A C-vine copula framework to predict daily water temperature in the Yangtze River

Abstract

The thermal regime of rivers plays a crucial role in chemical, biological and ecological processes. Effectively predicting water temperature is a key issue related to environment management. This study develops a probabilistic model based on C-vine copulas for water temperature prediction. The proposed framework is applied to forecast daily water temperature in the Yangtze River, considering long-term effects of preceding air temperature and daily discharge. The prediction performance of this framework is compared with the logistic regression model (LogR) and generalized regression neural network (GRNN). The results of this study indicate that the proposed C-vine copula framework and GRNN model provide better forecasting of stream temperature with weak human-related disturbances than LogR model. The outperformance of C-vine framework is reflected by its ability to accurately capture variations in the water temperature greatly affected by the Three Gorges Reservoir (TGR). This steady and reliable framework is further applied for the conservation of Chinese Sturgeon to estimate the range of suggested discharge, given the daily air temperatures, to adjust the water temperature within 18–20 °C at Yichang station during the spawning season. This application is verified to be more effective, providing indications for reservoir management to lower water temperatures by regulating river flow to ensure the occurrence of spawning activities. The results of this study may provide a scientific reference for the ecological operation of reservoirs in regulated rivers.