Institute of Meteorology and Geophysics, University of Vienna, Austria For the purpose of specifying the strength of atmospheric convection we review how correlation fluxes and phase transition terms are generated from first principles, i.e., from a thermodynamic set of equations. They may be used either for a forecast (prognostic mode) or for an analysis (diagnostic mode). We identify the total convective heat flux c as key parameter for quantifying convection. Other convective quantities can be derived from c. In the prognostic mode the convective quantities are parameterized. In the diagnostic mode they are indirectly inferred. Here we concentrate on the diagnostic mode. We review how c can be estimated from the routinely analyzed gridscale budget of moist enthalpy by solving a first-order linear differential equation, referred to as convection equation. A novel aspect of the present standard of the convection equation is that the ice phase has now been included. The pertinent algorithm runs under the acronym DIAMOD (diagnostic model). The difference to a LAM is that DIAMOD takes the gridscale tendency as observed input while in a LAM it is output. We demonstrate the use of c (and of the net condensation rate CON) through an observation simulation experiment. Gridscale budget data from a forecast run with the Deutschland-Modell are taken as perfect input for DIAMOD; we study the corresponding output for several approximations of the convection quation. The vertical profiles of c (and to a lesser extent those of CON) are relatively robust and precise (error margins of a few percent) while the accuracy of the other convective quantities is inferior. Conclusion is that parameterization schemes of different LAMs can be compared with each other as well as validated in terms of c and CON. Application of this technique has been demonstrated in Part I of this study.