Estimation of Soil Salinity Profile in Tabriz Irrigation and Drainage Network Using SaltMod and ANN Models

Soil salinity within plant root zone is one of the most important problems that cause reduction in yield in agricultural lands. In this research, salinity in soil profile was simulated in Tabriz irrigation and drainage network using SaltMod and Artificial Neural Networks (ANNs) models. Based on initial spatial distribution of salinity in soil profile, studying area was divided to 4 different soil and water groups and for two seasons in one year salinity was predicted. The SaltMod model was calibrated and then was applied to generate 2400 data sets for training ANN models. Some of the input data of SaltMod were used in ANN models including irrigation water depth, evapotranspiration, water table depth, rainfall, and initial soil salinity. Efficiency of genetic algorithm in training phase of ANNs was analyzed. The mean of correlation coefficient (R2) and root mean square error (RMSE) of estimated salinity in all groups was 0.8 and 0.032 respectively. In conclusion ANNs could perform well in simulation of soil salinity and it could be replaced SaltMod with enough accuracy. The results showed that overall performance of ANN models improve by applying genetic algorithm.