![]() | Discussion | ![]() |
Additionally, workshop participants identified a need to reduce 24 and 48 h intensity forecast errors of all landfalling tropical cyclones to 5 and 10 ms-1, respectively. While average forecast errors between 1988 and 1997 meet this goal (Mark DeMaria, personal communication), large errors remain common. The forecast improvements shown in the current sample could also lead to reaching the intensity forecasting goal, though the GFDL model has yet to regularly show forecast skill within the hurricane warning period.
The current results confirm those of Burpee et al. (1996) that the collection of operational dropwindsonde observations in the environment of tropical cyclones threatening landfall is an effective means of reducing track forecast errors. The forecast track improvements from dropwindsonde data are at least as large as those accumulated over the last 20 to 25 years ( McAdie and Lawrence 1993). Additionally, synoptic-scale observations, especially in the upper troposphere, can also improve intensity forecasts from dynamical models, in agreement with Tuleya and Lord (1997). Research into targeted observing strategies and better data assimilation methods suggests that even larger track forecast improvements than those shown are possible ( Aberson 1998b).