Spatial Variation Of Soil Organic Carbon And It's Influance Factors Of Cropland Around Poyang Lake

Soil carbon pool is the largest carbon pool of terrestrial ecosystems.Small changes in soil organic carbon(SOC)play an important role in greenhouse effect and global climate change.Because it is easily affected by natural environment changes and human activities,it has strong spatial heterogeneity.How to consider the influencing factors of spatial change and accurately estimate SOC is an important issue in the current research on global climate change.At the same time,SOC is also one of the important indicators for evaluating land quality,and its content directly affects soil ecosystem function and sustainable utilization capacity.Especially in the fields of agricultural pollution control,sustainable agricultural development,national food security,global carbon cycle simulation,SOC research has become an important basic condition.Therefore,intensive study of the influencing factors of SOC,accurate understanding of the spatial variation characteristics of SOC is an important prerequisite for accurate estimation of SOC.In addition,further study on the relationship between soil organic carbon and environmental factors,The spatial distribution pattern and mapping research of carbon driven by different environmental factors not only provide a theoretical basis for deepening soil carbon cycle research in farmland ecosystems,but also have important practical significance for improving the quality of cultivated land.This study takes 11 counties(districts)in the Poyang Lake District of Jiangxi Province as the study area.Based on the sampling point data of Agricultural Soil Testing Formula Project in Jiangxi Province in 2010,combined with remote sensing image data,terrain data and meteorological data.Firstly,in the ArcGIS10.1 geostatistical analysis module to generate a subset of data,the total data is 2973,randomly extract 80%of the sample points as the modeling data set(modeling set,2378),20%of the data as Validation subset(verification set,595).Used Spass 19.0 statistical software to analyze the correlation and significance of soil organic carbon related environmental variables.Used the stepwise regression method to establish the optimal linear model.Slope,aspect,annual average temperature,annual average precipitation and vegetation coverage optimal linear model factors were reserved.The GWR4.0 software is used to establish the Geographically Weighted Regression(GWR)model of the dependent variable SOC for the above five variables,and the spatial distribution pattern of each influance factors coefficient is obtained and suggestions for improving land quality are proposed for the distribution of SOC influance factors in different regions.Finally,using GS~+9.0 modeling and analysis software,combined with Kriging interpolation method to study soil organic carbon mapping,and comprehensively compared with the predicts differences in soil organic carbon results of Geographically Weighted Regression Kriging method(GWRK)and Ordinary Return Kriging(RK).The main results and conclusions of this study are as follows:(1)From the rationality of the factors affecting soil organic carbon,according to the correlation test results of impact factors,the terrain factors,vegetation factors and meteorological factors selected in this paper are related to SOC(P<0.01);However,in the subsequent modeling research,the optimal linear model is established by stepwise regression method.Finally,the five factors of slope,aspect,annual average temperature,average annual precipitation and vegetation coverage are the optimal linear modeling factors.It is also used for in-depth study as the modeling factors in GWRK and RK method in the paper.(2)According to the GWR modeling results,among the influencing factors of SOC,the slope,aspect,annual average precipitation and annual average temperature factors are positively correlated with SOC,that is,the aspect change are 1°,respectively,which affects the SOC by 0.001 g/kg;for every 1 mm increase in annual precipitation,the annual average temperature increases by 1°C,and the corresponding SOC content increases by 0.012 g/kg and 0.532 g/kg.There was a negative correlation between slope and vegetation coverage and SOC.The slope change are 1°,which decreased the SOC by 0.125g/kg;The vegetation coverage each additional unit,the SOC decreased by 0.104 g/kg.In the spatial distribution characteristics of the influencing factors,the relationship between each influencing factor and SOC presents a significant difference between positive and negative effects and factor combination in space.It is necessary to take targeted measures according to local conditions to effectively improve the land quality.(3)The spatial variation characteristics of SOC:The SOC values of 2378sampling points in the model data of this study was 5.220~33.930 g/kg,the average value was 17.549 g/kg,the standard deviation was 5.864 g/kg,and the skewness and kurtosis were 0.224 and-0.855.According to the K-S test,the SOC data conforms to the normal distribution and the coefficient of variation is 33.42%,which is moderately variable.The SOC content was calculated by different methods,the SOC content of the GWRK method was 6.35~31.93 g/kg,which is close to the value of the sampling point.The spatial variation characteristics of SOC is“northwest low,east and south high”.The high value area and the low value area are staggered,which also reflects the detailed information of SOC changes with natural and human factors.The RK method was5.40~29.01 g/kg,the overall spatial variation characteristics are similar to those of the GWRK.However,the RK method predicts a smoother result and there is a certain difference from the actual value of the SOC content.(4)Comparison and verification results of the mapping model:The GWR model has a significantly improved fitting effect.The goodness of fit of GWR R~2(0.50)is significantly higher than RK R~2(0.20).In the model accuracy evaluation,the Mean Absolute Error(MAE)of GWRK method is 3.63,which is lower than the RK method4.40.At the same time,the Root Mean Square Error(RMSE)of the GWRK method is4.58,which is smaller than the RK method 5.35.The prediction result obtained by the GWRK method is also close to the actual value of the SOC sample.Therefore,using the GWR model for SOC space mapping research has more advantages.