The efficiency of genetic programming model in simulating rainfall-runoff process (Case Study: Khorramabad river basin)

Predicting the river discharge is one of the important subjects in water resources engineering. This subject is of utmost importance in terms of planning, management, and policy of water resources with the aim of economic and environmental development, especially in a country like Iran with limited water resources. Awareness of the relation between rainfall and runoff of basins is an inseparable past of water design studies. Lack of sufficient data on rainfall-runoff due to the absence of appropriate hydrometric stations reveals the importance of using indirect methods and heuristic algorithms for estimating the basins' runoff more than before. In the present research, the genetic programming model has been employed to simulate the rainfall-runoff process of Khorramabad River basin, and in order to introduce the patterns and identify the best pattern dominating the nature of flow, all statistical data were divided into two groups of training and experiment (52 percent training and 48 percent experiment) and the program was implemented for 1000 replications using fitting functions and going through replication and developmental processes so as to find the optimal replication. Moreover, in order to evaluate the relations obtained from the simulator model, Root Mean Square Error (RMSE) and Mean Squared Error (MSE) indexes and Coefficient of Determination (R2) have been used. The investigations demonstrate that the employed equation 3 has the greatest relevance with the observational data. Therefore, it is recommended that the said equation be used for the rainfall-runoff studies of the abovementioned basin. Based on the results, the genetic programming model is an accurate direct method for predicting the discharge of Khorramabad River basin.