Spatial Prediction And Comprehensive Evaluation Of Soil Nutrients Based On Environmental Variables

The soil nutrient content directly affects the growth status,fruit yield and quality of the crop throughout the life cycle.The accurate prediction of soil nutrient by spatial prediction method based on finite sample points is the prerequisite for providing effective data support and theoretical guidance for agricultural production.Choosing the appropriate spatial prediction model is the key to improve the accuracy of soil mapping,but the different nutrients in the same area are affected by the surrounding environment and human activities to different degree.In order to ensure that the spatial prediction model can accurately reflect the mapping relationship between soil nutrients and explanatory variables,it is necessary for improving the accuracy of mapping and guiding farmland management scientifically to fit model by various types of methods and draw soil nutrient map by appropriate method.The main farming areas in Xianyang City was taked as an case study.Based on the soil sampling data of the soil testing and fertilization recommendation in 11 counties and the collected multi-source environmental variables data,the geostatistical method,linear regression algorithm of spatial non-stationary,data mining algorithm and hybrid interpolation algorithm combined with residuals are used to establish spatial prediction model.By comparing the prediction accuracy of different methods,the best prediction method is selected to draw spatial distribution map.Based on the results of model fitting and correlation analysis,the effects of soil properties and natural environment on soil nutrients were evaluated comprehensively.By comparing the average soil nutrient content under different irrigation and cropping system,the changing trend of fertilization amount and structure was analyzed,and the effect of human activities on farmland nutrient was explored.The soil nutrient synthetic index was calculated by entropy weight and TOPSIS method,and the synthetic index was spatialized by the cokriging to explore the spatial characteristics of soil nutrient comprehensive level.(1)From the perspective of spatial prediction mapping,different nutrient indicators may have linear or complex nonlinear relationship with the environment variables.Geographically weighted regression model(linear)and random forest model(nonlinear)were selected to fit this relationship.Considering the spatial autocorrelation of soil nutrients,the composite model combined with residuals(regression kriging,geographically weighted regression kriging and random forest regression kriging)was introduced to reveal the true distribution of soil nutrients from multiple perspectives,and to explore the adaptability of different methods,so as to provide references for related research on spatial prediction.In terms of synthesis evaluation,the TOPSIS method can reduce the subjective errors and ensure the accuracy and scientificity of the evaluation results.The geostatistical method was applied to visualize composite index and express the spatial characteristics of soil nutrient comprehensive level.From the three aspects of soil physical and chemical environment,natural environment and human activities,the factors affecting soil nutrient are comprehensively analyzed.(2)The results of spatial prediction showed that the best prediction models of organic matter,total nitrogen,available phosphorus and available potassium were random forest model,geographically weighted regression,geographically weighted regression kriging and random forest regression,respectively.The main reason was that the topography was complex in study area,the geographic weighted regression can take into account the spatial heterogeneity of soil nutrients,and the random forest can effectively capture the complex non-linear relationship between soil nutrients and the environment,which was closer to the actual situation of the study area.As a whole,the spatial distribution trend of organic matter was high in the south and low in the north,and its content change was obviously affected by the topography.The organic matter content in the plain area was higher than that in the plateau.The spatial distribution of total nitrogen was basically consistent with organic matter.The low value areas of available phosphorus were mainly distributed in the southeast,and the other areas were more uniform.The trend of available potassium was higher in Central than in other regions in the north-south direction,and the spatial distribution could obviously reflect the topographic changes.(3)The analysis of influence factors showed that topography and climate were the key factors affecting soil nutrients.In areas with high topography,large slope and high terrain,the soil was easily eroded to cause soil erosion and low nutrient content.The region of low temperature had a high nutrient content;precipitation was conducive to the accumulation of organic matter and nitrogen,and also the loss of available potassium due to soil erosion.The average soil nutrient content under irrigation areas were higher than that in non-irrigated areas.Under the four cropping systems of wheat-maize and wheat,maize and apple(one crop per-year),the overall level of soil nutrients in wheat-maize fields was higher than that of the other three,and the organic matter content in apple orchards was lower.Since 2000,the application rate of agricultural chemical fertilizers in Xianyang City has increased.The proportion of nitrogen and phosphorus fertilizers in chemical fertilizer structure has decreased,and the proportion of potassium fertilizers has increased.(4)Organic matter and total nitrogen content were relatively poor,available phosphorus content was moderate,available potassium content was abundant,and four nutrients were with moderate variability.The comprehensive level of soil nutrient was medium and relatively high level.The high level area was small,scattered in Liquan County,Wugong County,Xingping City and Jingyang County.The low and relatively low levels were mainly distributed in the eastern in the study area,central Changwu County,northern Yongshou County and northern Qianxian County.