Population census of a large Common tern colony with a small unmanned aircraft

Abstract: Small unmanned aircraft systems (UAS) may be useful for conducting high-precision, low-disturbance waterbird surveys, but limited data exist on their effectiveness. We evaluated the capacity of a small UAS to census a large (>6,000 nests) coastal Common tern (Sterna hirundo) colony of which ground surveys are particularly disruptive and time-consuming. We compared aerial photographic tern counts to ground nest counts in 45 plots (5-m radius) throughout the colony at three intervals over a nine-day period in order to identify sources of variation and establish a coefficient to estimate nest numbers from UAS surveys. We also compared a full colony ground count to full counts from two UAS surveys conducted the following day. Finally, we compared colony disturbance levels over the course of UAS flights to matched control periods. Linear regressions between aerial and ground counts in plots had very strong correlations in all three comparison periods (R2 = 0.972-0.989, P < 0.001) and regression coefficients ranged from 0.928-0.977 terns/nest. Full colony aerial counts were 93.6% and 94.0%, respectively, of the ground count. Varying visibility of terns with ground cover, weather conditions and image quality, and changing nest attendance rates throughout incubation were likely sources of variation in aerial detection rates. Optimally timed UAS surveys of Common tern colonies following our method should yield population estimates in the 93-96% range of ground counts. Although the terns were initially disturbed by the UAS flying overhead, they rapidly habituated to it. Overall, we found no evidence of sustained disturbance to the colony by the UAS. We encourage colonial waterbird researchers and managers to consider taking advantage of this burgeoning technology. About the data: Raw aerial photos were captured in JPEG format by a Canon Powershot S90 10-megapixel camera mounted on an Aerial Insight AI-Multi electric fixed-wing UAS. Overlapping photos of each of the two islands comprising the tern colony were then mosaicked using the PTGui panoramic stitching software. Photomosaics were then imported into ArcGIS and georeferenced with ground control points collected with a Trimble Pathfinder GPS at a series of yellow or orange plastic cones (visible in the imagery) marking the centre of plots in which ground nest counts were compared to aerial tern counts. Overall disturbance levels on each of the two islands were scored on a scale of 0–2 from a distance by an observer at 30-second intervals throughout UAS flights as well as "matched" control periods starting 10 minutes following landing. For more information about this research or the data, please contact: [e-mail]

摘要：小型无人航空器系统（UAS）或可用于开展高精度、低干扰的水鸟调查，但关于其应用效果的现有数据较为有限。本研究评估了小型UAS对大型（>6000个巢）沿海普通燕鸥（Sterna hirundo）种群的普查能力——该种群的地面调查极易造成干扰且耗时极长。我们在九天内分三个时段对整个种群分布区内的45个半径5米的样地开展航空摄影燕鸥计数，并与地面巢计数结果进行对比，以此明确变异来源并建立系数，以便通过UAS调查估算巢群数量。我们还将次日开展的两次UAS全种群计数结果与地面全种群计数结果进行了对比。最后，我们对比了UAS飞行过程中与匹配对照时段的种群干扰水平。
样地内航空计数与地面计数的线性回归分析显示，三个对比时段的相关性均极强（决定系数R²=0.972~0.989，P<0.001），回归系数介于0.928~0.977只燕鸥/巢之间。全种群的航空计数结果分别为地面计数的93.6%和94.0%。燕鸥可见度随地表覆盖、天气条件与图像质量变化，且孵化期内亲鸟巢栖率随时间改变，这些均可能是航空探测率出现变异的原因。按照本研究的方法选择最优时段开展普通燕鸥种群的UAS调查，其种群估算结果应可达到地面计数的93%~96%。尽管燕鸥最初会因UAS在头顶飞行而受到惊扰，但它们会快速对该设备产生习惯化。总体而言，我们未发现UAS对该种群造成持续干扰的证据。我们呼吁沿海水鸟研究与管理人员考虑利用这项新兴技术。

数据集说明：本研究采用搭载于Aerial Insight AI-Multi电动固定翼UAS的佳能PowerShot S90 1000万像素相机，以JPEG格式采集原始航空照片。随后使用PTGui全景拼接软件，对构成燕鸥种群分布区的两座岛屿的重叠航拍照片进行拼接。将拼接后的影像导入ArcGIS地理信息系统，并利用Trimble Pathfinder全球定位系统（GPS）采集的地面控制点进行地理配准——这些控制点以黄色或橙色塑料锥标记，对应地面巢计数与航空燕鸥计数对比样地的中心。
在UAS飞行过程中，以及着陆后10分钟开始的匹配对照时段内，一名观测者会在距离观测点处每30秒按0~2的评分标准对两座岛屿的整体干扰水平进行打分。
如需了解本研究或该数据集的更多信息，请联系：[e-mail]