Habitat suitability (HS) modeling relates a species' potential presence/absence to a set of environmental variables. Because of stochastic or demographic population fluctuations, relating abundance to environment is difficult and generally requires,, time-series of population-density data. Here, I propose an approach to compute relative capacity of a set of habitat classes. I defined relative capacity of a habitat class as being proportional to its carrying capacity, and replaced time-series of density measures with spatial replicates. A hypothetical environment was first divided into several habitat classes and population densities were measured in a sample of these habitats. Three methods of computing relative capacity were compared: (1) maximum density per habitat class, (2) coefficients of a Generalized Linear Model (GLM) of the habitat densities, and (3) slopes of isodars computed for all pairs of habitat classes. Accuracy of these methods was evaluated using spatially-explicit demographic Simulations. I investigated biological assumptions (species' sensitivity, reactivity and selectivity, regional stochasticity) and sampling design (sample size, distance between pairs of habitat classes. number of habitat classes, uncertainty over abundance). GLM and maximum-density methods provided accurate results when at least five site-pairs were available. Isodars outperformed other methods in landscapes with many habitat classes, but are best suited to large or noise-free data sets. Optimal performance was reached with more than 50 site-pairs and five habitat classes.