Balance in nutrient cycle is a key indicator that shows significant improvement or degradation of the soil system. Nitrogen is one of the most important elements affecting crop growth and soil fertility. Its deficiency is the major limiting factor in soil productivity and plant growth. Soil enzyme activity can be employed as a measure of soil biological potential. Soil enzymes are indicators of soil quality and fertility and they are considerd as the center of microbial activity and nitrogen transformation in soil. This study was conducted to predict three soil enzyme activities, potential mineralizable nitrogen (PMN) and soluble organic nitrogen (SON) using multivariate linear regression and artificial neural networks. Assessment of the efficacy of two modeling approaches and determining the most factors affecting the variability of the selected soil enzymes, PMN and SON were the other objectives of this study. The studied site with an area of 2400 ha in the Zargham Abad hilly region, located in Isfahan province, south Semirom was selected and soil samples were taken from 0-10cm depth at 125 sampling points. The elevation data were used to create 3×3 m digital elevation models (DEM) using ILWIS software. Then, primary and secondary topographical indices were generated from the DEM using ILWIS software. The soil was air-dry and ground to pass through a 2 mm sieve to remove gravel, roots and large organic residues for laboratory measurements. Urease, L-glutaminase and L-asparaginase activity, NMP, SON, and some soil properties including particle size distribution, soil organic carbon, total nitrogen, calcium carbonate equivalent, pH and EC were measured. Multivariate linear regression and artificial neural network modeling for prediction soil enzyme activities, NMP and, using topographic attributes and soil properties were conducted. In order to identify the most important terrain attributes and remote sensing data explaining the variability of SOC, sensitivity analysis was done using the Hill method The spatial distribution of enzyme activity, PMN and SON were explored by variography analysis and kriging technique. The results of study sowed that there nitrogen, topographic attributes artificial neural networks
