Figure 3: Remotely sensed snow and ice albedos for both basins during
mid-summer conditions. Light grey corresponds to masked areas due to
snow- and ice-free pixels, obstruction by clouds and shadows, or if the
BRDF retrieval was not possible for that landcover class.
3.2. Albedo DA Streamflow
Evaluation
DA and CTRL streamflow evaluation metrics had accurate results for most
analyzed years. In the heavily wildfire-impacted year of 2018, DA
outperformed CTRL for both basins. The KGE difference between DA and
CTRL was 0.18 and 0.20 for AGRB and PGRB, respectively. In 2019 (soot
algae growth) and 2020 (normal year), DA was only beneficial for PGRB.
In 2021, the year affected by heatwaves and a few light late summer
wildfires, DA did not improve streamflow predictions for either basins,
indicating that other mechanisms might have influenced streamflow
predictions during heatwave conditions or that the operation of the
albedo algorithm in CRHM could not be improved upon by assimilating
observations. The four-year overall evaluation revealed that albedo DA
substantially benefited streamflow predictions in PGRB (KGE improvement
of 0.12), but only a slight advantage was found for streamflow
predictions in AGRB (KGE improvement of 0.02) (Table 2). The four-year
overall evaluation NSEs for AGRB (0.74) and PGRB (0.78) were above the
mean of maximum values (0.64) found in 20 studies that predicted
streamflow with the CRHM model (Pomeroy et al. , 2022).
Table 2: Streamflow evaluation metrics for AGRB and PGRB. Metrics were
calculated for the combined four melt seasons and each melt season
individually (May 1 to Sept. 30), since streamflow rarely occurs outside
that period.