Prediction of rainfall – runoff relationship using artificial intelligence techniques

In order to determine the use, protection and economic life of water resources; It is important to analyze the precipitation-flow data and make estimates from the data obtained as a result of the analysis. For this reason, multiple linear regression (MLR) method, non-linear regression methods, artificial neural networks (ANN) method and generalized regression artificial neural networks (GRNN) method have been preferred for the prediction of precipitation-flow relationship with artificial intelligence technique. Analysis was performed by interaction regression (IR), pure-quadratic regression (PQR) and quadratic regression method (QR), which are among the multiple nonlinear regression methods. For this method, 1396 daily of temperature, precipitation and flow data between 2014 - 2018 of the station in Newaygo region in the Muskegon basin of the United States was used. Prediction study was done with artificial intelligence techniques by using hydrological and seasonal parameters of 1396 days belonging to the region. In the paper, 2 different analyzes (MLR1, MLR2; IR1, IR2; PQR1, PQR2; QR1, QR2; ANN1, ANN2; GRNN1, GRNN2) were selected for each method using flow and precipitation time series. While using these methods; Comparison was made using determination coefficient (R2), mean absolute error (MAE) and root mean of square errors (RMSE). When the measurement and model results are compared; although all the methods used gave good results in determining the relation of precipitation flow; ANN2 and GRNN2 models have the highest determination coefficients and lowest error amounts. The highest error of GRNN1 and MLR1 models showed lower performance among all models with the lowest determination coefficients. It was observed that ANN method gave the best estimation with low error and high determination in precipitation - flow modeling. For this reason, ANN method can be used as an alternative to classical methods in precipitation-flow modeling.