An empirical approach for estimating macroturbulent heat transport conditional upon the mean state

A stochastic specification for monthly mean wintertime eddy heat transport conditional upon the monthly mean circulation is proposed. The approach is based on an analog technique. The nearest neighbor for the monthly mean str eamfunction (at 850 and 300 hPa) is searched for in a library composed of m onthly data of a 1268-yr control simulation with a coupled ocean-atmosphere model. To reduce the degrees of freedom a limited area (the North Atlantic sector) is used for the analog specification. The monthly means of northwa rd transient eddy flux of temperature (at 750 hPa) are simulated as a funct ion of these analogues. The stochastic model is applied to 300 years of a paleosimulation (last int erglacial maximum around 125 kyr BP). The level of variability of the eddy heat flux is reproduced by the analog estimator, as well as the link betwee n monthly mean circulation and synoptic-scale variability. The changed boun dary conditions (solar radiation and CO2 level) cause the Eemian variabilit y to be significantly reduced compared to the control simulation. Although analogues are not: a very good predictor of heat fluxes for individual mont hs, they turn out to be excellent predictors of the distribution (or at lea st the variance) of heat fluxes in an anomalous climate.