with unusually dry conditions. However, in southern Africa, where governments prepared by building food reserves, the devastating droughts that had occurred during the 1982-1983 El Niño were not repeated in 1997-1998, and the rainy season (until June 1998) was relatively normal.
From a health perspective, the extreme weather events were associated with many disease outbreaks. In Latin America, flooding was associated with significant upsurges in malaria and cholera in Ecuador, Peru, and southern Brazil. Heavy rains in the Horn of Africa precipitated outbreaks of cholera, malaria, and Rift Valley fever. In Asia, drought was associated with poor water quality and cholera. The massive forest fires in Indonesia, as well as in Brazil, Mexico, Central America, and Florida, inflicted widespread respiratory illness. Poor air quality also affected trade and tourism, and fires in tropical forests have adversely affected wildlife and ecosystem functioning, as well as releasing additional carbon into the atmosphere (Epstein, 1998; Stevens, 1998).
In addition, high sea surface temperatures have taken an enormous toll on sea life, especially marine mammals. During 1997-1998, significant marine mammal mortalities were linked to El Niño on the Pacific coasts of the United States, Peru, Venezuela, and the Galapagos Islands and in the southeast Pacific, New Zealand in particular (Epstein, 1998; Stevens, 1998). These effects may have been caused by the migration of food sources, enhanced blooms of toxic phytoplankton, and/ or changes in the immune systems of marine mammals.
In sum, the 1997-1998 El Niño had major negative impacts on many people and regions and also brought significant benefits to other people and regions. The availability of accurate forecasts of extreme weather led some people and organizations to act in ways that spared them even worse damage. However, many others in these areas did not hear or respond appropriately to the forecasts, and, in other areas, forecasts were wrong and some prepared for forecast disasters that did not arise.
The experience of 1997-1998 strongly suggests that there is great potential social value in the developing ability to forecast climateaverages of temperature, precipitation, and the likemonths to a year or more in advance. Improved forecast skill, that is, accuracy beyond annual and seasonal averages,1 may open up a vast array of possibilities for the use of climate information to reduce the risk of damage from unfavorable cli-
1 The term ''forecast skill'' has precise meanings in meteorology. Commonly, skill is measured by the correlation between the forecast and actual values of an index of some weather or climatic event or by the average of the root-mean-square error over the length of a forecast (National Research Council, 1996a). The concept of forecast skill is described further in Chapter 2.