Accurate runoff forecasting is essential for flood risk management in a given area, although it is unpredictable and inaccurate; it is still a difficult task. Therefore, the primary objective of this research is to employ a various empirical models including Multi-layer perceptron neural network (MLPNN), Radial basis function (RBF), Autoregressive Integrated Moving Average (ARIMA) and response surface methodology (RSM) to predict the runoff at Kafr Kela Al Bab, Egypt. In this study, the global meteorological data are gathered for examination between 1985 and 2023 in order to achieve this goal. The findings demonstrated that, in comparison to other models, the RSME and ARIMA model performs better, with R2 = 0.9906 and 0.9902 for the validation phase. Additionally, the RSME and MAE show that, in terms of statistical parameters, the observed values and the obtained values of the RSM and ARIMA models closely agree with them.