Are Minnesota moose warming up to climate change? A validation of techniques for remotely monitoring moose behavior and body temperature.

Abstract

Climate change is causing rapid shifts in species distributions across the globe. Large-bodied endotherms, especially those at the edge of their bioclimatic range, are particularly vulnerable to these changes. It is critical that we understand the physiology, behavior, and energetic needs of these animals to develop effective management and conservation plans. Advancements in biotelemetry devices have greatly improved our ability to collect physiological and behavioral data from free-ranging animals; however, our understanding of how the data can be processed and used is still in its infancy. One species of conservation concern, the moose (Alces alces), experienced a 58% population decline in northeastern Minnesota between 2006 and 2017. To better understand behavioral and physiological responses of this species to increasing ambient temperature, the Minnesota Department of Natural Resources deployed two types of biotelemetry devices in moose throughout northeastern MN: 1) rumen boluses, known as mortality implant transmitters (MITs), capable of recording internal body temperatures, and 2) global positioning system (GPS) collars equipped with dual-axis activity sensors that detect and record changes in neck movements. The main goals of my research were to determine the accuracy of MIT-derived core body temperatures and test the efficacy of using dual-axis activity sensors for remotely predicting behavioral states of moose. Ten captive female moose (>2 years old) at the Moose Research Center in Kenai, Alaska with MITs were fit with vaginal implant transmitters (VITs) capable of recording internal body temperature, and GPS collars for 12 months starting in December 2014. A total of 384 hours of behavioral observations were collected during four, two-week windows distributed across seasons. I observed a notable effect of water intake on MIT-derived temperatures and developed an approach for censoring these observations. Using linear mixed-effects models, I predicted moose core body temperature (as measured by VITs) and found that on average, the difference between predicted and observed body temperatures was 0.05°C for winter and 0.33°C for summer, with >90% of prediction intervals containing the observed VIT-derived temperatures. Using a Dirichlet regression approach to analyze the dual-axis activity sensor data, I predicted the proportion of time individual animals spent resting, foraging, and moving during 5-minute intervals; these results were used to understand how behavioral states change as a function of habitat, ambient temperature, and time of day. I demonstrated that combining biotelemetry devices with modern statistical approaches allows researchers to examine the physiological and behavioral responses of moose to increasing ambient temperatures and changing landscapes at finer temporal and spatial scales than previously possible.