Evaluation of operational ECMWF surface heat fluxes: impact of parameterisation changes during 1986–1995

Abstract

Listen

Translate article

Monthly mean values of ocean surface fluxes produced at the European Centre for Medium-Range Weather Forecasts (ECMWF) between 1986 and 1995 are investigated. The variations in the fluxes over this 10-year period are clearly linked to changes introduced into the analysis and forecast system, and the present study aims at documenting and understanding the sensitivity of surface flux estimates to model parameterisations. Periods are identified when properties of the ECMWF analyses can be characterized and discussed in terms of their ability to provide a forcing function for ocean circulation models. ECMWF surface fluxes are found to exhibit a sufficiently coherent geophysical behaviour from 1986 to 1988 to be considered as representing a climate state, in spite of several deficiencies characterized by weak solar and latent heat fluxes in equatorial regions. Globally, the annual mean and the seasonal cycle agree fairly well with recent observation-based climatologies. Several changes made to the model's physical parameterisations in the 1989–1991 period had a dramatic impact on the radiative fluxes. Net surface shortwave radiation were obviously over-estimated by up to 40 W m −2, and the net heat flux was clearly far from the expected geophysical values. During the 1992–1994 period, additional modifications to the radiation parameterisation scheme brought back solar radiation to a more reasonable level. However, the net input of heat in equatorial oceans appears to be low, and the uptake of heat by evaporation is insufficient at mid-latitudes in the southern hemisphere, a deficiency noticed in ECMWF analyses since the changes in cloud and radiation parameterisations of 1989. ECMWF latent and radiative fluxes show significant improvements in 1995 after the introduction of prognostic clouds in early April. This new parameterisation increases and adjusts the value of the latent heat loss at mid-latitudes in the southern hemisphere. The latent heat flux is found to be globally larger than in recent climatologies derived from ship observations. Regional deficiencies remain in the high northern latitudes (input of heat too great) and in narrow tropical bands in the inter-tropical convergence zone (ITCZ) (cooling by evaporation too great and solar heating too small), and produce a significant bias in the meridional heat transport. Over the 10-year period, latent heat loss is increased through almost 40 W m −2 by the successive model changes, apparently in the right direction, although it might be too large at the end of the period. In preparation for a study of the freshwater flux at the ocean surface (evaporation minus precipitation) in a forthcoming paper by Béranger et al. [Béranger, K., Siefridt, L., Barnier, B., Garnier, E., Roquet, H., 1998. Evaluation of operational ECMWF surface freshwater fluxes over oceans during 1991–97, and possible thermohaline effects. Submitted to J. Mar. Systems], this paper studies the climatological average of the latent heat flux over the period 1991–95. This average is characterized by a larger loss than in previous years (late 1980s), which appears to be more consistent with the current opinion that most climatologies derived from observations made by the Voluntary Observing Ship (VOS) program underestimate this type of exchange. However, the averaging does not fully correct the low evaporation at southern mid-latitudes. To conclude, the paper contributes to understanding of the accuracy currently reached in flux calculations at the global scale in the ECMWF reanalysis project.