The level of success in meeting water needs for irrigation depends on the accuracy of water availability calculated by water engineers. Engineers estimate water availability in a watershed based on low-flow discharge data. However, the availability of discharge data is sometimes insufficient. Consequently, engineers need rainfall-runoff models to generate the estimated discharge to complete the data. This paper proposes a procedure to obtain the best model for generating low-flow data of a river. This paper utilized the proposed procedure to generate the low flow of the Matua River. The author uses rain, discharge, and climatic data from 2021 to 2023. The goodness of fit statistical test completes the discharge modeling testing process to get the best model. Based on the goodness of fit test results at the calibration and verification stage, the most suitable model for describing rainfall runoff for the low-flow data of the Matua River is the ANN model. The statistical test results for the ANN model were 0.966 of the correlation coefficient (r), 0.996 of the Nash–Sutcliffe Efficiency (NSE), and 0.0022% of the Percent Errors (PE).