This quote from a general discussion paper by Raupach and Finnigan is actually the title of the paper, and it succinctly summarizes one of the underlying compromises that must be made when developing models of earth system processes – convenience versus accuracy. Models are necessary for allowing us to proceed from observations to generalizations, and therefore from descriptions to projections. However, a model, by necessity, is an imperfect representation of a process. In earlier chapters we addressed the nature of models that have been developed at the biochemical and leaf scales. Here, we address models describing the turbulent exchanges between canopies or landscapes and the atmosphere.

If models are accurate in their depiction of transport processes, the area-specific turbulent flux that is predicted at the scale of a whole canopy should converge with the integral spanning all leaf diffusive fluxes determined within the canopy. Given this constraint, one could argue that knowledge of turbulence and its representation in a model is not necessary to determine surface-atmosphere fluxes; one need only determine the integral sum of diffusive fluxes in the underlying leaves. This type of bottom-up model, however, in which diffusive fluxes are simply summed to give a canopy flux, is incapable of describing how those fluxes distribute their respective scalar entities across the space within and above the canopy, and thus how variance in scalar concentration gradients can affect diffusive source or sink activity. In essence, modeled fluxes are uncoupled from dynamics in their associated concentration gradients. In order to dynamically link fluxes to gradients, atmospheric transport must be considered; mass or energy must be dispersed to or from the immediate vicinity of sources and sinks. One of the primary challenges facing modelers seeking to describe dynamic flux systems at the canopy scale is the inclusion of accurate scalar dispersion algorithms.