A hurricane's path is quite easy to forecast, but its intensity is not. A new model, which takes into account the interaction of the storm and ocean-surface temperature, gives remarkably accurate retrospective simulations.
On the night of 3–4 October 1995, Hurricane Opal turned into a forecaster's nightmare. This atmospheric vortex, which had wandered around the southern Gulf of Mexico since late September, had gradually strengthened to become a weak hurricane on 2 October. Then, under the influence of a low-pressure trough blowing eastwards over North America, Opal lunged abruptly towards the southern shoreline of the United States. During its northward careen, Opal crossed a deep pool of warm water that had detached from the Gulf Stream Loop Current and intensified dramatically. In just 22 hours, its strongest winds increased from 40 to 67 metres per second, taking it from category 2 to category 4 on the Saffir–Simpson hurricane scale (Table 1).
Few coastal residents who stayed up to watch late weather broadcasts worried about the hurricane as they went to bed that night. But the situation had changed by sunrise and, for a few hours, it looked as though there could be a repeat of the 1969 Hurricane Camille. (Camille, the only category-5 hurricane on record to strike the US mainland, killed 256 people.) The complication in 1995 was that Opal came upon them overnight, with little warning and essentially no time to evacuate. But then, even more suddenly than they had intensified, Opal's winds weakened to 50 metres per second before the eye passed onshore at Pensacola Beach, Florida. In the end, Opal took nine lives in the United States and caused an estimated $3,000 million in property damage. Like last month's Hurricane Floyd, Opal's main effect was inland flooding, which killed 50 people in Mexico and Guatemala during the storm's formative stage.
‘Hurricane’ is the term used in the Western Hemisphere for the general class of tropical cyclones, including western Pacific typhoons and similar systems, that are known simply as cyclones in the southern Pacific and Indian oceans. They are intense atmospheric vortices that derive their energy from the warm surface waters of the summertime tropical ocean. Their damaging winds are concentrated within about 100 km of their centres, just outside the clear central eye. When a hurricane threatens, everyone wants to know where it will go, so the most widely appreciated success in hurricane forecasting has been in more accurate track predictions. About 80–90% of a hurricane's motion depends on flow in the surrounding, undisturbed atmosphere1. Predictions of this ‘steering flow’ have improved in parallel with advances in the art of numerical weather prediction. If present trends continue, by the year 2005 track forecasts made 48 hours in advance will be as good, on average, as those made 24 hours in advance were in 1970.
On the other hand, forecasters claim little skill in predicting the intensity of hurricanes beyond simple extrapolation and climatological experience. For many hurricanes, extrapolation is good enough because they strengthen or weaken slowly. But in the case of Hurricane Opal, a phenomenon known as ‘rapid deepening’ occurred — the central, sea-level pressure fell suddenly as winds increased.
Rapid deepening is responsible for most typhoons and all Atlantic hurricanes with winds of greater than 50 metres per second (Saffir–Simpson categories 3, 4 and 5). Although only 20% of hurricanes fall into this class, they cause 80% of hurricane damage in the United States2. Moreover, the onset of rapid deepening is unpredictable — it can change a category 2 hurricane into a category 4 or 5 hurricane in less than a day. In other words, the intensities of the most damaging hurricanes are the most difficult to predict. Yet the paper by Kerry Emanuel on page 665 of this issue3 reports startlingly good retrospective simulations of hurricane intensity changes.
The thermodynamics of a hurricane can be modelled as an idealized heat engine4, running between a warm heat reservoir — the sea, at around 300 K — and a cold reservoir, 15–18-km up in the tropical troposphere, at about 200 K (Fig. 1). A hurricane's maximum intensity (that is, the strongest wind speed or lowest sea-level pressure) is proportional to the difference in absolute temperature between these two reservoirs. Nonetheless, most real hurricanes are weaker than predicted by models based on pre-existent sea-surface temperatures.
What Emanuel has done is to apply to hurricanes a simple, time-dependent model written with his student Lars Schade5. The model represents the hurricane as a completely circular vortex that moves along a pre-defined track. The model's initial condition is adjusted to match the observed intensity, and the rate at which this intensity changes, at the start of the forecast. Ocean-surface temperature and depth of the warm surface layer are represented on a coarse spatial grid. Turbulent heat exchange at the surface drives the hurricane model. It is proportional to surface wind speed and temperature from the oceanic grid points beneath the storm, averaged on circles around the storm's centre position. Momentum transport from the vortex to the ocean causes the warm surface layer to mix with the cooler water below.
The key to the model's success is the reduced heat supply as the surface water cools in response to the storm. Despite many simplifications, the model produces spectacular forecasts of maximum winds for memorable storms such as Hugo (1989), Andrew (1992) and Opal. Other hurricanes — such as Camille (1969), Gilbert (1988) and Gloria (1985) — where the forecasts are less successful point to considerable roles for atmospheric interactions or actual ocean-temperature structures that differ from the monthly climatic averages used in Emanuel's preliminary calculations.
In response to the near-disaster caused by Hurricane Opal, a symposium on intensity changes in tropical cyclones was held at the 1997 meeting of the American Meteorological Society in Phoenix, Arizona. There, a consensus emerged that intensity changes in a hurricane are a response to three factors — forcing by the surrounding atmosphere, the oceanic heat source and the storm's internal dynamics. The difficulty of reliable predictions was attributed to the complex interplay of these factors. Emanuel's result, if it stands the test of forecasting experience, will shift the consensus towards a dominance of oceanic forcing. Perhaps intensity forecasting will turn out to be a much less difficult problem than we thought just a few years ago.
Holland, G. J. J. Atmos. Sci. 40, 328–342 (1983).
Landsea, C. W. & Pielke, R. A. Weath. Forecast. 13, 621–631 (1998).
Emanuel, K. A. Nature 401, 665–669 (1999).
Emanuel, K. A. Annu. Rev. Fluid Mech. 23, 179–196 (1991).
Schade, L. R. & Emanuel, K. A. J. Atmos. Sci. 56, 642–651 (1999).
About this article
Development of a Convection‐Permitting Air‐Sea‐Coupled Ensemble Data Assimilation System for Tropical Cyclone Prediction
Journal of Advances in Modeling Earth Systems (2019)
Journal of the Atmospheric Sciences (2017)
Journal of Marine Science and Engineering (2016)
Journal of the Atmospheric Sciences (2009)
Applied Physics B (2009)