Section 4Many of the objective aids have peculiar strengths and weaknesses which may be exploited by forecasters at the warning centers (e.g., JTWC or NHC). Sometimes TC warning discussions will include descriptions of objective aid behavior and forecaster interpretation of threat behavior. For example:
Knowledge of the strengths and weaknesses of the objective aids can enhance a forecaster's ability to interpret the discussion section and then make forecast recommendations to the on-scene commander.
3.1 Extrapolation
The principal advantage of extrapolation is its ability to account for short-term trends of motion and intensification. The principal weakness is the limited time frame in which it remains valid. The validity of an extrapolation forecast quickly erodes as the environment around the TC changes.
3.2 Climatology and Analog
The principal advantage of a climatological model is that, with the exception of errors in the present storm position, they are generally insensitive to initialization problems caused by insufficient and misrepresentative data. The principal disadvantage of climatology is that it gives only the average behavior of storms. The analog approach attempts to minimize this problem by restricting the historical data base to a small subset that hopefully represents synoptic conditions that are fairly close to those influencing the present storm.
Climatological aids perform best if the storm is in an area with a large number of historical storms. Thus climatology and analog aids should not only be more accurate, but also show less variability from forecast to forecast during the active TC season. In the off-season, climatological aids are sporadic performers. An example of this is shown in Figure 5.2. Note the sudden change from a straight to recurve forecast.
Another problem occurs when the TC approaches the data sparse region over China. When a storm is moving toward land at latitude where recurvature occasionally occurs, the climatology forecast will likely predict recurvature (Fig. 5.3). One reason for this is that TCs usually dissipate over land; therefore, there are fewer straight moving tracks in the climatology data. This is especially true for 72-hour forecasts.
A similar bias exists for storms that are at or near recurvature and potentially within 48 hours of dissipation or extratropical transition. The 72 hour position given by climatology aids will likely contain a significant slow-speed bias, since only a slow moving historical storm would exist long enough to generate a 72 hour best track position.
3.3 Statistical
The principal advantage of statistical regression methods is that they attempt to correct for known and unknown systematic biases caused by data distribution. The principal disadvantage is that statistical regression methods produce forecasts that conform to the average behavior of storms in the dependent data set. The statistical-synoptic and statistical-dynamic models are also subject to the negative influences of erroneous or missing data that affect dynamic models.
Statistical regression aids tend to display unusual motion characteristics when they move into data-sparse regions, or have bad past motion inputs or erroneous intensity inputs. Figures 5.4 and 5.5 indicate how WPCLIPER (a statistical regression aid) responds to changes in time of year, latitude, longitude, past position, and initial intensity. WPCLIPER's response to time of year provides a good indication of where WPCLIPER expects the subtropical ridge axis to be.
Other statistical regression aids performance is affected more by the synoptic patterns governing the current motion. For example, CSUM (Colorado State University Model) performs well during periods when the storm is in one of three synoptic patterns, but does not predict the transitions between the patterns. For a storm that starts out equatorward of the subtropical ridge, CSUM tends to keep it moving west- northwest, but suddenly jumps to a poleward track whenever the past motion vector is between 330º and 030º (Fig 5.6). A similar jump takes place when a recurving storm's direction of motion first falls between 031 and 120 degrees.
3.4. Dynamic
The principal advantage of numerical methods is that they are sensitive to the current and future synoptic structure of the atmosphere, as represented by the model, and thus can better handle synoptic-scale variations. The principal disadvantage of dynamical methods is their sensitivity to insufficient or erroneous data (typically inducing a misplaced vortex), which results in forecasts that, even in the short term, are initialized with the wrong direction and speed (Fig. 5.7). The more sophisticated a numerical model is, the more sensitive it is to inaccurate initialization due to data limitations. The more sophisticated the model, the more rapidly its forecasts can depart from reality due to the growth of data-induced initialization errors. As a result of this weakness, numerical models often require initial TC data (e.g., position, intensity and movement) to insure the model vortex at least starts out in the right location and moves in the right direction.
Each numerical and dynamic aid used at JTWC and NHC has its own strengths and weaknesses. Numerical forecast aids can be based on tracking the movement of the TC vortex as represented in a global or regional numerical model, or the aids can be based on steering derived from such numerical models. Thus, the accuracy of the global model used to initialize the dynamic aids is a governing factor of the accuracy of the aid predictions.
3.5 Hybrid
Hybrid aids attempt to capitalize on the strengths of several types of aids by combining the outputs using statistical methods.
The primary consideration when using a hybrid aid that attempts to capitalize on climatology can be best described by the following example. To capitalize on the strengths of HPAC (Half Persistence and Climatology) a measure of the extent of deviation from climatology is the angle between XTRP and CLIM. If this angle exceeds 60º, then HPAC should be disregarded unless it has been performing well statistically on the current storm. Figure 5.8 shows an extreme example of degradation in a HPAC forecast. The degradation was caused by the combined effects of a significant slow speed bias in the 72-hour CLIM forecast and the fast 12-hour past motion of RUSS (31ºW, 1990) which, of course, is retained by XTRP throughout its forecast.
3.6 Empirical
The principal advantage of these techniques is the ability to exploit pattern recognition techniques that the forecaster develops with experience. These pattern recognition techniques can eliminate major errors associated with missed recurvature and acceleration of storms into the mid-latitudes. The major weakness lies in the amount of time required to gain the experience level necessary to accurately apply these techniques. The techniques require the use of tables and nomograms, which must be applied and adjusted to the current situations, based on the forecaster's judgement. Experienced forecasters make the best use of these aids, because they can quickly sense and evaluate changing synoptic conditions.
Chapter 5
