45 downwards because (j on the average increases with height; but this conclusion does not follow from (18.3) when the dependency of Kc upon ~o is taken into consideration. s 2 ERTELl and PRIESTLEY and SWINBANK have shown that the upward eddy­ flux of sensible heat must be larger than indicated by (18.3), because this formula does not account for the fact that rising eddies are systematically warmer than sinking eddies because of the effect of buoyancy. The reader is referred to the reviews by SUTTON [22], [23] and PRIESTLEY and SHEP­ PARD [15) for further details concerning eddy-flux of heat and turbulent diffusion. 19. RICHARDSON'S criterion. The right-hand side of (15.10) represents the rate of production of eddy energy. The last term represents energy loss by dissipation; in order that the eddy energy shall be maintained, it is therefore necessary that P div V" - (! V" v" . grad. v > O.