Local and Regional Grass Pollen Distribution Identified using HYSPLIT and Statistical Modelling Approaches

Background:

Amongst all aeroallergens grass pollen has the highest rate of allergenic sensitivity in the human population (Akdis and Agache 2014), with fundamental implications for productivity (Crystal-Peters et al. 2000) and quality of life (Šaulienė et al. 2016) for the sufferers. To avoid these issues during the grass pollen season sufferers are recommended to use OTC-drugs (Wallace et al. 2008), immunotherapy (Jutel et al. 2005) or to avoid going out on days with high atmospheric grass pollen levels. However, wind can transport pollen over long distances under optimal conditions (de Weger et al. 2016). Potential grass pollen source areas are normally identified using remote sensing, ground-based surveys or a combination of both. The identification of relevant grass pollen source areas can help to develop dispersal models (Skjøth et al. 2013) and to model pollen exposure (Rodríguez-Rajo et al. 2010). In this study we have investigated potential influences and source areas of grass pollen collected in urban and rural areas using grass vegetation source maps and atmospheric transportation modelling.

Methods:

Grass pollen was sampled using Burkard Hirst-type pollen traps in 2018 and 2019 from two locations (one urban and one rural) between May and September in the Worcester area of the United Kingdom. These two locations are located 6.5 km apart. From these samples bi-hourly grass pollen concentrations were calculated. Local temperature and precipitation were measured in both locations using the same model of meteorological station. Grass vegetation source areas were isolated from the ‘CEH Land Cover Plus®: Crop’ dataset and replotted to a grid-resolution of 100m. Atmospheric transport was calculated using HYSPLIT backwards trajectories of 2h resolution using the ARL format GFS0p25 dataset obtained from NOAA. Source areas likely responsible for each bi-hourly grass pollen datapoint were calculated using the HYSPLIT trajectories. All variables per location were then modelled and analysed using Generalized Linear Mixed Model (GLMER) approaches to identify important influences and distances to source areas.

Results:

The GLMER model statistics highlighted differential responses based on the development of the location (urban vs rural environments). For the urban location 20-30km distance had a positive very significant effect on bi-hourly grass pollen concentrations while closer source areas had no significant effect. Source areas located 30-40km away had negative very significant effects. For the rural location both 2-10km and 20-30km distances had positive very significant effects on the bi-hourly grass pollen concentrations. Micro-scale source areas (0-2km) had no effects on either location. The models had high predictability of bi-hourly grass pollen concentrations, with the models having R2 values of about 53%.

Conclusions:

The modelling approaches showcase that high-resolution grass vegetation maps can distinguish grass pollen source areas within a larger region. The combination of atmospheric trajectory modelling and source maps in a modelling approach suggests that source areas further than 30km away are not likely to contribute grass pollen. At the same time, it also suggests that there are differences in the source area contributions based on the development of the location.