Click here for
catalog of sonde drops.
Click here for zipped post-processed sonde data.
1. Synoptic situation
Tropical Storm Earl developed from an easterly wave that had been tracked across the Atlantic since 17 August. By 31 August, a tropical depression had rapidly developed into Tropical Storm Earl in the Bay of Campeche moving slowly northeastward toward the U. S. Gulf Coast. Due to the threat to this area, a synoptic surveillance mission was tasked for nominal time 02 September 1998 0000 UTC, when Earl was located about 450 km southwest of New Orleans ( Fig. 1). At that time, a longwave trough extended over the eastern half of the U. S., with a strong impulse moving southward near Wisconsin. Earl had already broken through the subtropical ridge axis. Hurricane Danielle was located off the North Carolina coast, Hurricane Isis was located near the southern tip of Baja California, and a spurious vortex in the model initial conditions was located just to the east of the Lesser Antilles.
2. General Assessment of dropwindsonde impact
A. GFDL model
Figure 2 shows the GFDL forecast tracks for Tropical Storm Earl, and Table 1 shows the errors and impact of the synoptic surveillance mission. The results are mainly negative, with improvement only at 12 h. The run with none of the dropwindsonde data is also qualitatively better than that with all the dropwindsonde data. The upper-tropospheric data improved the forecasts at all forecast times. Both the GFAL and GFP3 runs first predicted landfall near Pilottown, Louisiana, much further from the actual landfall position near Panama City 30 h into the forecast, than the prediction of Pensacola, Florida, provided by the run with no dropwindsonde data.
Figure 3 shows the VICBAR forecast tracks for Tropical Storm Earl, and Table 2 shows the errors and impact of the synoptic surveillance mission. The dropwindsonde data improved the forecasts at 36 and 48 h, those times near landfall. However, the forecast with the dropwindsonde data was qualitatively degraded compared to that without the additional data. The upper-tropospheric data improved the forecasts at all times until 48 h. The forecast of the landfall point was considerably degraded, mainly due to the forecast without the dropwindsonde data being excellent. VBNO predicted landfall near Sunnyside, FL, only 22 km from the actual landfall point near Panama City 30 h into the forecast. VBAL forecast landfall at Hurlburt Field, and VBP3 forecast landfall near Mary Esther, both about 80 km further to the west.
Figure 4 shows the GSM forecast tracks for Tropical Storm Earl, and Table 3 shows the errors and impact of the synoptic surveillance mission. The dropwindsonde data improved the GSM forecast track at all times except at 12 and 120 h. However, the forecast with the dropwindsonde data was further to the west, making the forecast qualitatively worse. The upper-tropospheric data improved the forecast at all forecast times. The forecast of the landfall point was degraded: All runs forecast multiple landfalls, one on the Mississippi delta, the other along the Mississippi coast, and both forecasts were particularly poor. AVNN forecast the landfall near Pilottown, Louisiana, whereas the other two runs forecast the landfall near Empire, further to the west, a degradation of almost 10%.
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 12, 36, 48, and 108 h. The upper-tropospheric data improved the forecasts at all times except 48 and 96 h. Tuleya and Lord (1997) also showed modest improvements to GFDL model intensity forecasts in the HRD synoptic flow cases.
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 Isis in the North Eastern Pacific Ocean. The area of relatively large perturbation centered southeast of Cape Hatteras was associated with Hurricane Danielle moving away from the U. S. east coast. The four perturbation maxima surrounding Danielle from south through northeast corresponded to weak upper-level features. Near Earl, the largest perturbation exists in the region close to Earl. A complex maximum extends along the axis of the subtropical ridge from southwestern Florida into the Bay of Campeche, and another maximum to the east of the anticyclone over eastern Cuba and the Central Bahamas. Of these, none were well-sampled. The southernmost part of the target corresponding to the axis of the subtropical ridge was not sampled due to the lack of rawinsonde data over southern Mexico. Earl itself was not well-sampled due to poor forecasts leading to the nominal time, so that the southwestern quadrant was not sampled. As a result, no targeting run has been made for this case.
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 is near the target corresponding to the subtropical ridge axis in the Bay of Campeche, and the second largest extends along the ridge axis further east. A third maximum is in the perturbation minimum in the northeastern Gulf of Mexico, and a fourth, smaller perturbation is on the northern side of Earl. Relatively large impacts extend away from the area of dropwindsonde locations, especially in the Bay of Campeche into the Gulf of Tehuantepec. 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 the two largest difference maxima are in areas of large ensemble perturbation, this aliasing may amplify in time and negatively impact the forecasts.
Figure 7 shows that, by 24 h into the forecast, the largest difference between the forecasts with and without the dropwindsonde data has decayed and moved into the Gulf of Tehuantepec. The maximum further to the east has remained approximately steady in location and amplitude. The maximum originally over the northeastern Gulf of Mexico has moved northeastward and decayed, and the small maximum to the north of Earl has amplified rapidly and moved slowly northward. These results generally confirm that the largest (smallest) ensemble perturbations correspond to amplifying (decaying) modes in the model.
The dropwindsonde data obtained during the synoptic surveillance mission for Tropical Storm Earl at nominal time 02 September 1998 0000 UTC has provided mainly negative results throughout the five day forecast of Tropical Storm Earl in all three models, especially in the forecast of the landfall point. The MRF ensemble forecasting system suggested that data surrounding Earl and in the subtropical ridge axis to the south and east of Earl would have the greatest impact on the Earl forecast. Due to logistical problems, none of these regions were effectively sampled, allowing for aliasing of data into unstable areas and a de gradation of the forecast.
|GFAL Error (km)|
|GFP3 Error (km)|
|12||167.||166.||( 1%)||166.||( 1%)|
|24||255.||268.||( -5%)||280.||( -10%)|
|36||343.||377.||( -10%)||382.||( -11%)|
|48||411.||486.||( -18%)||506.||( -23%)|
|72||577.||646.||( -12%)||675.||( -17%)|
|84||1532.||1603.||( -5%)||1632.||( -7%)|
|96||1616.||1697.||( -5%)||1759.||( -9%)|
|108||1357.||1420.||( -5%)||1502.||( -11%)|
|VBAL Error (km)|
|VBP3 Error (km)|
|12||81.||91.||( -12%)||91.||( -12%)|
|24||118.||139.||( -18%)||139.||( -18%)|
|36||140.||137.||( 2%)||146.||( -4%)|
|48||107.||69.||( 36%)||82.||( 23%)|
|72||403.||568.||( -41%)||529.||( -31%)|
|84||619.||747.||( -21%)||711.||( -15%)|
|GSAL Error (km)|
|GSP3 Error (km)|
|12||191.||199.||( -4%)||208.||( -9%)|
|24||357.||357.||( 0%)||375.||( -5%)|
|36||555.||533.||( 4%)||554.||( 0%)|
|48||763.||746.||( 2%)||784.||( -3%)|
|72||1766.||1753.||( 1%)||1804.||( -2%)|
|84||2854.||2863.||( 0%)||2918.||( -2%)|
|96||3572.||3557.||( 0%)||3586.||( 0%)|
|108||3777.||3755.||( 1%)||3814.||( -1%)|
|120||4011.||4106.||( -2%)||4107.||( -2%)|
|landfall||356.||387.||( -9%)||387.||( -9%)|
|GFAL Error (kn)|
|GFP3 Error (kn)|
|12||18||15||( 17%)||16||( 11%)|
|36||46||35||( 24%)||35||( 24%)|
|48||9||8||( 11%)||6||( 33%)|
|96||4||6||( -50%)||5||( -25%)|
|108||6||3||( 50%)||4||( -33%)|
Return to 1998 assessment page
Return to HRD home page