Section 54.1 Track Forecast Accuracy
Forecast error statistics at JTWC and NHC are computed as the absolute great circle distance between a forecast position and the corresponding post-analysis best track position. No correction is made for initial position error as described by Neumann and Pelissier (1981). The mean track forecast errors (MFES) in the western North Pacific for 1978 through 1993 are 116, 227, and 345 nm for 24, 48, and 72 hours, respectively. Track forecast errors for the western North Pacific have been declining at a slow and irregular rate. This slow improvement is mirrored in the Atlantic (Neumann, 1981; McAdie and Lawrence, 1993).
4.1.1 Western Pacific Track Forecast Accuracy
In order to evaluate JTWC performance, JTWC forecasts were adjusted with CLIPER (i.e., Climatology and persistence tropical cyclone track forecast model) to remove the annual variation in forecast difficulty level (FDL) (Sampson et al., 1991). FDL is defined as the forecast error of a CLIPER model run on best track initial data (Pike and Neumann, 1987). Figure 1.18 gives an assessment of JTWC's performance normalized with FDL for the period 1978-1990. While FDL can help quantify after-the-fact-forecast skill in terms of the inherent difficulty of a particular forecast situation, it is of little use to customers unless the warning agency can tell the customer ahead of time whether a particular situation will have a high or low FDL.
4.1.2 Indian Ocean and South Pacific Track Forecast Accuracy
JTWC forecast improvement in other basins has been much the same as in the western North Pacific. In the North Indian Ocean basin forecast improvement at 24 and 48 hours have occurred throughout the period from 1972 to 1990. Some forecast improvement initially occurred in the Southern Hemisphere (western South Pacific and South Indian Oceans) after JTWC began forecasting there in 1980, but that improvement appears to have reached a plateau similar to that seen in the western North Pacific.
4.1.3 Eastern Pacific and Atlantic Track Forecast Accuracy
Neumann and Pelissier (1981) found that unusually large forecast errors in the Atlantic were primarily related to recurvature situations. Although rare, their large magnitude had a profound effect on the Mean Forecast Error (MFE).
4.2 Results of Large Cross Track Forecast Errors
Neumann and Pelissier's analyses (1981) can be applied to both NHC and JTWC's AORS. In the western North Pacific, large forecast errors occur more frequently than that in the Atlantic. The impact of these large cross track forecast errors is a lack of confidence in the warning products in JTWC's AOR that effect military operations. Large variations in forecast performance dilute credibility with the customer. As operational meteorologists, a more thorough understanding of the warning process and terminology will greatly aid the military commander when making difficult operational decisions.
4.3 Basin Classification By Forecast Difficulty Level (FDL)
Pike and Neumann (1987) computed the FDL for the six worldwide TC basins using specially developed CLIPER models that had identical structural characteristics for each basin. According to the study the most difficult basin was the Southwest Pacific- Australian region while the least difficult was the North Indian Ocean (Table 1.6).
4.3.1 Effect of Forecast Interval on FDL
When considering the FDL as a function of forecast interval, it has been discovered that as the interval increases the FDL rapidly increases in all basins.
4.3.2 Effect of Latitude on FDL
Pike and Neumann (1987) stated that while many factors influenced FDL, the most important factor was probably latitude, since in each hemisphere, forecast errors increase with increasing latitude. The large number of low latitude TCs in the western North Pacific contributes significantly to that basin's lower FDL ranking. For TCs poleward of 15 degrees latitude the western North Pacific FDL ranks near the top. This difficulty arises from the frequent interactions of TCs with each other (Brand, 1968; Dong and Neumann, 1983) and/or with cyclonic cells in the TUTT (Tropical Upper Tropospheric Trough), and with the large-scale monsoon circulation anchored by the massive Asian landmass.
4.3.3 Erratic Movers and FDL
Regardless of the FDL rankings of the basins, erratic movement of tropical cyclone occurs in all basins (Fig 1.19). It is especially common in the western North Pacific, South China Sea, Southwest Pacific-Australian region, and the Southwest Indian Ocean.
4.4 Variability of Track Forecast Duration and Distance
The duration and distance traveled by TCs in the JTWC AOR and the Atlantic basin are highly variable. Some TCs last only a day and travel only a few hundred nautical miles. Others, such as Tropical Storm Winona (1989), which began southeast of Hawaii, and Typhoon Wynne (1987) may travel in excess of 5,000 nm. In the Southwest Indian Ocean area, tracks may exceed 2500 nm. TCs Elinor (1983) in the South Indian Ocean and Rita (1972) in the western North Pacific each lasted 20 days; a record 79 warnings were issued on Typhoon Rita (Annual Typhoon Report, 1972).
4.5 Intensity and Surface Wind Structure Variability
One of the major factors when determining how a tropical cyclone will react to the current environment is assessing how it compares with previous systems. If a system is much larger or smaller when compared with an average system it will react very differently to environmental forces. A very large cyclone will greatly modify its surroundings while a small cyclone will only be advected by it. Therefore it is very important to understand the range of systems you may encounter.
4.5.1 Largest Wind Radius
Tropical Cyclones are highly variable in terms of size and intensity. Weatherford (1985) found large variability between the maximum intensity of western North Pacific TCs and the radius of the cyclone's 30 and 50 kt winds. Typhoon Tip (1979) had the largest circulation pattern on record, 1379-nm radius (Dunnavan and Diercks, 1980).
4.5.2 Most Intense Tropical Cyclone
Typhoon Tip (1979) was the most intense TC on record with a central pressure of 870 hPa based on aircraft dropsonde data. The second most intense was Typhoon June with a central pressure of 876 hPa (Holliday, 1976) (Table 1.7). In the western hemisphere the most intense hurricane on record was Gilbert (1988) with a central pressure of 888 hPa.
4.5.3 Midget Tropical Cyclones
Small, or midget, typhoons are not uncommon in the western North Pacific or in the Southern Hemisphere basins. Typhoon Gay (1989) in the Gulf of Thailand, which attained an intensity of nearly 100 kts then reintensified to nearly 140 kts in the Bay of Bengal, had a diameter of 30 kt winds that did not exceed 125 nm (Table 1.8). The intensity of Hurricane Tracy (1974) was about 110 kts when it flattened Darwin, Australia, yet 30-kt winds extended to less than 30 nm from the center (Muffatti, 1975). The mechanisms that determine the size of TCs are poorly understood and need considerable study.
4.5.4 Super Typhoons
A TC which reaches an intensity (i.e., tropical cyclone surface wind speed or low level wind speed) of 130 kts or greater is classified as a "super typhoon". Super typhoons are most common in the western North Pacific where an average of about three occurrences are observed annually. This average value is somewhat misleading since there is a high degree of variability from year to year. For example in 1974, there were no cases, while in 1987 six cases occurred. Out of all the super typhoons that have occurred, the overwhelming majority occurred in the western North Pacific basin.
4.5.5 Rapid Developers
Most TCs that reach super typhoon intensity undergo a period of rapid deepening during which central pressure drops 42 hPa or more in a 24-hour period (Holliday and Thompson, 1979). Rapid deepening occurs most frequently in the western North Pacific and can be reliably estimated from meteorological satellite imagery using the Dvorak intensity technique (Dvorak, 1975, 1984). For example, as Typhoon Flo (1990) approached Okinawa, Japan, estimates from meteorological satellite imagery indicated a minimum sea-level pressure of 898 hPa and 145 kt winds based on a Dvorak T-number of 7.0. Within one hour, a subsequent eye wall penetration by a NASA DC-8 research aircraft provided a dropsonde measurement that indicated a sea-level pressure of 891 hPa. Application of the Atkinson and Holliday (1977) wind-pressure relationship yielded a wind estimate of 150 kts. Aircraft data were not available for prior calibration of the satellite data.
4.5.6 Intensity Estimate Variability From Different Forecast Centers
An important factor to consider when discussing TC intensity is what standards were used in determining the measurement. For example, large differences frequently exist between JTWC's maximum intensity for TCs greater than 100 kts and that of other warning agencies. JTWC's intensity is typically higher. Atkinson (1974) used several earlier wind studies to show the average 10-minute wind speed to be approximately 88 percent of 1minute averages. By using this technique JTWC has higher intensity estimates. To further complicate the problem of intensity comparisons, even the 1-minute to 10-minute conversions used by different warning agencies vary with some using 90 percent, others using 80 percent, and still others using some value in between. The Meteorological Satellite Center of the Japan Meteorological Agency (JMA) also uses a modified conversion from the Dvorak T-number to Dvorak current intensity (CI) number (Osano, 1989). As a result, TCs below typhoon intensity are stronger than the corresponding Dvorak technique would indicate and intense typhoons are weaker than the Dvorak values indicate. Consequently, super typhoon intensity ( 130 kts) is rarely attained using the JMA conversions.
4.5.7 Max Wind Estimates
With regard to peak wind gusts, Atkinson (1974) addressed studies by Myers (1954), Taniguchi (1962), and Kitaoka et al. (1971) to show peak gusts to be, on the average, 1.40 times the 10- minute average sustained wind speed, and 1.23 times the 1-minute average sustained wind speed. These gust factors are critically dependent on the sustained wind speed, the surrounding terrain and manmade obstacles. For unobstructed wind over open water, the gust factors would probably be 5% less than the values mentioned in the above.
***** The end of Section 4 *****
Chapter 1
