Somaiye Moghimi, Mohammad Hossein Mahdian, Yahya Parvizi, Mohammad Hasan Masihabadi
Department of Soil Science, College of Agriculture, Tehran Science and Research Branch, Islamic Azad University, Tehran, Iran
Soil Conservation and Watershed Management Research Institute, Tehran, Iran
Agriculture and Natural Resource Research Center, Kermanshah, Iran
Key words: Aspect, Curvature, hilly area, RBF, Slope.
Abstract
Soil organic carbon (SOC) is a source or sink of atmospheric carbon and importance of it has been increasingly recognized in soil physical, chemical and biological characteristics. The objective of this study was to predict and evaluate the effects of topographic attributes on the soil organic carbon content at a hilly pastureland in Mereg watershed, Iran. In this research, topographic attributes include the primary factors such as elevation, slope, plan and profile curvature, transformed aspect and secondary factors such as slope-aspect combinative index, wetness index and stream power. Multiple linear regression (MLR) and radial basis function (RBF) artificial neural network were employed. The comparison of model evaluation criteria demonstrates that the RBF model (R=0.954, RMSE=0.087%) provides more accurate predictions of SOC than the MLR model(R=0.528, RMSE=0.349%). The RBF model, with 15 neurons in hidden layer and 2 spread value was applied successfully and exhibited the more reliable predictions than the MLR model. Results showed that, SOC content were mostly sensitive to the profile curvature, plan curvature, transformed aspect and slope percent.
Get the original articles in Source: Volume 5, Number 2, August 2014 – JBES
Journal Name: Journal of Biodiversity and Environmental Sciences (JBES)
Published By: International Network for Natural Sciences