Estimating background error covariance parameters and assessing their impact in the OSTIA system

Abstract

Abstract To generate sea surface temperature (SST) analyses, observations are assimilated onto a background field to produce a gap-free map of SST. The spatial influence of an observation as well as its weight are determined by the error covariance matrices used in the data assimilation. The background error covariance parameters for an SST analysis have been estimated using observation-minus-background difference data from a long term SST reanalysis. Geographically varying estimates were obtained of both the error variance and the associated length scale of the background error which were decomposed into small-scale and large-scale errors. Both components of the background error variances were found to be seasonally variable. Minimal seasonality was observed in the error correlation length scales but they were found to be anisotropic. Observational errors were also estimated for the observation types used in the observation-minus-background calculations. The impact of updating the background error parameters in the Operational Sea surface Temperature and sea-Ice Analysis (OSTIA) was assessed. The updates improved the analysis accuracy compared to assimilated and independent reference observations. The covariance updates also improved the ability of the SST analysis to represent small-scale ocean features without introducing spurious noise. As OSTIA is used as boundary conditions in the numerical weather prediction (NWP) system at the Met Office the impact on the NWP forecasts was also assessed. The OSTIA changes were found to have a positive impact on NWP forecast skill.