Where will Hurricane IRENE go and how strong will it get? These are the questions forecasters at the National Hurricane Center (NHC) have been trying to answer ever since Irene became the 9th-named storm of the 2011 Atlantic Hurricane season, and which they will continue to try answering until this tropical cyclone finally ceases to be one.
Forecasting the center track and intensity of a tropical cyclone (TC), which is the technical designation for hurricanes, requires a prediction of what the global and regional state of the atmosphere will be in three or five days, and of how such conditions may continuously influence the behavior of one specific TC over time.
Hurricane forecasting is a critically important tool for emergency managers and civil protection officials, which use it to implement plans and make decisions to protect life and property in vulnerable communities, ahead of an approaching storm. Essential as it is, and despite considerable improvements in accuracy over the past 25 years, hurricane forecasting remains part science and part art, a process where arrays of instruments on board satellites and airplanes combine with supercomputers and complex forecast models to provide critical information, but which still depends on the expertise, experience and the acute eye of individual forecasters for the indispensable human factor.
Forecast models are tools used by forecasters to generate a prediction of the center track and intensity of a TC over time, which really means predicting what will happen in the regional atmosphere or area of influence of a moving tropical cyclone as well as what elements of a dynamic global atmosphere may become factors that affect at the regional level. Given the complexity of the physics of the atmosphere and highly complex structure of tropical cyclones, we must acknowledge that forecast models are only able to provide an approximate somewhat crude representation of the state of the atmosphere at any given time, however they are the best we’ve got.
There are several types of forecast models used by the NHC, which as a matter of practice uses them all individually or in combinations known as ensembles to arrive at a consensus prediction.
It is beyond the scope of this article to provide detailed descriptions of the various types of forecast model and how they are used, however I am including a list and brief description of these models for those who might be interested in researching this further:
(a) Statistical Models
These models use storm-specific information such as location (Latitude, Longitude), date and others, to compare it to the behavior of storms generated at the approximate same location and time of the year that are recorded in the historical records of cyclogenesis. Clearly this is just a statistical exercise where the caveat “past performance is no assurance of future results” is always present. On the other hand experience has shown these models can provide a good approximation of what may happen with a tropical cyclone currently under study. Also on the positive side, statistical models can be run rather quickly using little computing power, giving forecasters important assessments and general guidance to help in the prediction.
(b) Dynamical Models
These are highly complex models, which require considerable computing power and time to run, that set initial conditions based on observations of the state of the atmosphere either for the entire planet or just regionally for the area of influence of the specific tropical cyclone, or both. Such initial conditions and the projected interactions of various components of the atmosphere over time are then used to “run” the model using mathematical equations to represent the physics of the atmosphere to predict the center track and intensity of a TC.
Observations to set initial conditions come from data captured by satellites, hurricane hunter airplanes, surface ships and on land meteorological observatories. Lack of observations lead to deficiencies in data, which make the model initial state differ from the true state of the natural atmosphere and this in turn becomes a source of uncertainty introduced into the model run. Knowing this, it is clear that even a small starting error can become magnified, and grow quite large, the later you go in the forecast period. This is why the so-called cone of uncertainty used by the NHC to surround the TC central track starts out narrow and expands outward toward the end of a 3 o 5-day forecast. Relative to this, the margin of error for a predicted central track averages 400 kilometers either way by the end of a five-day forecast.
Despite these potential drawbacks, dynamical models are among the most trusted tools used by forecasters at the NHC.
(c) Statistical-Dynamical Models:
These type of models combine statistical information from the historical record with predictors of atmospheric steering mechanisms derived from dynamical models. When compared with results obtained from the dynamical models these are not as reliable, and consequently are used far less.
(d) Trajectory Models:
These are rather simple models, which can be run in minutes with little computing power, using prevailing flow data from the dynamical models to move the TC along. One weakness is that these models do not make the TC interact with the surrounding atmosphere, which limits their accuracy considerably.
(e) Ensemble and Consensus Forecast Models:
The most accurate predictions are obtained by combining multiple runs of a single specific model or runs from a collection of different models, in what is called an ensemble. What in essence emerges from running such ensembles, which requires enormous computational power or supercomputers and considerable time, is a consensus forecast that has over time proven more accurate than predictions from any given individual dynamical model.
The most recent phase in the evolution of ensemble consensus forecasting is the use of corrected ensembles, in which weights or multipliers are assigned to individual models to correct some of their weaknesses in an effort to make all of the models in an ensemble more similar in their performance.
In using the ensemble approach, but also the individual dynamical models, NHC scientists and forecasters need to constantly calibrate individual model performance by comparing it to actual results to then try an correct for deviations. This is an effort to have past performance be as representative of current performance as possible.
What this means is that using forecast models is a dynamic and interactive method of TC prediction, which must continue evolving and improving in the future.
A Delicate and Complex Dance in the Atmosphere!
Learning or knowing about the use of forecast models and realizing that some of the atmospheric flow and features used for predicting Hurricane IRENE’s track for example, began as atmospheric activity respectively over the Pacific and the northern Indian oceans some 20,000 kilometers ( half a world away) moving in opposite directions, is truly awesome.
I am referring to the waves of weather originating over the central Pacific ocean, which moved eastward toward North America where one or more interacted with the jet stream and land features to generate weather patterns that eventually lead to the trough over the southeastern USA, and the regions of high pressure straddling the southern USA and Gulf.
I am also referring to cells of disturbed weather that saw their genesis over the Indian ocean to then move westward over equatorial Africa, to become tropical waves as they emerged over the eastern Atlantic ocean near the Cape Verde islands to travel along hurricane alley where the influence of various factors and a favorable coupled ocean atmosphere resulted in the generation of the tropical cyclone that became Hurricane IRENE, which is now being steered on a path influenced by atmospheric conditions partially triggered by that over-the-Pacific-ocean event after a journey of more than two weeks!
The information we have shared here give us an approximate idea of how dynamic and complex the atmosphere is, the process of cyclogenesis, and the science and art that is involved in using forecast models to predict the center track of a hurricane five-days hence as all of the various pieces of this puzzle continue to individually evolve and move with respect to one another. The beauty and immensity of this feat become all the more outstanding when one considers that what started as separate cells in the atmosphere 16 days ago and 20,000 kilometers apart will soon coincide near the USA coast as Hurricane IRENE squeezes through, and that NHC forecasters using what amount to rudimentary sketches of the state of the atmosphere and some math have been able to predict it. Wow!
No Rest for the Weary!
Take all of the above in, enjoy it for a moment and now think about this: what has been described is just the successful prediction of the central track and intensity on just one tropical cyclone in the 2011 Atlantic Hurricane Season. But as we look east over the Atlantic and equatorial Africa there are already other tropical waves that warrant close monitoring for potential cyclonic development. and we still have more than half of the ‘official’ Atlantic hurricane season ahead of us. Also there continues to be tropical cyclonic activity over the eastern, central and western Pacific and over the Indian ocean and sub-basins off northern Australia and the Arabian sea. This activity will gradually shift southward as changes driven by the tilt in the Earth’s axis bring spring and then summer to the planet’s southern hemisphere and the tropical cyclone activity that goes with this.
In summary, tropical cyclone forecasting is a year-round activity requiring forecasters to work hard in between seasons, to review the results of their forecasting efforts, measure the levels of uncertainty and their accuracy in order to calibrate, enhance and improve to be ready for next year’s hurricane season. It never ends!