Identification Of Spatial Interpolation And Level Classification Of Cultivated Soil Nutrients

Soil nutrients affect not only crop growth,but also yield and quality.However,traditional soil testing methods are mostly based on random point sampling and obtain nutrient content through laboratory tests.This"point"instead of"face"method,due to the randomness and subjectivity,seriously affected the accuracy and integrity of cultivated soil fertility assessment.The development of precision agriculture and smart agriculture urgently requires the objective evaluation of soil nutrients.The earth observation technology represented by Remote Sensing(RS)and the spatial analysis technology represented by Geographic Information System(GIS)provide important technical guarantee for rapid,efficient and quantitative evaluation of soil nutrients.This paper selects Tongzhou District of Beijing as the research area.Statistical analysis and spatial interpolation of 402 soil nutrient measurement points were collected.The secondary development technology of ArcGIS Engine was used to design the soil nutrient spatial interpolation software,and five spatial interpolation algorithms were implemented;Based on the time series Landsat-7 and Landsat-8 OLI images,the spatial distribution and area of cultivated land at the parcel scale were extracted;According to the comprehensive index of soil nutrient grade evaluation in Beijing,the soil nutrient grade division at the parcel scale was realized.The main research contents and conclusions are as follows:(1)Comparison of Soil Nutrient Spatial Interpolation Methods and Software Design.Selected four interpolation methods:IDW(Inverse Distance Weighted),OK(Ordinary Kriging),Spline Function Method and Trend Surface Method to perform spatial interpolation on the slow-acting potassium.The ME(Mean Square)and the RMSE(Root Mean Square Error)are used to evaluate the interpolation accuracy.Among them,the semi-variation function type of OK selected three models of spherical function,exponential function and Gaussian function.Through comparison,it was found that the accuracy of the exponential function model was higher than the other two models,RMSE and ME were 176.71 and-1.13,respectively.The Spline includes a regular spline function method and a tension spline function method.The RMSE and ME of the tension spline function were the smallest,being 0.23 and 178.89,respectively.The end results showed that the OK interpolation method had the smallest RMSE,and the exponential function model had the best interpolation accuracy.Based on the ArcGIS Engine framework,a soil nutrient spatial interpolation software was designed to implement five spatial interpolation algorithms:IDW,OK,Spline,Trend and Natural domain method.OK interpolation includes spherical,circular,exponential,Gaussian,and linear interpolation models.Spline include two types of interpolation models:regular spline function and tension spline function.Trend uses polynomial regression analysis.There are two types of polynomial regression and logical trend surface.After the spatial interpolation was completed,the software could also count raster cell values and include auxiliary functions such as map browsing and right-click menus.(2)Remote sensing extraction of cultivated land blocks based on time series Lansat-7and Landsat-8.Based on time-series remote sensing images of Landsat-7 in 2009 and Landsat-8 in 2014 and 2019,they were pre-processed by radiation calibration,atmospheric correction,image fusion,etc.,and subjected to ISODATA classification and NDVI calculation.Multi-source remote sensing data was constructed by band combination,and land use classification was carried out using CART(Classification and Regression Trees)decision tree classification.The results showed that the overall accuracy in 2009,2014,and 2019 was 83.23%,85.54%,and 88.57%,and the Kappa coefficients were 0.7774,0.8171,and 0.8510,respectively.So that it could be used for the evaluation of soil nutrients grade in cultivated land.Statistic and analyzed of classified cultivated land.The results showed that the area of arable land in 2009was 462.37 km~2,and the area of arable land in 2014 was 463.907 km~2.The area of arable land has not changed much in five years.The reduction to 398.433 km~2 in 2019 indicated that the area of cultivated land has changed greatly.(3)Classification of soil nutrients at the parcel scale.Organic Matter(OM),Total Nitorgen(TN),Available Phosphorus(AP)and Available Potassium(AK)were selected for spatial interpolation of OK.Using the ArcGIS raster calculator to calculate the comprehensive index of cultivated soil nutrients,and using the cultivated plots extracted in 2009 as a mask,the soil nutrients in the study area were classified as low,medium,and high.The area proportions were 2.32%,92.02%,5.66%,respectively.The statistics of the soil nutrients in the arable land of each township and three levels of arable land were conducted.Concluded that Xiji Town had the highest OM content,with an average value of 19.04 g/kg;Majuqiao Town had the highest TN content,with an average value of 1.09 g/kg;Songzhuang Town had the highest AP content,with an average value of 70.56 mg/kg;the Ak content in Lucheng Town was the highest,with an average of 162.96 mg/kg.The low-grade arable land accounts for the largest proportion in Huoxian Town,with an area of 6.059 km~2,accounting for 7.05%of the town,and 1.31%of the total district;The largest proportion of medium-grade in arable land was Yongledian Town,with an area of 65.039 km2,accounting for 99.3%of the town and 14.06%of the total district;the largest high-grade arable land area was Lucheng Town,with an area of 8.581 km~2,accounting for 20.39%of the town,and 2.59%of the district.This article was based on soil nutrient sample data of cultivated land and remote sensing data of Landsat-7/Landsat-8 in Tongzhou District,Beijing.The spatial interpolation software for soil nutrients was designed by comparing the spatial interpolation methods of SAK;Based on the time series remote sensing data of Landsat-7/Landsat-8,the cultivated land plots in the study area were extracted;By calculating the comprehensive index of cultivated land nutrients,the spatial mapping of soil nutrients on the plot scale was carried out,and the spatial expansion of"point"to"face"of soil nutrient monitoring was realized.