Pollen allergies afflict over 35 million Americans, triggering both allergic rhinitis (hay fever) and asthma attacks. Despite the considerable public health consequences of allergenic pollen exposure, little is known about how pollen concentrations vary within cities. This is in part because there are few municipal-scale estimates of pollen production. This information could serve as the basis for pollen dispersion models, which are required for individual level estimates of allergenic pollen exposures.
In this study, we used a large street tree survey, LiDAR, high resolution imagery, and a geographic object based image analysis (GEOBIA) approach to classify several tree taxa across Detroit, MI, USA. Pollen production was estimated using allometric equations that we developed for several tree taxa, including Acer, Gleditsia, Platanus, and Quercus. The results of these two approaches were combined to provide spatially explicit estimates of pollen production across 344 km2.
Results/Conclusions
Of the ~200,000 street trees, the taxa with the highest relative basal area were Acer (47%), Ulmus (10%), Gleditsia (10%), Platanus (9%) and Quercus (6%). Using a random forest approach, we achieved overall accuracy rates of ~80%, with especially high accuracy rates for certain taxa that displayed distinct spectral characteristics in particular times of year. For example, Norway maple has a distinctive flower and foliage color in early spring and could be identified with 95% accuracy. Pollen production was well predicted by tree size and by canopy area and the R2 for each taxon ranged from 0.74 – 0.98. Total pollen production across the city varied considerably among taxa and spatial patterns in pollen production were idiosyncratic.
Combining remote sensing data and allometric equations offers a new opportunity for creating comprehensive spatially explicit predictions of pollen production across large areas. This approach provides a missing link required for developing high spatial resolution mechanistic models of allergenic pollen concentrations. These techniques could also be used to quantify floral resource availability for pollinators.