A premise is a constraint to be checked more or less strongly relating to the geographical environment favourable to a class. This premise can refer to:

• only one type of geographical data (``class c is at altitudes from 800 to 1000 meters'', ``class c is often on slopes facing north''...), or
• the combination of several data (``class c is situated only in irrigable areas'': the calculation of the irrigable areas in the image is defined from the distance to the rivers and from the slope steepness.

For each class, a production rule, connecting premises with conjunction and disjunction operators, is defined. The degree of realization of this rule for the current pixel determines if its context is favourable or not to the presence of a given class.

Moreover, certainty factors relate to each premise of a rule. Those express the degree of certainty assigned to each premise.

Using all this knowledge and these constraints, Desachy, Zahzah and Mascarilla calculated, for each pixel, a degree of realization taking its values in the interval [ -1, +1 ]:

• -1 means that the pixel is in an unfavourable context where no premise is carried out,
• 0 means that the pixel is in a context where no decision can be performed,
• +1 means that the pixel is in a favourable context where all the premises are carried out.

These degrees of realization were obtained by the evaluation of the various premises composing the rules (close to roads, altitude higher than 1850 meters...) weighted by the frequency of appearance attributed to these premises (only, mainly, seldom...).