Understanding Behavioral Responses of Wildlife to Traffic to Improve Mitigation Planning

Traffic noise and light measurements Sound pressure levels were recorded in A-weighted decibels (dbA) and C-weighted decibels (dbC) using digital sound level meter devices (TENMA 72-947 and PCE-322; 30-130 range, set to slow). To correspond to timing of crepuscular and night-time activities, we sampled sound levels for one evening (11pm – 2am) at 1-second intervals within a) the crossing structure entrance and b) the closest camera station in the background area. To characterize overall sound conditions at the structures and following the camera trapping period, we also collected dbA and dbC sound pressure levels at the crossing structure entrance for 1 week at 59-second intervals. Low-level light intensity as total lumine scence was measured along a 50m transect away from each of the 26 crossing structures (0m, 10m, 30m, 50m). We used a novel approach employing a camera with a very wide-angle lens to capture low light levels in collaboration with the Longcore lab at the University of California Los Angeles (Jechow et al., 2017). Habitat classification in the surrounding landscape We characterized habitat surrounding each structure using 16-class land cover data from the 2016 National Land Cover Dataset (NLCD, US Geological Survey), which has a spatial resolution of 30 m2 For each underpass we classified land cover within a 100 m2 radius and 1 km2 radius buffer in ArcGIS. This will determine whether habitat in the approach zone and background area respectively influences species-specific movement on a small and/or large scale and noise attenuation. Species detection at WCS in relation to background We compared species detections at WCS with detections at quiet ‘background’ camera stations at all 26 sites. We used the same model of camera traps across all sites (Bushnell Aggressor Trophy Camera). We set each of the cameras to capture still images and have a minimum of 3 seconds between trigger events, and one trigger event at a time. To avoid capturing the same individual multiple times, we classed a unique capture event of the same species as one image > 15 minutes apart. Four camera traps were positioned at the WCS 0.5 m to 1.0 m above the ground facing into or at an angle across the opening of structures (Figure 4). In order to measure background species detections and further examine the impact of noise on WCS use, we measured the distance to background noise levels from the nearest study WCS (~800 m) and established 4 bait stations with associated cameras, at >100 m intervals, for each of the sites. We used salt blocks, peanut butter, dried corn, grain, canned cat food, and chicken parts in an attempt to attract a wide range of species. We also included four non-baited cameras in these quieter areas, >200 m apart from the baited cameras. Cameras were positioned adjacent to areas with visible animal tracks. Cameras were set to have a 10-second delay between trigger events due to the high occurrence of false triggers caused by vegetation. Wildlife Activity To assess wildlife activity (hereafter referred to as “behavior”), Browning Dark Ops Pro cameras were set to video mode and deployed at and near to highway crossing structures (n = 2; “Mesa 2” site, SR 74; “PM24” site, I-80) and adjacent to the site of the proposed Liberty Canyon wildlife crossing structure. Based on preliminary data collection for deer and coyote, 20 types of behavior were extracted from videos as point events or state events (table 1). The activities were grouped into two categories of behavior (table 2). Species identification and behavior time budgets was extracted from all videos using the Behavioural Observation Research Interactive Software (BORIS; Friard and Gamba, 2016; figure 6). The number of humans and domestic dogs present were recorded for each video. In addition, for videos deployed at the highway crossing structures, we classified traffic within a video recording into one of three categories: 1) continuous traffic, 2) occasional, distinguishable traffic, representing between 1-5 clearly audible vehicles passing at random intervals, and 3) zero traffic.

交通噪声与光照测量
采用数字声级计（digital sound level meter）TENMA 72-947与PCE-322（量程30~130，设置为慢响应模式），记录A计权分贝（A-weighted decibels, dbA）与C计权分贝（C-weighted decibels, dbC）形式的声压级。为匹配晨昏与夜间活动时段，我们于一个夜间（23:00至次日02:00）以1秒间隔开展声级采样，采样点位分为两类：a) 穿越结构入口处；b) 背景区域内最近的相机监测站。为表征穿越结构整体声环境并匹配红外相机诱捕周期，我们还以59秒间隔在穿越结构入口处采集了为期1周的dbA与dbC声压级数据。

以总发光度表征的低强度光照，沿距离26处穿越结构各50米的样线（transect）进行测量，采样点分别设置在0m、10m、30m、50m处。本研究与加州大学洛杉矶分校Longcore实验室合作，采用搭载超广角镜头的相机开展低光照水平监测（Jechow等，2017），该方法为创新性研究手段。

周边生境分类
我们采用美国地质调查局2016年国家土地覆盖数据集（National Land Cover Dataset, NLCD）的16类土地覆盖数据，对每个穿越结构周边的生境进行表征，该数据集空间分辨率为30㎡。在ArcGIS地理信息系统中，我们分别对每个下穿式通道周边100米半径与1平方千米半径缓冲区范围内的土地覆盖类型进行分类。此举旨在分别明确近源区域与背景区域的生境，如何从小规模和/或大规模尺度影响物种特异性移动与噪声衰减效果。

野生动物穿越结构（Wildlife Crossing Structures, WCS）处的物种检测与背景对照
我们将WCS处的物种检测结果，与所有26处监测点的安静“背景”相机站的检测结果进行对比。所有监测点均采用同款型号的红外触发相机（camera trap）Bushnell Aggressor Trophy Camera。我们将每台相机设置为拍摄静态图像，触发事件间隔至少3秒，且单次触发仅生成一张图像。为避免重复记录同一动物个体，我们将同一物种的单次独立捕获事件定义为间隔超过15分钟的单张图像。

4台红外触发相机被安装在WCS处，离地0.5米至1.0米高度，朝向穿越结构入口或呈一定角度横跨入口（图4）。为测量背景物种检测结果并进一步探究噪声对WCS使用的影响，我们测量了距离最近的研究用WCS约800米处的背景噪声水平，并在每个监测点以大于100米的间隔设置4处诱站，每处诱站搭配一台相机。诱剂选用盐块、花生酱、干玉米、谷物、罐装猫粮与鸡肉制品，以吸引多种类的野生动物。此外，我们在这些安静区域还设置了4台未放置诱剂的相机，各相机间距大于200米。相机被安装在有明显动物足迹的区域附近。考虑到植被易引发误触发，我们将相机的触发事件间隔设置为10秒。

野生动物活动评估
为评估野生动物活动（以下简称“行为”），我们将Browning Dark Ops Pro相机设置为录像模式，部署在2处高速公路穿越结构处及其周边：分别为SR 74公路的"Mesa 2"站点与I-80公路的"PM24"站点，同时部署在拟建的Liberty Canyon野生动物穿越结构场地周边。基于针对鹿与郊狼的预实验数据，我们从视频中提取了20类行为事件，分为点事件与状态事件两类（表1）。这些活动被划分为两类行为大类（表2）。

我们通过行为观察研究交互软件（Behavioural Observation Research Interactive Software, BORIS; Friard与Gamba, 2016; 图6）从所有视频中提取物种识别信息与行为时间预算。每段视频均记录了在场的人类与家犬数量。此外，针对部署在高速公路穿越结构处的视频，我们将录制期间的交通状况划分为三类：1) 持续车流；2) 间断可辨车流，即随机间隔有1~5台清晰可闻的车辆通过；3) 无车流。

搜集汇总

背景与挑战

背景概述

该数据集由加州大学戴维斯分校的研究人员于2020年发布，旨在通过分析野生动物对交通噪声和光照的行为响应，优化野生动物通道结构的缓解规划。研究覆盖加州26个地点，收集了声音压力水平、光照强度、栖息地分类和物种检测等数据，并利用相机陷阱和视频分析野生动物活动，以支持交通生态学中的保护策略。