A context-calibrated framework based on uncrewed aerial vehicle imagery for power-line detectability and mitigation evaluation: a model system from Black-necked Crane wintering grounds on the Tibetan Plateau

Overhead power lines are a major source of avian mortality, yet quantitative and context-calibrated assessments of wire detectability and marker performance under variable field conditions remain scarce. Using 560 uncrewed aerial vehicle (UAV) images from three datasets collected in Black-necked Crane ( Grus nigricollis ) wintering grounds on the Tibetan Plateau, we quantified wire detectability with a Visibility Index (VI; target–background contrast and edge strength) and characterized background heterogeneity with a Visual Complexity Index (VCI; entropy, edge density, and colour variance at global and local scales). We integrate these metrics into a Collision Risk Index (CRI) as a relative, image-based proxy for detectability to support span prioritization and context-matched evaluation of mitigation devices. Within the warning-device datasets, which included sunny and cloudy conditions, sleeves provided more robust detectability gains, particularly at longer distances, whereas balls yielded smaller and less consistent improvements. Detectability varies strongly with weather, distance, and relative observation height, highlighting the need to compare mitigation options under matched contexts. We frame this high-altitude wintering landscape as a model system for collision-risk mitigation and provide practical guidance for adapting and recalibrating the workflow to other species, habitats, and linear infrastructures.