Why did the chicken NOT cross the road? Anthropogenic development influences the movement of a grassland bird

Movement and selection are inherently linked behaviors that form the foundation of a species space-use patterns. Anthropogenic development in natural ecosystems can result in a variety of behavioral responses that can involve changes in either movement (speed or direction of travel) or selection (resources used) behaviors which in turn may cause differential population level consequences including loss of landscape level connectivity. Understanding how a species alters these different behaviors in response to human activity is essential for effective conservation. In this study, we investigated the effects of anthropogenic development such as roads, power lines and oil wells on the greater prairie-chicken (Typanuchus cupido) in the post-nesting and nonbreeding season. Our first objective was to assess if greater prairie-chickens alter their movement behaviors or their selection patterns when encountering oil wells, power lines, or roads using integrated step selection analysis (iSSA). Our second objective was to determine if prairie-chickens avoided crossing linear features such as roads or power lines by comparing the number of crossing events in greater prairie-chicken movement tracks to the number of movements that crossed these features in simulated movement tracks. Based on the iSSA analysis, we found that greater prairie-chickens avoided oil wells, power lines, and roads in both seasons, but found little evidence for changes in speed or direction of movement at the population-level. However, at the individual level we observed individuals using a number of strategies near development including avoidance and increased rates of movements. Furthermore, prairie-chickens traveled across roads and power lines at much lower rates than expected. Consistent avoidance of development resulted in indirect habitat loss for greater prairie-chickens. These behaviors also resulted in a potential loss of landscape connectivity for this species. By considering both movement and selection we were able to develop an ecological understanding of how increasing human activity may influence the space-use of this species of conservation concern. This research provides insight into the decision-making process by animals when they encounter anthropogenic development by considering multiple behavioral responses. --

Methods
We captured prairie-chickens in March and April of each year from 2014 – 2019 using walk-in funnel traps at lek sites (communal courtship arenas). We aged and sexed all captured individuals based on plumage and the presence of secondary sex characteristics (enlarged air sacs and eye combs in males; Henderson et al. 1967), and we marked males and females with a uniquely numbered aluminum leg band to aid in identification. We attached rump-mounted, 22-gram ARGOS GPS transmitters (PTT-100, Microwave Telemetry, Columbia, Maryland, USA) to all captured female prairie-chickens. The GPS transmitters were programmed to record one GPS location every hour from 700 to 1900 from 1 March to 31 August and every two hours during the remainder of the year. All females were monitored remotely via data downloads from the ARGOS server as data became available. The GPS transmitters were estimated to have an average error of less than 20 meters (personal communication Microwave Telemetry).

We focused on two seasons for our analysis, the post-nesting and the nonbreeding season. We defined the post-nesting season as the period after a female concluded nesting activity for the year (typically late May to early June) until 14 September when the last broods are likely to have broken up prior to the fall/winter season. The start of the post-nesting period was determined separately for each individual based on when individual females were observed leaving the nest site (Londe et al. 2019). This period corresponds to the time when females are either raising newly hatched chicks or are recovering from previous reproductive efforts. Due to a lack of data about brood presence and survival in the early parts of our study, we combined data for all females during this period regardless of reproductive status. Previous studies suggest habitat selection patterns are likely similar between brooding and nonbrooding female prairie-chickens (Londe et al. 2019, Londe et al. 2021). The nonbreeding season was defined as the period from 15 September to 15 March of the subsequent year. This period corresponds to the fall and winter period when females are not engaged in any reproductive activity. We did not include telemetry locations from the period when females are attending leks (mid-March to late April) and nesting (early April to nest hatch or failure) as movements during these periods are limited and tend to be concentrated on leks or nest locations. For prairie-chickens that were monitored in multiple years in the study, we treated each year (hereafter, prairie-chicken year) as a separate individual to account for changing habitat conditions between years of the study (Hovick et al. 2015).

Integrated step selection analysis 

Integrated step selection analysis (iSSA) allows for the simultaneous modelling of selection and movement processes by comparing environmental and habitat variables observed during a step (straight line connecting two GPS locations) and variables that describe an animal’s movement pattern, such as step length and turning angles (change in direction of travel between two steps), to those same variables recorded on random steps (Fortin et al. 2005, Thurfjell et al. 2014). For our analysis, we used observed prairie-chicken steps where the beginning and ending locations were at GPS telemetry locations that were recorded 2 hours apart, and were part of a series of steps that included ≥ 3 telemetry locations (this is the minimum number of consecutive relocations required to calculate valid turning angles; Avgar et al. 2016). For each observed step, we generated 10 random steps that shared a starting location with the observed step, but where step length and turning angles of the random step was randomly selected from a gamma and von Mises distribution, respectively (Avgar et al. 2016, Signer et al. 2019). We assigned habitat attributes to each step by extracting environmental variables from GIS rasters at the beginning and ending location for each step. Each set of observed and ten random steps was defined as a choice set and were compared using conditional logistic regression models in program R using the survival package (Therneau and Lumley 2015). Prior to analysis, we removed six individuals from the post-nesting season analysis and nine individuals from the nonbreeding season analysis that had established home ranges in areas where there was no oil and gas development, and the only roads were private ranch roads that received little traffic.

运动与选择本就是紧密关联的行为，构成了物种空间利用模式的基础。自然生态系统中的人为开发会引发多种行为响应，包括移动（行进速度或方向）或资源选择行为的改变，进而可能造成差异化的种群水平后果，例如景观连通性丧失。理解物种如何响应人类活动调整这两类行为，是开展有效保护的核心前提。
本研究针对巢后与非繁殖季的大草原榛鸡（Typanuchus cupido），探究了道路、输电线与油井等人为开发活动对其的影响。我们的首要目标是，通过集成步选择分析（integrated step selection analysis, iSSA），评估大草原榛鸡在遭遇油井、输电线或道路时，是否会调整其移动行为或资源选择模式。其次，我们通过对比大草原榛鸡移动轨迹中的穿越事件数量，与模拟移动轨迹中穿越这些线性设施的运动次数，判断榛鸡是否会规避穿越道路或输电线等线性地物。
基于iSSA分析结果，我们发现两个季节里大草原榛鸡均会规避油井、输电线与道路，但未发现种群层面的移动速度或方向发生显著变化的证据。不过在个体层面，我们观察到开发区域附近的个体存在多种应对策略，包括规避行为以及更高的移动速率。此外，大草原榛鸡穿越道路与输电线的频率远低于预期值。对开发活动的持续规避间接造成了该物种的生境丧失，同时也可能削弱了其景观连通性。通过同时考量移动与选择两类行为，我们得以深入理解人类活动加剧如何影响这一受保护物种的空间利用模式。本研究通过整合多种行为响应，揭示了动物遭遇人为开发时的决策过程。

方法
2014至2019年每年的3月至4月，我们在求偶场（公共求偶竞技场）使用入口式漏斗陷阱捕获大草原榛鸡。依据羽毛特征与第二性征（雄性的膨大气囊与肉冠；Henderson等，1967）对所有捕获个体进行年龄与性别鉴定，并为雌雄个体佩戴唯一编号的铝制脚环以辅助识别。我们为所有捕获的雌性榛鸡安装了臀背式、22克重的ARGOS GPS追踪器（PTT-100，微波遥测公司，美国马里兰州哥伦比亚）。该GPS追踪器被设置为：3月1日至8月31日期间，每日7:00至19:00每小时记录1个GPS点位，其余时段每两小时记录1次。所有个体通过ARGOS服务器远程下载数据进行监测。据制造商反馈，该GPS追踪器的平均定位误差小于20米（个人通信，微波遥测公司）。

我们的分析聚焦两个季节：巢后季与非繁殖季。巢后季定义为雌性完成年度筑巢活动后（通常为5月末至6月初）至9月14日，此时最后一窝雏鸟大概率已离巢进入秋冬季节。巢后季的起始时间依据每只雌性离开巢址的时间单独确定（Londe等，2019）。该时段对应雌性抚育雏鸟或从过往繁殖活动中恢复的阶段。由于研究早期缺乏关于雏鸟存在与存活的相关数据，我们将该时段内所有雌性的监测数据合并分析，无论其繁殖状态如何。已有研究表明，育雏与非育雏的雌性大草原榛鸡的生境选择模式较为相似（Londe等，2019；Londe等，2021）。非繁殖季定义为9月15日至次年3月15日，对应雌性未参与任何繁殖活动的秋冬时段。我们未纳入雌性参与求偶活动（3月中旬至4月下旬）与筑巢（4月初至雏鸟出巢或繁殖失败）时段的遥测点位，因为该时期的移动受限，且多集中于求偶场或巢址附近。对于研究中多年被监测的个体，我们将每一年（下称“榛鸡年度”）视为独立个体，以消除研究年间生境条件变化的影响（Hovick等，2015）。

集成步选择分析
集成步选择分析（iSSA）通过对比步（连接两个GPS点位的直线段）期间观测到的环境与生境变量，以及描述动物移动模式的变量（如步长、步间转向角）与随机步的对应变量，实现选择与移动过程的同步建模（Fortin等，2005；Thurfjell等，2014）。本研究中，我们使用的观测步的起始与结束点位为间隔2小时记录的GPS遥测点位，且该步属于包含至少3个连续遥测点位的序列（计算有效转向角所需的最少连续定位数；Avgar等，2016）。对于每个观测步，我们生成10个共享该观测步起始点位的随机步，其中随机步的步长与转向角分别从伽马分布与冯·米塞斯分布中随机抽取（Avgar等，2016；Signer等，2019）。我们通过在每个步的起始与结束点位提取GIS栅格的环境变量，为每个步赋予生境属性。每组观测步与10个随机步构成一个选择集，使用R语言survival包（Therneau与Lumley，2015）中的条件logistic回归模型进行比较分析。分析前，我们移除了巢后季分析中的6个个体与非繁殖季分析中的9个个体，这些个体的家域位于无油气开发、仅存在低流量私人牧场道路的区域。