The hyperspectral bidirectional reflectance of snow: Modeling, measurement, and instrumentation

Seasonally snow-covered areas in the Earth's mountains are an integral component of the global hydrologic cycle. These regions account for the majority of fresh water resources over much of the mid-latitudes, affect regional climate, and are sensitive indicators of climate change due to the ephemeral nature of snow. Imaging spectroscopy in the solar spectrum allows us to analyze the spectral reflectance of the seasonal snow cover to infer snow properties such as snow covered area, grain size, albedo, and algal concentration. However, the models that retrieve these properties rely on a correct linkage between the spectral and angular distribution of reflectance and the physical properties of the snow surface. The characterization of the bidirectional reflectance of snow has been limited in terms of complete coupling of spectral range and resolution, angular range and resolution, and description of the snow stratigraphy. Moreover, we lack knowledge of the effect of anisotropic snow reflectance on the inference of snow covered area and albedo from imaging spectroscopy models.; This thesis presents a radiative transfer/spectral mixture model for measuring snow properties from imaging spectrometer data, a spherical robot for the measurement of the bidirectional reflectance factor of snow at fine spectral and angular resolution, comparisons of measurements of the bidirectional reflectance factor of snow for a range of particle sizes and solar zenith angles, and the sensitivity of imaging spectroscopy models to anisotropic snow reflectance. The model had accuracies for subpixel snow-covered area, grain size, and albedo of 4%, 74μm, and 2.5%, respectively. The spherical robot, coupled with a field spectroradiometer, facilitates rapid, repeatable measurements of the spectral bidirectional reflectance from snow into any direction. Measurements with the spherical robot showed significant wavelength-dependent changes in bidirectional reflectance with change in grain size and solar zenith angle. The sensitivity study showed that by assuming that snow reflects isotropically, inference of snow-covered area and albedo with current imaging spectroscopy models may have errors of up to 20% and 11%, respectively, for topographically-realistic solar and view geometries.