Optimization And Application Of Soil Samples For The County’s Cultivated Land Quality Evaluation

Cultivated land is the foundation of food production.With the development of economy,cultivated land resources have been consumed by the rapid expansion of urban construction,and its reduction has become one of the most critical problems in China.Therefore,China proposed the strictest cultivated land protection system in the world.The cultivated land management is more focus on the management of quality and quantity,not just the quantity management.After evaluating the cultivated land quality several times at a country scale,China has established a relatively well-developed cultivated land quality classification database.However,at present the efficiency,accuracy and refinement of cultivated land quality evaluation is well below the requirements for the management of land resources.There are still some problems of evaluate methods expecting to be further explored.Soil,as a key factor in cultivated land quality,the rational use of its sampling and data is an important approach for improving the efficiency and accuracy of cultivated land evaluation.In here,in order to improve the evaluation refinement of regional cultivated land quality,and to ensure the implementation of the regional quantity and quality management of cultivated land,the paper studied the optimization and application of soil sampling points and data.This paper takes Donghai County,Jiangsu Province which is the typical agricultural county of Yellow river-Huaihe river Plain region and the cultivated field quality monitoring pilot base of ministry of land and resources as the study area.A large number of soil sample data is collected from cultivated land quality grade monitor(140),multi-target geochemical survey,the land productivity investigation and other data.The aim of this research is to improve the efficiency and accuracy of cultivated land quality evaluation.Based on the spatial heterogeneity and density characteristics of soil samples distribution,the geostatistical method,soil-landscape model,simulated annealing algorithm and other methods are taken to research the optimization of soil samples and data of cultivated land quality evaluation with the support of RS,GIS and Matlab programming technology.Based on the analysis of the relationship between soil properties prediction accuracy and sample density,according to the sample density zones,the following three aspects are explored:forecast optimization of soil attributes of scarce soil samples region;selecting optimization of representative samples of soil attributes;assignment methods optimization of soil attribute data of evaluation units.Then,evaluated the cultivated land quality by comprehensively using different optimization methods,and analyzed the accuracy of the evaluation results.The main conclusions of this paper were as follows:(1)Based on the relationship between prediction accuracy of soil properties and sample density,we divided three sample density zones,which provided a basis for the method selection of optimization and application of soil samples.The integration of 1300 soil samples from different sources together can effectively improve the prediction accuracy of soil properties,but the prediction accuracy and sample density was unevenly distributed in space.There was a change inflection between the prediction accuracy and sample density,when sample density reached the inflection point value,the prediction accuracy will no longer improve with the sample density growth.The sample density inflection values SOM,pH,clay content and topsoil thickness are 0.24,0.20,0.30,and 0.16 samples per square kilometer.According to the inflection point values,the study area is divided into the sample intensive area,the sample infrequent area and the sample deficient area,and the area scale is 85.16%,9.51%,5.33%respectively,which build a foundation of soil samples and data optimization according to the partition of sample distributing characteristics.(2)Based on a large number of sample data from samples intensive areas,using soil samples deleted region-landscape model,it’s available to get soil attribute prediction data with high accuracy that arable quality evaluation required in samples scarce area.Soil-landscape model’s accuracy R2 predictions of four soil properties were all above 0.7000,in which prediction accuracy of soil organic matter was the highest,with precision fitting equation’s R2 reaching 0.8419 and clay content,topsoil thickness,pH’s R2 reaching0.7729,0.7627 and 0.7139 respectively.Comparing soillandscape model to geostatistical interpolation’s predictions,Soil-landscape model prediction accuracy was significantly higher than geostatistical interpolation in samples scarce area and relative error of prediction was smaller over 30%than geostatistical interpolation.However,the accuracy of geostatistical interpolation was slightly above soil-landscape model in samples intensive areas and relative error of prediction was smaller about 10%than soil-landscape model.(3)According to the simulated annealing algorithm which is consideration of the spatial variability of soil properties,we could attain a minimum number of soil samples and the optimal spatial layout that match the pattern of the spatial variability of soil properties in study area.Thus,the efficiency and accuracy of cultivated land quality evaluation could be improved.In order to improving the simulated annealing algorithm,we added the spatial variability of soil properties as parameters into the model and retained the sample points with strong spatial variability in the simulated process.After annealing optimization,the number of soil samples that can best represent the soil properties such as organic matter,pH,clay content and topsoil thickness in point-intensive areas was reduced from 1300 to 178,72.315 and 70,respectively.Though the number of the soil samples was significantly reduced,the statistical characteristics of the original data and the spatial distribution pattern characteristics of soil samples were remained.Compared to the original data,the precision of optimized samples was improved about 5%.(4)By selecting the optimal spatial interpolation method in conjunction with the optimized selection of unit assignment methods,we can compute the optimal assignment of soil property data in cultivated land quality evaluation from the point to the plane and from plane to the evaluation unit.By comparing the results of the four spatial interpolation methods,Co-Kriging,Ordinary Kriging,Radial basis function and Inverse Distance to a Power,we can determine the best spatial interpolation method for soil properties in the four methods:the best method for organic matter is Co-Kriging(elevation,exponential model),and the standard prediction RMSE is 0.8978,increased from 3.0%to 8.6%comparing with other methods;the best one of pH is Ordinary Kriging(spherical model),and its RMSSE is 0.7848,increased from 0.4%～7.3%comparing with other methods;the best method for clay content is Radial basis function interpolation(inverse multi-quadric functions),and its RMSSE is 0.6445,increased from 3.7%to 3.9%over other methods;the best of topsoil thickness is Radial basis function interpolation(completely regularized spline),and its RMSSE is 0.7785,increased from 9.2%to 25.8%comparing with other methods.On this basis,and then optimizing the selection unit assignment methods,we can further improve the accuracy of soil properties in cultivated land quality evaluation unit.Among them,the accuracy of assigned four soil properties by homogeneous unit assignment is 0.8118～0.8313,and the accuracy can reach 0.8697～0.9042 by area weighted evaluation method.(5)The comprehensive optimization prediction of soil properties in samples sparse area and deletions area,optimization select of representative samples in soil samples intensive areas,and optimization of the evaluation unit assignment methods,we can further improve the accuracy of arable land quality evaluation in the study area.Papers refers to "agricultural land quality grading regulations"(GB/T 28407-2012)natural quality evaluation methods and parameters systems,and evaluates the quality of arable land in the study area on the basis of the optimization and application of integrated soil samples and data,and the accuracy(R2)of the results can be up to 0.8958,significantly higher than the accuracy(R2)of the results of Supplement and Improvement the Quality Evaluation of cultivated land 0.7319,and it further improve the accuracy of cultivated land quality evaluation.