| A detailed knowledge of the volume of potential water present in snowpacks is vital. It is estimated that about three quarters of the world's terrestrial water reserves, used mainly for municipal and agricultural locked purposes, are locked in snow and ice. The principal objective of this study was to analyze the degree to which passive microwave radiation can be used to interpret snow depth and snow water equivalent. The methodology reflects the importance of isolating phenomena other than snow depth and water equivalent which may influence the microwave signal. These factors include: snow wetness; depth hoar; complex mountainous terrain; dense forest cover: atmospheric precipitable water; and mixed pixels incorporating combinations of open water, bare soil, shallow or deep snow.; Results from this study, using United States climate station data, suggest that the variability associated with sampling uncertainty and snow density overwhelm any relationship between snow depth and brightness temperature, producing non-significant regression models. Using snow water equivalent, however, which includes snow density information, the results are mostly significant. It is shown that for a non-forested, non-mountainous terrain, the snow water equivalent of a pack with no depth hoar and no melting snow can be estimated with 95 percent confidence within {dollar}pm{dollar}44 mm. The regression probability for error is highly significant at less than 0.01 percent. This 95 percent confidence interval doubles when there is depth hoar present ({dollar}pm{dollar}84 mm), and the relationship between snow water equivalent and brightness temperature reverses, as the scattering is greatest in shallow snowpacks with large depth hoar crystals. Significant results are obtained for melting snow conditions (probability for error equal to 4.4 percent), despite theory indicating the contrary, by using only the night-time satellite passes. Dense forest cover impacts the microwave signal but appears not to completely mask the influence of the underlying snowpack. Different combinations of brightness temperatures and polarizations compared with non-forested areas explain much of the snow variance (95 percent confidence interval equals {dollar}pm{dollar}41 mm). Lastly, it is difficult to collect ground data in regions of complex mountainous terrain which accurately represent snow conditions over a large area. However, some information can be obtained using high elevation SNOTEL sites from some the Rocky Mountains. The confidence intervals in mountainous regions range from {dollar}pm{dollar}163 mm to {dollar}pm{dollar}729 mm.; The results from this study represent a marked improvement compared with previous experimental data sets. Previously, little or no information regarding the confidence of the microwave-derived estimate of snow water equivalent was available. Furthermore, one algorithm was employed for all surface types and all snow conditions. Regionalization based on different snow characteristics and different terrain types produces unique and statistically significant regression models for almost every region. However, despite this progress, further work in reducing the magnitude of the confidence intervals associated with the estimates of snow water equivalent is essential, particularly with respect to the reduction of the sampling uncertainty. Until more accurate estimates can be made, passive microwave data can be used only as an indicator of the "probable" snow water equivalent, at the hemispheric and global scale. |
