The International Research Institute for Climate Prediction (IRI) was forme
d in late 1996 with the aim of fostering the improvement, production, and u
se of global forecasts of seasonal to interannual climate variability for t
he explicit benefit of society. The development of the 1997/98 El Nino prov
ided an ideal impetus to the IRI Experimental Forecast Division (TRI EFD) t
o generate seasonal climate forecasts on an operational basis. In the produ
ction of these forecasts an extensive suite of forecasting tools has been d
eveloped, and these are described in this paper. An argument is made for th
e need for a multimodel ensemble approach and for extensive validation of e
ach model's ability to simulate interannual climate variability accurately.
The need for global sea surface temperature forecasts is demonstrated. For
ecasts of precipitation and air temperature are presented in the form of "n
et assessments," following the format adopted by the regional consensus for
ums. During the 1997/98 Fl Nino, the skill of the net assessments was great
er than chance, except over Europe, and in most cases was an improvement ov
er a forecast of persistence of the latest month's climate anomaly.