Validation of assimilation products will ultimately be carried out through forecast/hindcast experiments. Of particular concern is the best way to use the state estimates for coupled forecast initialization. Many case studies have to be undertaken to give an indication of the statistical significance of differences in forecast performance. Unfortunately, the ocean observation database is not substantial enough to support a large number of case studies. However, an attempt will be made in this regard by conducting hindcasts from January and July (say) initial conditions taken from the 20-year retrospective global ocean analyses. The evaluation will assess the impact on forecast skill as a function of lead time. An important component of this theme will be the development of metrics to assess the differences between forecasts, in the same manner as for the intercomparison metrics. The interest is not only in the Niño SST indices, but also SST patterns, thermocline anomalies, etc. The ultimate goal of the consortium is to improve the ocean initialization for coupled forecasts; however, one of the difficulties in assessing different products/assimilation methods with coupled forecasts is the impact of coupling to the atmosphere. Hence some validation will be undertaken through forced-ocean hindcast experiments.
All consortium members will continue their routine coupled forecasts. LDEO will contribute a suite of forecasts with a coupled intermediate model (LDEO4) matching COLA GCM forecast runs. In addition, all consortium members will conduct the agreed upon case study hindcasts. COLA and the IRI will focus on how best to use assimilation products for coupled initialization, including high resolution products with a lower resolution ocean model configuration. All will collaborate on the design of appropriate metrics.