At the start of the MDSS demonstration, when it was believed that the WRF and MM5
had equally valid values, the RWFS was configured to weight each of them 50%. On 27
January 2004, an analysis showed that the WRF was underestimating the insolation
values on clear days so the weights were changed to 75% to 25% in favor of the MM5
model. On 16 February 2004, the WRF model insolation calculation was corrected and
the weight blending was returned to 50% for each model. The purpose for changing the
values was to provide the best input available to the road temperature model
(SNTHERM-RT) and to get a feel for the response.
A comparison of the modeled insolation (short wave radiation) values with the
pyranometer measurements for a 7 day period after the WRF was corrected is provided in
Fig. 11.4 and a separate figure of the same except for just the RWFS (Fig. 11.5). During
this period there were two clear days (18 and 22 February 2004) where the insolation
values were near the theoretical maximum (~630 watts per square meter), and several
partly cloudy days.
The comparison illustrates the differences between models and their respective physics
for determining cloud characteristics. For example, MM5 had a tendency to overestimate
insolation (underestimate clouds) on partly cloudy days while WRF tended to
underestimate insolation (overestimate clouds). Eta was closer to MM5 than to WRF. The
blending used by the RWFS provided the best overall results even though the weights
were fixed. Because the dynamic data blending process used in the RWFS works well
for verifiable parameters, it is believed that this would hold true for insolation. Real time
insolation data would need to be available over a broad region to prove this concept.
Results and Recommendations: The ability of models to predict insolation
varies greatly between models, particularly in partly cloudy conditions. Care
must be taken to ensure the model values compare well with measured values.
Changes to both research and operational models may impact insolation
calculations so routine comparisons should be made. Because insolation
measurements are critical for road temperature prediction, it is recommended
that insolation measurements be added to surface observing stations and be
provided in real time to weather service providers. Real time access to
insolation data would provide an opportunity for systems like the RWFS to
utilize the data and optimize predictions.