Abstract
Water quality early warning is crucial for protecting ecological security and controlling pollution in lakes and reservoirs. However, the traditional warning level may not provide accurate data for a specific area. Therefore, it is necessary to design an adaptive early warning threshold and identification system that conforms to the actual operating environment. This study monitored nine water quality parameters-water temperature (WT), pH, dissolved oxygen (DO), permanganate index (COD(Mn)), chemical oxygen demand (COD), five-day biochemical oxygen demand (BOD(5)), total nitrogen (TN), total phosphorus (TP), and ammonia nitrogen (NH(3)-N)-monthly from 11 sampling sites in the Danjiangkou Reservoir, i.e., the largest artificial lake in Asia, from 2017 to 2022. The reservoir was divided into three sub-areas by cluster analysis: Danku, Hanku, and Water intake. The Water Quality Index (WQI) was used for comprehensive spatiotemporal water quality evaluation, and a minimum WQI (WQI(min)) model was developed using multiple linear regression. Finally, a water quality risk early-warning model was proposed based on frequency analysis, categorizing water quality into six levels. The findings reveal that the water quality in each area maintains at "good" or "excellent" levels during the study period. The average WQI values, from lowest to highest, are Hanku (75.44), Danku (78.78), and Water intake (79.07). This result shows that the water quality of Danjiangkou Reservoir has been maintained at a good level due to the pollution control and management of Chinese government after the operation of the Middle Route of the South-to-North Water Diversion Project of China. The WQI(min) models for each area have different key parameters: WT, DO, TN, TP, and COD are common in all areas, whereas NH(3)-N is included in both Hanku and Danku models. BOD(5) and pH were unique to the Danku and Water intake models, respectively. TN and TP are identified as the key parameters affecting water quality safety in Danjiangkou Reservoir. The risk thresholds for TN and TP in Hanku are significantly higher than those in Danku and Water intake, indicating that the water quality in Hanku is the worst. These thresholds are dynamically revised through the early warning model as new data became available. The proposed risk assessment framework provides a robust tool for water quality risk early warning and offers a scientific and reliable reference for administrative departments to implement effective water environment risk prevention and management strategies.