Modeling the rainfall-runoff relationship with artificial intelligence methods

The need for accurate modeling of the rainfall-runoff process has increased rapidly in recent years. However, given the complexity and nonlinearity of hydrological systems, many models are still being developed to describe phenomena such as the rainfallrunoff relationship. It has become even more important to determine the rainfall-runoff relationship due to factors such as global warming and global climate change, which have been effective especially in recent years. Recently, artificial intelligence techniques such as Artificial Neural Network (ANN), Genetic Expression Programming (GEP), Adaptive Neural-Fuzzy Inference System (ANFIS) have become widespread for hydrological purposes such as rainfall-runoff modeling. In this study, three different artificial intelligence-based methods (ANN, GEP, ANFIS) for basin precipitation-flow modeling were compared with Multiple Linear Regression (MLR), which is a more classical statistical method. Precipitation data, one of the input parameters, consists of satellite data with a resolution of 4km x 4km. Precipitation data were obtained in the form of daily average precipitation height for the basin. The daily flow data of the EIE-2334 flow observation station (FOS) located at the exit of the Berta Water Basin were used as flow data. The flow data Q(t-1), Q(t-2), Q(t-3), Q(t-4) and precipitation data P(t), P(t-1), P(t-2), P(t-3) various input combinations are created with values and matched with Q(t) as output. Some of the inputs are reserved for model training and the remainder for testing. The performance of the models was evaluated by the coefficient of determination (R2) and the root mean square error (RMSE). In the study, it was seen that the results of ANFIS, GEP, ANN, and EDR were close. ANFIS-K1 was found to be the model that gave the highest R2=0.988 and the lowest RMSE=4.770 value for the Berta Water Basin. It has been concluded that artificial intelligence methods can be easily used to determine the rainfall-runoff relationship.