| Land resources play an important role in economic development and agricultural production,it's about human survival and social progress.In the ecosystem,land quality maintains biological productivity,promoting animal and plant health and so on.The document DD2008-06,?Technical Requirements for Geochemical Assessment of Land Quality(Implementation)?,of the Technical Standard for Geological Survey of China Geological Survey,points out that geochemical assessment of land quality should be carried out to quantify land quality and implement dynamic management so as to provide geochemical basis for land management and planning and serve for sustainable land use.Among them,soil pH values,soil organic carbon,nitrogen,phosphorus,potassium content are the main evaluation indicators of internal factors affecting land quality.As the major indicators of land quality geochemical assessment,soil pH,soil organic carbon,nitrogen,phosphorus and potassium content are affected by a variety of factors,such as complex topographic environment and agricultural production practice activities and so on.Under different geographic environment and geomorphological conditions,their spatial distribution and influencing factors of soil properties are different,and change with the research scale.Especially in mountainous and hilly areas,soil properties and spatial distribution at small watershed scale are more significantly affected by topography.Understanding their spatial distribution characteristics and its influencing factors clearly,is of great significance for efficient land quality assessment and soil management,rational agricultural regionalization and land resource utilization,and the implementation of precision agriculture.In this paper,a total of 102 samples were collected from the topsoil at a depth of20 cm to predict and map the spatial distribution of soil pH,soil organic carbon,nitrogen,phosphorus and potassium content over the Caotang River Basin in the Three Gorges Reservoir Area.After analyzing and measuring the soil characteristics based on soil parent material.terrain indicators,normalization difference vegetation index(NDVI)and land use types,four methods,which included the Multiple Linear Regression(MLR),Ordinary Kriging Interpolation(OK),Classification Regression Tree(CART),and Random Forest(RF),were applied to predict the spatial distribution of soil pH values,soil organic carbon(SOC),nitrogen(N),phosphorus(P)and potassium(K)content,respectively.The prediction accuracy of these algorithms were compared and the optimal model was selected for spatial distribution prediction mapping and influencing factors analysis with the relative importance of each influencing factor was evaluated,which provided a scientific basis for land quality evaluation,soil management,utilization and agricultural production management in the study area,and provided a scientific basis for watershed-scale soil.The prediction of spatial distribution of characteristics provides a reference for methods.The main results are as follows:(1)Soil pH in Caotang River Basin ranged from 5.08 to 8.59,with an average value of 7.35 and a coefficient of variation of 15.65%.The variation ranges of soil organic carbon,nitrogen,phosphorus and potassium were 0.47-2.10,0.58-1.79,0.21-0.52 and 1.20-2.84 g/kg,respectively.The mean values were 1.07,1.13,0.35 and2.01 g/kg,and the variation coefficients were 33.64%?23.01%?20%?17.91%,respectively.The results showed that soil pH values,soil organic carbon,nitrogen,phosphorus and potassium contents were moderately variable.(2)The study area includes two types of parent materials,quartz sandstone of Xujiahe Formation and limestone of Daye Formation of Triassic.The average values of soil pH value,soil organic carbon,nitrogen,phosphorus and potassium contents among different parent materials were compared.The results showed that there were significant differences in soil pH values,soil phosphorus and potassium contents among different parent materials(P<0.05),while there were no significant differences in soil organic carbon and nitrogen contents among different parent materials(P<0.05).Among them,the soil pH values,soil phosphorus and potassium content of the Triassic Daye Formation limestone were higher than those of the Triassic Xujiahe Formation quartz sandstone.(3)There were correlations between soil pH value,soil organic carbon,nitrogen,phosphorus and potassium content and topographic factors.Soil pH value was negatively correlated with elevation and vertical distance from river network,and positively correlated with Valley depth,with correlation coefficients of-0.583,-0.397and 0.412,respectively,indicating that the value decreased with elevation and the increase of vertical distance from river network,and increased with the increase of valley depth.Soil organic carbon content was positively correlated with slope gradient,slope length,and normalized vegetation index,with correlation coefficients of 0.257,0.255 and 0.234,respectively.Soil nitrogen content was positively correlated with slope gradient and slope length factors,with correlation coefficients of 0.259 and 0.287,respectively,but not with other topographic factors.Soil phosphorus content was negatively correlated with elevation and vertical distance from river network,and positively correlated with Valley depth.The correlation coefficients were-0.393,-0.341 and 0.310,respectively.There was no significant correlation with other topographic factors.Soil potassium content had no significant correlation with topographic factors.(4)Four models were used to construct the models of soil pH value,soil organic carbon,nitrogen,phosphorus and potassium content in the study area.The results showed that the predictive accuracy of multivariate linear regression for five soil properties(R~2)was 0.57?0.19?0.23?0.30 and 0.21,respectively;the predictive accuracy of ordinary Kriging interpolation for five soil properties(R~2)was 0.56?0.25?0.21?0.33 and 0.30,and the predictive accuracy of classification regression tree for five soil properties(R~2)was 0.73?0.64?0.51?0.36 and 0.78,respectively.The predictive accuracy(R~2)of random forest for five soil properties were 0.85?0.85?0.95?0.89 and 0.87,respectively.In mountainous and hilly areas,the prediction accuracy of random forest model was higher than that of multiple linear regression,classified regression tree and ordinary Kriging interpolation.It can be seen that the random forest prediction model based on environmental factors has high prediction accuracy and can be used as an effective method for predicting spatial distribution of soil characteristics at watershed scale.(5)The relative importance affecting five soil properties was analyzed.The results showed that the main environmental factors affecting soil pH were parent material,elevation and vertical distance from river network;the main environmental factors affecting soil organic carbon content were slope,slope length factor and elevation;the main environmental factors affecting soil nitrogen content were slope direction,topographic wetness index and slope length factor;and the main environmental factors affecting soil phosphorus content were slope direction,topographic wetness index and slope length factor.The main environmental factors affecting soil potassium content were slope height,elevation and vertical distance from river network,while the main environmental factors affecting soil potassium content were elevation,slope direction and vertical distance from river network.(6)The model with the highest prediction accuracy(R~2)was used to map the spatial distribution of soil pH,soil organic carbon,nitrogen,phosphorus and potassium in the study area.The results showed that the spatial distribution of soil pH values in the study area formed distinct southwest-northeast boundaries because of the different types of parent materials,which mainly manifested that the soil pH value in the northeast region was lower than that in the southwest region.Soil organic carbon content was higher in the north and southeast regions,but lower in the southwest regions,showing a trend of increasing gradually from south to north and from west to east.The spatial distribution of soil nitrogen content was lower in the northwest corner,southwest margin and southeast regional center,and increased along the north and south respectively.There was no obvious boundary,and the overall distribution was more balanced.The distribution of soil phosphorus content showed a trend of increasing gradually from north to south,showing obvious differences in different regions due to topographic differences,and showing obvious reverse differences with elevation.The distribution of soil potassium content was obviously lower in the northwest region and higher in the Southeast region,showing a trend of gradual increase from northwest to southeast. |
PDF Full Text Request