Hurricane Detection, Tracking, and Forecasting

A policy statement of the American Meteorological Society as adopted by the Executive Committee on 23 April 1993


  1. Introduction
  2. Scientific understanding of tropical cyclones
  3. Hurricane-related hazards
  4. The state of hurricane tracking and forecasting
  5. Prospects and priorities for the next decade

1. Introduction

Our ability to understand and deal effectively with hurricanes is tested annually in the United States and worldwide. In 1992, three major hurricanes — Andrew, Iniki, and Omar — caused extensive damage to the United States. As these hurricanes approached our coastline, the importance and need for accurate forecasting of hurricane movement and intensity and for improved emergency preparedness planning and response was made clear to the nation.

Hurricanes (typhoons in the western North Pacific) are a member of a class of systems referred to as tropical cyclones. The term hurricane is applied whenever the surface winds, rotating about a tropical cyclone center, or "eye," reach a sustained wind speed of 64 kt (74 mph or 33 m/s). Although there are some differences in strength and behavior of tropical cyclones in the various regions of the tropics, the dynamics of these disturbances are similar. This statement examines the hurricane problem from a research and scientific point of view. A separate AMS Policy Statement (Bull. Amer. Meteor. Soc.,May 1986, p. 537) addresses the hurricane warning system and preparedness efforts

Top of Document

2. Scientific understanding of tropical cyclones

Tropical cyclones form from relatively common tropical weather systems, referred to as cloud clusters. These groups of loosely organized, deep cumulus clouds occur in a variety of tropical weather situations, but in the Atlantic the most common pattern for storm genesis has historically been intensification of tropical waves that regularly move off the west coast of Africa during the Atlantic hurricane season. Most cloud clusters and tropical waves, however, do not evolve into tropical cyclones. In this sense, the hurricane is a rare phenomenon.

The initiation of a vortex with winds of moderate strength (cyclogenesis) can occur very rapidly, often in less than a day. The climatology of Atlantic tropical cyclogenesis suggests that formation is favored when a strong convective disturbance occurs in a region where the air is already "spinning" in a cyclonic (counterclockwise) direction. Other favorable factors, such as weak vertical wind shear, low-level inflow, and high-level outflow, have also been identified. Interactions between incipient disturbances and upper-tropospheric systems often contribute to cyclone development as well. Genesis almost always occurs over warm tropical waters. The dynamics of the initial stage of the tropical cyclone's life cycle is not well understood due to the lack of observations in the regions of storm genesis and the complexity of the interactions between the many scales of motion involved in formation.

Intensification of the weak circulation into a hurricane can be thought of as the evolution of a vortex in which the dominant forces are in approximate balance. The balance of forces near the sea surface is altered by the friction, causing moisture-rich air to move toward the storm center. Clouds near the center are organized into spiral-rainband structures by a complex, poorly understood interaction between the physics of the clouds, the strong rotation in the vortex, and the atmospheric conditions in the environment of the storm. The strengthening winds extract ever larger amounts of water vapor from the warm ocean. As this water vapor rises near the center, it cools and condenses; the latent heat thus released creates a warm central core, and air is drawn toward the center, contracting the vortex and further spinning up the winds. The reasons that some disturbances intensify to hurricane, while others do not, are not well understood. Also unknown are the reasons that some hurricanes become severe while others do not.

Although the small-scale details of the storm may change continuously, and sometimes rapidly, the tropical cyclone, as a whole, is a stable system that may persist for many days over the warm tropical ocean. During this time, a tropical cyclone moves in the general direction of the broad-scale wind patterns in which it is embedded. Tropical cyclones dissipate rapidly after landfall, due primarily to the loss of the surface moisture source. The vortex may retain some organization, particularly in the middle troposphere, for several days after landfall. Storms that move poleward over cold waters tend to weaken at a slower rate than those storms that move over land. In either case, the circulation center frequently interacts or combines with a midlatitude weather system and, in the process, loses its warm core structure. The transformed system can still produce substantial rainfall.

Top of Document

3. Hurricane-related hazards

In the coastal zone of the United States, extensive damage and loss of life are caused by the storm surge, heavy rains, strong winds, and hurricane-spawned tornadoes. Usually, when large loss of life occurs, it is due to the storm surge. The height of the storm surge varies from 3–5 ft (1–2 m) for weak systems to more than 20 ft (6 m) for strong storms striking areas with shallow water offshore. The dome of water associated with Hurricane Andrew reached a height of about 17 ft, the highest level recorded for the southeast Florida Peninsula, and with Hurricane Hugo (1989) reached a height of nearly 20 ft (6 m) about 20 miles northeast of Charleston, North Carolina. In the case of Hurricane Hugo, the surge exceeded 10 ft (3 m) over a distance of nearly 100 miles at the coastline.

In regions with good building codes, wind damage is typically not as catastrophic as storm surge damage, but affects a much larger area and can lead to large economic loss. For instance, winds associated with Hurricane Andrew produced over $20 billion of damage over the southern Florida and Louisiana area. Tornadoes occur in most hurricanes that strike the United States, but generally account for little of the total storm damage.

Although hurricanes are mainly coastal hazards, the weakening storm circulation, with its moisture-laden air, can produce extensive flooding hundreds of miles inland long after the winds have lost hurricane force. Occasionally the damage from inland flooding exceeds storm surge destruction. Although the deaths from storm surge and wind along Florida's coast from the remnants of Hurricane Agnes in 1972 were minimal, over 200 deaths were attributed to inland flash flooding over the northeastern United States. Not all hurricane-related phenomena are detrimental to humankind, however. Hurricane rainfall, for example, has often benefitted drought-stricken areas.

Top of Document

4. The state of hurricane tracking and forecasting

a. Hurricane tracking

Since tropical cyclones usually form far at sea and spend much of their existence over remote oceanic areas, detection and monitoring of these storms have traditionally posed serious problems to the forecaster. The advent of geostationary weather satellites has largely solved the detection problem and has improved the monitoring problem. However, the satellites are remote sensors and it is not unusual for position estimates to have errors of tens of miles or for wind speed estimates to be in error by tens of knots. Although advancements have been made using microwave imagery, it is still not possible to determine surface wind field distributions or detailed structural characteristics of tropical cyclones from present satellites. A combination of observing systems is necessary to provide the data required for accurate forecasts and warnings.

The principal sources of data in addition to weather satellites are reconnaissance aircraft, coastal radars, and measurements from ships, buoys, and land stations. Reconnaissance aircraft can measure details of a storm's structure when a tropical cyclone is within range of staging areas used by the aircraft. Theses specially instrumented aircraft provided accurate information on the storm's position and its current state of evolution. Surface winds can be measured from the aircraft using remote sensing techniques. Despite advancements of remote sensing capabilities from satellites, measurements from reconnaissance aircraft will be required for the foreseeable future to maintain the present level of accuracy for forecasts of landfalling tropical cyclones in the United States.

b. Hurricane track forecasting

The environmental flow in which the tropical cyclone is embedded is the main factor determining its track. Also important for track and intensity forecasts are the internal structure of the tropical cyclone and the interaction of this structure with its environment. Accurate prediction requires detailed measurements on scales that range from the storm's large-scale environment to its small inner core.

A recent assessment of the role of aircraft reconnaissance on tropical cyclone analysis and forecasting (Gray et al. 1991 ) concludes that center-fix aircraft reconnaissance is required for optimum short-range (0–24 h) forecasts of cyclone landfall and storm surge. Unfortunately, the present reconnaissance aircraft fleet and weather satellite information cannot provide the full three-dimensional data required for hurricane track forecasting. Omega dropwindsondes deployed from the aircraft can provide wind, temperature, and moisture information from flight level to the surface, and have been shown to have a positive impact on track forecast models. The aircraft are relatively slow, however, and the information derived from the sondes does not cover the important region above flight level. The remote-sensing satellite data are limited in accuracy and coverage, particularly at the critical middle-troposphere levels.

A tropical cyclone forecast involves the prediction of several interrelated quantities, but the fundamental element of the forecast is the future motion of the storm. Track prediction serves as the basis for forecasting other storm features, such as winds, rainfall, and storm surge, and, of course, the areas threatened. Normally, motion forecasts out to 72 h are issued every 6 h. These forecasts are based on guidance, ranging from simple climatological aids to complex dynamical and statistical–dynamical models. However, inconsistencies in quality and availability of these guidance products limit their utility. Thus tropical cyclone forecasting remains rather subjective, and forecaster skill and experience are important ingredients in the success of the forecast.

National Hurricane Center (NHC) forecast errors, the distance between a forecast and the subsequent observed position of the storm center, for the decade 1982–1991 averaged 54 (100), 104 (193), 206 (383), and 309 (573) n mi (km) for the 12-, 24-, 48-, and the 72-h forecasts, respectively. Using a combination of climatology and persistence as a basis for comparison, track forecast skill exists at time intervals out to 72 h, with the 48-h forecast showing the highest level of skill. Forecast errors show large spatial variation, averaging up to 30% greater than the mean in the central Atlantic, and up to 30% less than the mean over the Gulf of Mexico and the Caribbean Sea. These differences arise from better data availability in the latter areas, as well as from the different characteristics of hurricane motion within these areas.

A recent study shows that NHC 24-h forecast errors have declined about 14% over the past 20 years. This decline can be attributed to various factors, especially the improved ability, beginning in the early 1960s, to monitor and track these storms with satellite imagery. Recent improvements in dynamical and statistical–dynamical models used as forecast guidance have also contributed to decreased errors.

Consistent with current forecast accuracy, it is necessary to issue hurricane warnings for rather large coastal areas. Warnings issued 24 h before hurricane landfall average 300 n mi (560 km) in length. Normally, the swath of damage encompasses about one-third of the warned area, so the ratio of affected area to warned area is about one to three. In other words, approximately two-thirds of the area is, in effect, "overwarned." Such overwarning is not only costly, but also results in a loss of credibility in the warnings. The National Weather Service's Hurricane Probability Program was implemented as an attempt to quantify the uncertainty implicit in hurricane forecasts.

c. Forecasting hurricane intensity and coastal hazards

Numerical models can predict the storm surge inundation associated with a given hurricane with a reasonable degree of accuracy provided that the forecast of the hurricane's track and intensity are adequate. However, in view of the inherent inaccuracy in track forecasting, overwarning of storm surge flooding remains a problem.

Considerable improvement is needed in the understanding and prediction of tropical cyclone intensity changes. Present operational forecasts are only slightly better than objective forecasts that are based on persistence and climatology. Mean NHC absolute errors of maximum hurricane wind speed, most often based on satellite estimates for the decade 1981–1990, are 8.2 (4.2), 11.4 (5.9), 15.6 (8.0), and 19.1 (9.8) kt (m s-1) for the 12-, 24-, 48-, and 72-h forecasts, respectively. These errors are deceptively low, however, since they are heavily weighted toward the average condition where intensity changes are gradual and persistence forecasts work well. They do not reflect the occasional large misses that can occur with rapid strengthening or weakening of a storm. The inability to anticipate these changes for a storm that is less than 24 h from landfall is of great concern.

There is little skill in the prediction of hurricane-related rainfall. Areas of heavy rainfall can be monitored, however, from conventional radars along the United States coast. The prediction problem is complicated by terrain effects and uncertainties arising from the forecast of the track. Although in situ estimates by research aircraft of the precipitation distribution in a hurricane are becoming available, rainfall prediction remains rather subjective. Estimates of rainfall based on satellite imagery and numerical models appear to offer promising avenues for improvement.

Top of Document

5. Prospects and priorities for the next decade

Despite reductions in track forecast errors from objective dynamical and statistical–dynamical models in recent years, operational hurricane forecast errors have not decreased enough to solve the forecast problem. To achieve smaller errors, the models that provide objective guidance to the forecasters must become more reliable, and more data are needed. Until such improvements can be effected, forecasting methods will continue to be subjective; disaster preparedness officials and the public should be kept aware of current limitations in forecast accuracy.

The reliability and interpretation of the output from the dynamical forecast models is clouded by uncertainties in specifying their initial conditions. Therefore, obtaining an improved initial analysis should be a high priority. These improvements must come about primarily through increased data availability, but also through better analysis methodologies. Additional reliable data should be acquired in the hurricane and its environment throughout the depth of the troposphere. Reconnaissance aircraft with upgraded instrumentation is also an important component for improved data coverage in the hurricane core. Simulation experiments should be designed, as well, to determine the observations that are optimal for improved initialization of track forecast models. In the absence of true observations, interjection of synthetic data to represent a hurricane-like disturbance in the dynamical model initial conditions is an alternative strategy being pursued by several forecast centers. Improved representation of tropical cyclones in the numerical forecast models has recently become feasible because the analysis centers have begun distributing basic quantitative information on storm positions and intensity.

Global models provide boundary conditions for dynamical hurricane models with limited domains, and have sufficiently high resolution to provide competitive forecasts of hurricane tracks out to 72 h. The improved performance of 48- and 72-h forecasts during the past few years indicates that these models are getting better, but still perform relatively poorly in the data-sparse tropical oceanic areas where hurricanes form and move. Improvements needed included better representations for some physical processes, such as cumulus convection and air–sea interaction, which are of particular importance in the tropics. It is encouraging that the latest high-resolution hurricane-prediction models allow improved representation of the hurricane's inner core structure. These new models show promise for the ability to predict trends in the storm intensity since they can resolve mesoscale features (eye, eyewall, spiral bands) that are important for these changes. However, the forecast ability of the models will be limited by the sparsity of available data. Further observational studies are needed as well to improve our conceptual models of the physical processes through which tropical cyclones vary in intensity.

The continued advancement in computer technology, and progress in dynamical modeling and associated initial analysis methodology, has moved well beyond the ability of the available data to describe the initial state to the accuracy and resolution required by these models. For us to realize the potential of existing models and those on the near horizon, major improvements in the data available for the inner core of the tropical cyclone and its near and far environment throughout the troposphere are required. Such data acquisition seems possible with current technology using midsize jet aircraft providing in situ and remote sensing capabilities for use in the tropical cyclone and its near environment, coupled with improved satellite remote-sensing capabilities and perhaps unmanned aircraft providing in situ data. These data would be supplemented by coastal land-based Doppler radars. A potential for substantial improvements in tropical cyclone track and intensity forecasts would seem to be within reach if a substantially greater data coverage could be made available to the more advanced models on a regular basis.

We must also guard against any deterioration of our observing platform network. For example, the importance of meteorological satellites in hurricane detection and monitoring must not be underestimated. The current GOES weather satellite configuration is limited to one aging satellite that must support many operational needs. Replacement satellites may not be ready until the mid-1990s, resulting in a chancy reliance on the current satellite and contingency plans.

The primary goal of both research and operational groups is to minimize loss of life from hurricanes. Unfortunately, evacuation times for some communities now exceed what can reasonably be expected from present and projected forecast abilities. Thus, a concerted scientific effort to improve forecasts must be combined with community development and preparedness programs to reduce evacuation times. The hurricane problem is complex and difficult, but is not insurmountable.


Gray, W. M., C. J. Neumann, and T. L. Tsui, 1991: Assessment of the role of aircraft reconnaissance on tropical cyclone anlysis and forecasting. Bull. Amer. Meteor. Soc., 72, 1867–1883.

Top of Document Policy Statement Home Page AMS Home Page

© 1996 American Meteorological Society