Study On The Driving Forces Of Land Use Based On Spatial Autocorrelation Model With Different Spatial Matrices

Based on large quantities of the cases of coal mine water-inrush,theLand Use and Land Cover Change-LUCC , is the main causes and important components of global environmental change, It also provides important scientific decision-making for human behavior, It involves many aspects, such as the development and utilization of resources,the protection and development of cultivated land, the sustainable development of social economy, the protection and management of environment and so on. In the existing study of the land cover change patten, the traditional model analysis methods are based on the Data and the Data are always independent and normal distribution. However,the space Data can not be independent and there must be certain relevance between them which need to be solved witn some proper method. To this end, spatial autocorrelation analysis has become the focus of attention, and resulting in a number of spatial autocorrelation measurement parameters. Moran's I index is one of the earliest and most widely used parameters.For a clearer explanation of the spatial relationships between the overlying classes, the adjacency relation of each spatial phenomenon needs to be defined before starting study. Spatial weight matrix is the main expression mode to express the adjacency relation. With the development of spatial correlation and further research in this field,more definition about spatial weight matrix are brought. Meanwhile in the current research, most analysis result are based on only one matrix.To this degree, it is necessary to study the effects of different spatial weights on spatial correlation and spatial clustering.In this paper, Shandong province was selected as the research area,and more than one kind of spatial weight matrix are used into this analysis on correlation of different cover land types. There are three main weight matrices, including distance threshold weight, rook matrix and queen matrix. The Moran's I exponents of 6 geodetic classes are calculated and the spatial regression model of each land class was established by introducing the influence factors. Here are the conclusions:1) Every land class have spatial autocorrelation between them, no matter what kind of matrix are used.2) If the coefficients are different, even in the same space matrix,Moran's I index has different result. For the distance threshold weight,the distance selection does not produce the minimum value of the island phenomenon, the Moran s I index is the largest, and getting smaller and smaller with the increase of threshold setting. The Rook weights and Queen weights are in the first order to the maximum value.3) The residuals of the classical regression model have spatial autocorrelation , which indicates the necessity of using the autocorrelation model. In the study of autocorrelation model, the spatial lag model is more representative...