Abstract Author: Anthony Barnston
Abstract Title: Verification of IRI's seasonal climate forecasts, 1997 to present
Abstract: The skills of the first 10 years of seasonal precipitation and temperature probabilistic forecasts of the International Research Institute for Climate and Society (IRI) are presented. Several verification measures are shown, and the meanings of differences among them are discussed in terms of freedom from bias, discrimination ability, probabilistic reliability, and other aspects of accuracy.
As might be expected, skills are very modest when averaged over all seasons and locations, but are significant and of potential utility in certain regions and seasons influenced by known teleconnections to sea surface temperature anomalies. A few of these instances of higher skill may not have been expected, due to the inadequacy of observed data for statistical detection, in view of IRI's use of primarily dynamical prediction tools.
Factors such as global climate change and its implications in defining the current expected climatology, the format of probabilistic forecasts, the upper limit of predictability and user perceptions of nonzero but modest skill levels are brought into the discussion.