06d

• Board Meeting Date: 5/20/2008
• Description: CWCB Director's Report
• Board Meetings - Doc Type: Memo

If one compares SNODAS basin average SWE to SNOTEL, the ratios vary from 10% in low-elevation
(with plains) basins to 50% in high-elevation basins. The explanation for this is that the SNOTEL averages SWE
from a few high-elevation SNOTEL sites in or near the basin, whereas the SNODAS averages are from all the
lkm grid cells within a basin, including any low-elevation cells that may not hold snowpack. Therefore
SNODAS averages in high elevation basins like the Upper Colorado will agree more with SNOTEL averages.
Similarly, SNODAS averages in low elevation basins (e.g., including the plains) like the Arkansas Basin
disagree with the SNOTEL averages. This helps explain the calculations in the legend of the daily mapping
hosted by the USBR.
SNODAS LINK TO HYDROLOGIC MODEL -PHASE II
The SNODAS datasets provide detailed spatial information about the state's snowpack; however, use of the
information has mostly been limited to qualitative analyses. The purpose of this phase was to provide a
quantitative analysis of the SNODAS data as inputs to a distributed hydrologic model, with the objective of
improving hydrologic forecasts. The study focused on the Colorado River Basin. A data analysis was
performed over most of the Colorado River Basin within the state of Colorado to understand the large and small-
scale properties of the SNODAS data. Three sub-basins (Blue River below Dillon Reservoir (DIRC2), Colorado
River below Lake Granby (GBYC2), and Williams Forlc below Williams Fork Reservoir (WFDC2)) were
selected for detailed modeling using the National Weather Service research distributed hydrologic model
(ROHM). Snowmelt and rain on bare ground derived from SNODAS were used as input to ROHM in order to
assess the value of the detailed spatial SNODAS information in computing streamflow forecasts. The
distributed modeling results were evaluated and compared to results from the lumped modeling process used
operationally at the Colorado Basin River Forecast Center (CBRFC). The report is dated August 2007 and the
final Phase II project meeting was in November 2007.
RESULTS PHASE II
Although the results presented in this report are based on a very limited dataset, much insight was gained
concerning the potential usefulness of SNODAS data. Some of the findings include:
~ On an annual basis, the SNODAS total precipitation estimates are reasonably consistent with other estimates
of precipitation in the Colorado basin; however, the SNODAS total precipitation grids reflect less spatial
variability than expected because of the coarseness of the precipitation inputs from the RUC2 numerical
weather prediction model. There seems to be a tendency for the SNODAS total precipitation estimates to
underestimate the western slope and overestimate the eastern slope. In general, there is a tendency to
underestimate high elevation precipitation.
~ On a seasonal basis, the SNODAS total precipitation values underestimate the winter precipitation and
overestimate the summer precipitation. This will lead to an underestimate of SNODAS snow-water-
equivalent unless the data assimilationlupdating process compensates for it.
~ Once the distributed model parameters were calibrated, the SNODAS-ROHM coupled system modeled
these basins reasonably well; however, two of the basins still had a significant negative volume bias.
Apparently, the SNODAS updating process was not able to completely compensate for the underestimate in
winter total precipitation for the three years tested.
• Because of the uncertainty in the forcing data, it was difficult to assess if the SNODAS energy-balance
snow model provided better simulations of Snowmelt than CBRFC's operational lumped model.
RECOMMENDATIONS PHASE II
Recommendations from the report are:
Improved estimates of precipitation can improve SNODAS estimates of snow-water-equivalent and Snowmelt.
More aggressive updating of SNODAS snow-water-equivalent may be warranted, because of the uncertainty in
the precipitation estimates. Additional study is needed to assess the value of the SNODAS melt simulation
resulting from a distributed energy balance model. Because the SNODAS snow model is driven by the RUC2
model outputs, it is limited by the RUC2 estimation of precipitation and other meteorological variables. It would
be useful to compare the lumped and distributed modeling systems using consistent precipitation and
temperature inputs. A potential advantage of distributed models is their ability to provide information at internal
3