Depending on the manner in which it is assessed, the El Nino event of 1997-1998 might be the strongest recorded. Certainly it was of comparable magnitude to the event of 1982-1983 which brought drought to many areas surrounding the Indian Ocean and heavy rainfall to continental areas adjoining the eastern Pacific Ocean. Without doubt the 1997-1998 event was the most heralded in history, as it offered the first opportunity for modellers to test their El Nino predictions using data from the then newly completed Pacific tropical buoy array. Discussion continues, though, over whether or not the event was predicted ahead of its presence being detected by the observing systems (Barnston et al., 1999).
There is a standard diagram (Ropelewski and Halpert, 1987) that details climate anomalies likely (but not certain) to occur during El Niño events, and countries threatened with drought or flood according to this diagram took some considered actions through the latter months of 1997, which certainly mitigated negative impacts in places (World Meteorological Organisation, 1999; Glantz, 2001). In hindsight, however, some actions were inappropriate to the climate conditions experienced and the event provides an object lesson in many of the issues surrounding prediction on seasonal to interannual time scales. Examples from southern Africa will be used to illustrate the point.
Coincidentally a major new initiative was planned for the period of the 19971998 El Niño event, the world's first Regional Climate Outlook Forum (RCOF). Entitled SARCOF, with SA standing for Southern Africa, this was a pilot for the 30+ RCOFs that have subsequently been held around the globe. RCOFs have proven to be a prime instrument through which climate services based on seasonal to interannual predictions have been introduced to the developing world. Forums bring together regional climatologists, intermediaries, end users, policy makers, the media, and various experts in an assembly at which presentations are made in a capacity building context, a consensus forecast is created using all available empirical and numerical inputs, and the forecast is discussed and interpreted in actions appropriate to the forthcoming rainfall season. The standard output product from a Forum is a map on which are given probabilities for rainfall terciles in individual parts of a region (Basher et al., 2001). A text prediction is also provided.
The first meeting in the 1997-1998 SARCOF series was held in Kadoma, Zimbabwe, during September 1997. The concept of probability forecasts was new to most attendees and there were consequent difficulties in defining the interpretation of the forecast, which placed highest probabilities in the driest tercile in most areas likely to experience drought during an El Niño but retained modest probabilities in the other two terciles, including the wettest. Against a media and an Internet barrage (a medium being widely exploited for the first time in delivering seasonal forecasts) calling for a severe drought in line with the canonical El Niño response in southern Africa, the carefully-placed Forum forecast apparently carried little weight. The text accompanying the Forum forecast accepted that drought, possibly severe, might occur but also pointed out that, historically, drought does not necessarily accompany an El Niño event. At a Forum update meeting in Windhoek, Namibia, in December, several of the numerical model predictions had reduced likelihoods of drought as compared to the September situation, a change reflected in the Forum forecast. Despite these activities it is probably fair to state that the general perception across southern Africa, and the frequent planning position, was that a drought was inevitable.
In reality seasonal rainfall over southern Africa was generally around the long-term climatological normal, but with somewhat above-normal totals in favoured places. Certain areas experienced localised flooding. Departures from the canonical response also occurred elsewhere. Rainfall in East Africa was above average, as would be expected from the canonical response, but by an excessive amount. As in southern Africa, the droughts anticipated in India and Australia did not occur. However an unusually severe drought did happen in parts of South East Asia, and was responsible for creating suitable conditions for the haze that developed there.
Almost certainly one reason why these responses did not match the canonical expectation was due to the influence of sea surface temperature anomalies in the Indian Ocean, which were perhaps larger than any previously observed. While the El Niño was taking place in the Pacific Ocean, temperatures across the Indian Ocean resembled the equivalent of a La Niña event with warm anomalies in the west, a distribution that might have been linked with the El Niño itself (Yu and Rienecker, 1999). The warm waters in the western Indian Ocean probably assisted in increasing the rainfall across the whole of the eastern part of Africa. Most of the empirical forecast models used in Africa focused on Pacific Ocean sea surface temperatures as the main predictor, hence failing to incorporate any information from the Indian Ocean. Certain numerical models may have been picking up information from the Indian Ocean given their predicted reduced probabilities of dry conditions by December.
The credibility of seasonal forecasting suffered initially as a result of the 19971998 episode. In the view of the author the issue was one of understanding and of recognising the probabilistic nature of the forecasts. This message has been passed continually through ensuing Forums, although it is still probably fair to say that many are not yet comfortable with probabilities. Confidence returned to a certain extent in following years as an extended La Niña event occurred that was associated with generally canonical impacts. But in 2001 the possibility of a new El Niño event developing by the end of the year caused ripples in some parts of the Pacific region. No El Niño developed. At the time of drafting the original version of this paper a similar scenario to that of 2001 was being played out in early 2002, although against the background of El Niño signals certainly stronger than those during 2001. It is now clear, in late 2002, that an El Niño event has developed.
No final solution has yet been obtained in the creation of a general understanding of the limitations of seasonal to interannual forecasting. Many users meetings continue to call for improved forecast accuracy, which will certainly come as the models and understanding improves but which will not meet deterministic expectations. Education initiatives, together with clearer statements from climatologists on the limitations of the prediction systems, plus a willingness amongst users to examine how current predictions might be best used in particular applications, are amongst the requirements at the present time.
Was this article helpful?