Rainfall-runoff Modelling of Coastal Watersheds near Hormuz Strait Using Data Mining

Estimating runoff created by rainfall is a very important step in water resources planning, especially in ungauged River Basins. Therefore, research on models simulating the river flow with minimum error in the river basins is necessary. In this study, rainfall-runoff simulation of Minab watershed was done using data mining methods and their performance was compared to present the proper one. For this purpose, eight data mining algorithms including Model Tree (MT), Random Forest (RF), Support Vector Machines (SVM), Bayesian Ridge Regression (BRR), Gaussian Process (GP), Extreme Gradient Boosting (XGB), Artificial Neural Network (ANN), and Multivariate Adaptive Regression Splines (MARS) were used. Coefficient of determination (R2), root mean square error (RMSE), mean absolute error (MAE) and Taylor diagram were used to evaluate the model performance. The results indicated that the MARS model had the best performance among the all models to simulate the monthly discharge of the Minab watershed. Also, the SVM model with (RSME =7.73) has a good performance. The other models also performed relatively close to each other (The XGB model with 9.98 had the highest and the MARS model with 7.7 had the lowest RMSE). Then, by entering the values of sea level temperature (PGSST) in the simulation process, the effect of this parameter on the simulation results was investigated. The results showed that PGSST values did not improve the runoff simulation results in the study area.