A multi-metric movement model for identifying elk parturition events

For wildlife managers, identifying the time and location of ungulate parturition events is necessary to protect parturition habitat and inform population models. Current methods for identifying parturition events include neonatal capture and monitoring and visual analysis of location data, which are resource intensive and subject to biases. Recently, movement-based analyses using the spatial and temporal patterns contained in location data have shown promise for identifying parturition behavior; however, questions regarding efficacy and best practice have limited the adoption of such movement-based parturition models into management programs. We used location data collected between 2017-2018 from a California population of female Roosevelt elk with known parturition status to build a set of movement-based models. We compared each model's ability to accurately classify individual's parturition status and identify the timing of parturition. The top model showed that within 72 hours of a parturition event, elk made smaller movements, had smaller home ranges, and had longer visits to individual locations. Of the 16 individuals in our study with known parturition status, our approach to classifying an individual's status was 81% accurate in cross-validation, which increased to 100% accuracy when the complete dataset was used to train the top model. We then used the top, global and single-metric models to establish probability thresholds and identify parturition events for collared-elk with unknown parturition status. Our results align with a growing number of studies that have shown movement-based parturition models are a reliable method that could augment or replace field-based techniques for identifying parturition events and quantifying parturition rates in ungulate populations.