A comparison of surface-layer and surface turbulent-flux observations over the Labrador Sea with ECMWF analyses and NCEP reanalyses (Formerly Paper P1.9)

Comparisons are made between a time series of meteorological surface-layer observational data taken on board the R/V Knorr, and model analysis data from the European Centre for Medium Range Weather Forecasting (ECMWF) and the National Center for Environmental Prediction (NCEP). The observational data were gathered during a winter cruise of the R/V Knorr, from 6 February to 13 March 1997, as part of the Labrador Sea Deep Convection Experiment. The surface-layer observations generally compare well with both model representations of the wintertime atmosphere. The biases that exist are mainly related to discrepancies in the sea surface temperature or the relative humidity of the analyses.

The surface-layer observations are used to generate bulk estimates of the surface sensible and latent heat fluxes, and these are compared with the model-generated surface heat fluxes. The ECMWF surface sensible and latent heat flux time series compare reasonably well, with only a 15% overestimation of the bulk fluxes. In contrast, the NCEP model overestimates the bulk fluxes by over 50% and 30% for sensible and latent fluxes respectively. The differences between the bulk estimates and those of the two models are due to different surface heat flux algorithms. It is shown that the roughness length formula used in the NCEP reanalysis project is inappropriate for situations of large air-sea temperature differences and high wind speeds. Its failings are therefore most acute in areas of high sensible heat fluxes such as the Labrador Sea. The new operational NCEP bulk algorithm is found to be more appropriate for such areas. It is concluded that surface heat flux fields from the ECMWF are within the bounds of observational uncertainty and therefore suitable for driving ocean models. This is in contrast to the surface flux fields from the NCEP reanalysis project, where the application of a more suitable algorithm to the model surface-layer meteorological data is recommended.