Assimilating Multi-source Data To Model And Map Potential Soil Loss In China

Soil erosion by water is a major threat to agricultural productivity and environmental protection,particularly in China where up to one third of its territory is affected by water erosion.Estimate soil erosion quantitatively and analysis its spatial distribution objectively plays an important role in the soil loss prevention and land resources protection in China.Despite its importance,however,spatial heterogeneity between environmental factor(soil,climate,vegetation,land use,etc.)that related to soil erosion at different spatial and temporal scales,and the lack of field measurements make the assessment of soil erosion across China a challenging prospect.Conventionally,soil erosion has been quantified based on the standard ground area or runoff experiment.However,there are few data on soil erosion in the national scale,largely because erosion is difficult and expensive to measure and because of the vast areas of land with diverse soil and vegetation,variable climate,topography and land use and management.The scarcity of suitable surface environmental data to estimate soil erosion can be overcome using the latest satellite technology.Satellite platforms can help to quantify the spatial distribution of rainfall,vegetation,topography and land use over large areas,rapidly and relatively cheaply.However,satellite products are not without problems due to limitation of systematic accuracy and inversion algorithm.Therefore,the improvement of soil erosion estimation based on the combined multi-source data effectively has becomes the hot issue in the soil science.Digital soil mapping(DSM)produces reliable spatially explicit soil erosion information and expores its evolution spatio-temporally.The aim here were to use the latest multi-source data to estimate potential soil loss by water erosion using the Revised Universal Soil Loss Equation(RUSLE),and map potential soil loss in China with 1 km spatial resolution using DSM and data mining technology.The main results are as follows:(1)Rainfall erosivity estimation based on a two-dimensional non-parametric blending methodologyHigh-quality rainfall data is essential to estimate rainfall erosivity and potential soil loss.One of the aims of the present study is to improve estimates of spatial and temporal rainfall and thereby estimates of rainfall erosivity in China at a fine spatial resolution.To achieve this aim,a new dataset of merged daily rainfall based on a two-dimensional non-parametric blending methodology is calculated,and the mean annual rainfall erosivity value is estimated annually between 2002-2013.The analysis involves the following steps:?)merge rain gauge and satellite-based rainfall data to estimate daily rainfall maps across China,?)estimate mean annual rainfall erosivity across China based on merged daily rainfall data for each 0.25°×0.25° pixel across China.and iii)downscale estimated mean annual rainfall erosivity map to 1 km resolution using Geographically Weighted Regression(GWR).The daily rainfall estimated by merged data is superior to that estimated by using TRMM or rain gauges alone.The new mean annual rainfall erosivity map in this study is validated with a strong correlation(R=0.86)with data from the literature.The distribution of rainfall erosivity across china shows seasonal and zonal difference obviously.(2)Digital soil erodibility mapping based on decision treeSoil erodibility is a factor which is related to soil itself directly.Conventionally,the soil erodibility estimation is mainly based on the regional soil map.However,the widely used national soil map largely ignores the spatial heterogeneity of the soil,and treats soil as a whole within large areas.The applicability of these soil maps among regional scale deserves further discussion.This study used a DSM framework,to assimilate into soil erodibility a set of environmental predictors that might best complement the RUSLE factors to represent soil erosional processes across China.The soil erodibility factor was modelled using soil-environmental factors that describe the interactions between soil,climate,terrain,geology in space and time.This study collected 3758 soil profile data from the second national soil survey.The accuracy of the model in this study is acceptable(R2=0.52).Radiation,soil type and elevation are the most important environmental factors in the model.The accuracy of the model is controlled by the samples.According to the result,the primitive soil has the lowest soil erodibility value.Compared with the result which is calculated based on the HWSD(Harmonized World Soil Database),our result shows lower soil erodibilty value.(3)Estimation of cover management factor,slope length and steepness factor and support practice factor in the RUSLEThe cover management factor,slope length and steepness factor and support practice factor were calculated in this study based on MODIS NDVI?DEM?Landsat TM data.The spatial distributions of these factors were also analyzed.The results show that the C factor is inversely related to NDVI,and the spatial distribution trend is decreasing from northwest to southeast in China.The implementation of farming measures and engineering measures in the Northeast plain,the North China Plain,the Middle-Lower Yangtze plains,Guanzhong Plain,the Loess Plateau and the Sichuan Basin can reduce the risk of soil erosion effectively.(4)Spatial variability of the potential soil loss and the estimation of the potential soil loss in the futureThe annual potential soil loss and its spatial distribution and variation during 2003-2014 were estimated and analyzed in this study.The results show that the areas with the smallest amount of potential soil loss are mainly distributed in the desert areas while the areas with the highest potential soil erosion are mainly distributed in the southern Qinghai-Tibet Plateau and the Hengduan Mountains.Furthermore,this study estimated the potential soil loss in the 2050s using the estimated current potential soil loss and the regression Kriging(RK)model under different emission scenarios.The results show that the active responding to climate change can reduce potential soil loss in the 2050s compared to negative responding.When other factors are constant,the potential soil loss in the southeastern China has the largest change in the 2050s.This study completed the research content and achieved the expected research objectives.The progresses are made as follows:(1)Rainfall erosivity factor at different scales has been analyzed by several studies.These analyses were mainly carried out based on ground observation rainfall data.With the development of satellite technology,remote sensing data has been used in the estimation of rainfall erosivity.However,there are limited studies of rainfall erosivity estimation based on the merged gauge-satellite data.A two-dimensional non-parametric blending methodology is executed using rain gauge and TRMM data to improve the accuracy of the daily rainfall data in this study.A spatial downscaled methodology,GWR,was implemented to improve the spatial resolution of rainfall erosivity map across China.This study provides a new idea to estimate rainfall erosivity factor at large scale with high precision.(2)Conventionally,soil erodibilty factor at large scale was estimated based on the plate filling of soil map which can hardly represent local situation and can poorly applied national wide.Digital mapping of soil property factors with high-precision is a difficult issue in the field of soil science.This study used limited soil profile data and a DSM framework to assimilate into soil erodibility a set of environmental predictors.The high resolution soil erodibility factor in China was modelled using soil-environmental factors and data mining methodology.This study provides a new method for estimating soil erodibility and analyzing its spatial variation at arge scale in the future.(3)The change of soil properties caused by the change of future climatic conditions is the frontier of the soil field.Under the hypothesis that other factors are constant in the future,this study estimate the potential soil loss in 2050s using the estimated current potential soil loss and the RK model under different emission scenarios.The quantitative results of potential soil loss in 2050s provide guidance to the future environmental protection,soil erosion control and land use planning.