Application of Radar Technologies for Classifying Bird Flight Behavior at Dangermond Preserve, January 5-10, 2026

Seabirds are abundant in the U.S. west coast wind energy areas and some dynamic soaring species and several other groups are potentially vulnerable to collision with offshore wind (OSW) turbines. Regulatory permits and authorizations will be needed for OSW energy projects to be planned, constructed, and operated in ways to minimize environmental impacts. For protected species, proposed OSW projects will likely be required to generate collision risk models (CRMs) to estimate species-specific impacts and potential “take” anticipated over the permit term, and monitoring will be necessary to verify actual take once the project is implemented. CRMs require extensive species-specific metrics and are most sensitive to avoidance rate. We present a seabird flight classification framework that leverages multi-feature extraction from radar technologies coupled with a validation methodology using drones programmed to mimic natural flight kinematics. This effort leverages the SEABIRD (System for Environmental Assessment of Bird/Bat Interactions with Real-Time Detection) project that integrates multiple sensor technologies ─ marine biological radar, stereo thermal imaging, and blade-mounted impact sensors ─ into a common data collection and interpretation system that is capable of detecting, tracking, and identifying birds and bats over multiple scales (wind farm to individual turbine to blade), as well as in three-dimensional (3D) space. Our approach used drones to generate a ground-truthed dataset encompassing natural maneuvers including dynamic soaring (large-scale sinusoidal flight), transiting (lower speeds during commuting and higher speeds for feeding), foraging (hovering and turning), diving (higher-speed vertical descent for feeding and lower speeds for resting), aerial chasing, and fine-scale microscale avoidance (rapid, evasive shifts). By precisely controlling drone flight path and speed, we established a clean dataset that accurately reflects the kinematic space occupied by west coast seabirds across these behaviors. Multi-feature extraction techniques, e.g., sinuosity and the relationship between horizontal and vertical flight speed will be applied to data to increase the ability to classify flight behavior using radar technologies.

海鸟种群在美国西海岸风电区域分布广泛，部分动态翱翔类鸟类及其他类群或面临与海上风电（offshore wind, OSW）涡轮机组碰撞的风险。海上风电项目若要以最小化环境影响的方式开展规划、建设与运营，需取得相关监管许可与授权文件。针对受保护物种，拟建海上风电项目通常需构建碰撞风险模型（collision risk models, CRMs），以估算物种特异性影响及许可期限内预期的潜在“受伤害量”；项目投运后，还需开展监测以验证实际受伤害情况。碰撞风险模型需要大量物种特异性参数，且对规避率最为敏感。本研究提出一种海鸟飞行行为分类框架，该框架依托雷达技术的多特征提取方法，并结合以编程方式模拟自然飞行运动学的无人机验证方案。本研究依托SEABIRD（鸟类/蝙蝠与环境交互实时检测系统，System for Environmental Assessment of Bird/Bat Interactions with Real-Time Detection）项目，该项目整合了多种传感器技术——海洋生物雷达、立体热成像技术以及桨叶搭载式碰撞传感器——构建了一套通用的数据采集与解译系统，可在多尺度（风电场、单台涡轮机组至桨叶级别）及三维（3D）空间中实现鸟类与蝙蝠的检测、追踪与识别。本研究通过无人机采集了一套经地面真值验证的数据集，涵盖海鸟的各类自然飞行姿态，包括动态翱翔（大尺度正弦飞行）、迁飞（通勤阶段航速较低，觅食阶段航速较高）、觅食（悬停与转向）、俯冲（觅食时为高速垂直下降，休憩时航速较低）、空中追逐，以及精细微尺度规避行为（快速闪避机动）。通过精准控制无人机的飞行路径与航速，本研究构建了一套高质量洁净数据集，可准确反映西海岸海鸟在上述各类飞行行为中占据的运动学空间。本研究将应用多特征提取技术（例如弯曲度、水平与垂直航速的相关性）对数据集进行处理，以提升基于雷达技术的飞行行为分类能力。