An ensemble multi-model approach for long-term river flow forecasting in managed basins of the Middle East: Insights from the Karkheh River Basin

Abstract

In this study, we evaluated the accuracy of weather and river discharge forecasts for the Karkheh River Basin on the Iranian plateau. We utilized weather parameters from the North American Multi-Model Ensemble (NMME)—specifically precipitation and maximum and minimum temperature—for long-term weather forecasting and assessed their accuracy in runoff simulations using the Soil and Water Assessment Tool (SWAT). The primary aim of the study was to explore the potential improvements in forecast accuracy through the application of NMME models, both individually and in combination, to hydrological forecasting. To achieve this, we employed two statistical approaches (MLR and KNN), for spatial and temporal downscaling of the NMME models, respectively. The results revealed that the combination of NMME models outperforms individual models in robustly predicting precipitation and temperature. Specifically, precipitation forecasts showed better accuracy during spring (with correlation coefficients ranging from 0.79 to 0.89) and fall (correlation coefficients ranging from 0.43 to 0.79), while their performance was weaker during summer. Temperature forecasts exhibited high accuracy, particularly in warmer periods (with correlation coefficients ranging from 0.75 to 0.99). Given the importance of accurately predicting precipitation during rainy seasons for runoff predictions and precise temperature forecasts during warm seasons, the NMME system demonstrated satisfactory performance and proved to be a valuable input for hydrological models. Furthermore, we used SWAT to predict river discharge with lead times of 1 to 3 months. Notably, the runoff forecast with a 1-month lead time showed the highest performance, as indicated by a correlation coefficient of 0.61.