Data for behavioral state-dependent habitat selection analysis of translocated female greater sage-grouse, North Dakota 2018-2020

We equipped 48 female sage-grouse with rump-mounted Global Positioning System (GPS) Platform Transmitter Terminal (PTT) ARGOS-enabled tracking devices (23 g, GeoTrak, Inc.) scheduled to acquire 6 locations a day at irregular intervals. We resampled the data at regular 6-hour intervals using a Continuous Time Movement Model. For the first part of the analysis, we segmented the tracks into behavioral phases (corresponding to either an exploratory or a restricted state) using a Hidden Markov Model. For the second part of the analysis, we fit Integrated Step Selection models to individuals in each behavioral state. For each used step, we generated a set of 100 random steps drawn from the empirical population-level distribution of steps lengths and turning angles in the two behavioral states. We intersected each step with environmental covariates including aspect, percent sagebrush cover, percent herbaceous cover, sagebrush patch contiguity, slope, distance to roads, distance to well pads, and distance to mesic habitat. All habitat variables are scaled and centered. We removed geographical coordinates to fulfill our funding agreements not to disclose the location of our tracked individuals. We processed data in R (R Core Team, 2020) using the packages ‘amt’, ‘tidyverse’, ‘sf’, ‘raster’, and ‘lubridate’. For more details, see the associated manuscript (Picardi et al. 2021, Journal of Applied Ecology).本研究为48只雌性艾草松鸡（sage-grouse）佩戴了臀部搭载的支持ARGOS系统的全球定位系统（Global Positioning System, GPS）平台发射终端（Platform Transmitter Terminal, PTT）追踪设备（单台重量23克，由GeoTrak公司生产），设备原设定以非固定间隔每日采集6个定位点位。本研究采用连续时间运动模型（Continuous Time Movement Model）将原始数据以固定6小时间隔进行重采样。在分析的第一阶段，我们通过隐马尔可夫模型（Hidden Markov Model）将运动轨迹划分为两类行为阶段，分别对应探索活动状态与受限活动状态。分析的第二阶段，我们针对各行为状态下的个体拟合集成步选择模型（Integrated Step Selection Models）。针对每个实际利用的步段，我们从两类行为状态下的种群水平步长与转向角经验分布中抽取100个随机步段，构建对照样本集。我们将每个步段与一系列环境协变量进行空间叠加分析，协变量涵盖坡向、蒿灌覆盖率、草本覆盖率、蒿灌斑块连通性、坡度、距道路距离、距井场距离以及距湿生生境距离。所有生境变量均经过标准化与中心化处理。为遵守资助方关于不公开追踪个体位置的协议要求，我们移除了所有地理坐标信息。本研究使用R编程语言（R核心团队，2020）完成全部数据处理工作，调用的R包包括"amt"、"tidyverse"、"sf"、"raster"及"lubridate"。更多研究细节可参阅相关学术论文（Picardi等，2021，《应用生态学杂志》）。