Here is a description of the NCEP re-analysis 6-hourly and monthly-mean data for the SBI study domain for the years 2001-2004. These matlab files have been kindly provided by Kent Moore of the University of Toronto. Please acknowledge him when using the data. There are 2 files for each month: ncep_{01,02,...12}_{2001,2002,2003,2004}.mat contains the 6hourly data for month {01,02,...12} in year {2001,2002,2003,2004} ncep_mm_{01,02,...12}_{2001,2002,2003,2004} contains the monthly mean data. Please note that the 2004 data only goes up to the end of october (conlusion of the last SBI cruise). The variable names/units are as follows: curlms_m 4-D curl of the wind stress N/m3 icec_m 4-D icec concentration in % landg 23x46 land mask 1==land o==ocean latg 23x1 latitude of grid points (40N--80N) lhf_m 4-D adjusted latent heat flux (W/m2) lhfx_m 4-D raw latent heat flux (W/m2) long 46x1 longitude of grid points (160E--120W) nlwrs_m 4-D net long wave radiation (W/m2) nswrs_m 4-D net short wave (W/m2) prate_m 4-D precipitation rate (mm/hr) q2m_m 4-D 2m specific humidity (g/kg) shf_m 4-D adjusted sensible heat flux (W/m2) shfx_m 4-D raw sensible heat flux (W/m2) slpg_m 4-D slp (mb) t2m_m 4-D 2m air temperature (C) tsfc_m 4-D surface temperature (C) u10m_m 4-D 10m zonal wind speed (m/s) ums_m 4-D zonal wind stress (N/m2) v10m_m 4-D 10m meridional wind speed (m/s) vms_m 4-D meridional wind stress (N/m2) notes: 1) all fields are on a grid whose lat/lon are given in latg and long. Resolution is approx 2deg. 2) most oceanic fields have been masked out over land/ice (set to NaN). Some fields (prate for example) have not been masked out. 3) the sign convention for the fluxes is that fluxes out of the ocean are positive. 4) All fields are a direct extraction of the NCEP archive with the exception of the adjusted sensible and latent heat fluxes, and the curl of the windstress. For the heat fluxes, there are problems with NCEP's surface layer paramterization that yield model fluxes that are too large (but the surface fields themselves are fine). Hence the sensible and latent heat flux were re-calculated using bulk formulae validated against in-situ data from the R/V KNORR collected during the 1997 Labrador Sea convection experiment. Details are discussed in Renfrew et al. (2002)*. Regarding the windstress curl, this was calculated using a simple center difference scheme on the sphere. There may be issues in this field adjacient to coastlines/sea ice. *Renfrew, I.A., G.W.K. Moore, P.S. Guest, K. Bumke, 2002: A comparison of surface-layer heat flux and surface momentum flux observations over the Labrador Sea with ECMWF analyses and NCEP reanalyses. Journal of Physical Oceanography, 32, 383-400. 5) Some simple quality control was done and things look reasonable. Note that the monthly means in regions of variable sea ice (within a given month) will be suspect as the averaging routine was very conservative. 6) In the arctic the sparsity of observations means that the reanalysis is mostly model dependent and so one would expect some biases, as there are not enough data to pin the model down. There are probably issues with the winds as well. One known bias (at least 10 years ago) is that the NCEP air temperature tends to be 1-2 degrees too warm. This bias is most pronounced in the winter. In summer/fall when there is open water, the surface air temperatures are pretty much constrained by the SST which are probably well known. In the presence of ice, one needs to run a surface energy balance calculation to get the surface (ice) and air temperatures. Such a calculation is fraught with problems (you need the thickness of ice and snow to get the surface flux right...) and this is likely where the problem lies. Note, however, if everything else is constant, a 1 deg bias on a 20 deg air-sea temperature difference would introduce a 5% bias in the sensible heat flux. On a 10 deg air-sea temperature difference it would be 10% and on a 5 deg difference it would be 20%. The latent heat flux errors are probabaly more difficult to estimate because of the non-linear relation between saturation and temperature, but, if it is really cold, the latent fluxes are not that large anyway.