Multi-scale habitat assessment of pronghorn migration routes

We studied the habitat selection of pronghorn (Antilocapra americana) during seasonal migration; an important period in an animal’s annual cycle associated with broad-scale movements. We further decompose our understanding of migration habitat itself as the product of both broad- and fine-scale behavioral decisions and take a multi-scale approach to assess pronghorn spring and fall migration across the transboundary Northern Sagebrush Steppe region. We used a hierarchical habitat selection framework to assess a suite of natural and anthropogenic features that have been shown to influence selection patterns of pronghorn at both broad (migratory neighborhood) and fine (migratory pathway) scales. We then combined single-scale predictions into a scale-integrated step selection function (ISSF) map to assess its effectiveness in predicting migration route habitat. During spring, pronghorn selected for native grasslands, areas of high forage productivity (NDVI), and avoided human activity (i.e., roads and oil and natural gas wells). During fall, pronghorn selected for native grasslands, larger streams and rivers, and avoided roads. We detected avoidance of paved roads, unpaved roads, and wells at broad spatial scales, but no response to these features at fine scales. In other words, migratory pronghorn responded more strongly to anthropogenic features when selecting a broad neighborhood through which to migrate than when selecting individual steps along their migratory pathway. Our results demonstrate that scales of migratory route selection are hierarchically nested within each other from broader (second-order) to finer scales (third-order). In addition, we found other variables during particular migratory periods (i.e., native grasslands in spring) were selected for across scales indicating their importance for pronghorn. The mapping of ungulate migration habitat is a topic of high conservation relevance.  In some applications, corridors are mapped according to telemetry location data from a sample of animals, with the assumption that the sample adequately represents habitat for the entire population. Our use of multi-scale modelling to predict resource selection during migration shows promise and may offer another relevant alternative for use in future conservation planning and land management decisions where telemetry-based sampling is unavailable or incomplete.

Methods
The following databases were recorded using GPS-radio collars deployed on female pronghorn. Collars took GPS locations either every 2 or 4 hours. Important variables were intersected and added to both use and available locations for each scale of selection. Here are the names of the two databases:

Pronghorn_NSS_3rd_Order_Data_PlosOne_Paper

Pronghorn_NSS_2nd_Order_Data_PlosOne_Paper

本研究聚焦叉角羚（Antilocapra americana）在季节性迁徙期间的生境选择——这是动物年度周期中与大范围移动相关的关键阶段。我们进一步将迁徙生境的内涵拆解为大尺度与细尺度行为决策共同作用的产物，并采用多尺度方法，评估跨界北灌丛草原（Northern Sagebrush Steppe）区域内叉角羚的春季与秋季迁徙活动。

本研究采用分层生境选择框架，评估一系列已被证实会在大尺度（迁徙邻域）与细尺度（迁徙路径）上影响叉角羚生境选择模式的自然与人为干扰特征。随后我们将单尺度预测结果整合为尺度整合步长选择函数（scale-integrated step selection function, ISSF）图谱，以评估其在预测迁徙路径生境方面的有效性。

春季期间，叉角羚偏好原生草原、高牧草生产力区域（归一化差分植被指数，Normalized Difference Vegetation Index, NDVI），并规避人类活动干扰，即道路与油气井。秋季期间，叉角羚偏好原生草原、较大型溪流与河流，并规避道路。

我们发现在大空间尺度上，叉角羚会规避铺装道路、非铺装道路与油气井，但在细尺度上对这些特征无明显响应。换言之，迁徙中的叉角羚在选择迁徙所需的大尺度邻域时，对人为干扰特征的响应强度要高于在迁徙路径上选择单个步长时的响应。

研究结果表明，迁徙路径选择的尺度层级嵌套关系清晰，从更宽泛的二阶尺度延伸至更精细的三阶尺度。此外，我们还发现部分变量在特定迁徙时期（如春季的原生草原）在所有尺度上均受到偏好，这体现了这些变量对叉角羚的重要性。

有蹄类动物迁徙生境的制图是极具保护价值的研究课题。在部分应用场景中，迁徙廊道的制图基于从部分个体采集的遥测定位数据进行，其前提假设是该采样样本能够充分代表整个种群的生境需求。本研究采用多尺度建模方法预测迁徙过程中的资源选择，该方法展现出良好的应用前景，可为未来缺乏或无法获取遥测采样数据的保护规划与土地管理决策提供另一可行的替代方案。

## 研究方法
本研究通过部署于雌性叉角羚身上的GPS-无线电项圈采集相关数据库数据。项圈每2小时或每4小时记录一次GPS定位点。针对每一类选择尺度，我们将重要变量与使用点位及可用点位进行空间叠加关联，从而将变量信息添加至对应数据集中。本研究包含两个数据集，其名称分别为：
Pronghorn_NSS_3rd_Order_Data_PlosOne_Paper
Pronghorn_NSS_2nd_Order_Data_PlosOne_Paper