Raster-mode generalization of land use/land cover maps
The purpose of the research presented in this thesis was to apply the parameters of a raster-mode algorithm for nominal-level areal generalization to scale reduction of the national land use/land cover digital data, and to evaluate the results of the algorithm in terms of the quality of spatial expression. A secondary purpose was to establish the scale at which the algorithm will generate a satisfactory generalization for a given set of features. Cartographic generalization has not been easily adaptable to computer methodologies. In work at the U.S. Geological Survey, Monmonier developed algorithms for raster-mode, nominal-level areal generalization, intended for reduction of the 1:250,000-scale national land use and land cover map series to the 1:2,000,000 scale of the U.S. National Atlas. Monmonier's algorithms comprised nine parameters for generalization that concatenate the elements of feature selection, smoothing and shifting. The working hypothesis was that the display priority weights for each land use/land cover category, the gap interval and the priority rankings of land use and land cover types for gap bridging, and erode smoothing would be the most important parameters for production of a generalized map image. A window was selected from the U.S.G.S. Paducah land use/land cover quadrangle and formatted for the generalization programs in a series of procedures using GIRAS and DLG programs. Specifications for the nine parameters were tested, one parameter at a time, in three different scale reductions. For each generalization, statistics were obtained summarizing the changes in proportions of the land use/land cover categories. Plots of the generalized map were also produced. Statistics and plots were compared and evaluated. The importance of the display priority weights and the gap-bridging specifications were supported by the results of the tests. Erode smoothing was found to be less significant in itself, but could be regarded as an integral part of the gap-bridging parameter with which it is associated. The generalization parameters proved to be useful in all of the scale reductions tested.