Automated modelling of digital elevation models for predictive ecosystem mapping in GIS

摘要

This thesis is an exploratory analysis of automated mapping protocols that can be used to support Terrestrial Ecosystem Mapping and Predictive Ecosystem Mapping in British Columbia. This thesis employs neighbourhood analysis of elevation and its derivatives to discriminate the bioterrain elements defined by Terrestrial Ecosystem Mapping standards. In achieving these standards, discrimination beyond the basic topographic forms presented in current research is explored. The method developed strives to be - easily implemented by mapping projects employing standard GIS software ; flexible so that the extracted topographic forms can be tailored to varying project objectives ; compatible with the hierarchical procedure employed in Terrestrial Ecosystem Mapping ; efficient and accurate in that the process is advantageous over manual mapping methods. The effect of data quality is addressed through an assessment of DEM data interpolation techniques and classification accuracy. Random and systematic artifacts of the DEM that influence the quality of the derivatives are explored. The issue of scale-dependent shape is addressed by the constraints of objective-based mapping in which a map scale is specified and the most basic shape elements are aggregated into contiguous classes by a roving neighbourhood window. The results indicate that basic topographic elements are mapable from relief as well as first and second order elevation derivatives. These results give preliminary accuracy of 80% based on the three classes tested. The procedure requires decisions at every step, but it is felt that this complements the traditional mapping process in that it is hierarchical, and requires a synthesis of extensive knowledge of vegetation and landscape across many scales. Key Words: elevation, digital elevation model, topography, slope, aspect, curvature, Terrestrial Ecosystem Mapping, Predictive Ecosystem Mapping, scale, random, systematic error.