Every year, on average, 84 tropical storms (including 43 hurricanes) form around the world, with considerable variability by ocean basin. Basin size and location, proximity to continents (dry air), sea surface temperatures, wind shear, and relative humidity, to name several, all contribute to tropical cyclone variability by region. To study how these factors affect tropical cyclone formation, make seasonal and real-time forecasts, and assess potential losses in catastrophe modeling, researchers depend on best track data.
Best track data represent not only the "best" location (storm track) estimates for tropical cyclones throughout the lifetime of the storm, but also the intensity (central pressure and/or wind speed) estimates at each point along the track. Because these storms are complex and highly dynamic and observations are limited and subject to uncertainty, scientists need to conduct post-event analysis of all available data to ensure that each storm is appropriately represented in the best track data set. Best track data have a wide range of applications across disciplines. Wind engineers use best track data to inform the development of building codes; forecasters use them for model validation; and climatologists use them to study climate change. In catastrophe modeling, researchers use best track data to develop and validate probabilistic landfall frequency and intensity estimates to inform a long-term view of tropical cyclone risk.
This article discusses how these data are developed, sources of uncertainty in the data, and how the data are used in catastrophe modeling.
Regional Specialized Meteorological Centers
The World Meteorological Organization's (WMO) Tropical Cyclone Program (TCP) aims to minimize the loss of life and damage caused by tropical cyclones. As part of achieving this objective, the TCP is tasked with providing tropical cyclone track and intensity forecasts as well as best track data for historical tropical cyclones.
The WMO divides regions of tropical cyclone formation into seven basins, monitored by six Regional Specialized Meteorological Centers (RSMCs) and six regional Tropical Cyclone Warning Centers (TCWCs), which have similar responsibilities (see Figure 1 and Table 1). Note that tropical cyclones have been observed in the Mediterranean Sea and off the coast of Brazil; however, due to the infrequent occurrence in these regions, the WMO has not designated a forecast center for these regions. Best track data are the official responsibility of Regional Specialized Meteorological Centers (RSMC) and are updated each year with the previous year's tropical cyclone information.
In each basin, RSMCs and TCWCs are responsible for collecting and disseminating all tropical cyclone-related information, including advisories, warnings, and best track data. The National Hurricane Center (NHC; RSMC Miami) and the Japan Meteorological Agency (JMA; RSMC Tokyo) are two well-known and busy centers.
Table 1. RSMCs and TCWCs by ocean basin with tropical cyclone (TC), tropical storm (TS), and hurricane (Hurr) average annual frequencies.
||National Hurricane Center
||Central Pacific Hurricane Center
||Central Pacific (III)
||Japan Meteorological Agency
||Northwest Pacific (IV)
|RSMC New Delhi
||India Meteorological Agency
||North Indian (V)
|RSMC La Reunión
||South Indian (VI)
||Australian Bureau of Meteorology
||Southwest Pacific (VII)
||Indonesian Agency for Meteorology
||Southwest Pacific (VIII)
||Australian Bureau of Meteorology
||Southwest Pacific (IX)
|TCWC Port Moresby
||Papua New Guinea
||Southwest Pacific (X)
||Australian Bureau of Meteorology
||Southwest Pacific (XI)
||Fiji Meteorological Service
||South Pacific (XII)
||New Zealand Meteorological Service
||South Pacific (XIII)
Best Track Data Tools
The tropical cyclone tracking agencies all use remotely sensed, or satellite, data and land-based radar data to estimate storm intensity and location. The NHC is the only agency to also routinely fly aircraft into tropical cyclones to directly measure storm characteristics. One might assume that direct observation is superior to remotely sensed data, but that is not necessarily the case because both methods are subject to uncertainty for different reasons that are explained in the next section. The two methods are complementary, but aircraft reconnaissance is costly and not widely used outside of the U.S.
The NHC's aircraft reconnaissance teams, commonly referred to as "Hurricane Hunters," use dropsondes (instrumentation packets dropped from the aircraft) to measure a storm's characteristics within the vicinity of the plane, including wind speed and direction, pressure, temperature, and relative humidity. In addition, the planes use Stepped Frequency Microwave Radiometer (SFMR), a scatterometer that remotely estimates wind speed based on the ocean surface brightness temperature below the aircraft, and radar is used to remotely estimate precipitation intensity and wind speed and direction at various levels. All these tools provide unique information that allows meteorologists to estimate the current location, intensity, and movement direction of a tropical cyclone (Figure 2). All North Atlantic reconnaissance data can be found at http://www.nhc.noaa.gov/recon.php.
Aircraft reconnaissance data are subject to uncertainty for a number of reasons. It is not used consistently for all storms because of costs, different aspects of the entire storm cannot be sampled simultaneously, and planes need to return to base to refuel, resulting in data gaps. There is also uncertainty in peak wind speed measurements because it is difficult to determine where the strongest winds are occurring. Because the center of the storm circulation is typically easy to identify from satellite imagery or radar, there is less uncertainty in measuring the minimum central pressure of a storm using dropsondes, especially of intense storms with well-defined eyes.
Furthermore, how efficiently momentum is transferred from the mid-levels to the surface is constantly changing, which makes it difficult to know whether the peak wind was sampled. Because tropical cyclones are dynamic, there is also uncertainty in the assumption that the sampled part of the storm is representative of the entire storm. As a result, the NHC tends to be conservative in their estimates and errs on the high side for intensity estimates.
Satellite Remote Sensing
The majority of agencies that do not use aircraft reconnaissance heavily rely on the Dvorak technique4 to estimate tropical cyclone intensity. The Dvorak technique originated as a purely subjective cloud pattern matching technique in the late 1960s, then evolved to become a more objective assessment based on infrared satellite data in the late 1990s (Figure 3). The Dvorak technique uses satellite estimates of cloud-top brightness temperature (in general, the colder the cloud top, the more intense the storm).
Because storm size, environmental pressure, and sea surface temperatures (among other characteristics) vary by basin, basin-specific wind-pressure relationships were developed to estimate central pressure and wind speed.
Sources of uncertainty in the Dvorak technique relate to the inherent limitations in empirical methods, which allow for the possibility of subjective analysis and user error. Tropical cyclones are complex and dynamic systems; the relationship between cloud top temperature and surface wind speeds varies from storm to storm, even for ones that appear similar on satellite imagery.
HURDAT—The World's Best Known Best Track Data Set
The Atlantic hurricane database, or HURDAT, is maintained by the National Hurricane Center (RSMC Miami) and is arguably the most well-known and scrutinized best track data set in the world. Beginning in the late 1990s, the NHC has been reanalyzing and revising historical tropical cyclone data as part of the HURDAT Reanalysis Project. This effort improves upon earlier tropical cyclone frequency and intensity estimates and removes systematic biases and errors, some of which were the result of antiquated analysis techniques that are inconsistent with the latest understanding of tropical cyclones. Additional variables, including radius of maximum wind, environmental pressure, and various wind radii (distance from center to 34-, 50-, or 64-knot winds by quadrant), are also provided—if discernible in the data.