Medium Range Forecast Exercise for Tallahassee, Florida:
A Prototype for Forecasting along the Gulf Coast in Fall

VIEW DISCUSSION

Discussion

ZFP hits the T_max right on and MEX MOS is off by only a couple of degrees. Both are correct in predicting values well below climo, but not for the right reason. A cold, dry air mass was expected to drop T_max below the climo value, but it turned out to be a cloudy, rainy day that produced the lowered T_max.

MEX MOS is 23° too cold on T_min, and ZFP is too cold by 17° at best (taking low 40’s to be 44°). The observed T_min is 9° above climo. The much warmer T_min is due to the presence of a warm, moist air mass rather than the expected cold, dry air mass, in addition to cloudy skies overnight which reduced the radiational cooling.

Both MEX MOS and ZFP called for clear skies and no precipitation. The GFS completely missed the southern Texas/northern Mexico positively tilted trough, which propagated through northern Florida and brought TLH 1.17” of rain on 28 October.

VIEW DISCUSSION

Discussion

GFS Model

The GFS 120-hour (5-day) forecast from 0000 UTC 23 October 2003 was faced with a daunting task. To produce a successful forecast the model had to do all of the following:

• Propagate two very strong short waves in the Pacific through the eastern Pacific longwave trough, “over” the western North American longwave ridge, and into the next longwave trough over the east central U.S.
  
• Properly treat the longwave pattern over western North America and the eastern Pacific, which was undergoing significant retrogression
  
• Move the vorticity center and residual moisture of the eastern Pacific Tropical Depression Patricia (once a hurricane) northeastward, and bring this system into proper relationship with the two shortwave troughs moving into the east central U.S. longwave trough

We have seen that there was a degree of uncertainty in the magnitude and location of the two Pacific shortwave troughs and the vorticity center of Tropical Depression Patricia in the eastern Pacific, as was determined by comparing the 0000 UTC 23 October 2003 initializations of the GFS and UKMET (See Part I, Section I, Question A). Even small and subtle differences in the initial state of an NWP model can evolve into large and important differences in the model forecast 5 days later.

GFS MEX MOS

The MOS technique is designed to improve upon NWP model forecasts by removing known systematic errors, or biases, in the model’s performance, as measured over a period of several years. However, the MOS technique cannot correct for an extreme and uncharacteristic forecast of a model, such as a major “bust,” as we have seen in this scenario. Thus, the GFS MEX MOS forecast in this case largely reflects the raw GFS model output and correspondingly contains the same large forecast errors.

Forecaster (ZFPs)

We have seen that the day-5 ZFPs from 23 October 2003 are consistent with the GFS 120-hour forecast and the GFS MEX MOS forecast. The forecaster evidently did not see important reasons to challenge this objective guidance. Yet, there were a number of reasons on 23 October 2003 to view the objective guidance with caution:

• Two global models had different initializations of the two short waves in the Pacific
  
• The graphic showing the 5-wave pattern superimposed on the GFS initialization indicated the great intensity of the two short waves in relation to the longwave trough in the Pacific and, therefore, their potential importance in the evolution of the flow pattern down stream in the medium-range forecast
  
• There was a tropical depression in the eastern Pacific that had been a hurricane. Two global models had different initializations of this system, but both models showed eastward or northeastward propagation of the system toward the Gulf of Mexico
  
• The 5-wave graphic indicated important changes in the longwave pattern over North America and the eastern Pacific, i.e., a significant retrogression