| In this dissertation, we present a novel use of the Global Positioning System (GPS) signal for terrain classification. In contrast to current methods that rely on visible or near-infrared wavelengths, this new method utilizes the surface-reflected GPS satellite signal to identify terrain or land cover classes having a surface moisture component. The amplitude of the reflected L-band (1.57542 GHz) GPS signal is strongly influenced by the amount of moisture in the top 2--5 cm of the land surface, and is particularly sensitive to top-surface water features. This classification method leverages the existing constellation of GPS satellites, similar to a forward-scattered bistatic radar system, with the satellites serving as the transmitter and an airborne GPS Remote Sensor (GPSRS) as the signal receiver.; Two closely-related terrain classification approaches are presented which utilize characteristics of the surface-reflected GPS signal. The first method requires only GPS signal data, while the second method combines GPS signal data with imagery. GPS-derived signal measures were shown to be effective as terrain classification features, particularly for data acquired after precipitation events. Combining surface-reflected GPS data with visible wavelength imagery was shown to significantly increase the classification accuracy of visibly identifiable landcover/terrain classes as compared to utilizing image data alone.; This dissertation also includes development of a signal calibration method which ties the ratio of reflected-to-direct satellite signals to surface reflectivity---a geophysical parameter common to all remotely sensed surfaces---to allow quantitative comparison of temporally and spatially separated data sets. Calibration includes an approach to mitigating the effect of extraneous signal reflections ("multipath") in the direct satellite signal, which introduces errors into the surface reflectivity calculation. Calibration and multipath mitigation also enable traditional soil moisture retrieval via a physical surface model. Additionally, functional simulation and mathematical modeling of the current-generation, NASA-Langley GPS Remote Sensor, utilized for this research, directly supported development of the calibration and multipath mitigation techniques.; Results from this work have broad applicability in areas that include terrain and land use mapping, flood and natural disaster emergency management, precision agriculture in arid regions, and regional hydrology, among others. |
