Scale, Resolution and Resampling: Representation and Analysis of Remotely Sensed Landscapes Across Scale in Geographic Information Systems.
Date of Award
Doctor of Philosophy (PhD)
Geography and Anthropology
Nina S.-N. Lam
Earth system scientists are increasingly using the technologies of Geographic Information Systems (GIS) and Remote Sensing (RS) in their analyses of earth system processes and patterns. These investigations take place over a wide range of scales, from the local to the global. Global change researchers focus on both the physical and human dimensions of changes in the earth's landscapes, which occur across a range of scales and may be scale dependent. The way in which landscapes are represented in GIS and RS, using specific spatial data models and data spatial resolutions, affects the subsequent analyses that can be performed. Optimally those analyses are grounded in firm geographical and spatial analytical principles, so as to be appropriate and therefore meaningful interpretations of the data. This research investigates two specific issues of importance to research investigating landscape change across scale, those of resampling and analysis. Four different resampling algorithms, which are used to rescale remotely sensed pixels from higher to lower spatial resolutions, are investigated using Landsat TM data representing the Flint Hills region of Kansas. Two analytical methods for examining scale effects in RS data, local variance analysis and fractal analysis, are used to examine both the effects of the resampling methods on subsequent analyses and the performance of the methods in detecting potential "scales of action" in the landscapes. Results show differences in the resampling methodologies, which affect the subsequent analyses in different manners. The averaging and convolution methods performed comparably, and are the most reliable type of algorithm examined in this study. Their ongoing use in resampling processes is recommended, recognizing their limitations. The systematic sampling method is not recommended as a resampling procedure. The TM-to-MODIS algorithm, based on the optical properties of the two different resolution sensors, is potentially useful, although the algorithm behaved erratically at times. Both the fractal and local variance methods performed comparably to indicate scale effects in the data, with corresponding results to each other and to the statistical information on the images. As such both methods are deemed appropriate for examining landscapes across scale.
Weigel, Stephanie Jean, "Scale, Resolution and Resampling: Representation and Analysis of Remotely Sensed Landscapes Across Scale in Geographic Information Systems." (1996). LSU Historical Dissertations and Theses. 6169.