Research On Spatial Interpolation Methods Of Reclaimed Soil Organic Carbon And Optimization Of Monitoring Samples

Through optimizations of soil sampling plan and spatial interpolations, we can get the basic characteristics of reclaimed soil reconstruction in mine areas, and it has important theoretical and practical value to monitor and evaluate reclaimed farmland.Firstly, this study compared multiple linear regression, inverse distance weighting method, ordinary kriging method and regression kriging to predict the spatial distribution and variation of soil organic carbon in Pingshuo mining area. Predicted results were validated by Pearson’s Correlation coefficient, RMS error, Root mean Square Error and Accuracy. Results showed that:(1) The spatial variation coefficient of soil organic carbon is high in reclaimed dumps of opencast coal mine, reaching up to 145.408%, which belongs to thehigh level;(2) According to accuracy, regression kriging is the best method, its Pearson’s Correlation coefficient reaches 0.984; RMS error, Root mean Square Error and Accuracy are-0.012, 0.211, 0.991 respectively. Multiple linear regression takes second place; The effect of inverse distance weighting method and ordinary kriging method is the worst.(3) According to the expressed details, regression kriging is more meticulous, which can also depict the beating and uncertainty tendency of the data; Inverse distance weighting method and ordinary kriging method have obvious smoothing effects: the spatial variation is continuous.In addition, group the closely related environmental factors-slope and optimized vegetation adjustment index, and use Fuzzy C-Means to optimize the existing soil sampling points. Finally, 67 representative sample points are selected from 162 sample points. The total spatial distribution of soil organic carbon was generated of from the two groups of soil samples by the inverse distance interpolation. Predicted results were validated by Pearson’s Correlation coefficient, RMS error, Root mean Square Error and Accuracy. It shows that accuracy evaluation indexs are close, and the forecast accuracy is basically the same.(1)The RMSE and RMSE of the two groups of soil samples are highly consistent, the two values are equal to 1.125;(2) The correlation coefficient of Pearson using representative samples is 0.240, and the correlation coefficient of Pearson using all the training data was 0.246, which was slightly higher than that of the former;(3) The ME and AC using all the training data are respectively 0.252 and 0.318, the ME and AC using representative samples are respectively 0.271 and 0.303, the former is slightly higher than that of the latter, but the gap is very small.This indicates that Regression Kriging method is the most suitable interpolation method to predict soil properties in reclamation area and using fuzzy clustering is feasible to optimize the soil samples in reclamation area.