ALBERT

Description: To analyze the probability density distributions of surface turbulent heat fluxes, the authors apply the twoparametric modified Fisher–Tippett (MFT) distribution to the sensible and latent turbulent heat fluxes recomputed from 6-hourly NCEP–NCAR reanalysis state variables for the period from 1948 to 2008. They derived the mean climatology and seasonal cycle of the location and scale parameters of the MFT distribution. Analysis of the parameters of probability distributions identified the areas where similar surface turbulent fluxes are determined by the very different shape of probability density functions. Estimated extreme turbulent heat fluxes amount to 1500–2000 W m22 (for the 99th percentile) and can exceed 2000 W m22 for higher percentiles in the subpolar latitudes and western boundary current regions. Analysis of linear trends and interannual variability in the mean and extreme fluxes shows that the strongest trends in extreme fluxes (more than 15 W m22 decade21) in the western boundary current regions are associated with the changes in the shape of distribution. In many regions changes in extreme fluxes may be different from those for the mean fluxes at interannual and decadal time scales. The correlation between interannual variability of themean and extreme fluxes is relatively low in the tropics, the SouthernOcean, and the Kuroshio Extension region.Analysis of probability distributions in turbulent fluxes has also been used in assessing the impact of sampling errors in theVoluntaryObserving Ship (VOS)-based surface flux climatologies, allowed for the estimation of the impact of sampling in extreme fluxes. Although sampling does not have a visible systematic effect onmean fluxes, sampling uncertainties result in the underestimation of extreme flux values exceeding 100 W m22 in poorly sampled regions.