Multi-scale Spatial Autocorrelation Analysis And Simulation Of Land Use Patterns In Habahe County

As the major course and the principal constitution of the global environment change,changes in Land Use and Land Cover Change is also a factor from which we regulate our behavior and make proper decision. These involve a series of important issues.Including efficient exploitation of resource, protection of environment, maintenance of farmland, safety of food and sustainability of our social and economic development. So, in the fields of geography and others related, Land Use and Land Cover Change has been a hot spot recent years.Also,the problems of data and scale are very import in many all-around researches at all times. For now, it is critical to future development for us to research the relation between multi-scale of land use and their motivation and to unveil the mechanism.Habahe country which lays in the middle and lower reaches of the Ertysh River is a backward area with rich deposits but a fragile ecosystem. The sustained increase of economics there leads to a expansion in population, which cause several changes of amount, quality, type and structure in land use and have arouse more contradiction among land, stock and environment. Great devastation has been made to the brittle arid and semi-arid desert ecosystem .So, it very important to manage land use in a rational way. Still, only when we have a complete view of the process of land use change and thus expose the true driving forces can the dynamic model of vast spatial scale of land use been understood driving the research ahead.We choose Habahe country as the study area to analysis and simulate multi-scale spatial autocorrelation of land use patterns. We used the multi-scale spatial autocorrelation to analysis the space autocorrelation between land use types and land use driving factors at different scales in the study area. And also ,we achieved the study of using multi-scale spatial autocorrelation of statistical methods in land-use distribution.Secondly,Selected a certain simulation study of driving factors of land-use patterns, and to consider different scales of land-use change driving factors acting on different levels of land-use pattern, established a multi-scale Logistic regression model on the basis of a clear spatial location, comprehensive integrated multi-scale. At last, in determining the best analog scale in study area, we used the AutoLogistic regression model which is considering spatial autocorrelation to simulate land-use patternsThe main conclusions of this research are as follows:(1) Spatial autocorrelation degree of various land use types and land use driving factors reduced as the increasing of the weight; from the small scale to the large scale, space autocorrelation of Moran's I of different land use types in Habahe County increased by the enhancing of the study scale.(2) Through a multi-scale Logistic regression analysis showed that the impact of land use pattern-scale spatial factors showed effects of different analog scale, the selected variables are different. And variables for describing the land-use type and driving factors of theβcoefficient of the variable relations also have some differences. Found that when analog scale is in the 500m ,the goodness of fit of the simulation model are up to the highest. Therefore, we can determine the 500m×500m is the best-scale to build simulation model in the study area land use pattern.(3) In the 500m×500m scale,local class ROC values are: cultivated land is 0.914, forestland is 0.736, grassland is 0.742, other land is 0.855, waterbody is 0.931, construction land is 0.924, all the goodness of fit are greater than 0.7, showed that we have a better fitting results. We thought that the AutoLogistic regression model which is considering spatial autocorrelation has strong applicability and explanatory power, it is relatively reasonable to simulated pattern of regional land-use change, and it is also a simulation model of land use which is worthy to be applied and promoted.