Figure 4: AGRB (top) and PGRB (bottom) observed and simulated DA and CTRL streamflows for the four analyzed WYs.
3.3. Albedo and SWE Updating with DA
As stated in section 5.4.2, satellite albedo DA improvements were observed in years when DA streamflow predictions were lower than CTRL and bias was positive. The benefit of satellite DA was caused by a decrease in modelled glacier albedo combined with an increase in snowcover in spring and summer (Figures 5, 6, and Table 2). This mechanism occurs because the associated melt decreases considerably while snowcover covers glacier ice, considering the relatively high albedo snow generates less meltwater than relatively low albedo ice for the same shortwave insolation fluxes. The DA streamflow used for evaluation was the mean of all 20 ensembles. This mean is lower if more ensembles have prolonged spring and summer snowcover over ice. Since the benefits of DA stem from a decrease in streamflow, snowcover over ice has a larger influence on melt than the decrease in albedo introduced by DA. In AGRB, the ice albedo decrease introduced by DA was larger than in PGRB. At the same time, the DA snowcover over ice was smaller in AGRB (Figures 5 and 6). The integrated response was that AGRB streamflow was less sensitive to DA than in PGRB. These results indicate that the benefit of satellite albedo DA stems not only from the albedo itself, but also from associated changes in other model states and glacier configuration.