A multi-method approach to delineate and validate migratory corridors

Context Managers are faced with numerous methods for delineating wildlife movement corridors, and often must make decisions with limited data. Delineated corridors should be robust to different data and models. Objectives We present a multi-method approach for delineating and validating wildlife corridors using multiple data sources, which can be used conserve landscape connectivity. We used this approach to delineate and validate migration corridors for wildebeest (Connochaetes taurinus) in the Tarangire Ecosystem of northern Tanzania. Methods We used two types of locational data (distance sampling detections and GPS collar locations), and three modeling methods (negative binomial regression, logistic regression, and Maxent), to generate resource selection functions (RSFs) and define resistance surfaces. We compared two corridor detection algorithms (cost-distance and circuit theory), to delineate corridors. We validated corridors by comparing random and wildebeest locations that fell within corridors, and cross-validated by data type. Results Both data types produced similar RSFs. Wildebeest consistently selected migration habitat in flatter terrain farther from human settlements. Validation indicated three of the combinations of data type, modeling, and corridor detection algorithms (detection data with Maxent modeling, GPS collar data with logistic regression modeling, and GPS collar data with Maxent modeling, all using cost-distance) far outperformed the other seven. We merged the predictive corridors from these three data-method combinations to reveal habitat with highest probability of use. Conclusions The use of multiple methods ensures that planning is able to prioritize conservation of migration corridors based on all available information.