The effect of horizontal resolution on ocean surface heat fluxes in the ECMWF model

Abstract

Annual mean ocean surface heat fluxes have been studied as a function of horizontal resolution in the ECMWF model (cycle 33) and compared with Oberhuber's COADS (1959–1979) based empirical estimates. The model has been run at resolutions of T21, T42, T63 and T106 for 15 months with prescribed monthly varying climatological SST and sea ice. The T42 simulation was extended to 2 years, which enabled us to determine that many differences between the resolution runs were significant and could not be explained by the fact that individual realizations of an ensemble of years can be expected to give different estimates of the annual mean climate state. In addition to systematic differences between the modeled and the observed fluxes, the simulated fields of surface shortwave and longwave radiation showed much more spatial variability than the observed estimates. In the case of the longwave radiation this may be attributable more to deficiencies in the observations than to errors in the model. The modeled latent and sensible heat fields were in better agreement with observations. The primary conclusion concerning the dependence of ocean surface fluxes on resolution is that the T21 simulation differed significantly from the higher resolution runs, especially in the tropics. Although the differences among the three higher resolution simulations were generally small over most of the world ocean, there were local areas with large differences. It appears, therefore, that in relation to ocean surface heat fluxes, a resolution greater than T42 may not be justified for climate model simulations, although the locally large differences found between the higher resolution runs suggest that convergence has not been achieved everywhere even at T106.