| The traditional soil type methods, vegetation type methods and ecological simulation estimation methods fail to accurately reflect the spatial variability of soil properties. Therefore, these factors restrict the estimation precision. This study was conducted in the Gaofeng Forest, a hilly area under Eucalyptus plantations in Nanning, China. The conditioned Latin hyper-cube sampling method was used to select samples. And it built a digital elevation map of the study area, from extraction of terrain factors (elevation, slope, aspect, plane curvature, profile curvature and Topographic Wetness Index), and combined with digital soil mapping techniques (multiple linear regression models and regression trees models) to predicate soil properties (soil organic carbon content, soil bulk density and thickness of soil layer). Finally, according to the soil organic carbon density formula, the soil properties of each soil layer were superimposed, expecting to be able to accurate estimate soil organic carbon storage of Eucalyptus plantations. The main research results are as follows:(1) The conditioned Latin hyper-cube sampling method was used to select 41 representative samples. The results show that the representative performance of the sampling points obtained in this study can meet the requirements of this study.(2) In this area, due to the influence of other non terrain factors, the relationship of soil properties and topographies of linear is weak. Therefore, the regression tree model is better than the multiple linear regression prediction models on the spatial interpretation of soil properties. The accuracy of prediction and the stability of regression tree models are better than that of the multiple linear regression models(3) In conclusion, compared with the traditional estimation method, digital soil mapping technology can improve the accuracy of predicting soil organic carbon storage in the study area, especially for the deep soil. Therefore, this study that used the regression tree prediction model to estimate can improve the accuracy of the total soil organic carbon storage of Eucalyptus plantations in the study area.In this study, we use the regression tree models to predict the complete profile soil organic carbon storage of Eucalyptus plantations in the study area, about 33.72×10~6 kg. |
