Northern elephant seal UAS mass estimates

Unmanned aerial system (UAS) photogrammetry offers a method that is safer for both animals and researchers and is logistically simpler than traditional weighing methods (Fiori et al. 2017). Additionally, UAS photogrammetry facilitates larger sample sizes because it allows measurement at larger spatial scales, thereby increasing statistical power (Sweeney et al. 2015). However, UAS photogrammetry requires calibration and validation prior to use in order to assess the error relative to known mass measurements. Species-specific calibration of appropriate metrics (e.g., footprint area (Christiansen et al. 2016)) is necessary to account for body shape differences (e.g., "peanut head" syndrome) (Miller et al. 2012, Joblon et al. 2014). Further, the accuracies of photographs taken in overhead and oblique positions are rarely compared due to the challenge of weighing and repeatedly photographing uniquely identified individuals (Fearnbach et al. 2018). Northern elephant seals (Mirounga angustirostris) offer a unique opportunity to calibrate UAS photogrammetry to measure body mass. Elephant seals undergo dramatic changes in mass throughout the year and consistently haul-out to breed and molt at Año Nuevo Natural Reserve, California, USA (37.11° N, 122.34° W) (Le Boeuf and Laws 1994) (Figure 1). Our objectives were: 1) to evaluate the accuracy of UAS photogrammetry for estimating mass in adult female northern elephant seals and 2) to examine the effect of body position on UAS mass estimates. Our error estimates were comparable to other ground-based studies that used multiple photogrammetric measurements to estimate mass (Table 1) and slightly higher than the only other overhead drone study (Krause et al. 2017), possibly due to lower camera resolution. This error is also slightly higher than previous mass estimates derived from morphometric measurements obtained during manual procedures (2.8% for Weddell seals Leptonychotes weddellii (Shero et al. 2014) and 4% for elephant seals (Crocker pers. comm)) but results in significantly less disturbance. Errors in estimated mass were marginally better when we used dorsal footprint area (R2= 0.895) than lateral footprint area (R2= 0.822), suggesting that our UAS mass estimates are robust to changes in body position. In summary, photogrammetric measurements from a single, vertical image obtained using UAS provide a promising approach for estimating the body mass of pinnipeds and similar approaches can be used to find species-specific calibration equations. Mass measurements can inform ecosystem-based resource management (Boyd et al. 2006) by providing information about the inter-annual productivity of the ocean environment and in turn individual, population, and ecosystem-level health in marine mammals (Krause et al. 2017).无人航空系统（Unmanned aerial system, UAS）摄影测量技术为动物与科研人员提供了更安全的研究手段，且在作业流程上比传统称重方法更为简便（Fiori等人，2017）。此外，UAS摄影测量技术支持更大的样本量，因其可在更大空间尺度下开展测量，进而提升统计效力（Sweeney等人，2015）。但UAS摄影测量技术在投入使用前需进行校准与验证，以评估其与已知体重测量值之间的误差水平。适宜指标的物种特异性校准（例如足印面积（footprint area），Christiansen等人，2016）对于校正体型差异（例如花生头综合征，Miller等人，2012；Joblon等人，2014）而言是必要的。此外，由于对个体进行称重并重复拍摄的难度较高，目前鲜有研究对比垂直拍摄与斜拍照片的测量精度（Fearnbach等人，2018）。北象海豹（Mirounga angustirostris）为校准UAS摄影测量技术以测算体重提供了绝佳的研究对象。象海豹全年体重变化幅度极大，且会定期集群上岸于美国加利福尼亚州阿诺纽沃自然保护区（37.11° N, 122.34° W）进行繁殖与换毛（Le Boeuf与Laws，1994）（图1）。本研究的目标为：1）评估UAS摄影测量技术在估算成年雌性北象海豹体重时的精度；2）探究拍摄体位对UAS体重估算结果的影响。本研究的误差估算结果与其他采用多组摄影测量手段估算体重的地面研究相当（表1），且略高于仅有的另一项垂直拍摄无人机研究（Krause等人，2017），这一差异可能源于相机分辨率较低。该误差也略高于此前通过人工操作获取的形态学测量数据推导的体重估算误差（对威德尔海豹Leptonychotes weddellii为2.8%，Shero等人，2014；对象海豹为4%，Crocker 个人通信），但对动物的干扰程度显著降低。当采用背部足印面积（决定系数R²=0.895）时，体重估算误差略优于侧部足印面积（决定系数R²=0.822），表明本研究的UAS体重估算结果对体型体位变化具有较好的鲁棒性。综上，通过UAS获取的单张垂直拍摄照片的摄影测量数据，为鳍足类动物的体重估算提供了极具应用前景的方法，且可通过类似方法推导物种特异性校准方程。体重测量数据可用于支撑基于生态系统的资源管理（Boyd等人，2006）：其能提供海洋环境的年际生产力信息，进而反映海洋哺乳动物的个体、种群及生态系统层面的健康状况（Krause等人，2017）。