Species conservation initiatives play an important role in ecological studies. Occupancy models have been a useful tool for ecologists to make inference about species distribution and occurrence. Bayesian methodology is a popular framework used to model the relationship between species and environmental variables. In this dissertation we develop a Gibbs sampling method using a logit link function in order to model posterior parameters of the single-season spatial occupancy model. We incorporate the widely used Intrinsic Conditional Autoregressive (ICAR) prior model to specify the spatial random effect in our sampler. We also develop OccuSpytial, a statistical package implementing our Gibbs sampler in the Python programming language. The aim of this study is to highlight the computational efficiency that can be obtained by employing several techniques, which include exploiting the sparsity of the precision matrix of the ICAR model and also making use of Polya-Gamma latent variables to obtain closed form expressions for the posterior conditional distributions of the parameters of interest. An algorithm for efficiently sampling from the posterior conditional distribution of the spatial random effects parameter is also developed and presented. To illustrate the sampler's performance a number of simulation experiments are considered, and the results are compared to those obtained by using a Gibbs sampler incorporating Restricted Spatial Regression (RSR) to specify the spatial random effect. Furthermore, we fit our model to the Helmeted guineafowl (Numida meleagris) dataset obtained from the 2nd South African Bird Atlas Project database in order to obtain a distribution map of the species. We compare our results with those obtained from the RSR variant of our sampler, those obtained by using the stocc statistical package (written using the R programming language), and those obtained from not specifying any spatial information about the sites in the data. It was found that using RSR to specify spatial random effects is both statistically and computationally more efficient that specifying them using ICAR. The OccuSpytial implementations of both ICAR and RSR Gibbs samplers has significantly less runtime compared to other implementations it was compared to.

Converting exotic forages to native warm-season grasses (NWSG) such as big bluestem (Andropogon gerardii), little bluestem (Schizachyrium scoparium), and Indian grass (Sorghastrum nutans) offers a sustainable alternative because NWSG may yield comparative livestock gains with less fertilizer, as well as offer habitat for arthropods and declining grassland bird species such as Dickcissels (Spiza americana). In the Southeastern United States, the predominant forage species are exotics such as bermudagrass (Cynodon dactylon) and tall fescue (Schedonorus arundinaceus), so NWSG conversion could substantially improve sustainability and wildlife habitat on private lands in the region. In 2011 and 2012, I studied response of Dickcissels and arthropods to forage origin, diversity, and grazing at the Mississippi State University Prairie Research Unit in Monroe Co., MS, USA. Four treatments were established among 12 pastures representing a gradient in management intensity, including a grazed bermudagrass and tall fescue mix (GMEP), grazed Indian grass monoculture (GINP), grazed mixed native pasture (GMNP), and a non-grazed mixed native pasture (NMNP). Grazed treatments were stocked with steers from May through August each year. I also evaluated the economic implications of each grazing system. In general, there was a positive response to native grasses among Dickcissels and arthropods and a negative effect from grazing. Dickcissel productivity (fledgling/ha) was highest in NMNP and lowest in GMEP, whereas productivity was intermediate and similar among grazed native pastures. This pattern was attributable to availability of suitable nest sites because nest survival and brood size did not vary by treatment. Several arthropod taxa responded positively to greater vegetation density in NMNP, but communities were largely similar among grazed pastures irrespective of forage origin or diversity, suggesting little short-term response to NWSG. In spite of unfavorable growing conditions due to drought, cattle grazing NWSG consistently outperformed conspecifics in GMEP, with 45–72% higher average daily gain. Partial budget analysis indicated that NWSG pastures yielded up to 36% marginal rates of return despite establishment costs. These results suggest NWSG conversion can benefit tall grass specialists such as Dickcissels while offering competitive alternatives to exotic forages, resulting in net benefits for both conservationists and producers.

Rafinesque’s big-eared bats (Corynorhinus rafinesquii; RBEB) and southeastern myotis (Myotis austroriparius; SEM) are listed on IUCN Red List of Threatened Species. Limited information on roost sites exists. I conducted roost surveys for RBEB and SEM on public forest lands in central Mississippi during winter and spring 2010. During winter, RBEB and SEM roosted in cavity trees with greater trunk diameters. In spring, roost trees used by SEM were located in forested areas of lower elevation, less slope, and greater distances from roads. Because imperfect detection can affect occupancy estimates, I estimated detection probabilities under different survey methods. Detection probability ranged from 95 – 100% and 92 – 99% when one to 2 observers used repeated surveys and removal method, respectfully. When estimating for abundance, presence of ≤20 bats led to count errors of <4%. When >20 bats were present, count errors were 38.1%. Observers correctly identified species 91% of the time.

The Virginia northern flying squirrel (Glaucomys sabrinus fuscus) and the Carolina northern flying squirrel (G. s. coloratus) are geographically isolated subspecies of the northern flying squirrel found in montane conifer-northern hardwood forests the Appalachian Mountains of the eastern United States. Both subspecies were listed under the Endangered Species Act in 1985 as endangered, and accordingly, the Virginia northern flying squirrel and the Carolina northern flying squirrel are considered high conservation priorities by state and federal agencies. Although the listing prompted work to determine the broad distribution and habitat associations of both subspecies, numerous data gaps remain, particularly with regard to habitat management and development of efficient monitoring techniques. Regional interest in restoration of red spruce (Picea rubens) forests in the central and southern Appalachian Mountains, considered to be the flying squirrels' primary habitat, increases the importance of understanding habitat selection and managers' ability to detect squirrels at multiple spatial and temporal scales. I compared two novel survey techniques (ultrasonic acoustics and camera trapping) to a traditional technique (live trapping) to determine which method had higher probability of detection (POD) and lower latency to detection (LTD, number of survey nights to initial detection) of northern flying squirrels in the region. Both novel techniques performed better than the traditional techniques with higher POD and lower LTD. I found that ultrasonic acoustics and camera trapping had similar POD, whereas LTD was significantly lower with ultrasonic acoustics versus camera trapping. Additionally, the ability to distinguish between northern flying squirrels and the parapatric southern flying squirrel (G. volans) also is possible with ultrasonic acoustics, but not with camera trapping. This ultimately makes ultrasonic acoustics the most effective and efficient method to obtain detection/non-detection data. To better inform management decisions and activities (i.e., red spruce restoration), this method should be used in conjunction with existing traditional monitoring techniques that provide demographic data such as nest boxes. I assessed habitat selection of radio-collared Virginia and Carolina northern flying squirrels at multiple spatial scales with use-availability techniques. I analyzed field data from paired telemetry and random points and determined Virginia northern flying squirrels microhabitat (within-stand habitat) selection showed preference for conifer-dominant stands with deep organic horizons, a factor that might be directly linked to food (hypogeal fungi) availability. Similar to previous studies on the Virginia northern flying squirrel on the landscape- and stand-level using Euclidean distance based analysis, Carolina northern flying squirrels also selectively preferred montane conifer forests in greater proportion than their availability on the landscape. Additionally, Carolina northern flying squirrels did not select for or against northern hardwood forests regardless of availability on the landscape. Habitat preference of both subspecies indicates that red spruce restoration activities may be important for the persistence of Appalachian northern flying squirrels into an uncertain future, as anthropogenic climate change may cause further reduction of the quality and extent of high-elevation montane conifer forests in the region. / Ph. D.

Freshwater turtles are among the most threatened groups of taxa globally, and the bog turtle, Glyptemys muhlenbergii is among the most imperiled in North America. In Virginia, USA, bog turtles are restricted to occupying Appalachian Mountain fens. Fens are naturally small and fragmented wetlands characterized by elevated water tables and an open canopy. Although there is a strong need to document and monitor populations of bog turtles, efforts to do so are often limited by the low detection of the species. The first objective of this thesis was to assess proposed methodologies for locating populations of turtles on the landscape. My first chapter assessed a previously-developed habitat distribution model for bog turtles using an occupancy modeling approach. I conducted 216 surveys of 49 discretely predicted patches of habitat, recording conditions such as weather, size of wetland and time of year, hypothesized to affect detection during each survey. In addition, I assessed factors including stream entrenchment, grazing presence and surrounding impervious surfaces for each surveyed patch to identify data sources that could improve future models or better assess sites. I found that sites with larger total wetland area had higher detection per survey, possibly due to larger sites having higher densities of turtles (among other explanations), and that sites with higher amounts of impervious surfaces within their drainage were less likely to be occupied. In addition to the bog turtle, several plant species also occur in mountain fens. These species usually have a locally rare distribution or are disjuncts from a more northern latitude. Because of these traits, a high diversity of specialist plants may be indicative of a fen with a robust hydrology that has historically been less disturbed. Past site quality analyses have proposed using indicator diversity to assess sites, but no study has found if these species to tend to co-occur. My second chapter examines this hypothesis. I first chose a list of plant species that would most likely have habitat requirements similar to those of turtles. Then, at 12 sites, 6 with turtles and 6 without, I conducted a complete floral inventory. I first tested community-wide differences between the floral communities of these sites and found no difference, but when I narrowed my analysis to examining occurrence patterns of plant species determined a priori to be fen specialists and Glyptemys muhlenbergii, a pattern of co-occurrence was found. This lends support to the idea that indicator plants could be used as a tool to better evaluate sites that may have bog turtles. My last chapter investigated movement of bog turtles in a landscape impacted by anthropogenic development. Movement of turtles between adjacent sites is critical to maintaining genetic diversity and maintaining metapopulation integrity. Despite this importance, records of long distances movements among wetlands are scarce in the literature, likely due to the lack of long-term studies for areas with multiple adjacent sites. In Virginia, mark recapture monitoring has been done intermittently in a cluster of sites for over 32 years. To determine the prevalence of movement among sites for bog turtles, I examined the dataset for all instances of turtles found at sites different from their last capture. I calculated the straight-line distance for each recorded movement. I also examined the sex of the turtle to test whether sex influences movement the frequency and distance of movements. For a subset of movements, I calculated least-cost pathways to identify possible barriers to movement using a previously published resistance model. I found 21 instances where a turtle was caught at a different site than its last capture over 32 years of monitoring. Neither sex was more likely to move farther than the other. Although the study's observed rate of movement may appear low, it is likely an underestimate when detection and asymmetric sampling are taken into account. The least cost pathways analysis suggested that roads or driveways were likely crossed for a significant portion of movement events. Finally, to examine how movement may be affecting the current distribution of bog turtles, I described a method to test whether adjacency to known populations influences the probability of a new site being occupied by turtles. I prove the utility of the method by applying it to a map of bog turtle occurrences collected over this study and show that it can account for habitat differences and barriers to movement between sites as well. In spite of plausibility of the method, limitations in how occurrence data are currently collected prevent its immediate application. Together, this thesis will help managers not only find and assess wetlands on the landscape, it will also provide information about the network of connected patches on the landscape. Knowing where bog turtles are and what wetlands or sub-populations are potentially connected will allowed for a more directed and informed regional management strategy. / Master of Science / Freshwater turtles are facing population declines worldwide, and the bog turtle Glyptemys muhlenbergii is among the most imperiled in North America. Bog turtles occupy naturally small, specialized wetlands called Appalachian Mountain fens. The prevalence of fens on the landscape has declined over recent decades due to agricultural practices. Although there is a strong need to document and monitor bog turtle populations due to their threatened status, bog turtles are difficult to find due to their small size and ability to burrow completely into substrate. Thus, considerable effort must be expended to find populations and track their status. The first overall objective of this thesis was to assess methods for locating populations of bog turtles. My first chapter tests a habitat distribution model that uses publicly available landscape data such as topopgraphy and land cover to predict areas likely to contain turtles. To do this, I systematically surveyed 49 predicted sites multiple times each over 2 years. Simultaneously, I recorded variables such as the time of year, size of the wetland and the weather to determine whether any factor significantly explained the ability to find turtles on any given survey. In addition, I was able to record several variables relating to wetland quality and isolation that were not in the initial model. I found that larger wetlands were easier to search than smaller wetlands, possibly due to larger sites having more turtles, and that wetlands near more impermeable surfaces (such as roads and buildings) were less likely to have bog turtles. As another potential method to find bog turtles and assess sites, we tested the use of 'pristine indicator' plants as a metric for potential wetlands. Mountain fens have specific attributes, such as high groundwater influence and exposure to a large amount of sunlight. Several species, including the bog turtle, are specialized to these factors and are rarely found in the surrounding landscape. Because a distinct community exists for mountain fens in this region, sites with a higher diversity of fen specialist plants may be indicative of a higher quality site which can support more specialists, including the bog turtle. My second chapter tests this hypothesis. I first chose a list of species that would most likely have habitat requirements similar to those of bog turtles. Then, at 12 sites I documented every plant species I encountered within the wetland. I compared the plant community as a whole between bog turtle-occupied and unoccupied sites and found no significant difference between the two. When I narrowed my analysis to focus on plants I previously identified as sharing habitat requirements with the bog turtle, I found a strong pattern of their co-occurrence with bog turtles. This lends support to the idea that these 'pristine indicator' plants could be used as a tool to better evaluate sites that may have bog turtles. My last chapter investigates movement of bog turtles in a landscape impacted by human development. Movement of turtles between adjacent wetlands is critical to maintaining long term regional viability of the species, as it lets turtles colonize new sites and exchange genes. Despite the importance of these movements, records of turtles moving long distances between two wetlands is scarce in the literature, likely due to the lack of long-term studies for areas with multiple adjacent wetlands. One method of recording movements is by marking turtles with a unique ID and recording where it was encountered as wetlands are surveyed on the landscape. In Virginia, this procedure has been conducted at multiple sites over 32 years. To understand the prevalence of movement between sites for this species, I examined this dataset and examined all instances of a turtle being found at a site different from its last capture. I recorded the straight-line distance moved for each recorded movement as well as the sex of the turtle, to test if either sex was more or less likely to undertake these movements. Then, for a subset of movements, I calculated least-cost pathways, a metric that accounts for landscape features and plots the easiest route for turtles to move. This way, I could evaluate the prevalence of barriers to movement, such as roads or development, on the landscape. I found 21 documented movements among sites over 32 years of monitoring. Neither sex was more likely to move further than the other. Compared to studies looking at other freshwater turtles, the observed rate of movement appeared low, but this was likely an underestimate due to the difficulty of capturing specific individuals. I also found evidence of significant barriers to movement in 13 out of 17 evaluated least-costs paths, usually roads or driveways. Finally, to examine how movement affects bog turtle distribution, I describe a methodology of testing if adjacency to known populations influences the probability of a new site being occupied by turtles. I demonstrate the plausibility of the method by applying it to a map of occurrences collected over this study and show that it can account for habitat differences and barriers to movement between sites as well. However, limitations in my sampling scheme limit conclusions from my dataset. Together, these findings will help future managers find where turtles are and which sites may be connected. These results will help managers make more informed decisions for managing bog turtles at a statewide level.

The kit fox (Vulpes macrotis) is a species of conservation concern in western North America. Recent methods for monitoring populations of kit fox include using lures and remote cameras in an occupancy-modeling framework and habitat modeling to predict areas of occupancy. In chapter one, we tested the optimal lure and movement procedure for scent stations to maximize visits and detection of foxes, thereby improving estimates of occupancy. Between May 2015 and October 2016, we placed remote cameras at 522 random locations throughout nine study areas in the Colorado Plateau, Great Basin Desert, and Mojave Desert. Each location was randomly assigned one of three methods (Scented Predator Survey Disks, cotton swabs, or hollowed golf ball) to broadcast one of three lure types (Red and Gray Fox liquid lure, Willey liquid lure, and fatty acid lure). After seven nights, half of all stations were moved 100 meters within the same sample grid cell, while the others remained in the same location. Stations were then monitored for an additional week. We used Program MARK and AIC model selection to identify optimal lure types and broadcast methods and to estimate rates of occupancy. Detection of kit foxes differed by method of scent deployment; cotton swabs were associated with the highest rates of visitation. Detection of kit foxes did not differ by lure type. Relocating the scent station after one week did not influence detection probability. We suggest that the use of cotton swabs maximizes detection, and therefore, the precision of estimates of occupancy. For chapter two, we used resource selection functions to identify variables that best discriminated between locations where kit fox were detected and random locations. We then produced a habitat map that predicted the relative probability of kit foxes occurring across seven study areas throughout the state of Utah. We placed remote cameras at 458 randomly selected locations throughout the study areas in the Colorado Plateau, Great Basin Desert, and Mojave Desert. We detected kit foxes at 157 "use" points from these cameras between May 2015 and October 2016. We then compared the attributes of these "use" points to 14,742 available, randomly selected points located within the study areas using variables derived from a Geographic Information System (GIS). We used model selection and minimization of AIC values to determine key habitat characteristics that differentiated use and random locations. We identified slope, elevation, and soil type as significant variables (P < 0.05) in habitat selection of kit foxes. Kit foxes selected areas that were 1) less steep, 2) lower in elevation, and 3) classified as having silty soils. The identification of these specific variables from our modeling effort was generally consistent with kit fox ecology. Our study produced a habitat model that can serve as a foundation for future monitoring efforts of kit foxes in potential habitat across Utah.

Understanding species-habitat relationships for endangered species is critical for their conservation. However, many studies have limited value for conservation because they fail to account for habitat associations at multiple spatial scales, anthropogenic variables, and imperfect detection. We addressed these three limitations by developing models for an endangered wetland bird, Yuma Ridgway's rail (Rallus obsoletus yumanensis), that examined how the spatial scale of environmental variables, inclusion of anthropogenic disturbance variables, and accounting for imperfect detection in validation data influenced model performance. These models identified associations between environmental variables and occupancy. We used bird survey and spatial environmental data at 2473 locations throughout the species' U.S. range to create and validate occupancy models and produce predictive maps of occupancy. We compared habitat-based models at three spatial scales (100, 224, and 500 m radii buffers) with and without anthropogenic disturbance variables using validation data adjusted for imperfect detection and an unadjusted validation dataset that ignored imperfect detection. The inclusion of anthropogenic disturbance variables improved the performance of habitat models at all three spatial scales, and the 224-m-scale model performed best. All models exhibited greater predictive ability when imperfect detection was incorporated into validation data. Yuma Ridgway's rail occupancy was negatively associated with ephemeral and slow-moving riverine features and high-intensity anthropogenic development, and positively associated with emergent vegetation, agriculture, and low-intensity development. Our modeling approach accounts for common limitations in modeling species-habitat relationships and creating predictive maps of occupancy probability and, therefore, provides a useful framework for other species.

Snowy plovers (Charadrius nivosus) are small, ground-nesting shorebirds that are a species of conservation concern throughout North America. Despite increased efforts to understand factors contributing to the decline of snowy plover, little is known about habitat selection and breeding ecology of snowy plover for the large population found in the Great Basin. We tested hypotheses concerning the occupancy and nesting success of snowy plover. First, we identified factors influencing snowy plover nest survival at Great Salt Lake, Utah. We hypothesized that snowy plover would demonstrate differences in nest survival rates across years due to differences in habitat characteristics, predator abundance, human influence, resource availability, and fluctuating water levels. We conducted nest surveys at five sites along the Great Salt Lake to locate new nests or monitor known nests until nest fate was determined. We found 608 nests between 2003, 2005-2010, and 2012. The most common cause of nest failure was predation, followed by weather, abandonment, and trampling. Nest survival estimates ranged from 4.6 -- 46.4% with considerable yearly variation. There was no correlation between researcher activity (visits to nests and trapping of adults) and nest survival. Nests in close proximity to roads had lower survival than nests far from roads. Nests located on barren mudflats also had lower survival than nests in vegetated areas or near debris. We found that nests had a higher probability of survival as they increased in incubation stage. Because nesting areas around the Great Salt Lake host some of the largest concentrations of breeding snowy plover in North America, we suggest that managers consider measures to maintain suitable nesting habitat for snowy plover. Second, we determined factors affecting snowy plover occupancy and detection probabilities in western Utah between 2011 and 2012. We hypothesized that snowy plover would be associated with spring water flows and sparsely vegetated salt flats. We made repeated visits to randomly selected survey plots recording the number of snowy plover adults and habitat characteristics within each plot. We modeled the relationship between snowy plover detection probability and habitat and environmental characteristics. The detection probability was 77% (95% CI = 64 -- 86%) and did not vary by year. There was a positive relationship between ambient temperature and detection probability. Next, we modeled the relationship between snowy plover occupancy and individual habitat characteristics including distance to water, distance to roads, land cover types, and vegetative characteristics. Snowy plover occupancy did not vary by year and was estimated at 12% (95% CI = 7 -- 21%). Occupancy was best predicted by close proximity to water, playa land cover, and minimal shrub cover. We used habitat characteristics that best predicted snowy plover occupancy to generate a predictive habitat model that can help prioritize future snowy plover surveys and guide conservation efforts.

Bat populations across North America are either facing new threats from white-nose syndrome (WNS) and wind energy development or have already experienced precipitous declines. Accordingly, researchers and managers need to know how to best monitor bats to document population and distribution changes, as well as where to look for persisting populations. Landscape-scale WNS impacts to summer bat populations are not well understood, and although acoustic monitoring is commonly used to monitor these populations, there is limited information about differences among acoustic detectors and the implications to managers thereof. My objectives were to model the relationship between WNS impact, influence of available hibernacula, and environmental factors for summer nightly presence of three WNS-affected bats and to compare how multiple models of acoustic detectors perform in terms of detection probability and nightly recorded bat activity. I collected acoustic data from 10 study areas across Virginia, West Virginia, Ohio and Kentucky to describe changes in nightly presence of WNS-affected bat species during summer 2017. During the same period of time, I compared five types of acoustic detectors at Fort Knox, Kentucky. My results show the potential efficacy of using a WNS impact-year metric to predict summer bat presence, and highlight which environmental variables are relevant for large-scale acoustic monitoring. Additionally, my findings suggest that each of the detector types tested would suffice for most research and monitoring activities, but standardization of detector type within the scope of a project or study should be encouraged. / Master of Science / Bat populations across North America are either facing new threats from white-nose syndrome (WNS) and wind energy development or have already experienced devastating declines. Accordingly, wildlife biologists need to know how to best monitor bats to document population and distribution changes, as well as where to look for remaining populations. Landscape-scale WNS impacts to summer bat populations are not well understood, and although acoustic technology is commonly used to monitor these populations, there is limited information about differences among acoustic detectors and the implications to managers thereof. My objectives were to model the relationship between WNS impact, influence of available bat hibernation caves, and environmental factors for summer nightly presence of three WNS-affected bats and to compare how multiple models of acoustic detectors perform in terms of detection probability and nightly recorded bat activity. I collected acoustic data from 10 study areas across Virginia, West Virginia, Ohio and Kentucky to describe changes in nightly presence of WNS-affected bat species during summer 2017. During the same period of time, I compared five types of acoustic detectors at Fort Knox, Kentucky. My results show potential viability of a WNS impact-year metric to predict summer bat presence, and highlight which environmental variables are relevant for large-scale acoustic monitoring. Additionally, my findings suggest that each of the detector types tested would suit most research and monitoring activities, but standardization of detector type within the scope of a project or study should be encouraged.

Freshwater mussels (Unionidae) in the United States have experienced dramatic declines, and 25% species are listed as federally endangered. Hence, recovery plans for endangered species proposed a strategy of propagation of young mussels for release to natal rivers to augment declining populations. In this study, I conducted laboratory experiments, assessed site suitability for mussel restoration, and evaluated survival and growth rates of released mussels to meet the requirements of recovery plan. I conducted multiple experiments to develop an improved protocol for juvenile mussel propagation and culture. Significantly greater survival and growth rates were found in newly metamorphosed juveniles of the rainbow mussel (Villosa iris) reared in a substrate of fine sediment and one-month-old juveniles of wavy-rayed lampmussel (Lampsilis fasciola) fed on natural food in pond water. Bio-filter media greatly increased water quality by reducing the concentration of ammonia and nitrite. The negative impacts of flatworm predation and filamentous algae in juvenile culture were controlled, and juvenile escapement was prevented. Juvenile mussels were successfully produced and cultured to stockable size (>15 mm) for release. I released laboratory-propagated mussels at three historically important sites in Clinch and Powell rivers for the assessment of site suitability. Use of cages was the most effective method to determine site suitability because the free-released mussels (untagged, tagged) had low catchability. Mussels released at Horton Ford, Clinch River, exhibited significantly faster growth. Horton Ford is the most suitable site, while environmental conditions at Fugate Ford, Powell River, are deemed unsuitable for mussel restoration and recovery. To facilitate the detection of released mussels, I applied Passive Integrated Transponder tags to laboratory-produced juveniles of the endangered Cumberlandian combshell (Epioblasma brevidens) and released them near Brooks Bridge, Powell River. The detection probability increased above 98%. I developed a set of hierarchical Bayesian models incorporating individual variations, seasonal variations, periodic growth stages and growth cessation to estimate survival, detection probability and growth of released mussels in a changing environment. Mussels of E. brevidens exhibited great survival (> 99% per month) and growth, indicating suitable conditions for recovery of this endangered species at this site. / Ph. D.