A novel red-edge based water anomaly detection index and diagnostic framework.

The frequent occurrence of water anomalies has posed a serious threat to human habitats. Due to the complexity of aquatic environments and the diversity of anomaly types, satellite remote sensing methods still face challenges in accurately detecting and diagnosing water anomalies. In this study, two intermediate parameters, the anomaly water index (AWI) and the advanced water turbidity index (AWTI), were developed using the red edge band of Sentinel-2 imagery. Based on these parameters, we constructed a two-step decision-tree-based diagnostic framework (ADF) to determine types of water anomalies. The proposed indices and framework were comprehensively compared with existing spectral indices and classical supervised learning algorithms in eight globally distributed study areas. The results show that the AWI is effective for identifying multiple water anomalies across diverse aquatic environments, including lakes and oceans, and outperforms existing indices in four mixed cases. Compared to existing indices, the AWTI excels in distinguishing turbid water from algal water. The ADF achieved comparable performance to supervised learning algorithms, with satisfactory time-dynamic monitoring results across four case-study areas and F1 scores exceeding 0.76. In conclusion, this study provides a valuable theoretical basis in the field of water anomaly detection and classification.