Comparison of Surface Turbulent Flux Products Paul J. Hughes, Mark A. Bourassa, and Shawn R. Smith Center for Ocean-Atmospheric Prediction Studies & Department of Meteorology Florida State University, Tallahassee, FL Flux Product Information There are many types of products that include turbulent fluxes of latent heat, sensible heat, and stress. Reanalysis products, created from numerical weather prediction (NWP) models with fixed model physics (albeit with changing data products for assimilation), are often used because they provide a wide range of additional information at the surface and at various levels in the atmosphere. However, such models have poor representations of the atmospheric boundary layer, and questionable parameterizations of surface turbulent stresses. These deficiencies in the lower boundary condition decrease the credibility of these models for climate studies as well as for ocean forcing. The fluxes from the NWP models are compared to fluxes derived from satellite measurements and ship and buoy observations, which are not without their own shortcomings (e.g., regional/global biases and poor/inhomogeneous sampling). Satellite derived products include IFREMER and HOAPS. Products based on ship and buoy observations include FSU3 and NOC1.1 (see Fig. 3 for in situ coverage). Hybrid NWP model and satellite products include WHOI and GSSTF2. 1. Introduction Monthly averaged surface turbulent fluxes (stress, sensible heat, and latent heat) are compared for nine products: NCEPR2, JRA25, ERA40, WHOI’s OAFLUX, GSSTF2, IFREMER, HOAPS2, FSU3, and NOCSv1.1 (formerly SOC). The common period of March 1993 through Dec is examined. Input data are also compared when available. Each product has been regridded onto a 1x1  grid. To reduce problems related to land, data within two grid cells of land are not used in this comparison. Some satellite winds are included in this comparison. There are very large differences in the distribution of zonally averaged heat fluxes (Fig. 1). Curl of the wind stress (Fig. 2) also shows great differences among these products. In many cases the observing pattern (ships or TAO buoys; Fig. 3) can be easily identified. In other cases, ringing induced by orography modifies the spatial derivatives. Figure 3. Example VOS observation density for December, averaged from Atlantic Ocean Pacific OceanIndian Ocean Zonal stress Meridional stress Sensible Heat Flux Latent Heat Flux 3. Conclusions The reanalyses based on numerical weather prediction (NWP) stand out as outliers from the observation-based products. The various products have differences in zonally averaged latent heat fluxes for all ocean basins. These differences are large from the point of view of climate modeling, where biases of >10 Wm -2 (including radiation fluxes) are considered serious problems. Stresses, sea surface temperatures (not shown), and to a lesser extent air temperatures (not shown) are relatively similar between products. The greatest difference occur with wind speeds (a seemingly well observed quantity) and atmospheric moisture. These difference both contribute to the large product to product differences in the latent heat flux. Monthly averaged satellite winds are extremely similar in magnitude and pattern to the in situ based FSU3 and NOCS winds. The large differences between the latent heat fluxes of these two products are largely due to differences in the atmospheric moisture. These results are based on monthly averaged fluxes; therefore, they likely underestimate the issues with fluxes produced for shorter time scales. The large differences in fluxes, and in the spatial/temporal changes in these products indicates that there are still serious problems to be resolved in the construction of surface forcing fields for applications in climate and general oceanography. NCEPR2 JRA25 ERA40 IFREMER FSU3 NOCv1.1 Curl of the Stress Mean Curl of the Stress Standard Deviation Divergence of the Wind Mean Divergence of the Wind Standard Deviation Figure 2. Curl of the stress is particularly important for ocean forcing; divergence of winds is import for atmospheric work. The means show similar patterns, although the TOA buoy array is easily identified in the ERA40 product, and ship tracks are seen in the NCEP2, FSU3 and NOC products; however, the ship coverage is much better in the North Atlantic Ocean, where they are not easily identified in the FSU3 product. Orographically induced ringing is apparent in the NCEP2 and JRA25 products. The standard deviations also highlight problems with the objective techniques and the data assimilation. WHOI GSSTF2 HOAPS2 Latent Heat Flux Mean Latent Heat Flux Standard Deviation Figure 1. Distributions (5 th, 25 th, 50 th, 75 th, and 95 th percentiles) of contributions to zonally averaged fluxes. There are very large differences among products, and these difference are a function of the percentile examined. ProductLHFSHF Stress (x,y) Wind Speedu windv windTairQairSST Product Type Grid Spacing NCEPR2xxxxxxxxReanalysisGaussian (T62, 194x94) JRA25xxxxxxxxReanalysis(T106L40) ~120km ERA40xxxxxxxxReanalysis1 1/8 degrees WHOIxxHybrid1 x 1 degree GSSTF2xxxxxHybrid1 x 1 degree IFREMERxxxxxxxxxSatellite1 x 1 degree HOAPS2xxxxxSatellite0.5 x 0.5 degree FSU3xxxxxxxxxIn-situ1 x 1 degree NOC1.1xxxxxxIn-situ1 x 1 degree 4. Acknowledgements We thank the many people that made their products available for comparison, and who were involved in preliminary discussions of how comparisons could be made. We also thank the NOAA Climate Observation Division and NSF PO for supporting this effort. FSU