Projection of monthly surface flows by an optimized SWAT–MLP: a case study

Abstract

Recent investigations have noted that using a hybrid arrangement of Soil and Water Assessment Tool (SWAT) and multi-layer perceptron (MLP) has high efficiency in runoff prediction. In this research, in addition to using the SWAT and MLP models, an optimized algorithm called Mutated SunFlower Optimization (MSFO) algorithm has been proposed to predict better runoff, which improves the results of prediction runoff by decreasing the error percentage in the MLP model. For this purpose, first, runoff modeling is used to assess the efficiency of the SWAT system. The model's verification and calibration have been performed using data from the previous 30 years of statistics. Then, the flow stream simulated by the SWAT method is evaluated with the observational data and applied as the inputs to the MLP model, and finally, runoff is predicted through the MLP model, and MSFO is used in the MLP model to obtain better results for runoff prediction. The results show that the values of statistical indices R2, RMSE, NSE, and RE give satisfying agreement for runoff forecast in the SWAT–MLP/MSFO model with values of 0.83, 1.68, 0.51, and −0.1. HIGHLIGHTS Two versions of the SWAT model are utilized for forecasting runoff.; The SWAT–MLP model is based on Multi-Layer Perceptron networks.; The SWAT–MLP is optimized based on an improved metaheuristic.; Modified SunFlower Optimization algorithm is used for optimizing SWAT–MLP.; The SWAT–MLP\MSFO model simulates the runoff more accurately.;

Keywords

• artificial neural network
• calibration
• mutated sunflower optimal algorithm
• prediction
• swat
• verification