IMPACT ON HURRICANE TRACK AND INTENSITY FORECASTS OF GPS DROPWINDSONDE OBSERVATIONS FROM THE SYNOPTIC SURVEILLANCE MISSION INTO HURRICANE LINDA ON 14 SEPTEMBER, 1997.

Sim D. Aberson

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

4301 Rickenbacker Causeway
Miami, Florida


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


1. Synoptic situation

Hurricane Linda developed from an easterly wave south of Mexico on 09 September, and moved westward, strengthening rapidly to become the strongest hurricane ever measured in the Eastern Pacific Ocean by 12 September. Due to the rare potential threat to the Southern California coastline, a synoptic surveillance mission was tasked for nominal time 15 September 1997 0000 UTC. At that time, Linda was embedded in the westerlies to the south of the subtropical ridge, about 1100 km west southwest of Cabo San Lucas (Fig. 1). A strong trough was located just west of the U. S. west coast, and a break in the subtropical ridge to the northwest of Linda provided the opportunity for a strong, recurving hurricane to strike California. The vortex near 12°N 135°W was Tropical Storm Marty. The vortex between the two tropical cyclones was a spurious vortex in the National Centers for Environmental Prediction Global Data Assimilation System, which negatively impacted the track forecasts.

2. General Assessment of dropwindsonde impact

A. GFDL model

Figure 2 shows the GFDL forecast tracks for Hurricane Linda, and Table 1 shows the errors and impact of the synoptic surveillance mission. The dropwindsonde data substantially improves the forecast at all times. The upper-tropospheric data also improved the forecast at all times. The dropwindsonde data pushed the forecast further to the south, away from the best track; however, the data also slowed the forward motion, leading to improved forecasts.

B. GSM

Figure 3 shows the GSM forecast tracks for Hurricane Linda, and Table 2 shows the errors and impact of the synoptic surveillance mission. The impact was similar to that for the GFDL model in that the dropwindsonde data pushed the forecast to the south away from the best track. However, the data did not substantially slow the GSM forecast as in the GFDL, so the forecasts were degraded at all times except 12 h. The upper-tropospheric data improved the forecasts at all times except 48 h.

C. Intensity

Table 3 shows the GFDL intensity forecast errors and impact of the synoptic surveillance mission. The dropwindsonde data had a negative impact on the forecast only at 24 h, and the upper-tropospheric data seems to have degraded the intensity forecast at all times. This is likely because Linda was undergiong shear at the time, and the vortex did not extend through the troposphere as the model expected. Tuleya and Lord (1997) 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 4 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 perturbations correspond to Linda itself. Other large perturbations pertain to the subtropical ridge axis extending from near San Diego southwestward nearly 2000 km, and a number of features circling the large midlatitude trough off the U. S. west coast. The feature which was sampled best was centered near Los Angeles, and was sampled with only two dropwindsondes. Only spotty data coverage was obtained in the subtropical ridge, although the center to the west of Linda was sampled except for its far-western extent.

One set of model runs, the TG set, includes the dropwindsonde data taken within and around the perturbation corresponding to the subtropical ridge to the west of Linda (the dropwindsondes extending from 26°N 128°W to 25°N 132°W, or about one-third of the total dropwindsondes released during the mission. Results are shown in Tables 1-3 and Figs. 2-3. The TG run provided degraded forecasts compared to the run including all the dropwindsonde data at all forecast times in both the GSM and GFDL. The effect was to increase the southward displacement of the forecast from the run without the dropwindsonde data, and the tracks, especially in the GSM, were similar to those in which all of the dropwindsonde data were used.

Figure 5 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 difference is in the large trough off the California coast. Other differences extend along the axis of the subtropical ridge from just southwest of San Diego to 1000 km west of Linda. The difference closest to Linda extends more than 500 km southwest of the location of the dropwindsonde data, and this extension of the data forced the forecasts with the extra data southward. Since this extension is in an area in which the ensemble perturbations are large, the effect can be expected to amplify in time. Figure 6 shows that, by 24 h into the forecast, all the differences between the two forecasts have amplified. The large perturbation off the California coast has moved rapidly northeastward and did not impact the track of Linda. The large difference southwest of San Diego has amplified and moved slowly northeastward, also without impacting the track of Linda. The maximum difference initially located near 23°N 132°W has slowly amplified and moved westward, also without impacting Linda. The large difference initially just to the west of Linda has slowly amplified and erroneously pushed Linda to the south, due to the anomalous extension of the effect of the data in the assimilation. Since all the largest differences were in areas of large ensemble perturbation and amplified in time, these results confirm that the largest (smallest) ensemble perturbations correspond to amplifying (decaying) modes in the model. Sampling further to the southwest in along the subtropical ridge may have had a larger positive impact on the forecast track due to possible elimination of the erroneous impact spread.

4. Conclusion

The dropwindsonde data obtained during the synoptic surveillance mission for Hurricane Linda at nominal time 15 September 1997 0000 UTC has provided mainly negative results. The MRF ensemble forecasting system suggests that data obtained alont the axis of the subtropical ridge north of Linda would have the greatest impact on the Linda forecast. Dropwindsonde data obtained in this region was spread by the data assimilation to the southwest, pushing the forecasts with all dropwindsondes toward the south largely degrading the forecasts. The southwestern extent of this region was not sampled, suggesting that forecast improvements would have been possible.


Table 1
Track forecast errors for the no dropwindsonde GFDL control (GFNO), the all dropwindsonde run (GFAL), the lower-level dropwindsonde run (GFP3), and the run with only targeted observations (GFTG), and the percent improvement of the latter three over the control.
Forecast
time (h)
GFNO
Error (km)
GFAL Error (km)
(% Improvement)
GFP3 Error (km)
(% Improvement)
GFTG Error (km)
(% Improvement)
1295.82.(14%)83.(13%)83. ( 13%)
24140.122.(19%)150.( 7%)167. (-11%)
36207.182.(12%)189.( 9%)235. (-14%)
48239.205.(12%)219.( 8%)261. ( -9%)

Table 2
Track forecast errors for the no dropwindsonde GSM control (GSNO), the all dropwindsonde run (GSAL), the lower-level dropwindsonde run (GSP3), and the run with only targeted observations (GSTG), and the percent improvement of the latter three over the control.
Forecast
time (h)
GSNO
Error (km)
GSAL Error (km)
(% Improvement)
GSP3 Error (km)
(% Improvement)
GSTG Error (km)
(% Improvement)
12 106. 98.( 8%)106.( 0%)106.( 0%)
24 234. 244.( -4%)257.(-10%)273.(-17%)
36 400. 455.(-14%)441.(-10%)480.(-20%)
48 562. 593.( -6%)582.( -4%)645.(-15%)

Table 3
Intensity forecast errors for the no dropwindsonde GFDL control (GFNO), the all dropwindsonde run (GFAL), the lower-level dropwindsonde run (GFP3), and the run with only targeted observations (GFTG), and the percent improvement of the latter three over the control.
Forecast
time (h)
GFNO
Error (kn)
GFAL Error (kn)
(% Improvement)
GFP3 Error (kn)
(% Improvement)
GFTG Error (kn)
(% Improvement)
121313( 0%)13( 0%)12( 8%)
241012(-20%)11(-10%)12(-20%)
3677( 0%)6( 14%)7( 0%)
481111( 8%)10( 9%)10( 9%)


Return to 1997 assessment page
Return to HRD home page