We investigated the response of epiphytic lichens to air pollution, against the background of other ecological predictors in a prealpine heterogeneous area, using Non-Parametric Multiplicative Regression (NPMR). The best NPMR model for total lichen diversity according to N environmental predictors at tree level has a cross R(2)=0.709. It includes 10 variables, belonging to three different subsets of factors: two pollution-related factors (distance in meters from the road and from the cement factory); four stand-related (habitat, heat index, LAI and elevation) and four substrate-related factors (inclination, circumference and texture and tree species). Considering separately the effects of each subset on lichen diversity, substrate- and stand-related factors produce good models with similar cross R(2) (0.490 and 0.500, respectively), whereas pollution-related factors produce a model with a lower cross R(2) (0.340). Hence, we provide information to investigate the applicability of lichen biomonitoring to complex heterogeneous areas where standardized protocols are not reliable.