Movement behavior in a dominant ungulate underlies successful adjustment to a rapidly changing landscape following megafire

Background Movement plays a key role in allowing animal species to adapt to sudden environmental shifts. Anthropogenic climate and land use change have accelerated the frequency of some of these extreme disturbances, including megafire. These megafires dramatically alter ecosystems and challenge the capacity of several species to adjust to a rapidly changing landscape. Ungulates and their movement behaviors play a central role in the ecosystem functions of fire-prone ecosystems around the world. Previous work has shown behavioral plasticity is an important mechanism underlying whether large ungulates are able to adjust to recent changes in their environments effectively. Ungulates may respond to the immediate effects of megafire by adjusting their movement and behavior, but how these responses persist or change over time following disturbance is poorly understood. Methods We examined how an ecologically dominant ungulate with strong site fidelity, Columbian black-tailed deer (Odocoileus hemionus columbianus), adjusted its movement and behavior in response to an altered landscape following a megafire. To do so, we collected GPS data from 21 individual female deer over the course of a year to compare changes in home range size over time and used resource selection functions (RSFs) and hidden Markov movement models (HMMs) to assess changes in behavior and habitat selection. Results We found compelling evidence of adaptive capacity across individual deer in response to megafire. Deer avoided exposed and severely burned areas that lack forage and could be riskier for predation immediately following megafire, but they later altered these behaviors to select areas that burned at higher severities, potentially to take advantage of enhanced forage. Conclusions These results suggest that despite their high site fidelity, deer can navigate altered landscapes to track rapid shifts in encounter risk with predators and resource availability. This successful adjustment of movement and behavior following extreme disturbance could help facilitate resilience at broader ecological scales. Supplementary Information The online version contains supplementary material available at 10.1186/s40462-024-00488-4.

Table S3 -Starting parameter value ranges for deer behavioral state estimation at the Hopland Research and Extension Center, CA, USA via the two-state hidden-Markov model (HMM).We estimated two behavioral states using the HMM: 1) Resting and 2) Traveling.We randomly selected values from within these ranges for each state in 25 model iterations.We compared the Maximum Likelihood across models to ensure they converged similarly and selected the starting parameters from the model that had the best fit in terms of maximum likelihood.

Parameter Resting Traveling
Step   were randomly sampled from each deer within each time period to assess how robust our findings were to sample size (number of fixes).We find similar trends in changes to homerange size in this rarefied example as our analysis using all the collected GPS-fixes.The Mendocino Complex Fire burned July 27 th , 2018.These study periods include: 2017 Spring and 2018 Spring before the fire ("Prespring"), the summer season just before the fire burned ("Prefire"), directly following the fire ("Recently Burned"), the first spring following the fire ("First Spring"), and 1 full year post fire ("1 Year Post Fire") (from left to right).

Figure S1 -
Figure S1 -Pairwise plots of continuous covariates extracted at the Hopland Research and

Figure S2 -
Figure S2 -Pairwise plots of continuous covariates extracted at the Hopland Research and

Figure S3 -
Figure S3 -Home range size of black-tailed deer (O.hemionus columbianus) across five time

Figure S4 -
Figure S4 -Plotted beta coefficients of the Resource Selection Function model for black-tailed

Figure S5 -
Figure S5 -Histogram of step lengths and density of each predicted state from the best fitting

Figure S6 -
Figure S6 -Histogram of turning angles and density plots of each predicted state from the best

Figure S7 -
Figure S7 -Plotted histogram of step-length pseudo residuals from the fit hidden Markov

Figure S8 -
Figure S8 -Plotted relationship of forage quality, as represented by EVI (Enhanced Vegetation

Table S2 -
Number of collected GPS points collected from each collared deer (Odocoileus hemionus columbianus) at the Hopland Research and Extension Center before and after the Mendocino Complex Fire in 2018.Percent (%) collected displays the percentage of GPS fixes recorded of the maximum possible (n = 1464).

Table S4 -
Welch's t-test results of home range size comparisons across different time periods before and after the 2018 Mendocino Complex Fire.The 2018 Mendocino Complex Fire burned through the Hopland Research Extension Center July 27, 2019."Recently Burned" corresponds to deer home ranges estimated between August 1 st , 2018 -October 1 st , 2018."First Spring" corresponds to home ranges estimated between March 1 st , 2019 -May 1 st , 2019."1 Year Post Fire" corresponds to home ranges estimated between August 1 st , 2019 -October 1 st , 2019."Prespring" corresponds to two combined springs seasons that occurred before the date of the fire: March 1 st , 2017 -May 1 st , 2017 and March 1 st , 2018 -May 1 st , 208."Prefire" corresponds to deer home ranges estimated between May 25 th , 2018 -July 25 th , 2018.* denotes significant difference in home range estimates.

Table S5 -
Welch's t-test results of home range size comparisons using minimum number of GPS fixes (500 fixes per individual) for comparison.Five-hundred GPS fixes were randomly sampled from each deer within each time period to assess how robust our findings were to sample size (number of fixes).We find similar results in this rarefied example as our analysis using all the collected GPS-fixes.The 2018 Mendocino Complex Fire burned through the st , 2017 and March 1 st , 2018 -May 1 st , 208."Prefire" corresponds to deer home ranges estimated between May 25 th , 2018 -July 25 th , 2018.* denotes significant difference in home range estimates.

Table S6 -
Contingency table of deer behavioral states as estimated from the hidden Markov model for deer behavior at the Hopland Research and Extension Center.Behavioral states for each GPS-point were estimated using the "stationary" function of the "moveHMM" (v.1.8)package in R where State 1 = resting and State 2 = traveling.State probabilities for each GPSpoint were then summed across within each Time Period (Recently Burned, First Spring, or 1 Year Post Fire).