Application Of Simulation Scale Selection And Raster Map Similarity Evaluation Method

Subject on scale has been a hotspot, which has been paid close attention in landscape ecology, LUCC and GIS. Recently, with the update of theories and idea in many subjects, mutil-scale effects gradually emerge. It’s necessary to strength the study of those theories. The article summarizes all the practical problems that solved by existing mutil-scale theories and research status on mutil-scale method by reading plenty of articles at home and abroad. There are landscape metrics method, spacial statistic methods and "flow water model" method. Landscape metrics method is better applied to the landscape which is a combination of all kinds of batches with the properties of size and shape. Spacial statistic method more suit for the condition that the landscape is described by regular arrangement grids. And both methods cannot avoid the problem that one value may corresponding to one more landscape patterns. The "flow water model" can lighten the effect of this problem to some extent, but its object of study is same to the landscape metrics method’s. So there have not a method for the grid landscape on the premise that lighten or overcome the problem that one value may corresponding to one more patterns. Based on the limitations of these theories, a similarity evaluation method which is based on grid map is put forward in this paper. We do some studies on how to choose scales in research with this method both in theory and practice, whose purpose is to find the relationship between indexes of patches and domains of scale. All of these provide convenience for future research. The details are as follows:(1) We put forward a similarity evaluation method serves for grid data. The idea of this method is as follows: Overlap one map on another to get a new map whose attributes is the difference value between attributes of two maps, replacing the difference value with similarity value later. The similarity of whole map is the mean of all similarity values. We choose weighted mean here and the weight value is the area ratio of each similarity represents area and the total area, which obeys the principle that large area has a large contribution to the whole map.(2)We put the similarity evaluation method into mutil-scale choose. Grid data has a obvious advantage in application, with a great deal of grid data, data in which scale should be used to simulate the landscape, a combination of all kinds of batches with the properties of size and shape. This is the problem we should discuss. Take land use dynamic degree in fangshan district as an example in paper, some indexes are selected for the study of relationship between domain of scale and these indexes. These indexes are middle value of patch numbers, mode of patch numbers, average value of all area values of patches, middle value of all area values of patches. In general, the cases of patches distribution in fangshan are divided into uniform and no uniform. The result shows that:(1) we can choose domain of scale based on middle value of patch numbers and middle value of all area values of patches when patches are uneven distributed. (2) Average value of all area values of patches plays a important role when each patch is evenly distributed. This conclusion is reasonable and we can use these indexes to choose the domain of scale in the future research.