GEOMAR Helmholtz-Zentrum für Ozeanforschung Kiel
Tel.: 0431 600-0
Fax: 0431 600-2805
11:00 h, Gr. Konferenzraum, Düsternbrooker Weg 20
General circulation models (GCMs) are subject to systematic errors (biases) that have, to some extent, resisted improvement efforts and will likely remain a problem for the foreseeable future. It has long been assumed that these errors have negative effects on the ability of GCMs to accurately predict seasonal variability and long-term climate change. This is particularly expected in the tropical Atlantic and Indian Ocean, where models show little skill in predicting variability. Few studies, however, have attempted to examine the direct influence of biases on prediction skill. In this presentation we take a closer look at this question. The focus will be on atmosphere-only experiments with the SINTEX-F GCM, which is used for seasonal prediction at JAMSTEC. Sea-surface temperature (SST) biases from a free-running control simulation of SINTEX-F are used to deliberately introduce biases into the climatology of the SST boundary forcing while keeping the anomalies as observed. The results show that the ability of the model to reproduce surface wind and precipitation variability is only moderately affected by the SST biases. We will also discuss how the model drift (i.e. the transition from observation-based initial values toward biased model climatology during the forecast run) affects skill, and how precipitation biases distort the patterns of air-sea interaction in coupled GCMs.