Xiangbo Feng, University of Reading

Calculation of coupled error covariances from Coupled ECMWF Reanalysis

Keith Haines and David Mulholland


In the context of ocean and atmosphere modelling, coupled data assimilation has been developed at ECMWF, which is able to reduce the initialization shocks and better represent the coupled behaviours observed in the lower atmosphere and upper ocean. The Coupled ECMWF ReAnalysis (CERA) system has been built under ERA-CLIM and ERA-CLIM2 projects based on an ocean-atmosphere coupled system with data assimilation applied individually for each component. This system is classified as ‘weakly’ coupled data assimilation as no cross error covariances are used during the assimilation window. Now we are targeting the possibility of developing ‘strongly’ coupled data assimilation, where coupled error covariances must be used, based on the existing CERA system. We will present the coupled error covariances calculated by ensemble method, by which the spread of 10 ensemble members of forecast fields produced from CERA is treated as background error. First the features of ensemble spread are diagnosed with emphasis on lower atmosphere variables and SST. Correlations between different model fields are then calculated across the ensemble. The correlations are found not only exhibiting clearly regional/large-scale patterns over monthly time, but also showing strongly local/small-scale variations at daily time scales. Evolutions of both ensemble spread and coupled error covariances against different leadtime of forecast are discussed as well.

File available here