Research On Terrain Landscape Scale Similarity Characterizing Methods Based On Grid Dems--a Case Study In Shannxi Province

Analysis scale is one of the five foundmental scale issues in DEM data analysis, data interpretation and application. The methods for characterizing and analyzing such basic characteristics like self-similarity, hierarchy and multi-scale properties possed by the landform morphology become difficult and important issues. The now existing researches of multi-scale terrain analysis are either limited by the data sources or by the methods used for landform morphology analysis, thus these researches can not fulfil requirements of characterizing landform morphology scale similarity. In the thesis, a series of DEMs in the whole Shannxi province are selected as the source data. The DEMs cover the whole Shannxi province with671DEMs of25m-resolution. These DEMs guarantee the complexity, diversity, integrity and availability of the landform data. A series of Fourier frequency domain based methods are employed to analyze the terrain data sets. The methods include fouries spectrum analysis, continuous wavelet analysis and spatial point pattern analysis. These methods are investigated to analyze the similarity characteristics and their spatial distributions of the loess landforms, the spatial point patterns and their spatial distribution in the whole area over different scales. The paper adopts the similarity to landform classification and the accuracy evaluation.The main contents and achievements of the thesis are as follows:(1) Research on landform scale similarity by Fourier spectrum analysisProfiles are the most intuitive reflection of fluctuations of terrain features and landform morphology. The thesis proposes a method based on Fourier spectrum analysis and its second order statistics the profiles for landform scale similarity characterizing. Experiments in the whole area of Shannxi province show that:there is obvious linear relationship between the Fourier power spectrum and the frequencies of the landform profile. The linear relationship reflects the landform scale similarity. Experiments show that landform scale similarity is high in the northern area of loess plateau, Baoji and Hanzhong Basin belts and the Qin mountain zones of Shannxi province; while the landform scale similarity is low in the inbetweens areas of loess and sandy landforms in the northern area of Shannxi province, Guanzhong Basin and the Hanzhong Basin. This spatial distribution of landform scale similarity has high coincidence with the landform complexity.(2) Research on landform scale similarity by continuous wavelet analysisThe wavelet transform modules maxima reflect landform scale similarity. Firstly, this thesis designs a workflow for one dimensional profile data acquisizition and preprocessing. Then the thesis proposes a new model which could be used to characterizing landform scale similarity by construction and analysis of the wavelet transformation scale representation—scale space. Finall comparisons among different mother wavelets in landform scale similarity characterizing are investigated. The experiments in the Shannxi province show that there exists high correspondence by different wavelets in different test areas. The experiments show that the landform scale similarity characterized by two dimensional profiles Fourier spectrum analysis is more obvious the the one dimensionl cases. The mother wavelets used for landform scale similarity characterizing could be departed into two kinds. The first kind wavelets include mexican2d, esmex2d, wheel2d, dergauss2d, gaussz2d, dog2d, sdog2d, sqdog2d and endstop2; while the other kind of mother wavelets includes es2cauchy2d and endstopl. The first kind performs bettern than the second ones. The two dimensional wavelet transform modoule maxima not only characterize landform scale similarity, but their spatial scale as well.(3) Research on spatial distribution regularity of landform scale similarity in Shannxi provinceBy comparing the results obtained by the methods in the thesis, namely fourier,2D continuous wavelet transform (CWT), the thesis investigates the spatial distributions regularity of landform scale similarity in the whole Shannxi province. The experiments show that there are orderly spatial variations of landform scale similarity. There is high coincidence between results achieved by Fourier spectrum analysis and continuous wavelet transformation analysis. They all trend higher in the gully and hilly area in the northern loess plateau, low and hilly area in southeastern part of Shannxi province. While the landform scale similarity trends low in areas like Hanzhong&Ankang Basin, Guanzhong Basin and inbetween areas of sandy and loess areas. These characteristics reflect their terrain relief. Experiments show that Fourier and wavelet based methods perform better than the traditional statistical methods, the Fourier based method performs bettern than the wavelet based method in the plain area while the wavelet based method behaves better in the loess hilly area. The frequency domain based multi-scale analysis method reflects similarity, multi-scale characteristics of landform morphology in a new scope. There is high continuity in landform scale similarity quantitation and it benefits cognition of landform origins. The spatial distribution of landform scale similarity would offer helpful reference for landform classification and geographical comprehensive zoning. The frequency domain based landform scale analysis method would enrich and improve the methodology of digital terrain analysis.