorily considers the specific characteristics of dead -end roads, and, more importantly, the cumulative, landscape -scale effects of increasing numbers of dead -end roads. A nalyse s of changes in road networks and landscape fragmentation in the Santa Fe River watershed raise d important questions for road ecologists that were so succinctly, if inadvertently, brought up in the Science discourse: What are the differing ecological effects of a dead -end road and a through road? A nd how do we account for this difference in landscap e fragmentation and connectivity models? Most road effect modeling efforts assume that road endings are no more or less consequential than any other part of the road. Yet, dead -end roads are potentially associated with a range of ecological effects that di fferentiate them from through roads. They are, after all thresholds where humans, anima ls, plants, insects and microbes pass between the landscape and the road network. Dead ends are also nodes of land use activity, and, as such, are valued differently t han roadsides: they are the rural equivalent of a suburban cul -de -sac, where homes are more highly valued. As a distinct class of roads, the specific and cumulative ecological effects of dead -end roads are worthy of closer scrutiny. Patterns of road netwo rk development in frontier regions have shown that road penetration is associated with resource extraction, and that these particular kinds of roads are most associated with landcover change. This association suggests that the presence of high proportions of dead end roads may be a n important determinant of the threat to biodiversity due to habitat modification and loss. Due to their persistence and high visibility, road s are relatively easy to measure either by direct measurement or remote sensing. Becaus e there is a time lag between habitat destruction and species loss (MacArthur and Wilson 1967) detecting and measuring dead -end road development could be useful as a way t o anticipate
145 threats to habitat and biological diversity. To that end, n etwork analysis provides objective metrics for the description of ecologically relevant characteristics of road networks
146 APPENDIX A DETAILED DESCRIPTIONS OF ROAD NETWORK METRICS Detailed Descriptions of Map -Based Road Network Metrics Area related measures, including road density have been used to measure road network development as length per unit area (Go uld 1960) the number of intersections per unit area, or length per capita (Taaffe and Gauthier 1973) Road density is a simple and commonly used metric in statistical models of wildlife population characteristics and landscape fragmentati on (Mech et al. 1988; Clev enger et al. 1997; Heilman Jr. et al. 2002; Johnson and Collinge 2004; Platt 2004) Road densities above 0.6 km/ km2 tend to affect populations of certain large vertebrates and road density is an important and fundamental index to assess the ecological e ffects of roads (Forman et al. 1997; Forman et al. 2003) One problem with the road density measure is that its value depends on the scale of the source information used to map the roads A broader scale map will yield lower road density measures than a finer scale map biasing the result (Haggett and Chorley 1969; Hawbaker and Radeloff 2004) The distance to road measure is another index which often used as a variable in spatial analyses of the ecological effects of roads (Gelbard and Harrison 2003) and land use and land cover change (Angelsen and Kaimowitz 1999) Although using distance to road (DTR) as a proxy for the driving forces of land cover change may obscure some o f the subtleties of the change dynamic (Veldkamp and L ambin 2001) DTR is a persistent causal factor in models of land use conversion (Chomitz and Gray 1996; Kaimowitz et a l. 2002; Mertens et al. 2002) Distance to road (DTR) is also used to provide a broad -scale assessment of remoteness from roads. In an analysis of roads in the conterminous United States, Riitters and Wickham (2003) calculated that 83 % of the land was within 1061 m of a road. In most areas of the southeastern U.S., 60 % of the land is within 382 m of a road.
147 Finally, related to DTR roadless volume is a measure that can be used to show cumulative changes to a region as road networks develop, leaving fewer open spaces far from roads (Watts et al. 2007b) Roadless volume (RV) is calculated as the product of mean distance to road and the study area footprint. It is an indicator of the status of that space between roads, providing a quick measure of how roadless space, as an asset, is distributed across an area and, for temporal datasets, how it changes over time. A United States county map of per capita roadles s volume concisely captures the picture of the regional differences, with extremely low levels in some intense urban counties and extremely high levels in remote, rural counties. It represents the sense of how open space is distributed and how it changes. In the Colorado Front Range, RV has diminished by nearly 50 percent in the six decades from 1997 to 1937 (Watts et al. 2007b) Both the DTR and RV metrics are categorically different than others used in this st udy because they measure the areas between the road, and not the roads per se They are included here because their basic properties are derived from map of the road network and DTR is a measure commonly used in both econometric and ecological studies to describe the effects of road networks. The flow of traffic or the capacity is also an important basic characteristic of road networks Direct measures of flows are related to traffic counts and the capac ity of the networks. These can be expressed as the number of vehicles per unit time, or as tonnage, in the case of the movement of freight (Taaffe and Gauthier 1973; Lowe and Moryadas 1975) The amount of traffic in a roadway system describes the function of the network. Knowing how traffic varies over time gives information about the cyclical function of the netwo rk The capacity of the network describes its potential function Examples of the kinds of information used to describe
148 capacity include the number of traffic lanes available, the maximum speed limit of the lanes, or the type of surface of the lanes. Detai led Descriptions of Graph -Theoretic Based Road Network Metrics Measures of the aggregate network include the cyclomatic number ( ) and diameter ( ) of the graph The cyclomatic number ( ) is a summary of the maximum number of cycles or fundamental circuits in a graph In graph theory, this is more commonly known as the value of circuit rank (Weisstein 2008) It is the minimum number of edges that must be removed in order to make it free of circuits As such, for a tree graph (i.e. a graph with no circuits) 0 Graphs that are highly connected will have higher cyclomatic numbers As transportation networks develop and become more interconnected, is expected to increase (Kansky 1963) The diameter ( ) is the minimum number of edge s between the two most distant vertices in the graph, or the topological length of the longest path in the graph (Kansky 1963; Taaffe and Gauthier 1973) As a network expands and v increases, is expected to increase As a network becomes more connected and e increases while v remains relatively stable is expected to decrease. Since is not related to any real dimension of a network, transportation systems of vastly different overall lengths may have similar values Average path length measures the average of all the shortest paths between all pairs of vertices in the network As the network increases in complexity, the value for the average path length increases (West 2001) It is possible, and in fact likely, that there is more than one path in a graph. In the case of road networks, which have real dimensions in distances, the distance ( d ) is calculated as the average of the distances of the alternate diameter paths Indices describing the cumulative characteristics of graphs inclu de indices of connectivity and indices of relationships between the entire graph and its elements (Garris on and Marble 1962; Kansky 1963) The g amma index () measures the level of connectivity of a graph as a
149 ratio of the number of actual edges, e and the maximum number of edges possible in a planar graph, e(v -2) As a network ncreases Alpha ( ) is a measure of the circuitry of a graph It is a ratio of the number of actual circuits in the graph, expressed as e (v -1) and the maximum number of circuits possible in a planar graph, given as 2v 5 The theoretical upper limit of is one, and its lower limit is zero (a graph with no circuits) As such, when multiplied by 100, it is sometimes presented as a percentage of connectivity Alpha ( ) increases a s a network becomes more connected Taaffe and Gauthier (1973) associated the connectivity indices with three basic network configurations based on limits of their values The se classes used by engineers to describe planar networks include spinal, grid and delta structures, where a spinal stru cture is a minimally connected graph with no circuits, a delta structure was a maximally connected graph in which most nodes have three linkages, and a grid was an intermediate graph structure, in a transitional stage between spinal and delta configurations The graph types and their index limits are given in Table 3 3 Beta ( ) is a connectivity index expressed by the simple ratio of edges to vertices in the graph. For simple tree graphs and disconnected graphs < 1 For a graph with only one circuit, = 1 Higher values of are produced when network structures become more complex and the number of edges relative to vertices increase (Kansky 1963; Rodrigue et al. 2006) The shape index ( ) is so labeled because it relates the diameter of the network to the overall length ( L ) of the network (Kansky 1963) not unlike the relationship of the circumference to the diameter of a circle. In this case, the diameter measure is taken as the average distance of the alternate diameter paths ( d ), so it is a measure of the length of the network per diameter unit Kansky (1963) created this index as a way to measure the development of a network. A s a transportation network grows and develops i t is anticipated that will increase as the length
150 increa ses relative to its diameter. For less developed networks, the index approaches one, while higher numbers will be encountered for more complicated networks. This is first of the graph theoretic indices that are non topological, by relating to the mapped distance of the network. Another group of indices measures simultaneous characteristics of extent, structure and function (Garrison and Marble 1962; Kansky 1963) They are indices that describe complexities of the network and include the eta ( ), theta ( ) and iota ( ) indices These measures along with are specialized in their application to transportation networks While not widely used or cited in network analysis, they are useful as measures of transporta tion network structure The eta index ( ) expresses the relationship between the whole network and its edges, or the average edge length As e increases, relative to the overall length, the index decreases Kansky (1963) found a strong correlation between values for railroads and higher -order highways, and measures of economic development. In more highly developed countries, average edge length was s horter. The theta index ( ) relates the overall length of the network to v giving a measure of length per vertex It is a similar but complementary measure to capturing information about connectivity of the network. Finally, the structure index, or iota ( ), is a variation of It is a ratio between the overall length of the transportation network and its weighted vertices ( w ), or the network weight Since low order, terminal vertices have fewer functional connections than higher order vertices, network weight is determined by multiplying all third and higher order vertices ( o > 2) by two and summing their values (Kansky 1963; Rodrigue et al. 2006) Ideally the iota index is a ratio of the length of the network relative to the volume of traffic flow ( T ) The use of network weight as the denominator is intended to take the place of traffic flow This especially useful for networks where no information about traffic volume or flow is available, and presumes a strong
151 correlation between networks with a large number of higher order vertices and the capability of handling higher traffic flows Kansky (1963) made no attempt to define the correlation between network structure and flow, assuming that a network with a larger number of high order vertices will accommodate more traffic than one with a smaller number of high-order vertices. The theory suggests that, as iota increases the functional capacity of the network for handling more traffic volume increases Iota ( ) was intended to complement useful for examining functional differences in two networks of the same length and with the same values for To understand the relative changes of in a changing network, it is necessary to consider it in relation to the factors that influence the index, i.e. length, weight, number of vertices, and proportion of fi rst order vertices. By representing the graph as an adjacency matrixa binary matrix showing connections between vertices, one as connected and zero as unconnectedit is possible to analyze and rank the vertices by their level of connectivity or accessibil ity The order of the vertex is the most basic measure, computed by summing the rows of the adjacency matrix While this gives a sense of how well connected an individual vertex is, for road network applications it is limited because it fails to consider indirect connections, which are most often the case in transportation routes (Taaffe and Gauthier 1973) Indirect connections are taken into account by manipulating the adjacency matrix through matr ix multiplication The accessibility matrix is derived by powering the adjacency matrix to the value of the graph. Summing all the rows of the resulting matrix gives a column vector of the total number of connections for each vertex This is the level of accessibility for each node
152 APPENDIX B TRAINING SAMPLE DATA SHEET Santa Fe River Watershed Training Sample Data Sheet Alisa Coffin, Department of Geography, University of Florida PhD Dissertation Research AY 2006 2007 Training Sample #:_________________________________________ Long: _________________________ Lat: ________________________ (DD or DMS) Date:__________________________ Time:_______________________ GPS Reciever:_____________________ PDOP: ________________ Area name / Owner Name: _________________________________________________ Observation Type: Random pre -selected ground observation Aerial observation Opportunistic observation FLUCCS Code: _________________ Description: ______________________________ Canopy tree species: ______________________________________________________________________________ ______________________________________________________________________________ ______________________________________________________________________________ Understory tree/Shrub layer species: ______________________________________________________________________________ ______________________________________________________________________________ ______________________________________________________________________________ Herbaceous layer species: ______________________________________________________________________________ ______________________________________________________________________________ ______________________________________________________________________________
153 Notes and sketches: Canopy closure observations: Closure: ______/30 = 5 6 15 16 25 26 4 7 14 17 24 27 3 8 13 18 23 28 2 9 12 19 22 29 1 10 11 20 21 30 Movie/image filenames: ____________________________________________________
154 APPENDIX C FIELD NOTES DATABASE Object C 1. Field notes database
155 APPENDIX D EFFECTIVE MESH SIZES, IN HECTARES, FOR SUB BASINS OF THE SFRW SFRW Sub Basin Name FG1m eff FG2m eff 1975 1985 1995 2005 1975 1985 1995 2005 1 New River 1770.3 1768.8 1881.1 1076.0 1038.9 1137.3 1182.7 695.2 2 West Olustee Drain 323.1 305.3 294.7 248.7 253.5 272.2 195.6 226.4 3 Alligator Lake 353.7 252.2 325.5 156.5 145.2 78.1 42.5 48.3 4 Price Creek 267.6 250.3 247.5 104.8 95.0 102.5 46.5 45.2 5 Blue Lake 298.5 283.5 280.1 200.4 76.5 64.6 70.9 47.1 6 Cannon Creek 321.6 279.2 296.6 182.6 65.1 72.9 30.3 17.2 7 Bradford Turkey Creek 2711.4 2753.4 2685.1 2623.2 2313.4 2515.5 2347.9 2412.3 8 Birley Road Slough 453.0 398.5 395.9 258.6 89.7 46.1 40.3 31.7 9 Hwy 47/Hwy 441 Slough 388.7 352.9 355.7 242.8 55.3 52.1 28.6 20.8 10 Union Swift Creek 748.5 728.4 722.7 572.6 576.1 527.0 421.5 441.7 11 Columbia Rose Creek 382.5 374.1 370.7 282.4 110.6 104.0 48.2 58.4 12 Clay Hole Creek 432.8 404.6 391.3 231.1 110.9 71.5 40.8 30.2 13 Grannybay Drain 393.5 353.2 301.9 249.7 350.7 332.8 242.3 226.4 14 West Lulu 318.1 318.3 315.2 217.2 252.1 244.2 192.2 178.5 15 Bradford Olustee Creek 2475.8 2478.0 2380.5 545.4 1826.3 1919.1 2265.9 686.3 16 Swift Creek Pond 361.7 349.3 345.8 314.2 317.7 329.4 291.5 295.0 17 Cypress Lake Road 1008.7 1142.0 1115.9 702.6 404.4 291.9 52.5 36.3 18 Center Bay Drain 932.7 919.8 911.8 708.7 853.1 761.2 753.7 584.6 19 Swift Creek Drain #4 449.5 431.6 428.6 393.1 392.3 402.9 387.6 354.8 20 Richard Creek 1315.8 1261.3 1245.9 508.0 796.2 690.6 687.4 329.4 21 Ichetucknee Trace 813.1 834.1 788.2 517.9 248.7 130.7 114.6 40.6 22 Ebenezer Cemetery 293.1 293.5 281.9 251.2 118.6 106.3 70.4 104.8 23 Pounds Hammock 701.3 694.3 678.7 463.1 568.7 501.0 500.7 378.5 24 Lawtey Alligator Creek 1188.1 1168.5 1142.7 560.1 750.0 817.4 785.9 348.1 25 Swift Creek Drain #2 489.8 487.0 486.3 192.9 420.9 295.3 371.9 154.8 26 Piney Bay Drain 350.0 351.8 348.8 318.8 295.3 331.2 299.4 307.8 27 Olustee Creek 527.7 524.6 514.3 449.8 242.6 198.8 179.1 191.6 28 Swift Creek Drain #1 443.0 442.0 441.0 211.9 352.1 243.1 307.6 150.1 29 Swift Ck Swamp Drain 869.5 775.3 750.8 692.8 691.3 730.4 429.1 613.2 30 Washington Road 588.5 582.6 569.3 507.9 143.6 38.0 82.2 50.8 31 High Falls Road 766.6 762.1 554.5 495.3 672.2 371.7 396.0 412.6 32 New River Drain #2 1656.9 1626.1 1590.9 1087.7 761.7 675.1 694.6 388.8 33 Cliftonville 464.9 438.6 428.9 264.6 317.9 284.3 238.2 204.7 34 Mason Slough 404.8 404.5 389.1 342.1 61.8 50.7 33.7 55.5 35 Rogers Road 1370.6 1365.6 1254.9 1213.5 244.2 118.0 100.8 325.5 36 Lake Butler Outlet 509.3 399.4 395.0 349.4 401.2 320.4 160.1 272.7 37 Water Oak Creek 863.9 843.7 817.4 537.4 462.9 486.3 348.5 260.0 38 Cedar Hammock Drain 425.7 425.7 421.5 297.8 346.1 273.7 256.2 235.3
156 SFRW Sub Basin Name FG1m eff FG2m eff 1975 1985 1995 2005 1975 1985 1995 2005 39 Old Ichetucknee Road 996.3 987.3 947.4 661.7 209.6 114.4 163.3 59.6 40 Swift Creek Drain #3 672.4 672.4 665.7 550.3 484.5 408.2 330.7 433.4 41 Bradford Gum Creek 344.0 349.8 332.9 259.7 158.4 159.4 189.6 147.0 42 Fivemile Creek 999.6 994.4 977.3 828.2 458.2 334.2 287.0 305.0 43 Mckinney Branch 552.2 540.0 536.5 483.7 287.9 311.6 270.0 259.4 44 New River Drain #1 2378.2 2351.1 2338.7 634.8 1753.6 1396.0 1524.0 438.0 45 Bethany Cemetery 596.7 594.0 597.7 540.8 407.7 366.3 411.6 374.7 46 Fern Pond Drain 523.9 505.0 492.7 431.1 224.0 188.7 159.0 141.3 47 Browns Still Run 319.3 324.4 317.3 283.4 131.2 119.7 118.3 106.8 48 Hopewell Church 172.7 170.5 170.3 145.6 58.6 58.0 52.1 28.9 49 25th Street Basin 359.3 343.4 341.8 307.5 122.7 121.0 107.3 89.4 50 Cypress Run 639.8 561.8 555.2 488.3 577.9 502.6 442.0 434.9 51 Lake Crosby Outlet 421.2 294.4 316.9 287.2 245.6 239.1 254.4 196.7 52 Ichetucknee River 873.6 851.1 765.4 365.4 564.5 329.3 247.1 112.8 53 New River City 977.5 964.7 960.9 560.9 295.7 262.6 254.8 169.8 54 Burlington Church 235.8 234.1 161.8 109.5 83.8 48.0 12.0 16.6 55 Starke Alligator Creek 467.0 497.4 485.4 277.6 222.6 222.5 192.5 125.3 56 New River Church 1397.2 1388.8 1383.5 741.1 364.0 427.5 500.6 197.5 57 Harmony Church 450.7 450.3 442.5 383.2 41.5 45.5 37.8 50.0 58 Hammock Branch 2358.2 2272.6 2225.3 530.0 365.8 284.4 292.3 176.4 59 Sampson River 539.9 512.8 496.1 472.4 263.4 325.5 315.5 314.5 60 New River Drain #5 1172.7 1165.0 1159.7 636.0 373.4 286.3 251.8 166.1 61 Mined Area 817.2 679.0 648.3 453.1 111.6 103.7 245.3 196.0 62 Santa Fe River_2 1189.2 1210.8 1129.9 956.2 300.4 234.1 259.2 250.7 63 Little Springs Church 196.5 190.7 179.4 162.5 59.6 59.2 53.3 44.6 64 New River Drain #4 791.4 789.1 793.6 601.7 251.6 241.6 190.7 134.8 65 Clayno 389.6 390.6 388.8 356.3 252.4 301.9 267.4 263.6 66 Markey Cemetery 1070.0 1068.9 1063.8 861.9 486.8 354.1 279.0 240.6 67 Fort White 754.7 739.3 704.7 434.8 135.2 58.1 87.6 23.0 68 Wilson Spring 549.7 562.5 567.1 284.7 101.4 28.9 146.4 45.5 69 Near Graham 268.7 268.6 265.5 232.6 214.7 250.5 71.5 213.0 70 New River Drain #3 3645.4 3652.5 3620.0 3245.9 943.1 647.9 337.2 508.8 71 Old Bellamy Road 1460.9 1278.7 1260.7 896.6 414.7 331.0 315.5 278.4 72 West Brooker 3766.8 3777.3 3738.4 3468.1 539.5 477.6 460.7 430.8 73 Santa Fe Ranch Airport 1411.8 1651.7 1400.8 1357.2 186.2 46.5 33.4 71.0 74 Prevatt Creek 991.1 1059.1 1070.8 627.8 621.5 660.1 655.9 405.3 75 Robinson Sinks 579.4 582.0 571.8 471.5 19.3 10.7 20.4 28.4 76 West Hasan 1161.4 2255.4 1157.1 1075.4 185.1 116.9 58.5 47.1 77 Pareners Branch 972.7 967.7 957.7 771.4 221.2 226.9 222.8 190.3 78 Holder Branch 580.1 711.4 577.0 526.8 206.1 192.4 189.5 193.7 79 Double Run Creek 1063.0 1164.5 1135.4 924.9 583.5 630.8 602.6 467.9
157 SFRW Sub Basin Name FG1m eff FG2m eff 1975 1985 1995 2005 1975 1985 1995 2005 80 Rum Island 918.3 934.3 879.2 782.3 229.7 72.3 239.6 131.7 81 Hampton Ditch 819.5 811.7 429.1 298.0 408.7 195.4 169.9 142.4 82 Braggs Branch 671.6 677.0 653.1 620.8 294.3 171.3 127.2 108.1 83 Brooker 989.3 994.1 973.3 860.2 58.7 75.5 43.9 48.9 84 Hampton Lake Outlet 287.9 278.0 272.4 236.8 161.0 126.4 166.9 157.0 85 East Hasan 767.2 740.4 738.6 532.1 225.5 190.9 190.1 146.7 86 Theressa Slough 2231.8 2520.6 2473.9 1933.0 1078.1 1427.8 1153.4 854.0 87 Spring Hill 1003.8 977.4 936.4 781.0 249.9 208.0 213.8 218.7 88 Alachua Rocky Creek 846.7 808.5 816.2 695.0 304.8 248.5 226.3 272.3 89 Townsend Branch 1053.5 1066.3 926.1 909.7 285.1 232.9 222.7 255.9 90 Southeast 15th Street 4361.3 4251.3 3961.5 2334.1 2542.3 2594.7 2079.3 961.9 91 Mill Creek Sink 988.1 995.7 955.8 854.4 109.2 115.5 45.0 54.5 92 West Lacrosse 587.2 582.9 571.3 721.5 157.4 113.7 131.0 262.4 93 Sunshine Lake 896.0 985.2 1004.1 1072.9 105.5 176.6 177.8 210.5 94 Gilchrist Cow Creek 799.7 1212.2 1174.3 1000.9 447.5 588.6 497.3 532.8 95 North Alachua Drain 978.7 979.5 975.5 864.3 107.9 72.9 75.5 89.8 96 Hornsby Spring Run 1857.5 1867.0 1027.4 1539.9 284.0 276.2 211.5 299.6 97 Monteocha Creek 1368.1 1373.2 1405.2 971.1 855.5 663.6 742.0 348.6 98 Trout Pond Outlet 1222.7 1270.5 1271.2 1299.4 510.6 306.7 354.6 546.1 99 West Cow Creek Drain 4835.9 5730.6 5617.2 5228.0 3115.7 2467.2 2331.7 2689.4 100 East Louise 687.1 402.5 459.0 283.6 570.1 353.9 398.9 255.8 101 East Lacrosse 858.6 861.5 856.5 1002.5 307.7 307.2 240.9 554.8 102 High Springs 330.6 325.6 201.6 338.1 26.9 17.0 14.9 24.9 103 Little Monteocha Creek 1129.1 1163.6 1151.4 1105.1 473.1 492.2 397.0 487.7 104 Hainesworth Branch 641.7 642.3 636.1 567.9 191.8 202.2 251.5 350.7 105 West Flats Drain 1736.5 1595.1 1569.8 1381.7 1222.9 683.6 788.7 738.3 106 University Creek 489.7 464.0 459.9 396.6 209.3 197.2 115.7 211.5 107 East Waccasassa Flats 545.7 601.0 448.9 341.9 305.2 105.4 97.6 121.4 108 Rocky Creek Sink 610.3 514.8 603.1 561.8 185.0 160.8 198.5 192.9 109 Burnett Lake Drain 406.2 434.6 447.4 352.6 54.5 36.2 61.7 143.6 110 Alachua Slough 2303.1 1684.8 1687.3 1509.4 47.8 35.6 49.3 57.6 111 Dry Basin 613.2 628.1 594.3 523.8 150.9 63.4 55.1 43.7 112 Hague Branch 1236.2 696.9 660.5 508.9 242.2 165.3 154.1 124.8 113 Alachua Turkey Creek 2080.4 971.2 878.1 476.8 430.2 347.1 351.7 249.8 114 Sanchez Prairie 2743.5 2101.9 2068.8 1952.5 1225.2 1062.0 1120.4 1193.8 115 Blues Creek 3647.7 1820.9 1791.4 1455.9 1565.6 927.3 882.6 735.6 116 Santa Fe River_4 2658.7 2438.5 2417.8 1952.4 1270.4 1270.3 1021.3 812.6 117 Santa Fe River_3 2111.7 2104.1 2084.0 1961.7 847.3 752.6 744.0 811.8 118 Santa Fe River_1 1228.2 1045.2 967.4 824.7 329.3 287.3 238.3 237.9
158 LIST OF REFERENCES Abrahamson, W. G. & Hartnett, D. C. 1990. Pine Flatwoods and Dry Prairies. In Myers, R. L. & Ewel, J. J. (Eds) Ecosystems of Florida (pp. 103 149). Orlando: University of Central Florida Press. Agresti, A. 2002. Int roduction to Categorical Data Analysis. New York: Wiley. Agresti, A. & Finlay, B. 1997. Statistical Methods for the Social Sciences. Upper Saddle River, NJ: Prentice Hall. Altrichter, M. & Boaglio, G. I. 2004. Distribution and relative abundance of peccari es in the Argentine Chaco: Associations with human factors. Biological Conservation 116: 217225. Andrews, A. 1990. Fragmentation of habitat by roads and utility corridors: A review. Australian Zoologist 23: 130141. Angelsen, A. & Kaim owitz, D. 1999. Rethinking the c auses of d eforestation: Lessons from e conomic m odels. The World Bank Research Observer 14: 7398. Arnold, G. W. & Weeldenburg, J. R. 1990. Factors determining the number and species of birds in road verges in the wheatbelt of Western Australia. Biological Conservation 53: 295315. Barnes, R. F. W., Blom, A. & Alers, M. P. T. 1995. A review of the status of forest elephants Loxodonta africana in Central Africa. Biological Conservation 71: 125 132. Beckman, M. 1967. Principles of Optimum Location for Transportation Networks. In Garrison, W. L. & Marble, D. F. (Eds) Quantitative Geography Part 1: Economic and Cultural Topics (pp. 95119). Evanston, IL: Department of Geography, Northwestern University. Bellis, A. D. & Graves, H. B. 1971. Deer mortali ty on a Pennsylvania interstate highway. Journal of Wildlife Management 35: 232237. Bennett, A. F. 1988. Roadside vegetation: a habitat for mammals in Naringal, southwestern Victoria. Victorian Naturalist 105: 106113. Bennett, A. F. 1991. Roads, roadside s and wildlife conservation: A review. In Saunders, D. A. & Hobbs, R. J. (Eds) Nature Conservation 2: the Role of Corridors (pp. 99117). Chipping Norton, Australia: Surrey Beatty. Berry, B. J. L. An inductive approach to the regionalization of economic d evelopment. , 78107. 1960. Chicago, IL, University of Chicago, Department of Geography. Bhattacharya, M., Primack, R. B. & Gerwein, J. 2003. Are roads and railroads barriers to bumblebee movement in a temperate suburban conservation area? Biological Conservation 109: 3745.
159 Binford, M. W., Gholz, H. L., Starr, G. & Martin, T. A. 2006. Regional carbon dynamics in the southeastern U.S. coastal plain: Balancing land cover type, timber harvesting, fire and environmental variation. Journal of Geophysical R esearch 111: D24S92. Black, W. R. 1989. Transportation Geography. In Gaile, G. L. & Wilmott, C. J. (Eds) Geography in America (pp. 316332). Columbus: Merrill. Black, W. R. 1993. Transport route location: A theoretical framework. Journal of Transport Geogr aphy 1: 8694. Black, W. R. 1996. Sustainable transportation: A U.S. perspective. Journal of Transport Geography 4: 151 159. Black, W. R. 2003. Transportation: A Geographical Analysis. New York: Guilford Press. Brock, R. E. & Kelt, D. A. 2004. Influence of roads on the endangered Stephens' kangaroo rat (Dipodomys stephensi ): Are dirt and gravel roads different? Biological Conservation 118: 633640. Brody, A. J. & Pelton, M. R. 1989. Effects of roads on black bear movements in western North Carolina. Wildlif e Society Bulletin 17: 5 10. Brooks, C. P. 2003. A scalar analysis of landscape connectivity. Oikos 102: 433439. Bruinderink, G. G., Van Der Sluis, T., Lammertsma, D., Opdam, P. & Pouwels, R. 2003. Designing a coherent ecological network for large mammals in northwestern Europe. Conservation Biology 17: 549557. Bunn, A. G., Urban, D. L. & Keitt, T. H. 2000. Landscape connectivity: A conservation application of graph theory. Journal of Environmental Management 59: 265 -278. Burton, I. 1968. The Quantitative Revolution and Theoretical Geography. In Berry, B. J. L. & Marble, D. F. (Eds) Spatial Analysis: A Reader in Statistical Geography (pp. 13 23). Englewood Cliffs, NJ: Prentice -Hall, INc. Canaday, C. 1996. Loss of insectivorous birds along a gradient of hum an impact in Amazonia. Biological Conservation 77: 63 77. Canters, K., Piepers, A. & Hendriks -Heersma, D. 1997. Proceedings Habitat Fragmentation and Infrastructure. Delft, The Netherlands: Ministry of Transport, Public Works and Water Management. Carr, A. 1994. A Naturalist in Florida: A Celebration of Eden. New Haven and London: Yale University Press. Carr, L. W., Fahrig, L. & Pope, S. E. 2002. Impacts of Landscape Transformation by Roads. In Gutzwiller, K. J. (Ed) Applying Landscape Ecology in Biological Conservation (pp. 225243). New York: Springer -Verlag.
160 Chardon, J. P., Adriaensen, F. & Matthysen, E. 2003. Incorporating landscape elements into a connectivity measure: A case study for the Speckled wood butterfly ( Pararge aegeria L.). Landscape Ecology 18: 561573. Chomitz, K. M. & Gray, D. A. 1996. Roads, land use and deforestation: A spatial model applied to Belize. The World Bank Economic Review 10: 487512. Clevenger, A. P., Purroy, F. J. & Campos, M. A. 1997. Habitat assessment of a relict brown bea r Ursus arctos population in northern Spain. Biological Conservation 80: 17 -22. Cliff, A., Haggett, P. & Ord, K. 1979. Graph Theory and Geography. In Wilson, R. J. & Beineke, L. W. (Eds) Applications of Graph Theory (pp. 293326). London: Academic Press. Coffin, A. W. 2007. From roadkill to road ecology: A review of the ecological effects of roads. Journal of Transport Geography 15: 396406. Crist, M. R. & Wilmer, B. Roadless Areas: The Missing Link in Conservation. An Analysis of Biodiversity and Lanscap e Connectivity in the Northern Rockies. 2002. Washington, D.C., The Wilderness Society. Cristoffer, C. 1991. Road mortality of northern Florida vertebrates. Florida Scientist 54: 6568. Csrdi, G. & Nepusz, T. 2006. The igraph software package for comple x network research. International Journal of Complex Systems 1695. Davies, J. M., Roper, T. J. & Shepherdson, D. J. 1987. Seasonal distribution of road kills in the European badger. The Journal of Zoology 211: 525529. Davis, M., Grime, J. P. & Thompson, K 2000. Fluctuating resources in plant communities: A general theory of invasibilty. Ecology 88: 528534. Dean, C. E., Shilling, D. G. & Willard, T. R. Management of Noxious -Exotic Grasses on Highway Rights of -Way. 997007352. 1989. Gainesville, FL, Flori da Department of Transportation and Agronomy Department, University of Florida. Deckers, B., Becker, P. D., Honnay, O., Hermy, M. & Muys, B. 2005. Sunken roads as habitats for forest plant species in a dynamic agricultural landscape: Effects of age and is olation. Journal of Biogeography 32: 99 109. Delcourt, P. A., Delcourt, H. R., Morse, D. F. & Morse, P. A. 1993. History, Evolution, and Organization of Vegetation and Human Culture. In Martin, W. H. & Boyce, S. G. (Eds) Biodiversity of the Southeastern United States: Lowland Terrestrial Communities (pp. 47 79). New York: John Wiley & Sons, Inc. Demsar, U., Spatenkova, O. & Virrantaus, K. 2008. Identifying critical locations in a spatial network with graph theory. Transactions in GIS 12: 6182.
161 Denevan, W. M. 1991. Prehistoric Roads and Causeways of Lowland Tropical America. In Trombold, C. D. (Ed) Ancient Road Networks and Settlement Hierarchies in the New World (pp. 230242). Cambridge: Cambridge University Press. Develey, P. F. & Stouffer, P. C. 2001. Eff ects of roads on movements by understory birds in mixed -species flocks in central Amazonian Brazil. Conservation Biology 15: 14161422. Dijkstra, E. W. 1959. A note on two problems in connexion with graphs. Numerische Mathematik 269271. Dodd, C. K. Jr., B arichivich, W. J. & Smith, L. L. 2004. Effectiveness of a barrier wall and culverts in reducing wildlife mortality on a heavily traveled highway in Florida. Biological Conservation 118: 619631. Dooling, R. 2006. Estimating Effects of Highway Noise on the Avian Auditory System. In Irwin, C. L., Garrett, P. & McDermott, K. P. (Eds) Proceedings of the 2005 International Conference on Ecology and Transportation (pp. 30 31). Raleigh, NC: Center for Transportation and the Environment, North Carolina State Univer sity. Dunne, T. & Leopold, L. B. 1978. Water in Environmental Planning. San Francisco: W. H. Freeman. Ehrlich, P. R. 1988. The Loss of Diversity: Causes and Consequences. In Wilson, E. O. (Ed) Biodiversity (pp. 21 27). Washington, D.C.: National Academy Pr ess. Evink, G. L. Interaction between roadways and wildlife ecology: A synthesis of highway practice. NCHRP Synthesis 305. 2002. Ewel, J. J. 1990a. Introduction. In Myers, R. L. & Ewel, J. J. (Eds) Ecosystems of Florida (pp. 3 10). Orlando: University of Central Florida Press. Ewel, K. C. 1990b. Swamps. In Myers, R. L. & Ewel, J. J. (Eds) Ecosystems of Florida (pp. 281323). Orlando: University of Central Florida Press. Fahrig, L. 2001. How much habitat is enough? Biological Conservation 100: 6574. Fahrig, L. 2003. Effects of habitat fragmentation on biodiversity. Annual Review of Ecology, Evolution, and Systematics 34: 487515. Farmer, A. M. 1993. The effects of dust on vegetation: A review. Environmental Pollution 79: 6375. Findlay, T. & Bourdages J. 2000. Response time of wetland biodiversity to road construction on adjacent lands. Conservation Biology 14: 86 94. Folke, C., Holling, C. S. & Perrings, C. 1996. Biological diversity, ecosystems, and the human scale. Ecological Applications 6: 10181 024.
162 Forman, R. T. T. 1995. Land Mosaics: The Ecology of Landscapes and Regions. Cambridge, UK: Cambridge University Press. Forman, R. T. T. 1998. Road ecology: A solution for the giant embracing us. Landscape Ecology 13: III-V. Forman, R. T. T. & Alexande r, L. E. 1998. Roads and their major ecological effects. Annual Review of Ecology and Systematics 207 232. Forman, R. T. T. & Deblinger, R. D. 2000. The ecological road -effect zone of a Massachusetts (U.S.A.) suburban highway. Conservation Biology 14: 364 6. Forman, R. T. T., Friedman, D. S., Fitzhenry, D., Martin, J. D., Chen, A. S. & Alexander, L. E. 1997. Ecological Effects of Roads: Toward Three Summary Indices and an Overview for North America. In Canters, K., Piepers, A. & Hendriks Heersma, D. (Eds) P roceedings Habitat Fragmentation and Infrastructure (pp. 4054). Delft, The Netherlands: Ministry of Transport, Public Works and Water Management. Forman, R. T. T., Sperling, D., Bissonette, J. A., Clevenger, A. P., Cutshall, C. D., Dale, V. H., Fahrig, L. France, R., Goldman, C. R., Heanue, K., Jones, J. A., Swanson, F. J., Turrentine, T. & Winter, T. C. 2003. Road Ecology: Science and Solutions. Washington: Island Press. Forys, E. A., Allen, C. R. & Wojcik, D. P. 2002. Influence of the proximity and amount of human development and roads on the occurrence of the red imported fire ant in the lower Florida Keys. Biological Conservation 108: 2733. Gannon, M. 2003. Florida: A short history. Gainesville: University Press of Florida. Garrison, W. L. & Marble, D F. The Structure of Transportation Networks. 6211. 1962. Evanston, IL, Northwestern University, Transportation Center. Geist, H. J. & Lambin, E. F. What drives tropical deforestation? A meta analysis of proximate and underlying causes of deforestatio n based on subnational case study evidence. 4. 2001. Louvain la Neuve, Belgium, LUCC International Project Office. LUCC Report Series. Gelbard, J. L. & Belnap, J. 2003. Roads as conduits for exotic plant invasions in a semiarid landscape. Conservation Bio logy 17: 420432. Gelbard, J. L. & Harrison, S. 2003. Roadless habitats as refuges for native grasslands: Interactions with soil, aspect, and grazing. Ecological Applications 13: 404 -415. Getz, L. L., Cole, F. R. & Gates, D. L. 1978. Interstate roadsides a s dispersal routes for Microtus pennsylvanicus Journal of Mammalogy 59: 208212. Gilpin, M. 1996. Metapopulations and Wildlife Conservation: Approaches to Modeling Spatial Structure. In McCullough, D. R. (Ed) Metapopulations and Wildlife Conservation (pp. 1127). Washington, D.C.: Island Press.
163 Girvetz, E. H., Jaeger, J. A. G. & Thorne, J. H. 2007. Comment on "Roadless space of the conterminous United States". Science 318: 1240. Girvetz, E. H., Thorne, J. H., Berry, A. M. & Jaeger, J. A. G. 2008. Integrati on of landscape fragmentation analysis into regional planning: A statewide multi -scale case study from California, USA. Landscape and Urban Planning 86: 205218. Godefroid, S. & Koedam, N. 2004. The impact of forest paths upon adjacent vegetation: Effects of the path surfacing material on the species composition and soil compaction. Biological Conservation 119: 405419. Goodwin, B. J. 2003. Is landscape connectivity a dependent or independent variable? Landscape Ecology 18: 687699. Gordon, D. 1991. Steering a New Course: Transportation, Energy, and the Environment. Washington, D.C.: Island Press. Gorman, S. Nework Analysis of the Internet and its Provider Networks. 1999. University of Florida. 5 1 1999. Gould, P. R. The Development of the Transportation Pattern in Ghana. 1960. Evanston, IL, Northwestern University, Department of Geography. Northwestern University Studies in Geography, No. 5. Grant, S. B., Rekhi, N. V., Pise, N. R., Reeves, R. L., Matsumoto, M., Wistrom, A., Moussa, L., Bay, S. & Kayha nian, M. A Review of the Contaminants and Toxicity Associated with Particles in Stormwater Runoff. CTSW RT 03059.73.15. 2003. Sacramento, CA, Caltrans, California Department of Transportation. Groot Bruinderink, G. W. T. A. & Hazebroek, E. 1996. Ungulat e traffic collisions in Europe. Conservation Biology 10: 10591067. Gustafson, E. J. 1998. Quantifying landscape spatial pattern: What is the state of the art? Ecosystems 1: 143156. Haggett, P. 1965. Locational Analysis in Human Geography. London: Edward Arnold Ltd. Haggett, P. & Chorley, R. J. 1969. Network Analysis in Geography. New York City: St. Martin's Press. Harris, L. D. & Scheck, J. 1991. From implications to applications: The dispersal corridor principle applied to the conservation of biological diversity. In Saunders, D. A. & Hobbs, R. J. (Eds) Nature Conservation 2: The Role of Corridors (pp. 189220). Chipping Norton, Australia: Surrey Beatty. Haskell, D. G. 2000. Effects of forest roads on macroinvertebrate soil fauna of the Southern Appalachi an mountains. Conservation Biology 14: 5763.
164 Havlick, D. G. 2002. No Place Distant: Roads and Motorized Recreation on America's Public Lands. Washington, D.C.: Island Press. Hawbaker, T. J. & Radeloff, V. C. 2004. Roads and landscape pattern in northern W isconsin based on a comparison of four road data sources. Conservation Biology 18: 1233 1244. Hayasaki, E. Palin said yes to a road to nowhere. Los Angeles Times 2008. Los Angeles, Los Angeles Times. Heilman Jr., G. E., Strittholt, J. R., Slosser, N. C & DellaSala, D. A. 2002. Forest fragmentation of the conterminous United States: Assessing forest intactness through road density and spatial characteristics. Bioscience 52: 411422. Huijser, M. P. & Bergers, P. J. M. 2000. The effect of roads and traffi c on hedgehog ( Erinaceus europaeus ) populations. Biological Conservation 95: 111 116. Hunter, C., Farrington, J. & Walton, W. 1998. Transport and the Environment. In Hoyle, B. & Knowles, R. (Eds) Modern Transport Geography (pp. 97114). Chichester: John Wi ley & Sons. Hussey, B. M. J. 1999. The Flora Roads Survey: Volunteer Recording of Roadside Vegetation in Western Australia. In Saunders, D. A. (Ed) Nature Conservation 4: The Role of Networks (pp. 4148). Chipping Norton, Australia: Surrey Beatty & Sons Pr opriety. Jackson, J. B. 1984. A Pair of Ideal Landscapes. Discovering the Vernacular Landscape (pp. 9 56). New Haven: Yale University Press. Jackson, W. 1980. New Roots for Agriculture. San Francisco: Friends of the Earth. Jaeger, J. A. G. 2000. Landscape division, splitting index, and effective mesh size: New measures of landscape fragmentation. Landscape Ecology 15: 115130. Jaeger, J. A. G., Bertiller, R., Schwick, C., Mller, K., Steinmeier, C., Ewald, K. C. & Ghazoul, J. 2008. Implementing landscape fr agmentation as an indicator in the Swiss monitoring system of sustainable development (Monet). Journal of Environmental Management 88: 737751. Jaeger, J. A. G., Bowman, J., Brennan, J., Fahrig, L., Bert, D., Bouchard, J., Charbonneau, N., Frank, K., Grube r, B. & von Toschanowitz, K. T. 2005. Predicting when animal populations are at risk from roads: An interactive model of road avoidance behavior. Ecological Modelling 185: 329348. Jaeger, J. A. G., Fahrig, L. & Ewald, K. C. Does the Configuration of Road Networks Influence the Degree to Which Roads Affect Wildlife Populations? Irwin, C. Leroy, Garrett, Paul, and McDermott, K. P. 151 163. 2006. Raleigh, NC, North Carolina State University, Center for Transportation and the Environment. Proceedings of the 2005 International Conference on Ecology and Transportation.
165 Jaeger, J. A. G., Schwarz -von Raumer, H. G., Esswein, H., Mller, M. & Schmidt Lttmann, M. 2007. Time series of landscape fragmentation caused by transportation infrastructure and urban develop ment: A case study from Baden Wrttemberg, Germany. Ecology and Society 12: 22. Johnson, M. L., Pasternack, G., Florsheim, J., Werner, I., Smith, T. B., Bowen, L., Turner, M., Viers, J., Steinmetz, J., Constantine, J., Huber, E., Jorda, O. & Feliciano, J. North Coast River Loading Study: Road Crossing on Small Streams. Volume I, Status of Salmonids in the Watershed. CTSW RT 02040. 2002. Sacramento, CA, Caltrans, California Department of Transportation. Johnson, W. C. & Collinge, S. K. 2004. Landscape eff ects on black tailed prairie dog colonies. Biological Conservation 115: 487497. Jones, J. A., Swanson, F. J., Wemple, B. C. & Snyder, K. U. 2000. Effects of roads on hydrology, geomorphology, and disturbance patches in stream networks. Conservation Biology 14: 7685. Kaimowitz, D., Mendez, P., Puntodewo, A. & Vanclay, J. 2002. Spatial regression analysis of deforestation in Santa Cruz, Bolivia. In Wood, C. H. & Porro, R. (Eds) Land Use and Deforestation in the Amazon (pp. 41 65). University Press of Florida. Kansky, K. J. Structure of Transportation Networks: Relationships between Network Geometry and Regional Characteristics. 1963. Chicago, IL, University of Chicago, Department of Geography. Department of Geography Research Paper No. 84. Karl, T. R. & T renberth, K. E. 2003. Modern global climate change. Science 302: 17191723. Kauneckis, D., Brondizio, E., Greene, G., Schweik, C., Carlson, L., Slusher, B., Molnar, M., Summers, K., Shields, S., Ivie, S. & Day, A. 1998. CIPEC Indiana Group Training Sample Protocol Field Guide (Draft I.I). Kerley, L. L., Goodrich, J. M., Miquelle, D. G., Smirnov, E. N., Quigley, H. B. & Hornocker, M. G. 2002. Effects of roads and human disturbance on Amur tigers. Conservation Biology 16: 97108. Lachene, R. 1965. Networks and the location of economic activities. Regional Science Association, Papers 14: 183196. Lambin, E. F., Turner, B. L., Geist, H. J., Agbola, S. B., Angelsen, A., Bruce, J. W., Coomes, O. T., Dirzo, R., Fischer, G. & Folke, C. 2001. The causes of landuse and land -cover change: Moving beyond the myths. Global Environmental Change 11: 261269. Laurance, W. F. 2001. Tropical logging and human invasions. Conservation Biology 15: 45. Laurance, W. F. & Cochrane, M. A. 2001. Introduction. Conservation Biology 15: 14881489. Laurance, W. F. & Williamson, G. B. 2001. Positive feedbacks among forest fragmentation, drought, and climate change in the Amazon. Conservation Biology 15: 1529 1535.
166 Lillesand, T. M., Kiefer, R. W. & Chipman, J. W. 2004. Remote Sensing and Image Interpretation. New York: Wiley. Lonsdale, W. M. & Lane, A. M. 1994. Tourist vehicles as vectors of weed seeds in Kakadu National Park, Northern Australia. Biological Conservation 69: 277283. Lowe, J. C. & Moryadas, S. 1975. The Geography of Moveme nt. Prospect Heights, IL: Waveland Press, Inc. Lowery, L. & Perkins, J. 2002. Long dispersal of a red-cockaded woodpecker in central Florida. Florida Field Naturalist, 30(2): 42 43.2002. MacArthur, R. H. & Wilson, E. O. 1967. The Theory of Island Biogeogra phy. Princeton, NJ: Princeton University Press. Mader, H. J. 1984. Animal habitat isolation by roads and agricultural fields. Biological Conservation 29: 8196. Mader, H. J., Schell, C. & Kornacker, P. 1990. Linear barriers to arthropod movements in the la ndscape. Biological Conservation 54: 209222. Main, M. B. & Allen, G. M. 2002. Landscape and seasonal influences on roadkill of wildlife in southwest Florida. Florida Scientist.2002 Summer; 65(3): 149158. Martin, W. H. & Boyce, S. G. 1993. Introduction: T he Southeastern Setting. In Martin, W. H. & Boyce, S. G. (Eds) Biodiversity of the Southeastern United States: Lowland Terrestrial Communities (pp. 1 46). New York: John Wiley & Sons, Inc. McGarigal, K. 2002. Landscape Pattern Metrics. In El Shaarawi, A. H & Piegorsch, W. W. (Eds) Encyclopedia of Environmentrics (pp. 11351142). Sussex, England: John Wiley & Sons. McIntee, A. A Network and Security Analysis of the U.S. Internet Backbone Network. 2004. University of Florida. Meadows, D. H., Meadows, D. L., Randers, J. & Behrens III, W. W. 1974. The Limits to Growth: A Report for the Club of Rome's Project on the Predicament of Mankind. Washington, D.C.: Potomac Associates. Mech, L. D., Harris, S. H., Radde, G. L. & Paul, W. J. 1988. Wolf distribution and road density in Minnesota. Wildlife Society Bulletin 16: 85 87. Mertens, B., Poccard -Chapuis, R., Piketty, M. G., Lacques, A. E. & Venturieri, A. 2002. Crossing spatial analyses and livestock economics to understand deforestation processes in the Brazilia n Amazon: The case of Sao Felix do Xingu in South Para. Agricultural Economics 27: 269294. Mertens, B. & Lambin, E. F. 1997. Spatial modelling of deforestation in southern Cameroon: Spatial disaggregation of diverse deforestation processes. Applied Geogra phy 17: 143162.
167 Milanich, J. T. 1990. Florida historical contexts: North Florida. Milanich, J. T. & Borremans, N. 1990. Florida historical contexts: The Paleoindian period. Millenium Ecosystem Assessment Ecosystems and Human Well being: Biodiversity Synthesis. 2005. Washington, D.C., World Resources Institute. Moser, B., Jaeger, J., Tappeiner, U., Tasser, E. & Eiselt, B. 2007. Modification of the effective mesh size for measuring landscape fragmentation to solve the boundary problem. Landscape Ecology 22: 447459. Mumme, L. R., Schoech, J. S., Woolfenden, E. G. & Fitzpatrick, W. J. 2000. Life and death in the fast lane: Demographic consequences of road mortality in the Florida Scrub Jay. Conservation Biology 14: 501512. Munroe, D. K., Southworth, J. & Tucker, C. M. 2002. The dynamics of land-cover change in western Honduras: Exploring spatial and temporal complexity. Agricultural Economics 27: 355369. Myers, N. 1988. Tropical Forests and their Species: Going, Going ...? In Wilson, E. O. (Ed) Biodiversi ty (pp. 28 35). Washington, D.C.: National Academy Press. Nellemann, C., Vistnes, I., Jordhoy, P. & Strand, O. 2001. Winter distribution of wild reindeer in relation to power lines, roads and resorts. Biological Conservation 101: 351-360. Nepstad, D., Carv alho, G., Cristina Barros, A., Alencar, A., Paulo Capobianco, J., Bishop, J., Moutinho, P., Lefebvre, P., Lopes Silva, J. & Prins, E. 2001. Road paving, fire regime feedbacks, and the future of Amazon forests. Forest Ecology and Management 154: 395407. No rdlie, F. G. 1990. Rivers and Springs. In Myers, R. L. & Ewel, J. J. (Eds) Ecosystems of Florida (pp. 392427). Orlando: Central Florida Press. Norling, B. S., Anderson, S. H. & Hubert, W. A. 1992. Roost sites used by Sandhill Crane staging along the Platte River, Nebraska. Great Basin Naturalist 52: 253261. O'Neill, R. V., Krummel, J. R., Gardner, R. H., Sugihara, G., Jackson, B., DeAngelis, D. L., Milne, B. T., Turner, M. G., Zygmunt, B., Christensen, S. W., Dale, V. H. & Graham, R. L. 1988. Indices of landscape pattern. Landscape Ecology 1: 153162. O'Shea, T., Beck, C. A., Bonde, R. K., Kochman, H. I. & Odell, D. K. 1985. An analysis of manatee mortality patterns in Florida, 1976 1981. Journal of Wildlife Management 49: 111. Olff, H. & Ritchie, M E. 2002. Fragmented nature: Consequences for biodiversity. Landscape and Urban Planning 58: 83 92. Opdam, P. F. M., Verboom, J. & Pouwels, R. 2003. Landscape cohesion: An index for the conservation potential of landscapes for biodiversity. Landscape Ecol ogy 18: 113126.
168 Oxley, D. J., Fenton, M. B. & Carmody, G. R. 1974. Effects of roads on populations of small mammals. Journal of Applied Ecology 11: 5159. Parendes, L. A. & Jones, J. A. 2000. Role of light availability and dispersal in exotic plant invasi on along roads and streams in the H. J. Andrews Experimental Forest, Oregon. Conservation Biology 14: 6475. Pauchard, A. & Alaback, P. B. 2004. Influence of elevation, land use, and landscape context on patterns of alien plant invasions along oadsides in protected areas of South -Central Chile. Conservation Biology 18: 238248. Pearlstine, L. Florida Land Cover. Florida Cooperative Fish and Wildlife Research Unit (FCFWRU). 2002. Gainesville, FL, University of Florida, Geoplan Center. Florida Geographic Da ta Library. Peres, C. A. & Lake, I. R. 2003. Extent of nontimber resource extraction in tropical forests: Accessibility to game vertebrates by hunters in the Amazon basin. Conservation Biology 17: 521535. Platt, R. V. 2004. Global and local analysis of f ragmentation in a mountain region of Colorado. Agriculture, Ecosystems & Environment 101: 207218. Platt, W. J. & Schwartz, M. W. 1990. Temperate Hardwood Forests. In Myers, R. L. & Ewel, J. J. (Eds) Ecosystems of Florida (pp. 194229). Orlando: University of Central Florida Press. Puu, T. 2003. Mathematical Location and Land Use Theory: An Introduction. Berlin: Springer Verlag. R Development Core Team 2008. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistic al Computing. Reed, R. A., Johnson -Barnard, J. & Baker, W. L. 1996. Contribution of roads to forest fragmentation in the Rocky Mountains. Conservation Biology 10: 10981106. Ries, L., Fletcher, R. J., Battin, J. & Sisk, T. D. 2004. Ecological responses to habitat edges: mechanisms, models and variability explained. Annual Review of Ecology, Evolution, and Systematics 35: 491522. Riitters, K. H. & Wickham, J. D. 2003. How far to the nearest road? Frontiers in Ecology and the Environment 1: 125129. Road Eco logy Center at the University of California 2005. Events: Acoustic Ecology Winter 2005. Robinson, J. G. & Bodmer, R. E. 1999. Towards wildlife management in tropical forests. Journal of Wildlife Management 63: 113.
169 Rodrigue, J. -P., Comtois, C. & Slack, B. 2006. The Geography of Transportation Systems. Graph Theory: Measures and Indices. Rost, G. R. & Bailey, J. A. 1979. Distribution of mule deer and elk in relation to roads. Journal of Wildlife Management 43: 634641. Sabesan, A. Geo -spatial Assessment of the Impact of Land Cover Dynamics and Distribution of Land Resources on Soil and Water Quality in the Santa Fe River Watershed. 2005. University of Florida. Saeki, M. & Macdonald, D. W. 2004. The effects of traffic on the raccoon dog ( Nyctereutes procy onoides viverrinus ) and other mammals in Japan. Biological Conservation 118: 559 571. Saunders, S. C., Mislivets, M. R., Chen, J. & Cleland, D. T. 2002. Effects of roads on landscape structure within nested ecological units of the Northern Great Lakes Regi on, USA. Biological Conservation 103: 209225. Schmidt, W. 1989. Plant disperal by motor cars. Vegetatio 80: 147152. Schonewald Cox, C., Chambers, S. M., MacBryde, B. & Thomas, L. 1983. Genetics and Conservation: A Reference for Managing Wild Animal and P lant Populations. Menlo Park, CA: Benjamin/Cummings Publishing Company, Inc. Shimble, A. 1953. Structural parameters of communication networks. Bulletin of Mathematical Biophysics 15: 501507. Slabbekoorn, H. & Peet, M. 2003. Birds sing at a higher pitch i n urban areas: Great tits hit the high notes to ensure that their mating calls are heard above the city's din. Nature 424: 267. Smith, D. J. 1999. Identification and Prioritization of Ecological Interface Zones on State Highways in Florida. In Evink, G. L., Garrett, P. & Zeigler, D. (Eds) Proceedings of the Third International Conference on Wildlife Ecology and Transportation Tallahassee: Florida Department of Transportation. Smith, D. J. Ecological Effects of Roads: Theory, Analysis, Management and Plan ning Considerations. 2003. University of Florida. Smith, L. L. & Dodd, C. K. 2003. Wildlife mortality on U.S. Highway 441 across Paynes Prairie, Alachua County, Florida. Florida Scientist 66: 128 140. Smith, L. L., Smith, K. G., Barichivich, W. J., Dodd C. K. Jr. & Sorensen, K. 2005. Roads and Florida's herpetofauna: A review and mitigation case study. In Meshaka, W. E. Jr. & Babbitt, K. J. (Eds) Status and Conservation of Florida Amphibians and Reptiles Malabar, FL: Krieger Publications. Spellerberg, I F. 1998. Ecological effects of roads and traffic: A literature review. Global Ecology and Biogeography 7: 317333.
170 Stone, S. W. 1998. Using a geographic information system for applied policy analysis: the case of logging in the Eastern Amazon. Ecological Economics 27: 4361. Stout, I. J. & Marion, W. R. 1993. Pine Flatwoods and Xeric Pine Forests of the Southern (Lower) Coastal Plain. In Martin, W. H., Boyce, S. G. & Echternacht, A. C. (Eds) Biodiversity of the Southeastern United States: Lowland Terrestr ial Communities (pp. 373446). New York: John Wiley & Sons, Inc. Strittholt, J. R. & DellaSala, D. A. 2001. Importance of roadless areas in biodiversity conservation in forested ecosystems: Case study of the KlamathSiskiyou ecoregion of the United States. Conservation Biology 15: 17421754. Suwannee River Partnership 2007. Suwannee River Partnership: Background. Suwannee River Water Management District basinsusgs24kshp_santafe. 2008. Live Oak, FL, Suwannee River Water Management District. Taaffe, E. J. & Gauthier, H. L. 1973. Geography of Transportation. Englewood Cliffs, N.J.: Prentice Hall, Inc. Taaffe, E. J., Morrill, R. L. & Gould, P. R. 1963. Transport expansion in underdeveloped countries: A comparative analysis. Geographical Review 53: 503 529. T ewksbury, J. J., Levey, D. J., Haddad, N. M., Sargent, S., Orrock, J. L., Weldon, A., Danielson, B. J., Brinkerhoff, J., Damschen, E. I. & Townsend, P. 2002. Corridors Affect Plants, Animals, and Their Interactions in Fragmented Landscapes. Proceedings of the National Academy of Sciences of the United States of America 99: 1292312926. Theobald, D. M., Miller, J. R. & Hobbs, N. T. 1997. Estimating the cumulative effects of development on wildlife habitat. Landscape and Urban Planning 39: 25 36. Tigas, L. A. Van Vuren, D. H. & Sauvajot, R. M. 2002. Behavioral responses of bobcats and coyotes to habitat fragmentation and corridors in an urban environment. Biological Conservation 108: 299306. Tischendorf, L. 2001. Can landscape indices predict ecological proc esses consistently? Landscape Ecology 16: 235254. Tischendorf, L. & Fahrig, L. 2000a. How should we measure landscape connectivity? Landscape Ecology 15: 633641. Tischendorf, L. & Fahrig, L. 2000b. On the usage and measurement of landscape connectivity. Oikos 90: 7 19. Tischendorf, L., Grez, A., Zaviezo, T. & Fahrig, L. 2005. Mechanisms affecting population density in fragmented habitat. Ecology and Society 10: 7.
171 Trombold, C. D. 1991. An Introduction to the Study of Ancient New World Road Networks. In Tr ombold, C. D. (Ed) Ancient Road Networks and Settlement Hierarchies in the New World (pp. 1 9). Cambridge: Cambridge University Press. Trombulak, S. C. & Frissell, C. A. 2000. Review of ecological effects of roads on terrestrial and aquatic communities. Co nservation Biology 14: 18 30. Turner, M. G., Gardner, R. H. & O'Neill, R. V. 2001. Landscape Ecology in Theory and Practice: Pattern and Process. New York: Springer -Verlag. Tyser, R. W. & Worley, C. A. 1992. Alien flora in grasslands adjacent to road and t rail corridors in Glacier National Park, Montana (U.S.A.). Conservation Biology 6: 253 262. U.S. Census Bureau & Population Division 2008. U.S. Census Bureau. Population Estmates Archives. U.S. Department of Transportation & Federal Highway Administration 2000. Highway Traffic Noise in the United States Problem and Response. U.S. Environmental Protection Agency Indicators of Environmental Impacts of Transportation: Highway, Rail, Aviation, and Maritime Transport. EPA 230 R 96 009. 1996. Washington, D.C., Government Printing Office. U.S. Environmental Protection Agency Hydrologic unit boundaries of the conterminous United States in BASINS. 1998. Washington, D.C., USEPA/Office of Water/OST. U.S. Environmental Protection Agency 2001. Our Built and Natura l Environments: A Technical Review of the Interactions between Land Use, Transportation, and Environmental Quality. Washington, D.C.: Government Printing Office. U.S. Fish and Wildlife Service Florida Manatee Recovery Plan (Trichechus manatus latirostris) Third Revision. 2001. Atlanta, GA, U.S. Fish and Wildlife Service. Ullman, E. L. 1956. The Role of Transportation and the Bases for Interaction. In Thomas, W. L., Jr. (Ed) Man's Role in Changing the Face of the Earth (pp. 862 880). Chicago: The University of Chicago Press. Upchurch, S. B., Chen, J. & Cain, C. R. Trends of nitrate concentrations in waters of the Suwannee River Water Management District, 2007. 2007. Live Oak, FL, Suwannee River Water Management District. 5 302007. Urban, D & Keitt, T. 2001. Landscape connectivity: A graphtheoretic perspective. Ecology 82: 12051218. Van Dyke, F. G., Brocke, R. H. & Shaw, H. G. 1986. Use of road track counts as indices of mountain lion presence. Journal of Wildlife Management 50: 102109.
172 Veldkamp, A. & Lambin, E. F. 2001. Predicting landuse change. Agriculture, Ecosystems & Environment 85: 1 6. Vermeulen, H. J. W. 1994. Corridor function of a road verge for dispersal of stenotopic heathland ground beetles carabidae. Biological Conservatio n 69: 339349. Vitousek, P. M., Mooney, H. A., Lubchenco, J. & Melillo, J. M. 1997. Human domination of Earth's ecosystems. Science 277: 494499. Vuchic, V. R. 1999. How did we get here? Transportation policies and practices in the United States. Transpor tation for Liveable Cities (pp. 93 127). New Brunswick, NJ: Center for Urban Policy Research. Ware, S., Frost, C. & Doerr, P. D. 1993. Southern Mixed Hardwoord Forest: The Former Longleaf Pine Forest. In Martin, W. H., Boyce, S. G. & Echternacht, A. C. (Ed s) Biodiversity of the Southeastern United States: Lowland Terrestrial Communities (pp. 447-493). New York: John Wiley & Sons, Inc. Wartenberg, C. M. t. 1966. Von Thnen's Isolated State: An English edition of Der Isolierte Staat by Johann Heinrich Von Th nen. New York: Pergamon Press. Wasserman, S. & Faust, K. 1994. Social Network Analysis. Cambridge: Cambridge University Press. Watkins, R. Z., Chen, J., Pickens, J. & Brosofske, K. D. 2003. Effects of forest foads on understory plants in a managed hardwood landscape. Conservation Biology 17: 411419. Watts, R. D., Compton, R. W., McCammon, J. H., Rich, C. L., Wright, S. M., Owens, T. & Ouren, D. S. 2007a. Response to comment on "Roadless space of the conterminous United States". Science 318: 1240c. Watts, R D., Compton, R. W., McCammon, J. H., Rich, C. L., Wright, S. M., Owens, T. & Ouren, D. S. 2007b. Roadless space of the conterminous United States. Science 316: 736738. Weisstein, E. W. 2008. Circuit Rank. West, D. B. 2001. Introduction to Graph Theory. Upper Saddle River, NJ: Prentice Hall. West, E. 2006. Bioacoustic Profiles: Evaluating Potential Masking of Wildlife Vocal Communication by Highway Noise. In Irwin, C. L., Garrett, P. & McDermott, K. P. (Eds) Proceedings of the 2005 International Conferenc e on Ecology and Transportation (pp. 29). Raleigh, NC: Center for Transportation and the Environment, North Carolina State University. Wilkins, K. T. 1982. Highways as barriers to rodent dispersal. The Southwestern Naturalist 27: 459460. Wilkins, K. T. & Schmidly, D. J. 1980. Highway mortality of vertebrates in southeastern Texas. Texas Journal of Science 4: 343350.
173 Willard, B. E. & Marr, J. W. 1971. Recovery of alpine tundra under protection after damage by human activities in the rocky mountains of Colo rado. Biological Conservation 3: 181190. Willard, T. R., Hall, D. W., Shilling, D. G., Lewis, J. A. & Currey, W. L. 1990. Cogongrass (Imperata cylindrica) distribution on Florida (USA) highway rights of -way. Weed Technology. 4: 658660. Wilson, E. O. 1988. The Current State of Biological Diversity. In Wilson, E. O. (Ed) Biodiversity (pp. 3 18). Washington, D.C.: National Academy Press. With, K. A. 2003. The landscape ecology of invasive spread. The Journal of the Society for Conservation Biology 16: 11921 203. World Commission on Environment and Development 1987. Our Common Future. Oxford; New York: Oxford University Press.
174 BIOGRAPHICAL SKETCH Alisa Coffin was born in 1963 and grew up in Washington, D.C. After graduating from Woodrow Wilson High School in 1980, she attended Cook College at Rutgers, the State University of New Jersey. There she studied environmental science, with an emphasis on a gricultural ecology. As an undergraduate, she spent a semester in 1982 studying at the Land Institute in Salina, KS. After graduating in 1985, worked in the Washington, D.C. area in her parents landscape architectural firm, Coffin & Coffin, and for the U.S Forest Service in Roosevelt, UT. She also traveled to Australia and Europe. Then, in 1987, she lived in Israel for nearly three years, serving as a volunteer in the gardens operations of the Bah World Center in Haifa and Akko. Following this, she ret urned for a Masters of Landscape Architecture from the Harvard Graduate School of Design, where she was introduced to theories of landscape ecology. After graduating in 1993, she traveled to Bolivia with a Fulbright Fellowship, where she began a study of land ownership patterns in the Bolivian altiplano. In 1994, she was appointed to the position of Assitant Professor of Landscape Architecture at Ball State University in Muncie, IN. There she taught classes in environmental structures, site design and regi onal planning. In 1999, she moved with her husband and children to Gainesville, FL, to begin a doctoral program at the University of Florida. During her tenure there, she worked for four and a half years with the U.S. Geological Survey and for over two and a half years with Florida Sea Grant. In both of these positions, she worked on projects in Florida, using GIS for environmental and spatial analysis. She currently lives in Gainesville, FL, with her husband, Oswaldo and their three sons, Jasper, Beni and Noah, where they are active members of the Bah community. In spring of 2009, she is planning to move to Fort Collins, CO, where she will work with the U.S. Geological Survey as a research geographer in the Mendenhall Post -doctoral Research Fellowship pr ogram.