Abstract—Soil organic carbon (SOC) plays a key role in soil fertility, hydrology, contaminants control and acts as a sink or source of terrestrial carbon content that can affect the concentration of atmospheric CO2. SOC supports the sustainability and quality of ecosystems, especially in semi-arid region. This study was conducted to determine relative importance of 13 different exploratory climatic, soil and geometric factors on the SOC contents in one of the semiarid watershed zones in Iran. Two methods canonical discriminate analysis (CDA) and feed-forward back propagation neural networks were used to predict SOC. Stepwise regression and sensitivity analysis were performed to identify relative importance of exploratory variables. Results from sensitivity analysis showed that 7-2-1 neural networks and 5 inputs in CDA models output have highest predictive ability that explains %70 and %65 of SOC variability. Since neural network models outperformed CDA model, it should be preferred for estimating SOC. Keywords—Soil organic carbon, modeling, neural networks, CDA. I. INTRODUCTION OIL organic carbon (SOC) has an indispensable role in the ecosystems and acts as a buffer to global climatic change. It is therefore critical to maintain its quality for the sustainability of ecosystems [30]. Soil organic carbon is a vital component, as it plays a key role in soil fertility, hydrology, contaminants control and acts as a sink or source of terrestrial carbon content that can affect the concentration of atmospheric CO2. The terrestrial carbon reservoir is estimated to be between 3,400 to 3,500 Giga tons that SOC consists about 3,000 Giga tons of them [25]. Hence, soil can be considered as important sink and source for carbon sequestration and modifier of climate changes [6]. Appropriate management of SOC can substantially decrease the atmospheric carbon that has increased exponentially at a rate of 1.5% per year [7]-[15]. Also, soil is regarded as an important sink and source for carbon sequestration and modifier of climate changes. This can be done by management and land use activities [27]. Y. Parvizi is with the Agriculture and Natural Resource Research Centre of Kermanshah, Iran (corresponding author to provide phone: 0098-831- 8370070; fax: 0098-8318351022; e-mail: yparvizi@ut.ac.ir). M. Gorji Assistant Professor Tehran University, Iran, (e-mail: mgorji@ut.ac.ir) M.H. Mahdian ,Associate Professor, Agriculture Research and Education Organization, Iran. (e-mail: mahdian_1338@yahoo.com). M. Omid Associate Professor Tehran University, Iran. (e-mail: omid@ut.ac.ir) One approach in dealing with linear statistical methods in soil process modeling and SOC class estimation is not new. These methods include multiple linear regressions (MLR), logistic regression, and canonical discriminate analysis (CDA). However, there are characteristics of the models such as over simplification, ignorance of complex nonlinear interactions etc., which limit their use in accurately assessing the distribution of the C across the landscapes. Another approach in dealing with nonlinear systems is to use artificial Intelligence (AI) modeling paradigm such as artificial neural networks (ANNs). ANNs has been successfully used in classification, prediction, and pattern recognition problems [8]- [14]-[31]. The potential benefits of ANNs include greater prediction reliability, cost-efficient estimation and solving complex problems involving nonlinearity and uncertainty. ANNs are inspired by biological neural networks [2]-[29]. ANNs learn from training examples, adjusting weights to reduce the error between the measured and the predictions. ANNs have been successfully applied in various soil studies [26]. These applications include predictions of soil structure [17], pedotransfer functions [1]-[16]-[21, pedometric use in soil survey [13], environmental correlation of three- dimensional soil spatial variability [19], and prediction of SOC from soil parameters [10]-[11]. This study was conducted to model SOC variability at watershed scale across agricultural rainfed land use types in a semi-arid condition of Iran. We employed ANNs and CDA to investigate the effect of climatic, topographic and soil properties and also management variables on SOC. Various ANNs topologies were designed and tested. Since the estimation model was data based, selection of appropriate input and output variables is important. Thus, sensitivity analysis on exploratory variables was carried out to select the best input combinations for modeling and estimating SOC. II. MATERIALS AND METHODS Site Description Merek watershed from Karkheh river basin with area about 24200 ha, was selected for this study, because in this watershed we can find appropriate diversity in soil, topography and semi- arid climatic conditions. The elevation of this site ranged from 1450 to 19174. It has a semi-arid and cold climate with an annual precipitation of about 500 mm. The land use is mainly: agricultural land that covers about 14500 ha. In this site, soil texture is clay or salty clay and soil structure is blocky. Soils, in mountains and highlands, were covered by about 25-60% Sensitivity Analysis for Determining Priority of Factors Controlling SOC Content in Semiarid Condition of West of Iran Y. Parvizi, M. Gorji, M.H. Mahdian, M. Omid S International Journal of Environmental and Earth Sciences 1:3 2010 142