Data for: Temporal variations in female moose responses to roads and logging in the absence of wolves

Description

Notes

Methods

Telemetry data were collected on 20 moose (2 males, 18 females; 151,029 GPS locations) between 2017 and 2019. Capture and handling took place in February and March 2017, and our protocols were approved by the ministère de l'Environnement, de la Lutte contre les Changements Climatiques, de la Faune et des Parcs (hereafter MELCCFP; wildlife management permit SEG # 2017-02-10-010-01-S-F) and by the Animal Welfare Committee of the Université du Québec à Rimouski (hereafter UQAR; certificate CPA #68-17-183).

We delineated 5 biological periods (winter, spring/green-up, calving, summer/rearing, fall/rut) and 2 day phases (day, dusk-night-dawn) in order to consider the temporal variations in moose behavior associated with these factors. We determined the cut-off dates of the biological periods by identifying breaks in the distribution of mean movement rates in function of Julian days and using the available knowledge on moose ecology (Hundertmark, 2007; Leblond et al., 2010); we did so for each individual-year combination. We defined day phases using the official sunrise and sunset times (National Research Council Canada, 2021): the day was bounded by the 60-minute period following sunrise to the 60-minute period preceding sunset, and dusk-night-dawn was bounded by the 60-minute period preceding sunset to the 60-minute period following sunrise. 

We defined landcover types using 1: 20,000 ecoforestry maps published by the ministère des Ressources naturelles et des Forêts (hereafter MRNF) and combined information from two mapping exercises (4^th and 5^th decennial inventories) to create updated annual maps to account for anthropogenic disturbance, and fit the GPS data collected from our collared moose (from 2017 to 2019). We regrouped the map polygons into a total of 8 landcover types relevant to moose ecology based on stand cover, composition, height, age, disturbance, land types, and representativity.

We estimated movement rates (in m/h) for each individual and for each step (i.e., the trajectory linking two successive locations spaced by a 2 h interval) using Euclidean distances. We delineated seasonal home ranges using the kernel method based on Brownian bridges (Horne et al., 2007).  We characterized moose habitat selection patterns using resource selection functions (hereafter RSF; Manly et al., 2002) with the different landcover types and other covariates (elevation, slope, day phase, presence of forest and paved roads in buffer zone) for each biological period. For each biological period, we retained only combinations of individual ID – year for which we had data for the entire or nearly the entire duration of the biological period (i.e., ~4% of the ID – year were removed from the dataset for the statistical analyses). We compared moose space use patterns (movement rates and home-range sizes) between biological periods using an analysis of variance (ANOVA) with repeated measures followed by a multiple comparison test (Tukey). The RSF we used to describe the habitat selection patterns was a mixed logistic regression contrasting GPS locations (coded 1) with random points (coded 0) with different combinations of the following independent variables: landcover types, topography variables, day phases, and the presence of forest and paved roads in the buffer zone around each location.