The component models should make predictions for intensity (I) and linear polarisation (Q, U).

The aim of the models is to provide the raw materials for producing simulations. A simulated channel map is typically produced using the following steps:

1. In each pixel, the emission from each component is integrated over the band of the detector using the spectrum provided by the model,
2. The contributions of all the components are added together,
3. The map is convolved with instrumental beam,
4. Instrumental noise is added to map.

The models must predict emission at a given frequency at a particular point on the sky. In order to be used in the process outlined above, the predictions must be:
 not integrated across a spectral band,
 not convolved with a beam,
 noise-free.

If the input data do not allow this, then their deficiencies should be remedied as best as possible. For example, if the input data are at low resolution then the resolution could be boosted by adding artificial small-scale structure.

The predictions must be made in units of MJy/steradian

Each model should provide a set of input data (for example, a set of maps or a source catalogue) with an algorithm for predicting the flux at a particular frequency in a particular pixel or at a particular point on the sky. The algorithm need not be an actual piece of code, it should however provide enough information for the component model to be implemented in software. The models must include a "readme" file that gives a brief description of the data and algorithm.

If maps are needed for a model then they must be:

 HEALPix maps
 pixel size: N_side = 2048 (1.72’)
 coordinates: Level-S currently needs maps in ecliptic coordinates. Maps in galactic coordinates are useful too if simulated channel maps are produced without using timelines.
 polarisation conventions: HEALPix 1.2