Spatial Partition Method Of Soil Environmental Quality Based On Gaussian Mixture Model

Soil environmental quality reflects the joint action of natural processes and human activities,showing that soil environmental quality is similar within regions and different between regions.Soil environmental quality monitoring survey data is a sampling of regional environmental quality,which contains the common or interactive effects of natural factors and human activities.Clustering method is often used to classify and partition the environmental quality of sampling points,which is the mapping of regional environmental quality after classification in spatial domain.The probability distribution of heavy metal content in sampling points often shows the characteristics of right skewness distribution.In the traditional clustering analysis process,random initialization of clustering centers is easy to fall into local optimal solution,which leads to poor regional environmental classification and zoning effect.Based on this,this paper puts forward a method of soil environmental quality zoning based on Gaussian mixture model.Taking the research area of Shunyi District,Beijing,the main research work and achievements of this paper are as follows:(1)The natural and man-made influencing factors related to the formation of soil environmental quality were analyzed,the comprehensive characteristics of soil environmental quality were obtained by principal component transformation at sampling points,and the main factors/factors affecting environmental quality were extracted by geographic detectors.(2)Initialize the Gaussian mixture model based on the classification of the above factors,improve and construct the Gaussian mixture model based on EM algorithm,and realize the comprehensive zoning method of soil environmental quality based on attribute characteristics according to the strategy and solution of preliminary zoning adjustment of soil environmental quality.(3)Drawing lessons from landscape ecology,comparing and analyzing Simpson evenness landscape pattern index and landscape aggregation index of industrial and mining land and high-intensity agricultural land under traditional K-means classification and self-organizing neural network(SOFM)clustering zoning methods,and exploring the spatial pattern between environmental quality zoning and leading environmental influencing factors.