The analysis of chemical elements in the barks of trees is an alternative procedure to access spatial heterogeneity of traffic related air pollution. However, the role of tree species in the characterization of the variability of airborne pollution is poorly known. We present an observational study conducted in São Paulo, Brazil, based on the analysis of 498 trees from three common species: Tipuana tipu, Poincianella pluviosa, and Ligustrum sp.. We considered ANCOVA models to compare the concentrations of Al, Fe, Zn, Cu, Mn, Ba, and S in the bark (periderm) of trees located close to streets with different levels of traffic intensity controlling for the extension of nearby green areas. The expected trend of increasing elemental concentration in the bark of trees located near streets with greater traffic intensity or close to smaller green areas was only fully evidenced by T. tipu. For instance, the concentrations of Zn, Fe, Al, and Ba increase by 200, 350, 230, and 280% respectively, for trees of this species located near arterial streets when compared to those observed near local streets. On the other hand, the concentrations of Zn, Fe, Al, and Ba are reduced by 41, 45, 50, and 30%, respectively, for trees located near green areas. For P. pluviosa, the capacity to suggest an association between the tree bark concentration of chemical elements with increasing levels of air pollution and presence of green areas was only fully observed for Zn and Cu. For Ligustrum sp., weaker and sometimes non-expected associations between bark concentrations of the chemical elements and either street classification or green area extension were observed. Our results indicate that the choice of species is a key element in the use of tree barks as a biomonitoring tool in urban landscapes. Species like T. tipu, with rough and highly porous bark, are the most appropriate for such purpose.