predicting spatial variability of soil salinity and clay content using geostatistics and artificial neural networks methods

in order to comparing geostatistics and artificial neural networks (ann) methods in predicting soil salinity and clay content this experiment was conducted in sistan region land. 121 soil samples were taken at the depth of 0-30 cm within the grid of 750 750 m and soil ec and clay percent were determined. 105 samples were used for training and 16 samples were used for test in both models. different models of geostatistics and ann were fitted and the best models were selected. results showed that ann estimated better with determination coefficient of 0.67 and rmse of 6.18 for soil clay content in comparison to geostatistics with 0.62 for determination coefficient and 8.20 for rmse. also in predicting soil salinity, ann had a determination coefficient of 0.59 and rmse of 15.8 in compression to geostatistics with 0.54 for determination coefficient and 19.20 for rmse had more accuracy. management effect on soil salinity decreased the prediction accuracy in both models in study area.