Study on the Method of Areal Differentiation Based on SOFM

郝成元,吴绍洪,李双成
2008-01-01
Abstract:In the south of Yunnan Province in China,there are Mt.Laobie,Mt.Bangma,Mt.Wuliang and Mt.Ailao,which influenced its climate and vegetation.And the study area is commonly controlled by the Indian monsoon system and East Asian monsoon system,i.e.,it is one of the main areas where the two summer monsoon systems merge.Therefore,the obvious regional difference of spatio-temporal heterogeneity of vegetation is induced by the complicated topographical terrain and the monsoon climate system,which cause various river hydrologic characteristics,soil types,vegetation types,etc.That is,these unique microhabitats formed by such kinds of combination of complicated topography and subtropical warm-humid air current are beneficial to differentiation and variation of local species to a certain extent,thus leading to even more diverse vegetation type.Clustering analysis based on self-organizing feature mapping(SOFM) network is a new unsupervised clustering method that develops from neutral networks.In this paper,the neural network has been trained to perform complex functions in various fields of application including elevation,temperature,precipitation,wind speed,active accumulated temperature,evapotranspiration potential and enhanced vegetation index(EVI) at 30 meteorological stations.The result reveals that Ailao Mountain is such a firm barrier blocking the cold air coming from northern into southwest mountain region,and such a possible boundary between summer southwestern monsoon and winter northeastern monsoon of our country that it becomes a demarcation boundary of climate sort between west and east regions.This way,SOFM network is used in areal differentiation study of ecological geographical,is a rather good comprehensive physical regionalization method.For it can reflect the similarities and differences of different areas near the basic boundaries,and reveal a continuous process from quantitative changes to qualitative changes.
What problem does this paper attempt to address?