Home | About Us | Contacts | Help | Feedback | SEARCH Global | Australia | NSW | Vic. | Qld | WA | SA | Tas. | ACT | NT | Ant. | Weather & Warnings | Hydrology | Climate | Numerical Prediction | About Services | Learn About Meteorology | Registered User Services |

Global Guide
Main Menu

Global Guide to Tropical Cyclone Forecasting:
CHAPTER 5: SEASONAL FORECASTING

5.5 NORTH ATLANTIC SEASONAL FORECASTS

The north Atlantic experiences the largest interannual variability in tropical-cyclone activity of any region (Gray 1985). This strong interannual variability suggests that large-scale climate factors acting on seasonal and longer-term time scales are involved and that some degree of seasonal predictability may be possible. Recent research (Gray 1984a,b; Gray et al., 1992, 1993a,b) indicates that there are signals that allow skilful forecasts to be made as early as November the previous year. These signals have been exploited by Gray and collaborators to provide forecasts from Colorado State University of seasonal cyclone activity from the previous December, June, and August. Because this is the only region where regular seasonal forecasts are made to the public, the techniques are described in some detail in this section.

Figure 5.11: Locations of the meteorological parameters used for North Atlantic basin seasonal forecasts of tropical cyclone activity.

5.5.1 December Forecast Scheme

This extended range prediction is made during the late fall of the previous year (Gray et al. 1992). The basis of the forecasts are regression relations between several different predictors and six predictands covering the number of: Named Storms (NS), Named Storm Days (NSD), Hurricanes (H), Hurricane Days (HD), Intense Hurricanes (IH), Intense Hurricane Days (IHD). Also predicted is a normalised measure of the square of the maximum wind-speed for all hurricanes, called the Hurricane Destruction Potential (HDP):

where the and "a" parameters are empirically derived coefficients and:

U₅₀ and U₃₀ are winds extrapolated to September the next year for the 50 and 30 hPa level, to provide an indication of the QBO;

R_s is standardised rainfall in the western Sahel (Fig. 5.11) during August and September of the prior year; and

R_g is standardised rainfall in the Gulf of Guinea (Fig. 5.11) during August through November of the prior year.

The relationship is based on observations (Gray et al., 1992) that the phase of the QBO and African rainfall are strongly related to seasonal hurricane activity in the north Atlantic basin (Figs. 5.12 and 5.13). Using a cross-validated jackknife analysis for hindcast predictions of cyclones between 1950 and 1990, Gray et al. (1992) showed that Eq. 5.1 can explain 45-50% of the variance in all seven tropical cyclone predictands. This indicates that significant seasonal forecast skill is available for cyclones in the North Atlantic.

 Figure 5.12: Tracks of intense hurricanes during 15 y of lowest extrapolated 50-hPa zonal winds and wind shear to 30 hPa (top) compared to the tracks for the 15 y of highest values (bottom). The total period of record extended from 1950-1990 (Gray et al., 1992).

Figure 5.13: Tracks of intense hurricanes following the ten wettest August-November Gulf of Guinea seasons (top) compared to those for the ten driest August-November seasons (bottom). The period of record extended from 1949-1989 (Gray et al., 1992).

5.5.2 June Forecast Scheme

Gray et al. (1993b) have developed a regression relation for prediction of the aforementioned tropical cyclone parameters form early June in the following form:

where, in addition to the parameters described for Eq. 1 (see Fig. 5.11):

_xP, and _xT are the west African zonal pressure and temperature gradient anomalies from February to May;

SLPA and ZWA are the Sea Level Pressure and Zonal Wind Anomalies in the lower Caribbean during April-May;

SOI and _tSOI are the SOI in April-May and its change from January-February to April-May;

SSTA and _tSSTA are the SST Anomaly in NINO3 (Fig. 5.11) during April-May and its change from January-February to April-May.

Using a cross-validated jackknife analysis for hindcast predictions of cyclones between 1950 and 1990, Gray et al. (1993) showed that Eq. 5.2 can explain 50-70% of the variance in all seven tropical cyclone predictands. All but one (NS) of the predictands were better than 60% and 71% of Hurricane Destruction Potential was explained.

5.5.3 August Forecast Scheme

The prediction equation for 1 August is similar to Eq. 5.2 except that June-July rainfall and other meteorological parameters are used. Only a marginal improvement in skill from the 1 June forecast is obtained from this approach.

5.5.4 Discussion

These North Atlantic seasonal forecast provide useful information on the overall occurrence of tropical cyclones, and of hurricane damage potential. The North Atlantic basin best illustrates the potential for seasonal predictions and should be the standard by which seasonal forecast skills of other global basins are measured. A decade ago no one would have imagined that pre-seasonal climate signals would have been so well related to the hurricane activity. It is likely that all North Atlantic predictive signals have not yet been fully exploited.

The encouraging degree of skill with seasonal tropical cyclone forecasts for the North Atlantic does not, unfortunately, appear to carry over to the other cyclone basins. The North Atlantic seems to be special because:

• 1. It has marginal cyclone formation conditions, high interannual variability, and a short season (Fig. 1.22);
• 2. Seasonal cyclone activity in the North Atlantic is more strongly affected by slowly varying ENSO and QBO conditions than is the activity in other basins;
• 3. Most cyclone form from easterly-wave type disturbances in the trade winds, so that west African conditions have a major role, and westerly upper-wind anomalies can act as a strong inhibitor.

One major difficulty with developing seasonal tropical cyclone prediction equations for other basins lies in the lack of good data. Except for the western North Pacific, direct measurements of tropical cyclone intensity (eg., by aircraft reconnaissance) are not available and the resulting data base has to be considered unreliable (Holland, 1980). The use of satellite estimates since the 1970s has improved the quality enormously, but questions remain on the reliability of analyses of intense tropical cyclones.

Contents Chapter 6