Lessons from automatic airborne pollen monitoring by holography at three Australian sites

Abstract Automatic airborne pollen monitoring offers potential to enable real-time, sustainable environmental health information to assist with management of allergic respiratory disease. Automatic systems, however, have not been widely deployed in the Southern Hemisphere, where there is considerable biodiversity. An automatic holography-based device combined with a classification algorithm was trialled in Australia for 15 months over the main pollen seasons in subtropical Brisbane, and temperate Canberra and Sydney. Common airborne pollen taxa were monitored in parallel using Hirst-type volumetric traps, with manual observations serving as a reference. In Brisbane, the monitoring campaign coincided with the most extreme grass (Poaceae) pollen season ever observed. The automatic system assigned daily Poaceae pollen concentrations with high accuracy (ρ = 0.84, p < 0.001; Mann–Whitney U, p = 0.14). In Canberra, Poaceae and Betula were detected with a moderate to high correlation to manual observations (ρ = 0.67, p < 0.001 and ρ = 0.77, p < 0.001) respectively. Betula and Cupressaceae were, however, mistakenly assigned to Poaceae, and Betula was also misclassified as other Betulaceae taxa: Corylus and Carpinus . Pinaceae pollen was well recognized by the automatic system at this site (ρ = 0.59, p < 0.001; Mann–Whitney U, p = 0.35). The automatic system performed poorly in Sydney, where daily pollen concentrations were low, with high misassignment. While automatic pollen monitoring reduced the need for site visits and expedited daily forecasts, inconsistencies in classification and concentration levels between methods highlight the need to calibrate automatic systems to previously trained and locally relevant, regularly monitored pollen and their levels.