Assessment of short-range forecast error atmosphere-ocean cross-correlations from the Met Office coupled numerical weather prediction system

Wright, A., Lawless, A. S. ORCID: https://orcid.org/0000-0002-3016-6568, Nichols, N. K. ORCID: https://orcid.org/0000-0003-1133-5220, Lea, D. J. and Martin, M. J. (2024) Assessment of short-range forecast error atmosphere-ocean cross-correlations from the Met Office coupled numerical weather prediction system. Quarterly Journal of the Royal Meteorological Society, 150 (762). pp. 2783-2797. ISSN 0035-9009 doi: 10.1002/qj.4735

Abstract/Summary

Operational data assimilation systems for coupled atmosphere -ocean prediction are usually ‘weakly-coupled’, in which there is no explicit interaction between the atmosphere and ocean within the data assimilation step. Explicitly allowing for cross-correlations between the ocean and the atmosphere may have potential benefits in improving the consistency of atmosphere and ocean analyses, as well as allowing a better use of observations at the interface. To understand whether such correlations are significant on the timescales of numerical weather prediction, we investigate the atmosphere-ocean cross-correlations of short term forecast errors from the Met Office coupled prediction system, considering their temporal and spatial variability. We find that significant correlations exist between atmosphere and ocean forecast errors on these timescales, and that these vary diurnally, from day to day, spatially and synoptically. For correlations between errors in the atmospheric wind and ocean temperature, positive correlations in the North Atlantic region are found to be synoptically dependent, with correlation structures extending into the ocean throughout the deep mixed layer, beyond a depth of 100m. In contrast, negative correlations over the Indian Ocean are very shallow and are associated with the diurnal cycle of solar radiation. The significance and variability of cross-correlations indicates that there should be a benefit from including them in data assimilation systems, but it will be important to allow for some flow-dependence in the correlations. Furthermore, the differing vertical extents of the cross-correlations in different regions implies the need for situation-dependent localization of ensemble correlations when including them in coupled data assimilation systems.