Co-authors Andrew Staniforth and Jean Cote.

NWP and climate models are viewed as consisting of two distinct

modules: the adiabatic, inviscid equation set (the dynamics); and the parametrised diabatic and viscous forcings (the physics). Each module is generally considered in isolation. The inevitable coupling between them is often done in a seemingly ad hoc manner - any justification usually being argued on physical grounds rather than founded in numerical analysis.

Recent developments mean that the dynamical components of the models are generally very accurate and numerically stable while considerable effort continues to be expended on improving the accuracy, and also the scope, of the various physical parametrisations. However, the recent work of Caya, Laprise and Zwack (1998) [CLZ98] well illustrates that the way the parametrisations are coupled to the dynamics can significantly limit the

accuracy and numerical stability of the model as a whole. Additionally they show that the choice of coupling can impact dramatically on the accuracy of the steady state that the model can achieve (which is generally the most understood and proven aspect of the physical parametrisations). It seems clear therefore that if the community is to reap the full benefits of its investment in improved dynamics and physics modules, more attention must be paid to analysing, understanding and improving the strategy for the physics-dynamics coupling.

The complexity of the physics schemes and the non-linear nature of the coupled system makes this a non-trivial task. However, CLZ98 have shown that useful progress can be made by judiciously reducing the problem to its essence. They presented a highly simplified model, or paradigm, of a physics-dynamics coupling and used it to diagnose the source of a problem in their global model. We have extended their paradigm to allow for the effects of uniform advection and any number of temporally and spatially varying forcings. These extensions permit the study of rather more complex physics-dynamics coupling strategies (as well as numerical issues such as spurious semi-Lagrangian orographic resonance) but in an analytically tractable framework. This extended framework will be presented as will results of its application to four coupling strategies in the context of a given spatio-temporal forcing coupled with a diffusive process.