Soil and groundwater pollution with heavy metals is one of the most important problems in densely populated and industrial areas around the world. Mobility, bioavailability and toxicity of heavy metals in soils are strongly dependent on sorption of metal ions by soil constituents. Direct measurement of the sorption coefficients of metals in soil is difficult and time consuming. Hence, in the recent years, great efforts have been made to estimate sorption parameters with reasonable accuracy by using the readily available properties of soil. The purpose of this study was to evaluate the feasibility of using pedotransfer functions, i.e. Artificial Neural Network (ANN) and Multiple Regression (MR), to estimate the sorption parameters of Nickel (Ni) using readily available soil properties. For this purpose, Ni sorption isotherm experiments were carried out on 102 soil samples from Mazandaran Province and various sorption parameters were obtained from fitted Langmuir and Freundlich equations. Distribution confidents (Kd) at different Ni concentrations were also calculated as single-point indexes of Ni sorption. Results showed that Ni sorption parameters have significant correlations with soil properties such as pH, clay, silt, sand, organic matter, Calcium Carbonate Equivalent (CCE) and CEC, confirming the role of these soil characteristics on Ni sorption in the soils. A lot of multiple regression models were developed for estimating of Ni sorption parameters. The most important soil properties entered to regression models were CEC, CCE, organic matter and clay.ANN sensitivity analysis also showed that the clay, organic matter, and CCE contents had the most influence on Ni sorption parameters in the studied soils. Also, the results showed that the discrimination of soil samples based on landuse, improves the accuracy of sorption parametrs estimated by regression but reduced those estimated by the ANN. In conclusion, the results of this study showed that the pedotransfer functions can be satisfactorily used for estimation of different Ni sorption parameters from soil properties in Mazandaran province. Keywords : heavy metals, adsorption isotherm, multiple regression, Artificial Neural Network, Pedotransfer Functions, readily available soil properties.
