Simulating the discharge of Nahavand karstic springs based on neural intelligent models

ABSTRACT In this research, the discharge of five karstic springs, which are located in the Nahavand Plain of west of Iran, were simulated. The monthly discharge data for the years 1994–2020 were obtained. Three intelligent models, ANN, PCA-ANN, and CANFIS, were used to simulate the relevant data. In the present study, hydraulic parameters such as porosity, hydraulic conductivity, and specific yield were measured by frequent field visits and precipitation data of five meteorological stations and elevations were used as input data. The three intelligent models used are based on the error measurement indicators. Based on the analysis of results, Levenberg-Marquardt's learning rules were preferred as the best learning rules for all three models. Bearing the highest accuracy, MBE = −1.3 (l/s) and the highest value correlation coefficient (0.654) with nRMSE equal to 9.4% were preferred as the best intelligent models for simulating the discharge of these springs. The maximum average percentage of the simulated discharge increases compared with the value observed in the test phase. This was observed for May in the period of 27 years, which has a value of 24.20%. The lowest fluctuation percentage was observed, with a value of −24.61% for December. It is concluded that the hydraulic parameters were preferred instead of spring discharge lag-time for simulation.