<p>Many decisions take place in geographic contexts. This paper focuses on how to compare maps representing the state of a region under different evolution scenarios or before and after the application of a certain policy. In such maps, that we call decisional maps, each geographic unit is assessed with respect to some criterion. In our setting, this criterion can only take a finite number of different values. For instance, it may result from a qualitative evaluation of the geographic units or of the application of a multicriteria sorting (or classification) method assigning each geographic unit to a category. A crucial hypothesis is that the maps can be partitioned in homogeneous subregions in which the area distribution of geographic units into categories only matters, not the precise location of each geographic unit. Under this hypothesis, our model is characterized, leaning on expected utility theory. This model is compared with other models proposed for comparing decisional maps. We also describe procedures for eliciting its parameters, which we illustrate with a risk assessment example.</p>