Comparison of spatial interpolation methods for estimating heavy metals in sediments of Caspian Sea

摘要

This study aims to estimate the spatial distribution patterns of six heavy metals: Arsenic (As), Cadmium (Cd), Copper (Cu), Mercury (Hg), Plumbum (Pb), Zinc (Zn) in the sediments of Caspian Sea. Ordinary kriging (OK), genetic algorithm based on artificial neural network (GA-ANN), adaptive network fuzzy inference system (ANFIS), and conditional simulation (CS) have been used for spatial distribution modeling. A total number of 80 surface sediment samples were collected in the year 2007 in Caspian Sea and Volga Delta in framework of the Caspian Ecosystem Program (CEP) which focuses on contaminants survey. As part of these samples, five countries of Iran (18 samples), Azerbaijan (16 samples), Turkmenistan (21 samples), Kazakhstan (13 samples), and Russia (12 samples). Results indicate that the CS realizations yields interpolation values such that the parsimony principle can not be kept. Simulated maximum and minimum values based on the CS method, is less and more than corresponding observed values, respectively. The OK realization smoothed out spatial variability and extreme measured values between the range of observed minimum and maximum values for all of the contaminants. The GA-ANN model has been capable of simulating the minimum values of contaminants as well. ANFIS, GA-ANN and OK are capable simulate the average values of contaminants, as well, except Cd and Hg. The results of spatial distribution modeling of Cd, Cu, Hg, Pb, and Zn show that the maximum concentrations of these contaminants are distributed in the south of Caspian Sea, near the boundary of Azerbaijan and Iran. In the case of As, maximum concentration is found in the north and south of the study area. Finally, comparison between the four interpolated techniques, GA-ANN model is the best model in keeping the statistical characteristics of the observed data for all contaminants, however ANFIS model is the best model with least simulation errors.