عنوان مقاله [English]
Since Soil organic carbon (SOC) has a key role in production and soil productivity, it has also important role in sustainable soil and crop specially in our country. Among soil particles, SOC variability in medium time scale is a management based practice. This research was conducted to investigate the effects of physical and management variables on SOC variations and to quantify the relative importance of these variables on SOC distribution in a rainfed watershed by use of artificial neural networks (ANNs) technique. Sampling was done based on randomized systematic method and simulation, analysis of physical and management factors impacts and sensitivity analysis of predictor variables were don using artificial neural networks (ANNs). Results indicated that among ANNs that applied to simulate SOC, those with 2 hidden layers and two nodes in each layer with than transfer function have high efficiency with MSE=0.04 – 0.07 in SOC prediction. ANN models with management exploratory variable were able to distinguish SOC contents more efficient. Best ANN model with all 31 exploratory variables can estimate about 78 percent of SOC variability. Sensitivity analysis showed that tillage, crop residue and rotation related prediction variables by controlling about 24, 24 and 18 of SOC variability, respectively, have greatest effects.