Extreme air-sea turbulent heat fluxes over the global oceans: determination, implications and mechanisms

Extreme surface turbulent heat fluxes affecting convective processes in subpolar ocean regions may amount to 1000-3000 W/m². Their quantitative estimation is critically important for many oceanographic and meteorological applications. Extreme turbulent fluxes are largely responsible for the vertical mixing in the ocean, especially in the subpolar latitudes where deep convection forms intermediate waters. Over western boundary currents and their extension regions very strong turbulent heat fluxes may result in local responses in the lower atmosphere on time scales from several hours to days and spatial scales from several kilometers to several tens of kilometers. Accurate estimation of extreme turbulent fluxes also strongly relates to the sampling problem especially for the poorly and irregularly sampled regions. Estimation of extreme turbulent fluxes is thus requires knowledge of probability distribution of fluxes. We suggest a concept for determination of extreme surface turbulent heat fluxes based upon theoretical probability distributions, which allow for accurate estimation of extreme fluxes. In this concept the absolute extremeness of surface turbulent fluxes is quantified from the Modified Fisher-Tippett (MFT) distribution. Further we extend MFT distribution to a fractional distribution, quantifying the so-called relative extremeness, representing the fraction of surface flux accumulated during continuous time (e.g. months, season) due to most intense surface fluxes (e.g. the strongest 1% of flux events). We provide explicit form of the fractional distribution and effective algorithms for parameter estimation.

Further we demonstrate applications of the concept for the global ocean using reanalyses and Voluntary Observing Ship (VOS) data for the period 1979 onwards. Global climatologies reveal that the regions with the strongest relative extremeness are not collocated with the strongest mean fluxes. Moreover, interannual variability of the absolute and relative flux extremes is not necessarily correlated with variability of mean fluxes. Growing mean flux may result in both increase and decrease of absolute and relative extremes that has implications for estimates of linear trends which may have different signs for mean fluxes, absolute and relative flux extremes – the situations found for the western boundary currents and major convections sites. Suggested concept has also profound implications for comparative assessments of surface turbulent fluxes from reanalyses (ERA5, ERA-Interim, CFSR, MERRA2, NCEP-DOE, JRA55) and satellite (IFREMER, J-OFURO, HOAPS, SEAFLUX) products. High mean fluxes in some products are not necessarily associated with the strongest absolute and relative extremeness in the same products and vice versa. Also a new concept allows for accurate estimation and minimization of sampling biases in VOS and satellite flux products. Finally, we analyzed the mechanisms responsible for forming extreme surface turbulent fluxes associated with cold air outbreaks in the rear parts of midlatitudinal cyclones under locally high winds and strong air-sea temperature gradients.