The difference between reliability numbers in WRF (90.9%) and MM5 (89.7%) is not
considered to be indicative of anything beyond random effects in dropout events.
Percent of expected WRF and MM5 model
runs received by NCAR from FSL -- 2004
Fig 11.28. Daily percentage of expected model outputs that were successfully
transmitted from FSL to NCAR/RAP during the 2004 MDSS Demonstration.
Mesoscale Model Performance
For the 2004 MDSS demonstration in Iowa the FSL model ensemble consisted of two
different mesoscale, or regional, models (MM5 and WRF) with lateral boundaries from
the Eta model provided by the NWS National Center for Environmental Prediction. The
models were initialized each hour with updated satellite, radar, and RWIS data each hour
and run out to 15 h. The RWFS module uses time-lagged ensemble prediction
techniques; for example, an 8-h RWFS forecast uses as inputs the current 8-h forecasts
from the regional models along with the previous runs' 9-h forecasts and the 10-h
forecasts from the runs before that.
The ensemble modeling domain is illustrated in Fig. 11.29. This is an image of forecast
surface temperature with wind barbs plotted and contours of relative humidity overlaid.
This is an example of how the model outputs were presented in real time on FSL's web