Quantification of Scales Not Constrained by Observation Using Ensembles

Abstract Numerical models for ocean prediction typically resolve finer scales than can be constrained through assimilation of satellite measurements. As a result, variability not constrained by observation contributes to model forecast errors. Using an eddy‐permitting, ensemble ocean prediction system we demonstrate the ensemble mean can be used to filter out the variability not constrained by observation, also showing how the length scales associated with this unconstrained variability can vary globally. In the process it shows that the limit of length scales not constrained by observation is a product of the system and not imposed through ensemble perturbations. Finally, it is shown the removal of length scales not constrained by observation through the use of the ensemble mean reduces error in surface currents relative to a single simulation.