Quantifying coupling errors in atmosphere-ocean-sea ice models: A study of iterative and non-iterative approaches in the EC-Earth AOSCM

Abstract. The atmosphere, ocean, and sea ice components in Earth system models are coupled via boundary conditions at the sea surface. Standard coupling algorithms correspond to the first step of an iteration, so-called Schwarz waveform relaxation. Not iterating is computationally cheap but introduces a numerical coupling error, which we aim to quantify for the case of a coupled single column model: the EC-Earth AOSCM, which uses the same coupling setup and model physics as its host model, EC-Earth. To this end, we iterate until a reference solution is obtained and compare this with standard, non-iterative algorithms. Understanding the convergence behavior of the iteration, as well as the size of the coupling error, can inform model and algorithm development. Our implementation is based on the OASIS3-MCT coupler and allows to estimate the coupling error of multi-day simulations. In the absence of sea ice, SWR convergence is robust. Coupling errors for atmospheric variables can be substantial. When sea ice is present, results strongly depend on the model version. In the latest model version, coupling errors in sea ice surface and atmospheric boundary layer temperature are often large. Generally, we find that abrupt transitions between distinct physical regimes in certain parameterizations can lead to substantial coupling errors and even non-convergence of the iteration. We attribute discontinuities in the computation of atmospheric vertical turbulence and sea ice albedo as sources for these problems.