MainIn addition to the errors inherent in the forecast process itself, errors in the prediction of TC intensity can arise from uncertainties about the current storm maximum wind speed and the radial wind distributions in the various storm quadrants. This deficiency is also related to wind averaging times (sustained winds vs. gusts). These factors are discussed in this section.
2.1 Measurement of Tropical Cyclone Winds
Forecasters should be aware that the maximum wind speeds in TCs, as well as the radial decrease of wind from the radius of maximum wind speed, are seldom measured with a degree of precision implied by wind definitions. Tropical cyclones rarely pass directly over measurement devices. When they do, the devices are often incapacitated by the strong wind speed. Thus, there is heavy reliance on indirect estimates of surface TC wind speeds and directions such as wind data provided by aircraft, satellite cloud imagery, radar, etc. However, Gray, et. al. (1991) have shown that intensity estimates from these platforms are subject to various degrees of error.
Forecasters will note that intensity estimates are often stated in terms of pressure rather than wind speed. There is a strong empirical and theoretical relationship between these two parameters, thus they are used interchangeably to describe intensity. In the western North Pacific, the Atkinson-Holliday (1977) statistical relationship between wind and pressure is used by Navy personnel. Some solutions to their wind/pressure regression equation are given in Table 6.1.
Martin (1988) found that the most accurate intensity estimates (other than those from a ship or land site actually within a storm) are provided by reconnaissance aircraft observations. These types of observations are only available in the Atlantic and central Pacific. The mean absolute intensity observation error by aircraft is approximately 5 hPa. The mean error is approximately 15 hPa when the primary fix platform is satellite and 12 hPa when the primary fix platform is radar.
Using satellite imagery for intensity estimates can sometimes cause a serious underestimation of storm intensity at the critical decision levels (50 to 70 kt). For example, a satellite intensity estimate is reported to be 40 kt, which, as shown in Table 6.1 , is equivalent to 995 hPa central pressure. Subtract the mean error associated with satellite estimates (15 hPa) and the central pressure becomes 980 hPa which is equivalent to 60 kt. Decisions made based on a 60-kt forecast are much different than for those based on a 40-kt forecast. Therefore, great caution should be used when making recommendations based on satellite fixes alone. Satellite fixes are used extensively in the Pacific and Indian oceans.
2.2 Wind Speed Averaging Times
Since surface winds and gusts can change dramatically over short time intervals, it is necessary to define the length of time over which the winds are to be measured. For a cyclone of some given intensity, longer wind averaging times will yield lower maximum winds. Unfortunately, different meteorological services use different averaging times. Following World Meteorological Organization (WMO) guidelines, most regions use a 10-minute average. However, the Joint Typhoon Warning Center (JTWC), Guam, and WMO Region IV (United States and Caribbean area) use a 1-minute standard average. A TC defined as a typhoon using a 1-minute standard may not be defined as a typhoon using a 10-minute standard. Winds averaged over periods of at least 1 minute are referred to as sustained winds.
By examining a large number of recorded wind speeds vs. time traces and damage reports, conversion factors for going from one averaging time to another have been derived (Fujita, 1971; Simiuand and Scanlon, 1978; Krayer and Marshall, 1982). Depending on the methodology, there are some small differences in recommended conversion factors. For US Navy interests, the factor 0.88 is used in going from a 1-minute system to a 10-minute system such that TEN-MINUTE MEAN = 0.88 * ONE-MINUTE MEAN or ONE-MINUTE MEAN = 1.14 * TEN-MINUTE MEAN. Appendix A provides two tables for converting between 1-minute wind speeds and 10-minute wind speeds. These conversion factors should be considered as average rather than absolute conversions. There are many variations depending mainly on the frictional characteristics of the surface area and the atmospheric stability.
2.3 Gusts
Sudden brief increases and decreases in the wind speeds are called gusts and lulls, respectively (Huschke, 1959). Gusts or lulls can be much higher or lower than the sustained wind speed. Since gusts can substantially increase the damage potential from a TC, expected gusts are included in TC warnings. Conversion factors have been derived between sustained wind and gusts. The U.S. standard used at JTWC is given in Table 6.2.
2.4 Intensity Forecast Errors
As discussed earlier, the physical processes associated with TC intensity are complex and not well understood. Operational statistical and numerical weather prediction models have some difficulty in improving over forecasts based on simple climatology and persistence models. Forecasts based on simple statistical models do not adequately address the occasional rapid deepening and filling of TCs. Thus, forecast errors of TC intensity are rather high.
Table 6.3 , with data taken from the annual JTWC Typhoon Report, summarizes the JTWC maximum wind speed forecast errors for a recent 5-year period for the western North Pacific. The bias is simply the algebraic average of the forecast maximum wind speed errors, in the sense of forecast minus observed. Thus, positive bias indicates that the forecast more often called for maximum wind speeds higher than observed while negative bias indicates that the maximum wind speed forecast was not high enough. The average errors (Table 6.3) positive and increase with forecast period. Average wind forecast error is considered a poor error estimation since the positive and negative errors compensate each other.
Average absolute error gives a better indication of error. These errors (Table 6.3) are 11.2, 17.9 and 24.5 kt for 24-, 48- and 72-hour forecasts, respectively. Note that, for the 72-hr forecast, absolute wind forecast errors of at least 25 kt are seen to occur about 47% of the time. Most (27.8%) of these 72-hr errors of at least 25 kt are associated with the maximum wind being forecast too high (27.8%).
Additional details of the 72-hr error pattern are provided in Figures 6.1a and 6.1b. The former presents a frequency distribution of specific error values. Here, the small bias towards positive values is quite apparent. Data from Figure 6.1a were used to prepare the cumulative percentage frequency (CPF) plot shown in Figure 6.1b. Here, the errors are presented without regard to whether the forecast winds were too high or too low.
Some of the very high wind forecast errors depicted in 6.1a and 6.1b are often related to motion forecast errors. Intensifying TCs, if expected to remain over a favorable marine environment, would logically be expected to continue deepening. However, an incorrect motion forecast with the storm moving over a land area, rather than remaining at sea, could result in large intensity forecast errors. This event often occurs in connection with TCs moving northwestward off the east coast of Luzon where the intensity is heavily dependent on whether the storm passes over Luzon or remains at sea.
While the intensity errors noted in Table 6.3 and in 6.1a and 6.1b may appear to be rather high, intensity forecasts for other basins show similar error patterns. At the National Hurricane Center (NHC), for example, long term North Atlantic absolute wind speed forecast errors have averaged 11.5, 16.4 and 20.7 knots at 24-, 48- and 72-hours, respectively. The slightly lower Atlantic error at 72h is due to the fact that Atlantic systems typically do not reach the severity of the western Pacific TCs.
***** End of SECTION 2 *****
Main
Chapter 6
Section 1
Section 3
Section 2