Abstract Aim We addressed four objectives: (1) Determine the regional responses of species, size classes and a vegetation type to climate and parent material predictors, including their distributions in environmental space and the relative contributions of the predictors to explained variation. (2) Determine whether size classes of a species respond similarly to climate and parent material. (3) Assess the extent to which the predicted regional distribution of a vegetation type can be approximated by the distribution of its diagnostic species and vice versa. The establishment of a consistent relationship between the distribution of a vegetation type and its diagnostic species would facilitate change detection, management and conservation planning by allowing the use of one distribution to generate the other when data availability is limited. (4) Examine landscape-scale environmental variability in predicted species and vegetation type distributions. Location South-western USA (Arizona, New Mexico and southern Colorado). Methods Ecological response surface models were developed using a data base of 1409 vegetation plots to analyse biotic–environmental relationships of (1) Pinus ponderosa P. & C. Lawson and Abies concolor (Gord. & Glend.) Lindl. Ex Hildebr. size classes, (2) P. ponderosa, A. concolor and Quercus gambelii Nutt. combined size classes, and (3) a P. ponderosa forest type widely distributed in the south-western USA. Results and main conclusions Pinus ponderosa and A. concolor models generally were judged to be successful. Quercus gambelii models were judged unsuccessful, which may result from the influence of variables not modelled, such as soil moisture, disturbance, biotic factors and other site limiting factors. Size classes differed in the range of environmental conditions associated with high occurrence probabilities within and between species, reflecting differences in the effects of climate variability and anthropogenic changes, such as fire suppression, on the distribution of each size class. Pinus ponderosa alliance was predicted to be distributed over a narrower range of environmental conditions than P. ponderosa species models, therefore limiting the use of this vegetation type as a surrogate for the distribution of the dominant species, and vice versa. Maps of combinations of environmental variables that produced a high probability of P. ponderosa occurrence showed that some landscapes predicted to contain the species exhibited diverse environmental conditions over short distances. The use of regional environmental relationships to characterize areas with high local environmental variability may facilitate identification of areas of potential rapid biotic change.