Development of a hyperspectral method for detecting the radiative impact of desert dust on alpine snow.

Springtime dust storms regularly deposit hematite-rich dust upon the San Juan Mountains of southwestern Colorado. Dust deposited upon the snow surface reduces the visible albedo of snow. The high spectral resolution of AVIRIS (NASA-JPL Airborne Visible Infrared Imaging Spectrometer) makes it applicable to quantitative detection of the grain size and visible albedo of snow, and thus an effective tool for evaluation of the radiative effects of dust. An inversion technique, ND-Model, derives the spatial distribution of optical snow grain radii via continuum normalization of the 1.03 micrometer ice absorption feature. The Scaled Integral Dust Index (SIDI) provides an index of the shortwave radiative impact of dust by taking the integral of the spectral reflectance of dusty snow's deviation from the visible spectral reflectance of pure snow of the same spherical optical grain size as modeled by DISORT (Discrete Ordinate Radiative Transfer Model). Because dust is hypothesized to increase snow grain size, it was expected that retrievals values of dust index (SIDI) and snow optical grain size would be positively correlated. Initial applications of ND-Model and SIDI to AVIRIS data from the San Juans instead yielded a negative correlation between grain size and dust index. The presence of hematite-rich dust at the surface may violate the clean-snow, pure-pixel assumptions of ND- Model, changing the scaled area of the 1.03 μm ice absorption feature and retrieved optical snow grain size. Such a problem within the grain size model would affect the detected correlation between dust index and optical grain size. It is hypothesized that, in the presence of a critical amount of dust within the optical depth of the snowpack, the signals for grain size and dust concentration effectively are crossed. Here, we investigate the possible invalidation of ND-Model via spatially explicit topographic radiation modeling, analysis of in-situ observations, and GIS.