However, in each of the three case studies we have detailed above we have found
strengths and weaknesses in the RCTM. The lack of logic for dealing with blowing snow
conditions in the RCTM continues to be a noticeable weakness in the system causing the
recommended treatments to be below what might be needed. Even incorporating the
direction of potential blowing snow might be an improvement. Based on the cold rain
case of January 16-17 and the heavy rain / warm road case of March 15-16 it appears that
the road temperature predictions are sufficiently accurate during storms to allow a
tightening of the freeze-point threshold. But, the rigidity of a fixed freeze-point threshold
may also need to be addressed, perhaps by making the freeze-point threshold dependent
on the duration of the below freezing temperatures. Additionally, modifications made to
the RCTM and improvements in the RWFS forecast have yielded recommended
treatment strategies are fairly consistent from MDSS run to run. Enhancements made to
suppress pre-treatments under warm road and cold road/ blowing snow conditions are
effective. Finally, the added ability to protect the wet road surface after the storm
precipitation ends was also effective.
Results and Recommendations: Significant improvements to the Rules of
Practice module in 2004 resulted in treatment recommendations that better
matched DOT operations. In some cases, the treatment recommendations were
implemented without modification with excellent results. The Rules of Practice
code (as provided in MDSS Release-3.0), although not perfect, provides a solid
starting point for further development and tailoring to specific road operating
authorities by the private sector.
11.4
Mesoscale Modeling System Reliability
One of the major concerns of those who would consider running NWP models for their
own operational applications (such as MDSS system implementations) is the labor
intensiveness of the process and the system reliability. The first aspect is addressed here
by simply stating that the entire system runs automatically without any human
intervention, unless a failure occurs, in which case a system restart brings everything
back to life. The second aspect will be illustrated by statistics comparing the number of
forecast frames promised for delivery to NCAR's RWFS system to the number actually
delivered. This number was 90.3% for the entire period of the experiment, which is
probably not adequate for actual operations. However, a day-by-day depiction of
reliability (Fig. 11.28) reveals that improvements are within reach simply by applying the
usual full-time scrutiny to operational computing systems, which was not done during
this experiment. In such an application, the head-node crash would have been corrected
in a few hours by installing a spare processor, and the hack attack would have been
detected and repaired within an hour or two of its occurrence. The hacker had about a
full day to cause damage with lasting repercussions in this case.
129
Previous Page
