An urbanized version of MM5 (uMM5) was used at a 500 m horizontal grid-resolution to study effects on morning urban mixing depths and near roof-top stability from use of extensive green roofs in Mexico City, which is characterized by large Bowen ratios and high building storages. The model uses urban-morphology data, while building hydrothermal uMM5 input parameters were obtained from measurements over green and nearby conventional roofs. Evaluation of uMM5 predicted values against rooftop and planetary boundary layer (PBL) observations from extensive field measurement campaigns showed that the model performed reasonably well. Additional simulations were carried assuming that the roofs in entire urban neighborhoods were greened. Predicted mixing depths from these simulations, along with observed air pollution concentrations, were then used in a simple box model to evaluate potential green roof impacts on concentration. Results showed that green roofs produced an early morning (7–10 LST) cooling of up to 1.2 °C at rooftop levels, which reduced mixing depths during that period. Effects were greater on a day with weak synoptic forcing that on one 48 h later with strong synoptic forcing. The mixing-depth decreases produced increased box-model pollutant concentrations. While the green roofs did not elevate the observed concentrations of CO, SO2, and NO2 above World Health Organization (WHO) health standards, they did increase PM10, values (which were already above its standard) by as much as 8% from 7 to 9 LST, when local populations are normally exposed to peak concentrations. This study has applications in the analyses of building energy efficiency.