Reliable estimates of species composition that forest sites are capable of supporting – specific to ecosystem mapping units across landscapes – are useful for many purposes in forest science and management. Like forestry agencies in numerous countries, the U.S. Forest Service has invested in ecological land classification (termed terrestrial ecosystem survey [TES] in the study region of Arizona) that includes ecosystem-explicit species lists taken to be estimated potential natural vegetation (PNV). Using multivariate community analyses, PNV in the TES was compared to measured species composition on 66 sites representing among the least-disturbed vegetation (considered this study’s measured PNV) spanning 11 ecosystem types on a Pinus ponderosa P. & C. Lawson landscape in northern Arizona, USA. Agreement between the TES PNV and measured species composition was lowest for forbs and shrubs (compared to graminoids), and species composition differed significantly between the TES and this study for at least one plant lifeform in 73% of ecosystems. Reasons for differences between the TES and this study are difficult to resolve, but in some cases appear to result from identification of different species pools in the region. This study suggests that the TES is a useful starting point in understanding vegetation-environment relationships, but further work is needed to refine species lists and more thoroughly account for the influences of fire, grazing, and climate that can influence both PNV and current vegetation. Refining and updating ecosystem-specific species lists may benefit existing forest site classifications and could be planned for when new site classifications are developed, especially with changing climates.