Preliminary Assessment of the Impact of Lakes on Passive Microwave Snow Retrieval Algorithms in the Arctic

The retrieval of snow water equivalent (SWE) and snow depth (SD) information from passive microwave brightness temperatures is theoretically straightforward: as the depth and/or density of snow increases, so too does the amount of volume scatter of naturally emitted microwave energy. Shorter wavelength energy (i.e. 37 GHz) is more readily scattered than longer wavelength energy (i.e. 19 GHz), so the difference in scatter between these two frequencies (19 GHz–37 GHz) has been exploited to estimate SWE and SD. In reality, the relationship between snow depth, density, and microwave scatter is complicated by the physical structure of the snowpack (for example, ice lenses, the presence of liquid water, snow grain size variability) and the microwave emission and scattering characteristics of vegetation. The imaging footprint for spaceborne passive microwave data is large so these complicating factors are compounded by considerable within-grid cell variability in snowpack structure and any overlying vegetative cover. Snow surveys conducted during late winter of 2003 and 2004 in the Coppermine River basin of the Northwest Territories indicate that SSM/I derived estimates of SWE significantly underestimate actual SWE when utilizing an algorithm developed for SWE retrievals in open prairie environments, which we will refer to as the Goodison algorithm hereafter (Derksen and Walker, 2004). Lakes (ice-free or snow-covered) pose a challenge due to their unique microwave emission characteristics compared to terrestrial surfaces. In the Arctic tundra water bodies comprise a significant portion of the surface yet fractional lake area is presently not accounted for in any passive microwave SWE or SD retrieval algorithms. In the northern Hudson Bay Lowland, Canada, mostly shallow lakes occupy as much as 41% of the landscape. Other circumpolar high latitude regions such as Alaska, northern Scandinavia and northern Russia share this substantial areal coverage by lakes. On the North Slope of Alaska, for example, there are thousands of lakes that typically cover 20% of the land in most places, and as much as 40% near the coast. The need to examine the impact of lakes on SWE and SD estimates in lake-rich regions, and to reevaluate current retrieval algorithms, is further reinforced by published works on passive microwave remote sensing of freshwater ice (Hall et al., 1981; Chang et al., 1997) and recent field measurements of snow cover characteristics, including SWE and SD, on land and lakes in the tundra region (Sturm and Liston, 2003). In contrast to other land surfaces, the brightness temperature over lakes is higher at 37 GHz than 19 GHz during both the ice-free and ice-covered periods. As a result, the brightness temperature difference algorithms typically used to estimate SWE and SD over land result in negative values over lakes for much, if not all, of the winter