A Grid-Based Lstm Framework for Runoff Projection and Uncertainty in the Source Region of the Yellow River

Abstract

Long-term runoff projection and uncertainty estimates can provide both the changing trends and confidence intervals of water resources, provide basic information for decision-makers and reduce risks for water resource management. In this paper, a grid-based runoff projection and uncertainty framework was proposed through input selection, long short-term memory (LSTM) modelling and uncertainty analysis. We simultaneously considered dynamic variables such as precipitation and temperature and static variables representing the characteristics of the underlying surface, such as the mean elevation, in the candidate input datasets. Different input combinations were compared. We employed LSTM to develop the relationship between monthly runoff and the selected variables and compared it with the MLR, RBFNN and RNN models. The uncertainty sources originating from the parameters of the LSTM models were considered, and the Monte Carlo approach was used to provide uncertainty estimates. The framework was applied to the source region of the Yellow River (YRSR) at the 0.25° grid scale to better show the temporal and spatial features. The results showed that extra information about static variables can improve the accuracy of runoff projections. Annual runoff tended to increase, with projection ranges of 148.44–296.16 mm under the 95% confidence level, under various climate scenarios.

Keywords: runoff projection, uncertainty, source region of yellow river, LSTM, CMIP6