Thiessen Polygon Coupling Interpolation To Optimize The Spatial Prediction And Sampling Strategy Of Site Contamination

At present,there is a large amount of contaminated land in our country,and it is very costly to carry out the investigation and remediation of the contaminated land.An important part of site survey is to identify site pollution status and remediation scope by collecting soil pollution points and spatial interpolation method is often used in this process.Due to the large external interference of contaminated sites,the collected pollution data usually have high bias and high dispersion.The commonly used geostatistical interpolation methods and deterministic interpolation methods have their limitations when applied to the spatial prediction of contaminated sites.At the same time,with the deepening of site investigation,scientific and reasonable sampling strategy design will play an increasingly important role in the process of site survey.It is necessary to put forward a set of methods suitable for site sampling from the aspects of sampling density and sampling procedure.Based on the highly biased pollution data of contaminated sites in the study area,this study proposed a grid composite sampling method combined with Thiessen polygon coupling interpolation method for site pollution spatial prediction,making the prediction result in a small number of sampling points,not only can identify the accurate pollution range,but also maintain good prediction accuracy.The main conclusions of this study are as follows:(1)The existence of a few high peaks and over-standard points in the study area caused high bias in the overall data,and Arsenic pollution has a strong discrete characteristic in the study area.The average value of 164 As pollution data in the study area is low and the range is large.When 20 mg/kg is used as the screening value,there are only 7 points that exceed the standard,but the skewness and kurtosis reached to 6.41 and54.75,and the K-S test value was less than 0.05.The overall data had high bias.After removing the 7 over standard points,the skewness and kurtosis decreased to 0.72 and 1.56,respectively,and the coefficient of variation decreased to 36.26%.The K-S test value was greater than 0.05,which conformed to the normal distribution.From this,it can be judged that the high bias of the pollution data is caused by a few over-standard points.At the same time,due to the high dispersion of site data,different sampled data sets will produce different spatial correlation analysis results,so the data structure characteristics must be analyzed and evaluated before interpolation analysis,especially geostatistical interpolation analysis.(2)Through single interpolation analysis of as pollution data in the study area,it is found that the Ordinary Kriging method(OK)has the highest cross-validation accuracy,but its smoothing effect is strong,so it can not be used as a tool for pollution range prediction.The deterministic interpolation method can predict the corresponding pollution range in the study area,but the prediction accuracy is lower than that of ordinary kriging(OK).Based on the pollution spatial distribution and cross-validation results,the inverse multiquadric function method(RBF＿IMQ),the completely regularized spline function method(RBF＿CRS),and the inverse distance weight method(IDW)are suitable as interpolation methods for the study area,and RBF＿IMQ has the best prediction effect,among which RBF＿IMQ has the best prediction effect.(3)Using Thiessen polygons to partition the study area,which is divided into the polygon area where the non-exceeding points are located and the polygon area where the over-standard points are located.The non-exceeding points area retains the ordinary kriging(OK)interpolation results,and the over-standard point areas retains the deterministic interpolation results.Through cross-validation and pollution spatial distribution analysis,it is found that the use of Thiessen polygons for partition coupling interpolation can effectively improve the prediction accuracy and cross-validation accuracy of pollution spatial distribution in the study area.The best coupling interpolation is the OK-RBF＿IMQ method formed by the combination of ordinary kriging(OK)and the inverse multiquadric function method(RBF＿IMQ).(4)Grid plus professional judgment placement method supported by coupled interpolation method can obtain better pollution prediction effect with fewer sampling points.In this study,sampling data at different grid sampling intervals was obtained through spatial simulation sampling.Through the analysis of data structure,it is found that40 m × 40 m is the minimum spacing for grid sampling.The best coupling interpolation method OK-RBF＿IMQ method is used to interpolate the sampling points of different grids.Comprehensive prediction accuracy and economic cost,35 m×35 m is the best initial grid point spacing.On this basis,the professional judgment is carried out to increase the distribution of points density,and the coupling interpolation analysis is carried out again.The prediction accuracy and spatial distribution prediction results are better than the results of all the points in the study area using the RBF＿IMQ method interpolation alone.At this time,only 93 points are used.Compared with the 164 points in the study area,61 points are reduced,and the reduction rate is 37.20%.