Initial comparison of pollen counting methods using precipitation and ambient air samples and automated artificial intelligence to support national monitoring objectives

Abstract

Given the endemic nature of pollen throughout the environment, the impact upon human health, and the need for more extensive and better measurements of pollen in the USA, a preliminary project within the National Atmospheric Deposition Program’s (NADP) National Trends Network (NTN) was developed. Pollen was measured in ambient air by several methods and in precipitation wet deposition samples at three monitoring sites in the NTN. A method for counting pollen on filters was developed and provided pollen counts for NADP atmospheric wet-deposition samples and high-volume ambient air samplers (HVAS) for comparison with co-located traditional National Allergy Bureau microscopy samples and a commercially available pollen sensor (PS) counting method during the 2021 pollen season. The goals of this project were to test the potential of available air-monitoring infrastructures to obtain improved spatial measurements of aeroallergens, compare pollen counting results from the various methods, and to determine the suitability of using wet deposition samples for pollen collection. The onset and senescence of pollen seasons for general categories of genera compared favorably for each method at each site, indicating that pollen monitoring using wet-deposition and ambient air sampling filters could provide useful information to inform scientific studies, but not likely for public health objectives. Pollen counts were log transformed for Pearson product moment correlation. Tree pollen counts were correlated at all sites for daily PS data and traditional counting data (R = 0.69–0.84), but statistical correlations between methods for grass and weed pollen were weak (0.40 < R < 0.60) or considered not correlated (R < 0.40). Total pollen counts in NADP precipitation samples were correlated with traditional and PS counts at only one of three sites. Pollen counts for the weekly HVAS filter samples were correlated with PS counts for trees (R = 0.62) and with NAB counts for trees (R = 0.68) and weeds (R = 0.72). Correlations in the data between methods suggest that, given further methods development, a variety of techniques could be integrated to expand and enhance existing pollen monitoring networks. Improved ambient air and atmospheric deposition sampling methods specifically targeted for pollen capture and analysis could support the collection of accurate and efficient meaningful aeroallergen data from existing atmospheric monitoring networks.