Prediction Of Soil Nutrients Spatial Distribution In Black Soil Area At The Field Scale

Cultivated land resource is precious resource that human beings rely on for survival,the carrier of food production,and the important material guarantee for social prosperity and development.A correct and comprehensive understanding of spatial heterogeneity and spatial distribution pattern of nutrients in cultivated land is of great significance for the protection of cultivated land environment,improvement of cultivated land productivity,efficient and sustainable utilization of land resources,development of precision agriculture and digital soil mapping.In this paper,soil nutrients in Zhaoguang's farm in black soil area at the field scale were studied,through the laboratory sample design,wild field sampling and determination of soil nutrient analysis,obtained 0-20 cm soil organic matter(SOM),total nitrogen(TN),total phosphorus(TP),total potassium(TK),available nitrogen(AN),available phosphorus(AP)and the content of available potassium(AK)in the study area.The classical statistical analysis of each nutrient was carried out by the traditional statistical method,and then the spatial autocorrelation and heterogeneity of the soil nutrient indexes were analyzed by using the semi-variance function model of geostatistical method.,Finally,based on the "3S" technology,the Ordinary Kriging,Regression Kriging and Random Forest Kriging were used to simulate the spatial distribution of some nutrients in the study area.The main conclusions are as follows:(1)SOM of surface soil in zhaoguang's farm minimum value is 3.975%,maximum value is 7.158% and average value is 5.990%.The contents of soil TN ranged from 1.562g/kg to 3.378g/kg,with an average value of 2.521 g/kg and a standard deviation of 0.334.The contents of AN ranged from 106.753mg/kg to 420.840mg/kg,with AN average of 273.800mg/kg.Soil TP content ranged from 0.900 mg/kg to 1.576mg/kg,with an average value of 1.251mg/kg.Soil AP ranged from 23.058mg/kg to 70.243mg/kg,with an average value of 46.650mg/kg.TK content in surface soil ranged from 16.290mg/kg to 20.730mg/kg,with an average value of 18.569mg/kg.The content of AK ranged from 119.204mg/kg to 202.796mg/kg,with an average value of 153.844mg/kg.In general,the content of main soil nutrients was at a rich and extremely rich level,and the coefficient of variation of main soil nutrients was expressed as: AN > AP > TN > SOM > AK > TP > TK,in which all the other soil nutrients were moderately variable except TK.(2)The geostatistical analysis of nutrients in the study area shows that the optimal semi-variance fitting model of AN,AP,TK is exponential model,and the optimal semi-variance fitting model of SOM,TN,AK is Gauss model.The spherical model is the best semi-variance function fitting model of TP,and the determination coefficient of each soil nutrient semi-variance function fitting model is more than 0.671,and the fitting effect is good.According to the nugget effect value,soil nutrients are all spatially autocorrelated to varying degrees,but they are all between 0.25 and 0.75,belonging to the moderate spatial autocorrelation.The spatial autocorrelation distance of 7 soil nutrient indexes ranged from 65.602 m to 205.046 m.From the view of the total variability of the system reflected by the base values,the highest heterogeneity is AK,and the lowest is TK.(3)There was a significant or extremely significant correlation between SOM and soil nutrients except AK,and the correlation between TN and SOM was the highest,reaching 0.773,followed by TP,AN,AP and TK.There were significant or extremely significant correlations between some topographical factors extracted by high precision DEM and soil nutrients.There were significant or extremely significant correlations between the spectral indices and the spectral indices extracted from Sentinel-2A images and the main soil nutrients(SOM,TN,AN)in the study area.The correlation between spectral index and soil nutrient was higher than that of single band reflectivity.These two indexes can be used as effective auxiliary variables to predict the spatial distribution of soil nutrients.(4)By comparing the spatial distribution of main soil nutrients(SOM,TN,AN)predicted by Ordinary Kriging,Regression Kriging and Random Forest Kriging,the most suitable spatial simulation method in this study area is Random Forest Kriging method.Both Random Forest Kriging and Regression Kriging spatial interpolation can overcome the defects of patch distribution in the traditional Kriging interpolation,and the prediction results contain the spatial information of various auxiliary factors.Random Forest Kriging can more accurately capture the complex relationship between soil nutrients and quantitative environmental factors.In vision,the maximum degree of soil nutrient distribution is close to the natural distribution of soil nutrients,so it is more accurate.