Performance of Recent Multimodel ENSO Forecasts
AbstractThe performance of the International Research Institute for Climate and Society “ENSO forecast plume” during the 2002–11 period is evaluated using deterministic and probabilistic verification measures. The plume includes multiple model forecasts of the Niño-3.4 index for nine overlapping 3-month periods beginning the month following the latest observations. Skills decrease with increasing lead time and are highest for forecasts made after the northern spring predictability barrier for target seasons occurring prior to the forthcoming such barrier. Forecasts are found to verify systematically better against observations occurring earlier than the intended forecast targets, an effect that becomes greater with increasing lead time. During the study period, the mean forecasts of dynamical models appear to slightly (and statistically insignificantly) outperform those of statistical models, representing a subtle shift from earlier studies. The mean forecasts of dynamical models have overall larger anomalies but similar errors to those of statistical models. Intermodel spread is related to forecast error in an average sense with changes in forecast error due to changes in lead and verification season being properly reflected in changes in spread. The intermodel spread underestimates the forecast error variance, to a greater extent for statistical forecasts than for dynamical ones. Year-to-year changes in plume spread provide little additional information relative to climatological ones.