Over the past decade or so, a consensus has grown up among meteorologists that there are predictable cyclical variations in the temperature of the surface waters of the equatorial Pacific, which are linked with extremes of weather around the globe. The only problem is that no one predicted the floods in the Mississippi valley.
There was a significant warming of large areas of the Pacific in 1991 and 1992, and the consensus was that this would be followed by a cooling in 1993. One consequence of this was a forecast that the central US would experience a severe drought this summer. Instead, rain and flooding have caused billions of dollars of damage. This could be explained by a renewed surge of warming in the tropical Pacific.
A record-breaking warming of the Pacific in 1982 and 1983 triggered meteorologists' interest. For centuries, fishermen in Peru had been aware that, in certain years, warm water spread southwards along the coast, shutting off cold, up-swelling, nutrient-rich water and dramatically reducing fish stocks. Because this occurred around Christmas, they called the event El Nino (The Child) to associate it with the Nativity.
These periodic changes are part of a great shift in the weather patterns in the Tropics. The warm water first appears off the coast of Peru and during the next year spreads across the Pacific. At the same time, atmospheric pressure over the eastern Pacific falls and rises over Australia and the Indian Ocean. With that, the easterly trade winds along the Equator reverse.
Sir Gilbert Walker, an eminent British meteorologist, noted in the Twenties: 'When pressure is high in the Pacific Ocean, it tends to be low in the Indian Ocean from Africa to Australia, and vice versa.' He termed this behaviour the Southern Oscillation. For this reason, meteorologists seek to combine these far-flung effects and refer to El Nino Southern Oscillation (ENSO).
It is relatively easy to see the connection between shifts in the weather in equatorial regions and El Nino. The area of heavy rainfall that is normally found over Indonesia moves eastwards to the central Pacific. The heaviest rainfall over Amazonia moves west of the Andes, bringing torrential rains to the arid coast of Peru and Ecuador. The region of ascending air over Africa is replaced by a descending motion connected with the frequent droughts in the Sahel and southern Africa of the past two decades.
At higher latitudes the connections are more complicated. There seems to be clear evidence that El Nino delays and reduces the strength of the Indian monsoon and causes widespread drought across Australia. Over North America, El Nino years are associated with cold winters in the east.
More dramatically, in El Nino years the global average temperature is markedly higher. The scale of all these effects is such that considerable efforts have been made to forecast El Nino accurately. This has involved developing computer models of how the atmosphere and ocean interact. These have to address the basic conundrum about how the system can switch from one extreme to the other. During an El Nino, the more extensive warm water strengthens westerly winds along the Equator, which in turn drive more warm water towards South America. Conversely, during non-El Nino conditions, the cold water that extends westwards across the Pacific reinforces the easterly winds, which drive more cold water away from South America.
The key appears to lie in the surface layer of the ocean. In the tropical Pacific this consists of a layer of warm water some 100m (330ft) deep, on top of cold deep water. Disturbances caused by changes in the wind fields along the Equator create waves that travel in this surface layer east and west across the Pacific basin. Those travelling eastwards are relatively fast-moving and can cross the ocean in two to three months. Those moving westwards take between nine months and several years to make the journey.
Computer models suggest that the slow waves in the Pacific may interact over the years so that the sea's surface temperatures oscillate back and forth in a quasi-cyclical manner over a period of three to five years. Although this process is not strictly predictable, it did seem to justify extrapolations forward for a few years.
When they successfully predicted the warming in 1991 and 1992, modellers' euphoria rose. Some questioned early records of El Nino lasting two to three years (as in 1939 to 1941), arguing that the event was inconsistent with their models.
None of the models predicted the warming in 1993. So the modellers will have to go back to their drawing boards, while those who seek evidence of predictable cycles in the weather will see yet more confirmation for the one certain rule: the moment a cycle is identified with sufficient confidence to be used to make a forecast, it disappears.