Quantification of Backwater Effect in The Jingjiang Reach Caused by the Confluence of Dongting Lake Using a Machine Learning Model

The backwater effect caused by tributary inflow can significantly elevate the water level profile upstream of a confluence point. However, it still remains unclear that how the backwater effect in a river reach is influenced by the mainstream and confluence discharges. In the current study, various hydrological data measured in the Jingjiang Reach were collected. Using the statistical analysis method, the backwater degree and range were then determined under three representative mainstream discharges in the Jingjiang Reach. The results showed that the backwater degree increased with a higher mainstream discharge, and there was a positive relationship between runoff ratio and backwater degree under a certain representative mainstream discharge. The backwater effect in the Jingjiang Reach decreased after the Three Gorges Project operation. For example, the mean values of backwater degree for low, moderate, and high mainstream discharges were 0.83, 1.61, and 2.41 m during the period 1990-2002, whereas they reduced to 0.30, 0.95, and 2.08 m in 2009-2020. In terms of backwater range, it extended upstream with the mainstream discharge increasing from 7000 to 30000 m 3/s. Moreover, a random forest based machine learning model was proposed to quantify the backwater effect under different mainstream and confluence discharges, which can consider the effects of multiple influencing factors, and the impacts of mainstream discharge, confluence discharge, and channel degradation on the backwater effect in the Jingjiang Reach. Taking Jianli station as an example, a decrease in the mainstream discharge during the flood season led to a 7%-15% increase in monthly mean backwater degree, while an increase in mainstream discharge during the dry season led to a 1%-15% decrease in monthly mean backwater degree. The increase in confluence discharge from the Dongting Lake during the periods June to July and September to November resulted in an increase of 11%-42% in monthly backwater degree. The continuous channel degradation led to a decrease of 6%-19% in monthly mean backwater degree. Influenced by a combination of these three factors, the monthly mean backwater degree varied from a decrease of 53% to an increase of 37%.