For this case the start and stop times for both events were very well forecast. The total
snowfall accumulation predicted for the Ames area was also very close to the actual
reported snowfall total. There is some evidence in this case that poor cloud cover
predictions may have caused some problem in the air temperature forecasts, though both
are still fairly well forecast to within a few degrees Celsius most of the time. Visibility
forecasts were quite poor during times of snowfall, indicating that the visibility algorithm
may need to be revisited.
11 OVERALL PERFORMANCE RESULTS
In this section performance results are described for the entire winter 2003-2004 Iowa
field demonstration for specific components of the MDSS. All aspects of the system that
were examined showed improvement over the first field demonstration.
11.1.1 Configuration
The RWFS was configured to utilize and integrate ten different forecast modules for the
winter 2003-2004 demonstration. Models that were ingested into the RWFS included the
Eta, GFS, MM5, RUC and WRF. MM5 and WRF were initialized using the NWS Eta
model for boundary conditions and run hourly. MM5 and WRF forecasts from the current
run and previous two runs (valid at the same time) were utilized. Aviation model Model
Output Statistics (MAVMOS) were also used as input, and Dynamic Model Output
Statistics (DMOS) were calculated within the RWFS for each of the model inputs. The
ten weather forecast modules that were used to predict the weather parameters for each
MDSS forecast point were:
1) NWS Aviation MOS (MAVMOS)
2) Eta DMOS
3) GFS DMOS
4) MM5 Eta4 (run from 4 hours previous)
5) MM5 Eta3 (run from 3 hours previous)
6) MM5 Eta (run from 2 hours previous, the most current run)
7) WRF Eta4 (run from 4 hours previous)
8) WRF Eta3 (run from 3 hours previous)
9) WRF Eta (run from 2 hours previous, the most current run)
10) RUC DMOS
The RWFS integration process independently optimized the forecasts based on recent
skill at each prediction site for each parameter and forecast lead-time. The forecast
modules with the most skill were weighted more heavily. A final consensus forecast was
the ultimate output of the RWFS.
93
Previous Page
