be even more significant. Studies of these diabatic errors leave little 

 doubt about the fact that they may be, on occasion, large and important, 

 and justify the inclusion of diabatic processes in NWP. 



Winston's (1955) study of the February 1950 cyclogenesis in the Gulf 

 of Alaska, recently re-examined and extended by Pyke (1965), is an early 

 example of the efforts to evaluate the sea-air transfer from prediction 

 errors. A more compi^ehensive study of NWP errors by Martin (1962) demon- 

 strated even more clearly the probably importance of the sea as an energy 

 source for the atmosphere. Martin's computations were in remarkably good 

 agreement with the independent computations of Manabe (195^) for the winter 

 195^-55 cold outbreak over the Japan Sea (with transfers of more than 

 1^+00 ly day"-*-), and also with computations by Petterssen for the North 

 Atlantic . 



Petterssen, Bradbury, and Pedersen (1962) in a diagnostic study of 

 cyclone development over the North Atlantic Ocean have attempted to develop 

 the classical Norwegian cyclone models into a more complete dynamical model 

 by computing the energy transfers from the sea to the air. The results have 

 been somewhat disappointing. In the first place, the computed heat transfers 

 show, as expected, the major heat transfer in the cold air mass to the rear 

 of the cyclone -- but with no physical account of how (or if) this energy 

 contributes to the cyclone development. Secondly, the paper essentially 

 bypasses the question of how the heat source affects the circulation, because 

 only the thickness (i.e., temperature) tendency is computed. Inclusion of 

 the heat source term in the tendency calculation improves the predicted 

 thickness tendency, as expected, but unfortunately, tells us nothing about 

 the effect on the 500-mb circulation. 



C. Diabatic Prediction Models 



The effects of sea-air heat transfer on atmospheric circulation systems 

 are so complex that it appears likely that nothing less than time integration 

 of the complete diabatic system of equations can really tell us much about 

 these phenomena. Such experiments have been attempted with models of vary- 

 ing degrees of complexity. Bushby and Hinds (1955) were probably the first 

 to incorporate sea-air heat transfer in a numerical weather prediction model, 

 followed a few years later by Reed (1958), who employed Fj^rtoft's graphical 

 (La^rangian) method. 



yiy own experiments (Spar. I962), including heat of condensation as well 

 as sea-air heat transfer, employed a somewhat less constrained model, but 

 still only a two- level (vertically integrated) baroclinic model -- 

 geographically limited and geostrophic. Despite these constraints, the 

 experimental results may be of some interest. 



In the computations with my prediction model, I have used the empirical 

 transfer formulas (equations (l) and (2)) to compute (effective) sensible 



