IMPACT ON HURRICANE TRACK AND INTENSITY FORECASTS OF GPS DROPWINDSONDE OBSERVATIONS FROM THE SYNOPTIC SURVEILLANCE MISSION INTO HURRICANE GEORGES ON 25 SEPTEMBER, 1998.

Sim D. Aberson

Hurricane Research Division
Atlantic Oceanographic and Meteorological Laboratories
National Oceanic and Atmospheric Administration 4301 Rickenbacker Causeway
Miami, Florida


Click here for N42RF transmitted sonde data.
Click here for N42RF HSA sonde data.
Click here for N43RF transmitted sonde data.
Click here for N43RF HSA sonde data.
Click here for N49RF transmitted sonde data.
Click here for N49RF HSA sonde data.


1. Synoptic situation

Hurricane Georges developed from an easterly wave that moved off the coast of Africa 13 September, and developed into a tropical depression two days later. The system moved westward, strengthening rapidly into a hurricane by 17 September. Due to the potential threat to the Gulf coastal states, a synoptic surveillance mission was tasked for nominal time 26 September 1998 0000 UTC, with a follow-on mission the next day. At that time, Georges was moving west-northwestward and was located near Key West, to the south the subtropical ridge axis (Fig. 1). A cold low was located in the western Gulf of Mexico. The weak circulation to the east of Bermuda was a subtropical system which eventually developed into Hurricane Karl.


2. General Assessment of dropwindsonde impact

A. GFDL model

Figure 2 shows the GFDL forecast tracks for Hurricane Georges, and Table 1 shows the errors and impact of the synoptic surveillance mission. The results are mixed, though early forecasts are generally improved, and later forecasts are degraded. However, the forecast of the landfall point was improved by the assimilation of the dropwindsonde data. The dropwindsonde data caused the forecast track to be further west, away from the best track initially but closer later in the forecast, than the track without the dropwindsonde data.

Georges made landfall 59.5 h into the forecast at Biloxi, MS. The forecast without the dropwindsonde date made landfall 95 h into the forecast at Bitmore Beach, FL, 309 km to the east. The forecast with the dropwindsonde data made landfall at Sunnyside, FL, 289 km to the east, a slight improvement. Both versions forecasted landfall almost two days after it occurred.

B. VICBAR

Figure 3 shows the VICBAR forecast tracks for Hurricane Georges, and Table 2 shows the errors and impact of the synoptic surveillance mission. The dropwindsonde data degraded the forecast at all times except 120 h. The dropwindsonde data pushed the storm further to the south away from the actual track.

Georges made landfall 59.5 h into the forecast at Biloxi, MS. The forecast without the dropwindsonde date made landfall 96 h into the forecast at Jamaica Beach, TX, 582 km to the west. The forecast with the dropwindsonde data made landfall at Galveston, TX, 594 km to the west, a slight degradation. Both versions forecasted landfall almost two days after it occurred.

C. GSM

Figure 4 shows the GSM forecast tracks for Hurricane Georges, and Table 3 shows the errors and impact of the synoptic surveillance mission. The dropwindsonde data had a large positive impact on the GSM forecast track. The dropwindsonde data pushed the forecast further to the east, toward from the best track.

Georges made landfall 59.5 h into the forecast at Biloxi, MS. The forecast without the dropwindsonde date made landfall on the mainland 60 h into the forecast at St. Bernard Parish, LA, 97 km to the west. The forecast with the dropwindsonde data made landfall at the actual landfall point. Both versions forecasted landfall just a few hours after it occurred.

D. Intensity

Table 4 shows the GFDL intensity forecast errors and impact of the synoptic surveillance mission. The dropwindsonde data had a positive impact at all times except 48 h. Tuleya and Lord (1998) also showed modest improvements to GFDL model intensity forecasts in the HRD synoptic flow cases.


3. Targeting

The atmosphere has long been recognized as a chaotic system (Lorenz 1963), e.g. very small perturbations to initial conditions result in increasingly large differences in the evolution of the atmosphere with time. Since the exact state of the atmosphere can never be measured, all analyses contain errors whose magnitudes can only be estimated. An indeterminate number of initial conditions consistent with the observational data can therefore be used in numerical weather prediction, and single model runs at any synoptic time only give one possible solution to the evolution of the atmosphere. Many operational forecast centers around the world, therefore, now employ ensemble forecasting as a means of quantifying the uncertainty in the evolution of the atmospheric system. Small perturbations from a best "control" state are calculated and added to and subtracted from this control to allow for different integrations starting from theoretically equally likely initial states. These perturbations are designed so as to mimic the fastest growing modes in the model and to create the largest envelope of possibilities in the forecast. Therefore, they generally correspond to locations where large analysis errors will most impact the forecast. Those features corresponding to perturbations which most impact the tracks of tropical cyclones must be found, properly sampled, and thoroughly tested, to prove the efficacy of targeting techniques.

Figure 5 shows the variance of the size of the perturbations in the National Centers for Environmental Prediction (NCEP) global model ensemble forecasting system (Toth and Kalnay 1993). A series of three maxima extending in an arc from east central Florida to western Cuba are associated with Hurricane Georges. The second largest perturbation is associated with the weakness in the subtropical ridge to the northeast of Georges. Another large perturbation is associated with a vorticity maximum moving through the longwave trough over West Virginia, and another is related to another vorticity maximum moving northeastward through the Plains States. The mission did not surround Hurricane Georges with data, so the only features which were sampled during the synoptic surveillance mission was the weakness in the ridge and the one maximum over Central Florida.

An additional model run has been performed. The TG run includes the dropwindsonde data taken within and around these two perturbation maxima (the first eleven dropwindsondes of the mission), or just over one-third of all the dropwindsondes released. Results are shown in Tables 1-4 and Figs. 2-4. The TG run provided better forecasts than the run including all the dropwindsonde data within 36 h of the initial time and at 120 h in the GFDL. The TG run provided better forecasts than the run including all the dropwindsonde data at all times in VBAR, but only at 36 and 48 h in the GSM. The TG run improved the landfall forecast in VBAR only.

Figure 6 shows the difference in the 850 - 200 hPa averaged winds between the runs in which all the dropwindsonde data and none of the dropwindsonde data are included. The largest differences are north of the Yucatan peninsula and in the west Central Gulf of Mexico. Another large difference is off the U.S. East coast near the weakness in the subtropical ridge. The differences in the Gulf of Mexico extend a few hundred kilometers away from the location of the dropwindsonde data, and the largest maximum appears to be centered between dropwindsonde locations. This may help to identify the cause of the mixed results in this case. However, these areas are in regions in which the ensemble perturbations are small and therefore the differences are expected to decay. Figure 7 shows that, by 24 h into the forecast, the differences between the forecasts with and without the dropwindsonde data in the two western regions have decayed and remained in about the same location. The easternmost difference has moved southward along the east side of the anticyclone, and the amplitude has remained about the same, though the area of 1 ms-1 impact has become smaller. Very small initial differences over central Florida associated with Georges have amplified rapidly. These results seem to confirm that the largest (smallest) ensemble perturbations correspond to amplifying (decaying) modes in the model. The dropwindsondes surrounding the two largest ensemble perturbations are the only large perturbations well-sampled during this synoptic surveillance mission, and are expected to have the largest positive impact on the model forecasts. The results of the GSM TG model runs confirm that those few dropwindsondes have almost the same impact as all the dropwindsonde data, although this does not extend to the secondary models.


4. Conclusion

The dropwindsonde data obtained during the synoptic surveillance mission for Hurricane Georges at nominal time 26 September 1998 0000 UTC has provided mixed results. The MRF ensemble forecasting system suggests, and the model runs confirm, that data near the weakness in the subtropical ridge off the U.S. East Coast, and around Georges in Central Florida would have the greatest positive impact on the Georges forecast. There seemed to be little problem with spreading of the data impact far away from the location of the dropwindsondes in this case, and that may account for the small differences in the forecasts.


Tables

Table 1
Track forecast errors for the no dropwindsonde GFDL control (GFNO), the all dropwindsonde run (GFAL), and the run with only targeted observations (GFTG), and the percent improvement of the latter two over the control.
Forecast
time (h)
GFNO
Error (km)
GFAL Error (km)
(% Improvement)
GFTG Error (km)
(% Improvement)
1215.24.(-60%)24.(-60%)
2437.45.(-22%)30.( 19%)
3649.45.( 8%)45.( 8%)
4845.56.(-24%)59.(-31%)
72174.152.( 13%)196.(-13%)
84189.167.( 12%)210.(-11%)
96150.141.( 6%)180.(-20%)
120111.128.(-15%)89.( 20%)
Landfall309.289.( 6%)329.( -6%)

Table 2
Track forecast errors for the no dropwindsonde VBAR control (VBNO), the all dropwindsonde run (VBAL), and the run with only targeted observations (VBTG), and the percent improvement of the latter two over the control.
Forecast
time (h)
VBNO
Error (km)
VBAL Error (km)
(% Improvement)
VBTG Error (km)
(% Improvement)
1222.35.(-59%)22.( 0%)
2483.87.( -5%)83.( 0%)
36155.157.( -1%)155.( 0%)
48249.264.( -6%)242.( 3%)
72462.483.( -5%)446.( 3%)
84629.643.( -2%)612.( 3%)
96774.792.( -2%)758.( 2%)
108927.936.( -1%)899.( 3%)
1201109.1102.( 1%)1079.( 3%)
Landfall582.594.( -2%)548.( 6%)

Table 3
Track forecast errors for the no dropwindsonde GSM control (GSNO), the all dropwindsonde run (GSAL), and the run with only targeted observations (GSTG), and the percent improvement of the latter two over the control.
Forecast
time (h)
GSNO
Error (km)
GSAL Error (km)
(% Improvement)
GSTG Error (km)
(% Improvement)
1250.40.( 20%)50.( 0%)
2470.60.( 14%)69.( 1%)
3654.39.( 28%)31.( 43%)
4851.35.( 31%)35.( 31%)
72106.0.(100%)44.( 58%)
84200.86.( 43%)105.( 48%)
96383.293.( 23%)314.( 18%)
108630.543.( 14%)566.( 10%)
Landfall97.0.(100%)53.( 45%)

Table 4
Intensity forecast errors for the no dropwindsonde GFDL control (GFNO), the all dropwindsonde run (GFAL), and the run with only targeted observations (GFTG), and the percent improvement of the latter two over the control.
Forecast
time (h)
GFNO
Error (kn)
GFAL Error (kn)
(% Improvement)
GFTG Error (kn)
(% Improvement)
12-7-4( 43%)-7( 0%)
24-16-11( 31%)-14( 13%)
36-14-13( 7%)-15( -7%)
48-20-21( -5%)-15( 25%)
721917( 11%)19( 0%)
844140( 2%)40( 2%)
963432( 6%)35( -3%)
1203533( 6%)36( -3%)


Return to 1998 assessment page
Return to HRD home page