Exploration for shallow groundwater in an arid environment using directed factor analysis of Landsat scanner data
Abstract
Shallow groundwater in arid regions is found in groundwater discharge zones, which are ideal sites for water-supply wells. Areas of shallow groundwater are characterized by dark-colored soils and lush, riparian vegetation, both of which can be observed at the surface. In addition to the location of discharge areas, the surficial distribution of geologic materials is necessary for understanding groundwater flow patterns and exploring for water resources. Because of its low cost, high speed, and unusual ability to allow accurate and detailed mapping even in flat areas of fluvial sediments, remote sensing techniques were chosen as a primary reconnaissance tool. Mapping surficial geology even in partially-vegetated terrains along with mapping of vegetation density required development of two new remote sensing techniques: Directed Factor Analysis (DFA), find the surficial composition from composite spectra; and a Modified Regression Intersection Method (MRIM), to remove the offset in digital imagery due to haze. DFA is a new spectral unmixing technique combining features of traditional factor analysis and normative analysis. Composite spectra are assumed to be weighted averages of the spectra of a small number of constituents. DFA is an interactive, data-dependent method to design a linear transformation to convert composite spectra into compositions. MRIM locates the radiometric origin (haze offset vector) by finding, in M dimension, the points of closest approach of lines defined by spectrally-homogenous but variably-illuminated training fields. Thematic images depicting the abundances of three soil endmembers and riparian vegetation were generated with DFA of a Landsat Thematic Mapper image of southeastern Arizona. Comparison with soil and geologic maps indicates that these images provide a means of producing more spatially accurate and more compositionally detailed maps than present techniques. Comparison with topographic maps shows that the thematic image delineated more groundwater discharge areas than had been found using traditional photo-interpretation and field-checking. Thus, both the surficial geology and the location of groundwater discharge were successfully mapped.
Degree
Ph.D.
Advisors
Levandowski, Purdue University.
Subject Area
Geology|Hydrology|Geophysics|Remote sensing
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.