An Assessment of Uncertainty in the ECCO Global Ocean-Sea Ice State Estimate Due To Atmospheric Forcing Uncertainty

Abstract

The Estimating the Circulation and Climate of the Ocean (ECCO) state estimate is the result of adjusting a set of controls comprising atmospheric forcings, initial conditions, and mixing parameters to reduce model-data misfits. Despite this, uncertainties remain in the solution. Among others, small amplitude perturbations to the optimized controls may yield differences in the estimated state without notably increasing the misfits, providing distinct but equally acceptable solutions to the inverse problem. We focus on the impact of uncertainty in the atmospheric controls via ensemble perturbation. Our multivariate empirical orthogonal function (EOF) approach to construct the ensemble perturbations accounts for the covariance of control variables. Furthermore, it provides new insights into the space-time characteristics of ECCO's atmospheric adjustments. The two leading EOFs of these adjustments show a seasonal cycle dominated by high-latitude adjustments and a decadal component. Removing the time-mean of the adjustments results in large model-data misfits and thus unacceptable estimates. Ensemble perturbations in time-varying adjustments incur uneven uncertainties in oceanic metrics, for example, in global meridional heat transport (0.03 PW), the Atlantic meridional overturning circulation at 26°N (0.7 Sv), or ocean heat uptake (15 ZJ). These are an order of magnitude smaller than the uncertainty evaluated via ocean reanalysis intercomparisons and forward perturbation ensembles. The relatively weak impacts result from the relatively small amplitude of estimated atmospheric uncertainty in the ECCO release, out of sufficient consideration of a massive set of observational constraints. Future work should assess the impact of other sources of uncertainties.