当前世界环境

There is, unfortunately, a lack of exhaustive qualitative and quantitative information about Iran groundwater resources. That is why various models are used in estimation of qualitative and quantitative groundwater parameters. The present paper presents a comparison of the hybrid of Adaptive Neuro Fuzzy Inference System (ANFIS) with Genetic Algorithm (GA) model and L-moments regarding their power and efficiency in regional and at-site anticipation of salinity of groundwater at Kerman plain. In doing so, electrical conductivity is considered the dependent variable, while, through regression analysis, total cat ions, magnesium ion, sodium percentage, and level of groundwater are assumed to be independent parameters. The correlation coefficient between input values and anticipated ones is the criterion the study takes into account in comparisons as well as in the election of the optimum model. Wells of study area were classified into three homogenous regions. Hass-King Heterogeneity and Incongruity Criterion were calculated for each site. The best result for regional analysis is achieved in well No.17 with correlation coefficient (C.C) 0.9958 whereas the best result for at-site analysis is calculated in well No.2 with C.C 0.9787. Results showed that, in regions with lower heterogeneity criterion, ANFIS-GA regional anticipations were slightly more accurate than at-site anticipations.