Using seasonal landscape models to predict space use and migratory patterns of an arctic ungulate

• Published in: Movement Ecology
• Main Authors: A. P. Baltensperger, K. Joly
• Format: Article in Journal/Newspaper
• Language: English
• Published: BMC 2019
• Subjects: Alaska; Caribou; Climate; Ecological niche model; Infrastructure; Machine learning; Biology (General); QH301-705.5; Arctic; Climate change; Sea ice
• Online Access: https://doi.org/10.1186/s40462-019-0162-8; https://doaj.org/article/330fe3b330024f2da79647fdfe7f8dfe

Abstract Background Caribou in the Western Arctic Herd undertake one of the longest, remaining intact migrations of terrestrial mammals in the world. They are also the most important subsistence resource for many northern rural residents, who rely on the caribou’s migratory movements to bring them near for harvest. Migratory geography has never been static, but subsistence harvesters have reported recent shifts in migration away from areas where they traditionally occurred. The reasons behind these changes are not well-understood, but may be related to rapid climate change and anthropogenic disturbances. Methods To predict changes in distribution and shifting migratory areas over the past decade, we used GPS telemetry data from adult females to develop predictive ecological niche models of caribou across northwestern Alaska. We employed the machine-learning algorithm, TreeNet, to analyze interactive, multivariate relationships between telemetry locations and 37 spatial environmental layers and to predict the distributions of caribou during spring, calving season, insect-harassment season, late summer, fall, and winter from 2009 to 2017. Model results were analyzed to identify regions of repeated predicted use, quantify mean longitude, predict land cover selection, and track migratory changes over time. Results Distribution models accurately predicted caribou at a spatially-explicit, 500-m scale. Model analyses identified migratory areas that shifted annually across the region, but which predicted 4 main areas of repeated use. Niche models were defined largely by non-linear relationships with coastally-influenced, climatic variables, especially snow-free date, potential evapo-transpiration, growing season length, proximity to sea ice, winter precipitation and fall temperature. Proximity to roads and communities were also important and we predicted caribou to generally occur more than 20–100 km from these features. Conclusions Western Arctic Herd caribou were predicted to occur in warmer, snow-free and treeless ...