MainA significant amount of effort has been expended to develop pattern recognition techniques for motion forecasting. Many of these methods were developed for specific regions of the world and may not be applicable to other regions. These methods are subjective in nature and hence it is difficult to evaluate their utility. In spite of this, a few of these methods are described in this section.
3.1 Slow or Looping Versus Fast Moving Tropical Cyclones
Key features of idealized 500 hPa and corresponding surface patterns for slow or looping versus fast moving TCs have been compiled for the western North Pacific and Atlantic oceans (Xu and Gray, 1982). The idealized 500 hPa and corresponding surface patterns are shown, in schematic format, in Figs. 4.7 and 4.8.
Two rules determined from the key features schematics are:
a. Looping or Slow Motion - Looping or slow TC movement is associated with a large amplitude 500-hPa trough to the northeast (about 20-30 ° longitude) of the TC. The higher the latitude, the further east the trough is relative to the looping storm.
b. Fast Motion - A trough west (about 10-20 ° longitude) of a recurving TC is associated with fast TC movement. TCs which are equatorward of a well-defined subtropical ridge experience fast westward movement.
3.2 Recurvature Decisions Based on 200 hPa Charts
Composites for recurving and non-recurving TCs were developed for the western North Pacific using 200 hPa synoptic wind patterns by George and Gray (1976). Generally, when strong westerly winds (greater than 50 knots or 25 m/s) are seen within 20 degrees of latitude north of the storm, the storm is forecast to recurve. If the winds are weak or easterly, recurvature will not occur. A modification of these rules has been done in which storms are stratified by season and region (Guard, 1977). The modifications were done based on evaluation against an independent set of TC data. Appendix A is an abridged version of this technique. One advantage to using the 200 hPa level is that data are more plentiful than at the mid-levels. One disadvantage is that storm motion is better correlated with deep layer mean or mid-level steering flow.
3.3 Acceleration Prediction
Tropical cyclone forecast errors are based on the distance between a forecast storm position and the later observed position. Therefore, for a given directional error, storms with higher translational speeds can expect higher forecast error. This is reflected in forecast error statistics.
Failure to anticipate the acceleration will yield even higher forecast errors, which is a long-standing problem associated with recurving storms. In the fear of observing large forecast errors, forecasters are reluctant to accelerate storm motion.
In Fig. 4.9, it can be noted that equatorward of about 24N, the average motion of TCs is rather constant. Although short-period accelerations do occur in connection with synoptic-scale surges in the trade-wind circulation, these are fairly short lived. Unanticipated accelerations at these latitudes are typically not associated with large forecast error. It can also be noted in the figure that most TCs are within these latitudes. Accordingly, forecasters become conditioned to the lower translational speeds associated with low-latitude storms.
Poleward of 25N, however, the acceleration of TCs is quite marked. Even with a reasonably good direction forecast, large along-track errors can result. For example, Table 4.1 gives the 48h forecast errors for the Western Pacific CLIPER model used at JTWC (WPCLPR). To illustrate the effect of increasing storm latitude (translational speed), this simple model, in a non-operational mode, was activated on all storms over the period 1946 through 1993. Although best-track input data were used, errors were normalized to operational WPCLIPR errors (247 nm) over the five-year period 1988-1992.
In Table 4.1, it can be noted that WPCLPR forecast errors continually increase from south to north. Errors at 40N are about double those at 15N, with the largest rate of increase being near 40N. These large errors are caused by the rapidly increasing translational speeds as depicted in Fig. 4.9. Thus, even a model such as WPCLPR, where the accelerations are known to the model, is subject to larger forecast errors due mainly to the higher translational speeds.
It can be shown that the errors of any motion prediction model or of the official operational forecast, although typically averaging less than CLIPER errors, exhibit a similar poleward error gradient. Thus, higher forecast errors associated with poleward moving storms are typical of any model. However, as mentioned earlier in this section, failure to anticipate these accelerations can result in even larger errors.
In an effort to minimize large forecast errors associated with TCs entering the westerlies, the TAPT forecast technique was developed. This 200-hPa pattern-typing technique, developed for western North Pacific TCs, gives an assessment of the magnitude and duration of acceleration associated with northward moving TCs that enter the mid-latitudes. An abridged version of this technique is provided in Appendix B.
3.4 Satellite Image Interpretation
Interpretation of storm related cloud features can infer short 12- to 24-hour motion of TCs. Changes in the direction of movement of the TC have been correlated with rotational changes in the storm's gross features (Fett and Brand, 1975). The angular rotation of the gross features during the past 24 hours is added to a 24-hour persistence forecast to yield the new 24-hour forecast. Examples of gross features and rotated gross features are shown in Fig. 4.10. This method was developed and tested for western North Pacific TCs, but similar methods were developed for other regions.
Lajoie and Nicholls (1974) developed a method, which uses a cloud model (Fig. 4.11) and the following rules to help discern 12-hour direction of motion in the Australian region:
(1) A tropical cyclone will not continue to move, nor curve in a direction towards a cumulonimbus-free sector. If it is moving towards such a sector, it will curve rapidly away from that direction.
(2) A tropical cyclone having a single outer cloud band will move or curve within twelve hours of picture time towards a line joining the present position of the vortex center to the present position of the most developed cumulonimbus cluster at or near the downstream end of the outer cloud band.
(3) When a tropical cyclone has two outer cloud bands and at or just prior to picture time and is moving generally towards the most developed cumulonimbus cluster near the downstream end of one outer cloud band, it will curve within twelve hours of picture time towards a line joining the present position of the vortex center to the present position of the most developed cumulonimbus cluster near the downstream end of the other outer cloud band.
Note that the Lajoie and Nicholls (1974) method apparently gives good results in the Australian region, but requires satellite image interpretation skills and a TC in which features shown in Fig. 4.11 are visible.
3.5 A Systematic Approach to Tropical Cyclone Track Forecasting An overview of a systematic approach to TC track forecasting in the western North Pacific region is presented in Appendix C. This systematic approach provides the forecaster with a track-forecast methodology that combines traditional concepts with dynamical insights into how TCs interact with their environment and thus affect their track. JTWC forecasters tested this approach during the latter half of the 1994 typhoon season.
***** End of SECTION 3 *****
Main
Chapter 4
Section 3