A statistical-dynamical scheme for reconstructing ocean forcing in the Atlantic. Part I: Weather regimes as predictors for ocean surface variables

The links between the observed variability of the surface ocean variables estimated from reanalysis and the overlying atmosphere decomposed in classes of large-scale atmospheric circulation via clustering are investigated over the Atlantic from 1958 to 2002. Daily 500 hPa geopotential height and 1,000 hPa wind anomaly maps are classified following a weather-typing approach to describe the North Atlantic and tropical Atlantic atmospheric dynamics, respectively. The algorithm yields patterns that correspond in the extratropics to the well-known North Atlantic-Europe weather regimes (NAE-WR) accounting for the barotropic dynamics, and in the tropics to wind classes (T-WC) representing the alteration of the trades. 10-m wind and 2-m temperature (T2) anomaly composites derived from regime/wind class occurrence are indicative of strong relationships between daily large-scale atmospheric circulation and ocean surface over the entire Atlantic basin. High temporal correlation values are obtained basin-wide at low frequency between the observed fields and their reconstruction by multiple linear regressions with the frequencies of occurrence of both NAE-WR and T-WC used as sole predictors. Additional multiple linear regressions also emphasize the importance of accounting for the strength of the daily anomalous atmospheric circulation estimated by the combined distances to all regimes centroids in order to reproduce the daily to interannual variability of the Atlantic ocean. We show that for most of the North Atlantic basin the occurrence of NAE-WR generally sets the sign of the ocean surface anomaly for a given day, and that the inter-regime distances are valuable predictors for the magnitude of that anomaly. Finally, we provide evidence that a large fraction of the low-frequency trends in the Atlantic observed at the surface over the last 50 years can be traced back, except for T2, to changes in occurrence of tropical and extratropical weather classes. All together, our findings are encouraging for the prospects of basin-scale ocean dynamical downscaling using a weather-typing approach to reconstruct forcing fields for high resolution ocean models (Part II) from coarse resolution climate models.