Study On Spatial Prediction Model And Sampling Point Optimization Of Soil Organic Matter In Disturbance Area Of Coal Mining

Taking the coal mining disturbance area in Changhe River Basin as an example,by determining the soil sampling scheme and the series of sampling points in the study area,combining the two simulation methods of Ordinary-Kriging and Sequential Gaussian Simulation,the spatial interpolation optimization of organic matter for each sample series under each sampling mode is carried out,and different interpolation models are verified.The precision of the simulation results and the reasonable sampling number under different sampling modes are important for the optimization of spatial prediction model of soil nutrients in coal mining disturbance area,the rational selection of sampling points,and the implementation of dynamic monitoring management and ecological restoration.1.The descriptive statistical characteristics of soil organic matter content in different sample sizes under different sampling modes are similar,but the spatial information expressed by sample sizes under different sampling modes is significantly different,and the complete expression of local spatial information is also significantly different.With the change of sample size,the variation structure of organic matter content is irregular.Latin hypercube sampling(LHS)can better express spatial variability of soil nutrients than SG(system grid sampling),SR(simple random sampling)and VQT(variance quad-tree sampling).Therefore,the optimal soil sampling pattern may be more important than simply increasing the sample size.2.Based on the different interpolation simulation methods adopted by different sampling modes and sample series,after eliminating the salient features of some sample series,the RMSE of the four sampling modes after interpolation has little difference,basically simple random sampling and system grid sampling are larger than variance quad-tree sampling and Latin hypercube sampling.Taking the size of RMSE value as the precision standard,72 and 42 sample series are less than 117,28 and 21 as a whole.At the same time,the change characteristics of MRE value and RMSE value tend to be consistent.Comprehensive analysis shows that 28 and 21 sample series are difficult to meet the accuracy requirements of organic matter spatial prediction map,while 72 and 117 sample series are more complete in the expression of spatial information.At the same time,the different sampling points selected by different interpolation simulation methods in different sampling modes have significant differences in the expression of spatial information in the study area,and there is no optimal sample size in different sampling modes.3.According to the predicted distribution map,the local variation is not obvious with the decrease of the number of sample points under the condition of Gauss sequential simulation,which is consistent with the spatial variation expression of the original data.At the same time,the optimal sampling number of each sampling mode is also different.According to the analysis of RMSE and MRE values,the Gauss sequential simulation method has more advantages than the ordinary Kriging interpolation method in expressing local spatial information with fewer samples.