Urban greenery plays an important role in reducing air pollution, being one of the often-used, nature-based measures in sustainable and climate-resilient urban development. However, when modelling its effect on air pollution removal by dry deposition, coarse and time-limited data on vegetation properties are often included, disregarding the high spatial and temporal heterogeneity in urban forest canopies. Here, we present a detailed, physics-based approach for modelling particulate matter (PM10) and tropospheric ozone (O-3) removal by urban greenery on a small scale that eliminates these constraints. Our procedure combines a dense network of low-cost optical and electrochemical air pollution sensors, and a remote sensing method for greenery structure monitoring derived from Unmanned aerial systems (UAS) imagery processed by the Structure from Motion (SfM) algorithm. This approach enabled the quantification of species- and individual-specific air pollution removal rates by woody plants throughout the growing season, exploring the high spatial and temporal variability of modelled removal rates within an urban forest. The total PM10 and O-3 removal rates ranged from 7.6 g m(-2) (PM10) and 12.6 g m(-2) (O-3) for mature trees of Acer pseudoplatanus to 0.1 g m(-2) and 0.1 g m(-2) for newly planted tree saplings of Salix daphnoides. The present study demonstrates that UAS-SfM can detect differences in structures among and within canopies and by involving these characteristics, they can shift the modelling of air pollution removal towards a level of individual woody plants and beyond, enabling more realistic and accurate quantification of air pollution removal. Moreover, this approach can be similarly applied when modelling other ecosystem services provided by urban greenery.