Strategy For Optimal Interpolation Method And Efficient Sampling Of Upland Soil Carbon And Nitrogen Based On Spatiotemporal Variation Of The Soil

The spatial distribution of Soil Organic Carbon?SOC?and Total Nitrogen?TN?has time-series differences.To clarify the response of dynamic changes of Soil Carbon and Nitrogen to the appropriate interpolation method and sampling quantity under the time scale is the basis for the formulation of efficient field investigation strategies.In this study,a tract of upland?3.93×10⁴ km²?in North Jiangsu was delineated and selected as a case area,using the measured sample data of the second national soil survey in 1980 and the soil formula fertilization project of the ministry of agriculture and rural affairs in 2008,combined with 1:50000 high-precision soil vector database,analyze the prediction accuracy of 8 deterministic interpolation methods?Inverse Distance Weighted?Global Polynomial Interpolation?Local Polynomial Interpolation?Thin-plate spline?Spline with tension?Completely regularized spline?Multiquadric function?Inverse multiquadric function?and 4 geostatistical interpolation methods?Ordinary Kriging?Simple Kriging?Universal Kriging?Disjunctive Kriging?for soil organic carbon and total nitrogen content?0?20 cm?in different periods,and select the optimal interpolation model.On this basis,20 random sampling sites with different number of points were set up,and the prediction accuracy of each sample density data set of organic carbon and total nitrogen in different periods was evaluated by using the optimal interpolation model,in order to quantify the relationship between prediction accuracy and sample number.The results of this study can provide a theoretical basis for the formulation of a reasonable field soil investigation program and the selection of mapping methods for the future third soil survey in north jiangsu and China.The main results are as follows:?1?Under different interpolation methods,correlation coefficient?r?and root mean square error between the predicted value and the measured value of soil organic carbon in the drylands of northern Jiangsu in 1980 varied in the range of 0.186?0.570 and 2.077?2.564 g·kg^-1,and in 2008 in the range of 0.360?0.595 and 2.173?3.344 g·kg^-1,respectively,the optimal interpolation methods for the two periods are Inverse multiquadric function and Ordinary Kriging.The correlation coefficient?r?and root mean square error between the predicted value and the measured value of the total nitrogen content in the drylands of northern Jiangsu in 1980 were between 0.145?0.485 and 0.293?0.656 g·kg^-1,and between 0.205?0.878 and 0.133?0.611 g·kg^-1 in 2008,respectively,the optimal interpolation methods for the two periods are Completely regularized spline and Ordinary Kriging.This indicates that for the same region,the optimal interpolation methods for soil organic carbon and total nitrogen at different time scales are different,and the suitable interpolation methods for soil organic carbon and total nitrogen at the same time scale are also different.Therefore,in the future spatial prediction of soil properties,it is necessary to comprehensively consider the spatial and temporal variation characteristics of soil properties and the calculation principle of the interpolation model,so as to select the appropriate interpolation model.?2?Based on the optimal interpolation method selected in the above different periods,correlation coefficient?r?and root mean square error between the predicted value and the measured value of soil organic carbon in the drylands of northern Jiangsu in 1980 from a data set of 20randomly sampled sampling sites varied in the range of 0.083?0.568 and2.077?2.636 g·kg^-1,and the variation of total nitrogen was between0.041?0.485 and 0.292?0.335 g·kg^-1,respectively.When the number of soil organic carbon and total nitrogen samples is greater than 75%and 95%respectively,the prediction accuracy is high and can reach a relatively stable level,and the optimal sampling number is 563 and 585,respectively.Under 20 sets of random sampling sites,correlation coefficient?r?and root mean square error between the predicted value and the measured value of soil organic carbon in the drylands of northern Jiangsu in 2008 were 0.041?0.458 and 2.109?2.620 g·kg^-1,respectively.In the meantime,the variation range of total nitrogen is between0.023?0.629 and 0.216?0.289 g·kg^-1.When soil organic carbon and total nitrogen are at 70%and 85%,the number of samples can reach a relatively stable level,and the optimal number of samples is 526 and 524respectively.This shows that the spatial prediction accuracy of soil organic carbon and total nitrogen in different periods responds differently to the change in the number of samples.The greater the spatial autocorrelation of soil properties,the more sensitive the prediction accuracy is to the number of samples,and the fewer the number of samples needed to reach saturation of the spatial information.In addition,it was also found in this study that to set up enough sampling sites in key areas either high or low in soil organic carbon content is also one of the important means to improve effect of the spatial prediction of soils.