1. Synoptic situation
Hurricane Bonnie developed from an easterly wave midway between the Cape Verde Islands and the Leward Islands on 19 August, 1998. It moved westward, strengthening slowly into a major hurricane four days later. Due to the potential threat to the Carolina coastline, a synoptic surveillance mission was tasked for nominal time 24 August 1998 0000 UTC, with a follow-on mission the following day. The G-IV provided all the synoptic data in this case, with inner-core data provided by one NOAA P3 aircraft flying an air-sea interaction experiment. At the nominal time, Bonnie was moving slowly northwestward about 500 km east of the northern Bahama Islands. A broad region of westerlies extended to the north of 35°N. The remnants of a small cold low was located over the northeastern Gulf of Mexico (Fig. 1). The cyclonic circulation in the northeastern Caribbean is associated with a strong tropical wave that developed into Hurricane Earl in the Gulf of Mexico one week later. This feature was erroneously initialized in the GDAS due to lack of data.
2. General Assessment of dropwindsonde impact
A. GFDL model
Figure 2 shows the GFDL forecast tracks for Hurricane Bonnie, and Table 1 shows the errors and impact of the synoptic surveillance mission. The results are mixed, with large improvements before the landfall (through 36 h), and large degradations later in the forecast, except at 120 h. The forecast with the dropwindsonde data was qualitatively better than that without until 72 h. A large speed bias was the source of the forecast degradation. The upper-tropospheric data improve the forecasts only at 12 and 120 h. The forecast without the dropwindsondes made landfall near Oak Island, NC, about 62 km away from the landfall point near Wilmington. The forecast with the dropwindsondes made landfall near Salter Path, NC, about 51 km away from the landfall point, an 18% improvement in the landfall forecast. The upper-level data helped the landfall forecast, with the landfall about 30 km up the coast in the run excluding these data.
Figure 3 shows the VICBAR forecast tracks for Hurricane Bonnie, and Table 2 shows the errors and impact of the synoptic surveillance mission. The dropwindsonde data improved the forecasts at all times, except 48 and 72h, though the differences were mainly small, and the errors were large. Even with the dropwindsonde data, the model failed to forecast the landfall of Bonnie on the North Carolina Coast 52 h into the forecast. The upper-tropospheric data degraded the forecast at all times except 108 h, likely because Bonnie was a weakening hurricane with mainly shallow convection throughout the forecast period.
Figure 4 shows the GSM forecast tracks for Hurricane Bonnie, and Table 3 shows the errors and impact of the synoptic surveillance mission. The dropwindsonde data degraded the GSM forecast track at all times before landfall, with slight improvements only at 84, 96, and 120 h. The upper-tropospheric data improved the forecast at all forecast times after 36 h, those times around landfall. The forecast without the dropwindsonde data predicted landfall near Folly Beach, SC, or about 292 km away from the landfall near Wilmington, NC. The forecast with the dropwindsonde data predicted landfall near Wadmalaw Island, NC, or about 306 km away, for a degradation of about 5%. The upper-level data improved the forecasts, with the GFP3 run predicting landfall near Hilton Head Island, SC, almost 400 km from the landfall point.
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 between 48 and 108 h. The upper-tropospheric data improved the forecasts at all times except after 84 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 the cull region between Bonnie and the spurious cyclone in the eastern Caribbean. Other large perturbations correspond to the circulation of Bonnie itself, and an extension of the subtropical ridge behind the spurious cyclone. A complex set of smaller perturbations includes a north-south vorticity maximum between Bonnie and the subtropical ridge cell to Bonnie's east (along about 68°W), a dip in the westerlies off the New Jersey coast, and the cull region between Bonnie and the westerlies to the north. These features are only characterized by weak maxima in the perturbation variance fields, and may be stronger in a higher-resolution model. A strong vorticity maximum is seem as a large perturbation centered over Southern Ontario. The G-IV was only able to sample the complex regions to the north and east of the center of Bonnie. Removal of the spurious vortex in the Eastern Caribbean with another mission in this region may have provided further forecast improvements in this case. The southern half of the perturbation associated with the circulation of Bonnie was not sampled due to the logistical difficulty of sampling around both sides of the Greater Antilles with only one aircraft.
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. Because the southern portions of the perturbation corresponding to Bonnie was not sampled, only those observations within the complex region to the east and north of Bonnie were included (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 Table 1, Table 2, Table 3, and Table 4 and Figure 2, Figure 3, and Figure 4. The TG run provided better forecasts than the run including all dropwindsonde data from 48 to 108 h in the GFDL, including a substantial improvement in the landfall forecast. The TG run provided a better forecast than the run including all dropwindsonde data until 96 h in the GSM, also with a substantial improvement in the forecast landfall location. However, the TG run provided a degraded forecast at all times in VBAR.
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 near the coll region between Bonnie and the westerlies to the north, and in the north-south vorticity maximum to the east of Bonnie. A lesser maximum further east is along the subtropical ridge axis. The impact extends more than 1000 km from the nearest data to the south of the flight pattern, suggesting that the data assimilation has allowed the dropwindsonde data from surrounding areas to influence the initial conditions there. Since much of this is in an area of large ensemble perturbation associated with spurious features, this aliasing may amplify in time and negatively impacted the forecasts. Figure 7 shows that, by 24 h into the forecast, the differences between the forecasts with and without the dropwindsonde data in the largest two maxima have revolved around the circulation, with the leading difference located over the northern Bahamas to the south of Bonnie, the other located off Hatteras to the north of Bonnie. After amplifying for the first 12 h (not shown) both differences have started to decay. The maximum in the subtropical ridge axis, also located in an area of large ensemble perturbation, has moved slowly northeastward and amplified. The large differences to the south of the data have also amplified into small maxima surrounding the spurious vortex center that strengthened in the forecast. 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 Hurricane Bonnie at nominal time 25 August 1998 0000 UTC has provided mainly positive results before the landfall of Hurricane Bonnie on the coast of North Carolina, with mainly negative results afterwards, in both the GFDL and VBAR models, and mainly negative results from the GSM. The MRF ensemble forecasting system suggested that data surrounding Bonnie and a spuriously-analyzed cyclone in the northeastern Caribbean, and in the area between Bonnie, the subtropical ridge cell to the east and the westerlies to the north would have the greatest impact on the Bonnie forecast. Due to logistical difficulties, only the northern half of the Bonnie circulation, and none of the second circulation, was sampled during the synoptic surveillance mission. This led to aliasing of the data into these data-sparse regions, possibly causing the negative impact of the data in the models at some forecast times.
|GFAL Error (km)|
|GFP3 Error (km)|
|GFTG Error (km)|
|12||45.||35.||( 22%)||56.||( -24%)||45.||( 0%)|
|24||53.||24.||( 55%)||24.||( 55%)||53.||( 0%)|
|36||91.||0.||( 100%)||0.||( 100%)||67.||( 26%)|
|48||64.||86.||( -34%)||78.||( -22%)||14.||( 78%)|
|120||1076.||965.||( 10%)||992.||( 8%)||982.||( 9%)|
|Landfall||62.||51.||( 18%)||80.||( -29%)||43.||( 31%)|
|VBAL Error (km)|
|VBP3 Error (km)|
|VBTG Error (km)|
|12||95.||88.||( 7%)||81.||( 15%)||88.||( 7%)|
|24||123.||110.||( 11%)||105.||( 15%)||119.||( 3%)|
|36||178.||169.||( 5%)||151.||( 15%)||189.||( -6%)|
|48||347.||365.||( -5%)||328.||( 5%)||386.||(-11%)|
|72||626.||643.||( -3%)||602.||( 4%)||684.||( -9%)|
|84||1903.||1912.||( 0%)||1869.||( 2%)||1962.||( -3%)|
|96||2343.||2334.||( 0%)||2301.||( 2%)||2381.||( -2%)|
|108||2545.||2493.||( 2%)||2495.||( 2%)||2528.||( 1%)|
|GSAL Error (km)|
|GSP3 Error (km)|
|GSTG Error (km)|
|12||78.||89.||(-14%)||83.||( -6%)||33.||( 58%)|
|72||518.||532.||( -3%)||597.||(-15%)||493.||( 5%)|
|84||648.||641.||( 1%)||735.||(-13%)||592.||( 9%)|
|96||824.||783.||( 5%)||950.||(-15%)||838.||( -2%)|
|120||1346.||1304.||( 3%)||1526.||(-13%)||1421.||( -6%)|
|Landfall||292.||306.||( -5%)||371.||(-27%)||114.||( 61%)|
|GFAL Error (kn)|
|GFP3 Error (kn)|
|GFTG Error (kn)|
|24||6||16||( -167%)||1||( 83%)||16||( -167%)|
|48||12||19||( -58%)||10||( 17%)||21||( -75%)|
|72||2||1||( 50%)||1||( 50%)||0||( 100%)|
|84||5||2||( 60%)||3||( 40%)||6||( -20%)|
|96||13||12||( 8%)||11||( 15%)||13||( 0%)|
|108||9||8||( 11%)||7||( 22%)||5||( 44%)|
|120||4||8||( -100%)||7||( -75%)||7||( -75%)|
Return to Main page