The forecast was prepared by first making a long lead forecast of the tropical Pacific sea surface temperature anomaly (SSTA) using an anomaly coupled model of the Earth’s climate. More information on the anomaly coupled model and the latest forecast of Pacific SSTA are available. The SSTA forecast of the anomaly coupled model is statistically projected to estimate a global SSTA (Figure 1) which is then applied as a lower boundary condition to the COLA global atmospheric general circulation model (AGCM). The COLA AGCM, which has been shown to exhibit skill in making seasonal climate forecasts for previous cases, is used to make the winter 1998 forecast.

The accompanying figure (Figure 2) depicts the geopotential height at 500 hPa which is an indicator of the upper tropospheric pressure and can be used to deduce the associated storm tracks and surface weather patterns. The figure shows the geopotential height for the January - March 1998 average in thin isolines and the departures of this mean from the normal winter distribution in shaded contours. The blue tones correspond to lower than normal heights (pressure) and the yellow-red tones correspond to above normal heights.

The forecast calls for generally lower than normal heights (pressure) in a contiguous band extending from eastern Siberia and the Arctic Ocean through the Gulf of Alaska, across North America and into the North Atlantic. Above normal heights were forecast for Hudson Bay and surrounding parts of Canada as well as southern Greenland and western Europe. Smaller regions of above normal heights were forecast for northern Africa and Japan while below normal pressure was predicted for central Asia in the vicinity of the Caspian Sea.

In order to verify this prediction, analyzed observations of the 500 hPa geopotential height may be compared. The following figure (Figure 3) shows the average observed height and its departure from normal for the three month period January - March 1998. It may be seen from this figure that several of the forecast features verified well, in particular, the large anomalies in the region of North America and adjacent oceans. The forecast has its largest errors over the Arctic Ocean where both the model used to make the forecast and the observations are less reliable.

The global atmospheric model calculation made to produce the six-month lead forecast was repeated recently with observed SST, rather than predicted SST, as a lower boundary condition. Exactly the same procedure was employed, except that the observed atmospheric state in mid-December 1997 and the observed SST lower boundary condition in JFM98 were used. We use the terminology "hindcast" to distinguish a calculation made using observed input data from a "forecast" in which the only input data are from times prior to the verification time. We have compared the analyzed observations of 500 hPa geopotential height with both the six month lead forecast and the hindcast of that quantity in Figure 4. As shown in Figure 4, the forecast and hindcast are quite similar to each other and to the verification. It should be noted that the hindcast was a very good simulation of the observed 500 hPa heights, and major errors in the forecast over North America - overprediction of the trough in the Gulf of Alaska and underprediction of the low pressures over the southern tier of the U.S. - were largely eliminated in the hindcast. This is an indication that improving the six month lead forecast of SST can improve the forecast of the extratropical circulation.