An Improved Approach for Evaluating Ocean and Climate Model Errors in Mean Dynamic Sea Level

Observationally based reference products for time-mean dynamic sea level (MDSL) are useful for evaluating the representation of sea level and ocean circulation in climate models. They also serve as constraints for ocean reanalysis products. However, in previous analyses, little attention has been paid to consistently quantifying and interpreting various sources of model–data differences. Such an effort is especially important near coastlines, where data uncertainties are expected to be relatively large, and model representation of sea level is critical. Here, we find that two new MDSL reference products exhibit differences well beyond their provided uncertainties, especially in western boundary currents and along coastlines. We thus implement a consistent approach to evaluating reference product and model errors arising from different sources, including mistimed internal variability, that can be extended to accommodate different models and/or reference datasets. In a demonstration evaluation of a GFDL climate model, we find that, along the U.S. East Coast, uncertainties in reference products are generally smaller than biases in the climate model, suggestive of misrepresented dynamics and/or the influence of coarsened bathymetry. Model errors due to internal variability play a less important but nonnegligible role.