The radiation model is an additional energy transport mechanism; it does not create any additional sources of energy (except where the model provides for increased heat flow at boundaries). However, it does not follow that switching on the radiation model results in conservation of total energy within the system. The difference between the incident radiation and emitted radiation represents a heating or cooling effect. There is, therefore, a quantifiable energy "storage" (either positive or negative).

Suppose, for example, a combusting flow is solved with the radiation model initially set to None, and then radiation is turned on. Where the fluid is already hot (for example, owing to combustion), the default initial guess introduces a large amount of radiant energy into the domain. Note that incident radiation scales with the fourth power of the local temperature. This extra energy can take a long time to diffuse out of the domain. The default initial guess can therefore be poor for combusting flows.

By its nature, the incident radiation is more uniform throughout the domain than the medium temperature . Thus, a uniform value of incident radiation everywhere might be a better initial guess. It is recommended that the chosen value for the incident radiation result in as small a change as possible in the energy storage within the flow domain. A suitable level can be obtained by integrating the radiant energy equation over the entire flow domain (): For details, see The P1 Model in the CFX-Solver Theory Guide.

(10–1)

where is that part of the boundary where an "emissivity-specified" boundary condition is applied. Hence, an average constant value of may be evaluated from:

(10–2)

where:

(10–3)

and is the total volume.