Revealing distribution patterns of river obstructions in China via deep-learning and satellite imagery

• The deep learning-based method can accurately identify river obstructions. • Over ten thousand river obstructions have been identified in rivers across China. • The number of river obstructions in non-humid regions is underestimated. Rivers are crucial for supporting biodiversity, regulating climate, and supplying freshwater resources. However, the construction of river obstructions has significantly disrupted river connectivity, impacting biodiversity and hydrological processes. Despite the importance of identifying the distribution patterns of river obstructions for effective river restoration, especially given the extensive fragmentation of rivers in China, generating large-scale datasets of these obstructions remains challenging. In this study, we introduced a novel methodological framework that integrates advanced deep learning techniques with remote sensing imagery to automatically identify and classify a wide range of river obstructions across China. This innovative approach led to the development of the China River Obstruction Database (CROD), comprising 10,246 obstructions. The spatial distribution reveals that river obstructions are primarily concentrated in the southeastern region of China, with the low-head dam being the most prevalent type. Topography and precipitation may be key factors influencing the distribution of obstructions in humid regions. Additionally, the number of river obstructions in non-humid regions was underestimated previously. The deep learning-based classification method demonstrates significant advantages in accuracy and efficiency over traditional approaches, offering a powerful tool for large-scale river monitoring. The findings of this study can facilitate planning and management efforts aimed at restoring rivers and mitigating the impacts of human activities on aquatic ecosystems.