IMPACT ON HURRICANE TRACK AND INTENSITY FORECASTS OF GPS DROPWINDSONDE OBSERVATIONS FROM THE SYNOPTIC SURVEILLANCE MISSION INTO HURRICANE GEORGES ON 19 SEPTEMBER, 1998.

Sim D. Aberson

Hurricane Research Division
Atlantic Oceanographic and Meteorological Laboratories
National Oceanic and Atmospheric Administration

Click here for N42RF catalog of sonde drops.
Click here for N42RF zipped post-processed sonde data.
Click here for N43RF catalog of sonde drops.
Click here for N43RF zipped post-processed sonde data.
Click here for N49RF catalog of sonde drops.
Click here for N49RF zipped post-processed sonde data.

1. Synoptic situation

Hurricane Georges developed from an easterly wave which moved off the west African coast on 13 September, and developed quickly into a tropical depression two days later. By 19 September, Georges was strengthening rapidly into a hurricane to the east of the Leward Islands. Due to the potential threat to the Virgin Islands and Puerto Rico, a synoptic surveillance mission was tasked for nominal time 19 September 1998 0000 UTC, with a follow-on mission the next day. On the second day, Georges was moving west-northwestward about 650 km east of Dominica, to the south of a strong subtropical ridge (Fig. 1). A cold low was located in the central Caribbean Sea, and Tropical Storm Hermine was making landfall on the Central Gulf coast, just to the east of a strong midlatitude trough over the west-central United States.

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 positive, with large improvements at all times. However, the upper-tropospheric data degraded the forecasts at all forecast times.

Georges made landfall five times during the forecast period. The first, 28.5 h into the forecast at Antigua, was not forecast by the run without dropwindsonde data. Both runs with the dropwindsonde data forecast landfall on Barbuda, less than 100 km away.

The second landfall was on the island of St. Kitts 32 h into the forecast. Again, the run without the dropwindsonde data failed to forecast landfall. The run with all the dropwindsonde data forecast landfall on Anguilla, and the run without the upper-tropospheric data forecast landfall on St. Martin, both just over 100 km away from the actual landfall position.

The third landfall, near Humacao, Puerto Rico, 46 h into the forecast, was only forecast by the run without the upper-tropospheric data. That run forecast landfall near Anegada, about 162 km to the northeast of the actual landfall point.

As the GFDL forecasts began a recurvature of the hurricane, they failed to predict the last two landfalls near San Rafael del Yuma, Dominican Republic 60.5 h into the forecast, and near Guantanamo, Cuba, 93.5 h into the forecast.

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 improved the forecasts at all times through 48 h. The upper-tropospheric data provided better forecasts than the run with all the data at these same forecast times.

Georges made landfall five times during the forecast period. The first, 28.5 h into the forecast at Antigua, was forecast to be near Barbuda by the run without dropwindsonde data. Neither run with the dropwindsonde data forecast landfall, since the tracks slipped between the islands, though the forecasts were all excellent.

The second landfall was on St. Kitts 32 h into the forecast. The forecast with no dropwindsonde data forecast landfall on St. Martin, and the forecast with all the dropwindsonde data forecast landfall on St. Eustatius, a substantial improvement. The forecast without the upper-tropospheric data did not forecast landfall on any island, although that forecast was quite good.

The third landfall, 46 h into the forecast near Humacao, Puerto Rico, was also well-forecast. The forecast with no dropwindsonde data forecast landfall on Virgin Gorda, Virgin Islands. The forecast with all the dropwindsonde data was almost perfect, near Humacao, and the forecast without the upper-tropospheric data was near El Fajardo, Puerto Rico. Both of the forecasts with the dropwindsonde data were substantial improvements over the forecast without the dropwindsonde data. The fourth landfall 60.5 h into the forecast, was near San Rafael del Yuma, Dominican Republic. The landfall forecast of the run without the dropwindsonde data was near Samana, Dominican Republic. The landfall forecast of the run with all the dropwindsonde data was nearly perfect, and the landfall forecast of the run without the upper-tropospheric data was near El Macao, Dominican Republic. The latter two were again substantial improvements over the run without any dropwindsonde data.

The forecasts with the dropwindsonde data were somewhat less successful on the last landfall near Imias, Guantanamo, Cuba, 93.5 h into the forecast. The run without the dropwindsonde data was perfect, whereas the landfall forecast of the run with all the dropwindsonde data was near Port Maria, Jamaica, more than 300 km to the south. The run without the upper-tropospheric dropwindsonde data slipped between Cuba and Jamaica without making landfall.

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 improved the GSM forecast track at all times, and the upper-tropospheric data improved the forecasts at all times as well.

Georges made landfall five times during the forecast period. None of the runs, either with or without the dropwindsonde data forecast landfall the first or second landfalls near Antigua 28.5 h into the forecast and St. Kitts 32 h into the forecast, since the tracks slipped between the islands, though the forecasts were all very good.

The third landfall, 46 h into the forecast near Humacao, Puerto Rico, was forecast by the runs with the dropwindsonde data. All three forecasts remained north of Puerto Rico, though the run with all the dropwindsonde data forecast landfall over St. Thomas, Virgin Islands, and the run without the upper-tropospheric data forecast landfall over St. John, Virgin Islands.

None of the runs forecast either the fourth landfall 60.5 h into the forecast, near San Rafael del Yuma, Dominican Republic, or the fifth landfall 93.5 h into the forecast near Imias, Guantanamo, Cuba.

D. Intensity

Table 4 shows the GFDL intensity forecast errors and impact of the synoptic surveillance mission. The dropwindsonde data had a mixed impact, improving the forecasts at 72, 84 h, and 120 h. The upper-tropospheric data improved the forecasts only at 84 and 96 h. Tuleya and Lord (1997) 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). The largest perturbation is associated with Hurricane Georges. The areas of relatively large perturbation size off the Central Atlantic coast is associated with weak vorticity maxima moving rapidly eastward in the midlatitude westerlies. Other small maxima are associated with Tropical Storm Hermine in the Gulf of Mexico, and a series of vorticity maxima rotating through the long-wave trough to the northeast of Hurricane Georges. The forecast of Hurricane Georges the day before the synoptic surveillance mission was very poor, more than 500 km too slow. When this error is recognized, the two maxima near Hurricane Georges are considered well-sampled.

Two sets of model runs have been performed. The first, the TG run, includes the dropwindsonde taken in and around well-sampled areas of large ensemble perturbation. Since no observations were taken near the large perturbations off the Central Atlantic coast, only those dropwindsonde data near Hurricane Georges are considered in this run (all the dropwindsondes represented by closed circles in Fig. 5). The other, the NT run, includes the complement of the TG run, with dropwindsondes represented by open circles in Fig. 5. Results are shown in Tables 1-4 and Figs. 2-4. The TG run provided better forecasts than the run including all dropwindsonde data at all times except 24 h in the GFDL, and from 72 h onward in VBAR. The landfall forecasts from the TG runs were comparable to or better than those from the runs with all the dropwindsonde data in the GFDL model for all five landfalls, for the first and fifth landfall in the VBAR model, and all but the third landfall in the GSM.

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 form a wavenumber two pattern about 400 km in radius around Georges. Two more maxima are located about 750 km to the south and southeast of the center of Georges. Other maxima are located along the ridge axis along 65°W, in the subtropical ridge to the northwest of Georges, and in the cold low to the north of Hispaniola. The two large maxima to the south of Georges, and the maximum to the north of Hispaniola are not co-located with the data, and a large area of large impact extends northward more than 1000 km from the location of the northeasternmost dropwindsonde location. The spread of the impact away from the dropwindsonde locations suggests that the data assimilation has allowed the dropwindsonde data from nearby areas to influence the initial conditions there. Since the difference maxima to the south of Georges are in an area of large ensemble perturbation, this spread may amplify in time and negatively impact the forecasts.

Figure 7 shows that, by 24 h into the forecast, the differences between the forecasts with and without the dropwindsonde data have rotated around Georges and have started amplifying after 12 h into the forecast. The maximum initially in the southeastern Caribbean Sea has moved southwestward toward Venezuela, and the maximum originally north of Hispaniola has moved southward. Neither have amplified. Other weaker maxima have remained almost steady. These results generally confirm that the largest (smallest) ensemble perturbations correspond to amplifying (decaying) modes in the model.

4. Conclusion

The dropwindsonde data obtained during the synoptic surveillance mission for Hurricane Georges at nominal time 20 September 1998 0000 UTC has provided mainly positive results throughout the five day forecast of Hurricane Georges in all three models. The MRF ensemble forecasting system suggested that data surrounding Georges would have the greatest impact on the Georges forecast. Even larger improvements to the forecasts have been shown by using only data in these regions.