Spatial Variability Change Modeling Of Soil Organic Carbon At A County Scale In Hilly Area Of Mid-Sichuan Basin

Soil organic carbon (SOC) content and dynamic changes information affect the soil quality and the cycle of global carbon directly. Accurately accessing SOC content at regional scale has been very important in regulating soil carbon and global environmental changes. Therefore, the study of soil organic carbon dynamics and accurately imitating and predicting the spatial and temporal changes of SOC in regional and regional evolution were concerned and explored widely. Based on the data gathered during the Second National Soil Survey in 1981 and the measured data of 555 soil samples collected in 2012, exploration was done of characteristics of the temporal and spatial evolution of SOC and it's driving factors in the Purple Soil Hilly area of Mid-Sichuan Basin over the past 30 years. Furthermore, with the aid of constructing High Accuracy Spatially Modeling of Topsoil Organic Carbon Spatial Variability Change, different properties of spatial data were analyzed. By means of Multiple Linear Regression (MLR)> Radial basis function neural network model (RBF)?High accuracy surface modeling (HASM) with embedding environmental factors.Results indicated that:(1) The mean content of SOC in the topsoil layer was 6.41 g.kg-1 in 1981and 13.46 g.kg-1 in 2012, with coefficient of variation being 72.59% and 48.87% respectively, the mean content of SOC increased by 109.98% while the coefficients of variation showed a decreasing trend.(2) The spatial structure analysis reveals that over the past 30 years, SOC was affected jointly by structural and random factors, but tended to be more affected by random factors. Soil parent material did not have much impact on SOC, while soil type, topography, vegetation covers and land use did (p<0.001). land use type was becoming the dominant factor affecting spatial variability variation of SOC. while soil type was decreasing in its role but still cannot be overlooked as suggested in the structural analysis conducted here. The impact of topography was declining, while that of vegetation cover was rising over the past 30 years.(3) The model of High accuracy surface modeling for soil properties combined with radial basis function neural network model based on land use typ unit (LU-RBF-HASM) obtain a much lower prediction bias by taking into account the relationship of space unbalanced and non-linear. Comepared with ordinary kriging (OK), High accuracy surface modeling for soil properties combined with land use type unit(LU-HASM), High accuracy surface modeling for soil properties combined with Multiple Linear Regression based on land use type unit(LU-MLR-HASM), the mean absolute error (MAE)x the mean relative error (MRE) and the root mean squared error (RMSE) of LU-RBF-HASM are reduced by 6.39%?36.47%. LU-RBF-HASM suggests a high accuracy with the most effective accurate prediction.(4) The LU-RBF-HASM map shows that SOC contents increased in most parts of the study area, especially in areas of low mountains and deep hills. And it presents more realistic details as well as macro pattern of the spatial variability variations of SOC which lead the effect of the accurate prediction consistent with the real situation.In summary, the accurate prediction method of LU-RBF-HASM obtained the high accuracy spatially modeling of top soil organic carbon's spatial variability dynamic change across at a county scale in hilly area of Mid-Sichuan Basin, and also provides an efficient approach to accurately estimate the SOC stock and study the spatial variability dynamic change of soil properties.