Species-specific PM retention by urban tree bark is associated with functional traits under controlled exposure.

Urban tree bark plays a critical yet underappreciated role in mitigating atmospheric particulate matter (PM) pollution, particularly during leaf-off periods when foliage is absent. However, the relative contribution of bark versus other plant organs and the functional traits governing interspecific variation in bark PM retention remain poorly understood. In this study, we implemented a trait-based modeling framework utilizing modified open-top chambers to minimize external meteorological disturbances and reduce environmental variability. Results showed that: (1) at the unit-area scale, bark retained significantly more PM than leaves or branches across 11 of the 12 core calibration species (p < 0.05), with Cinnamomum camphora exhibiting the highest capture efficiency (364.5 ± 9.3 μg cm-2), whereas Juniperus formosana exhibited the lowest (112.1 ± 13.6 μg cm-2); (2) upscaling to the whole-plant scale revealed that aggregated bark PM2.5-10 retention was 1.40 times that of leaves, while PM2.5 retention between bark and foliage was comparable, indicating a distinct size-selective filtration advantage of bark; (3) Spearman correlation analysis and mathematical modeling indicated that interspecific bark PM variation was associated with specific micro-topographic traits (p < 0.05), with lenticel density, surface roughness, and groove prominence showing consistent positive associations with bark PM retention, and the resulting predictive models showed acceptable consistency with the limited independent validation data; and (4) an exploratory application to twenty-six dominant urban greening species in Hangzhou suggested a 4.83-fold interspecific variation in potential bark PM retention under controlled exposure. These findings suggest that neglecting bark retention may lead to underestimation of tree air purification capacity, providing a basis for incorporating bark-specific parameters into dynamic vegetation models.