Long-term river water temperature reconstruction and investigation: A case study of the Dongting Lake Basin, China

Abstract

Water temperature is a fundamental property of river habitats and aquatic ecosystems. In some cases, assessing water temperature changes is restrained by the availability of historical observation. This study proposed a framework to address the water temperature data scarcity in thermal regime assessment by incorporating the long short-term memory (LSTM) network and the time series technologies. The factors affecting water temperature included air temperature, evaporation, and streamflow. Driven by the historical observation of these factors, the LSTM network produced long-term water temperature series. The time series technologies unraveled the temporal dynamics of the reconstructed water temperature. The trend, periodicity, and complexity were investigated using linear regression, trend-free pre-whitening Mann-Kendall test, Sen's slope, wavelet transform, and moving sample entropy. The framework was applied in the Dongting Lake Basin, China, which drains four tributaries: Xiangjiang, Zishui, Yuanjiang, and Lishui Rivers. The LSTM network reconstructed the monthly water temperature series from 1960 to 2020, which was good surrogate data for thermal regime evaluation. The annual water temperature of the four tributaries increased significantly, with an average warming rate of 0.15 ℃ per decade. Heterogeneity of tendency and changing rates existed across sub-basins and months, and the warming water temperature dominated the variation. A significant oscillation of a one-year period across six decades was identified in the Dongting Lake Basin. The water temperature complexity decreased significantly in the Xiangjiang, Zishui, and Yuanjiang Rivers, while it increased significantly in the Lishui River. The case study indicated that the framework was practicable and flexible with explicit structure. It could be applied in other basins to address water temperature data scarcity and support the comprehensive riverine thermal regime assessment.