ESTIMATION OF RAINFALL-RUNOFF RELATIONSHIP USING SOFT COMPUTING TECHNIQUES

The estimation of rainfall-runoff data is crucial for managing basins, preventing or mitigating flood disasters, making better use of water resources, and designing hydraulic structures.In this work, 1802 days of precipitation, runoff, and temperature data from the Quinapoxet River, which supplies the Wachusett Reservoir in the state of Massachusetts, USA, were used to develop models.The rainfall-runoff relationship was predicted using this data as input for Multiple Linear Regression (MLR), Artificial Neural Network (ANN), Support Vector Machines (SVM), and M5 Decision Tree (M5T) techniques.The results of each model were compared with the site measurement data separately.The models are compared with each other according to the three criteria, namely, the root-mean-squared error (RMSE), mean absolute error (MAE), and the correlation coefficient (R).It has been found that Artificial Neural Network (ANN) models perform better in runoff prediction compared to the other models.