Integrating Multiple Databases to Evaluate the Impact of Weather on Animal Movements.

摘要

Wildlife ecologists strive to correctly model and predict complex relationships between animals and their environment and facilitate management decisions of delicate ecosystems. Spatiotemporal patterns of wildlife provide an interface for understanding movements and interactions with their environment. Recent advances in monitoring systems across scientific domains have made it possible, yet challenging, to combine and use detailed output from various sources to address questions spanning multiple disciplines.;To address these complex questions, WolfScout wildlife and weather tracking system was designed to automate integration of the latest technological advances in GPS collars, weather stations, drought conditions, severe weather reports, and animal demographic information (Chapter 1). Also, WolfScout spatial database management system stores a variety of landscape maps and calculates distance between animal locations and landscape characteristics. The distance data, environmental conditions, and animal information are fully downloadable through a secure website. WolfScout was designed to promote interoperability of data between researchers and software applications while standardizing analyses of animal interactions with their environment. However, analyses of this intimate view of wildlife and the environment presented problems from a lack of statistical methods that handle "big data" questions. To overcome these challenges, a recursive partitioning method (i.e., random forest) was used to identify variables of importance and condense datasets to manageable sizes for traditional regression methods.;To demonstrate WolfScout's capabilities, GPS locations from eastern coyote (Canis latrans; Chapter 2) and white-tailed deer ( Odocoileus virginianus; Chapter 3) residing in the Sandhills region of North Carolina were streamed into the system and processed. From WolfScout, distances to the closest water centerline, all available variables from closest weather stations, multi-sensor precipitation estimates, vapor pressure deficits (VPD), and drought conditions per animal location were downloaded. The centerlines provided the only areas of dense cover in the longleaf pine-wiregrass ecosystem, otherwise characterized by porous, sandy soils and sparse vegetation. For each species, variables of importance were identified with random forest methods and stepwise Akaike information criterion and variance inflation factors were calculated to identify top regression models of weather influence on distance to lowlands. Both species were relocated in bottomlands during day and early evening hours. The animals were closest to the water centerlines during denning/breeding for coyotes and late gestation/parturition for deer. Increasing VPD and dry conditions correlated with lowlands for coyotes and deer. Coyotes were in lowlands during high winds, low pressure, and low temperatures with high relative humidity and dew point temperatures. In winter, deer were located in areas sheltered from northeasterly high winds at times of high visibility, pressure, temperature, and humidity. In spring and summer, deer were in lowlands during precipitation, increased chance of precipitation, and impending storms.;To complement these results, the influence of severe weather events, weather station variables and drought conditions on human-mesocarnivore interactions in the Sandhills area were analyzed (Chapter 4). In general, most interactions occurred during severe drought and high VPD. Specifically, non-rabid fox encounters were at times of high VPD and wind gusts, whereas human-rabid fox interactions occurred during moderate droughts. High temperatures, relative humidity, precipitation and wind speeds from the southwest, and parturition time characterized non-rabid raccoon encounters. Human-rabid raccoon interactions occurred during severe weather events, high pressures, high dew point temperatures and high wind speeds. In general, fox were encountered during dry conditions and raccoons were encountered during times of severe weather and precipitation.;Through the design and implementation of WolfScout wildlife and weather tracking system, climatic variables have shown to influence wildlife locations, including encounters with humans. This weather information should be included in subsequent analyses aiming to predict the spatial-temporal dynamics of wildlife.