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. Plain Language Summary Ocean reanalysis and state estimation seeks to bring an ocean model into consistency with available observations via data assimilation, providing a complete reconstruction of the time‐evolving ocean state to support forecasting efforts and climate research. Uncertainty remains in the resulting products, however, arising from uncertainties in the underlying model, applied atmospheric forcings, assimilated data constraints, and assimilation method. Whilst computational challenges prevent comprehensive uncertainty quantification, this information is valuable for all applications (e.g., forecast initialization; robust climate change detection). To address this issue, we have sought to provide the first uncertainty estimate for the latest release of the ECCO global ocean and sea ice state estimate, spanning the period 1992–2017. Using an ensemble perturbation approach, we explore the impact of uncertainties in applied atmospheric forcings. Our ensemble design is advantageous in accounting for the joint variations between different atmospheric variables and in providing new insights into the space‐time characteristics of adjustments made to these terms as part of the ECCO data assimilation procedure, highlighting the relative role of time‐mean and time‐variable adjustments. Our perturbed ensemble yields distinct solutions to the inverse problem with moderate changes in climate‐relevant metrics, including ocean meridional overturning circulation, heat transport and heat uptake. Key Points Using a perturbed ensemble we explore the impact of uncertainty in atmospheric forcings on the ECCO ocean and sea ice state estimate The ensemble yields alternative estimated solutions (with acceptable model‐data misfits) if time‐meanforcing adjustments are retained These alternative solutions are accompanied by moderate changes in climate relevant metrics, including ocean overturning and heat uptake