From monitoring to modelling: harnessing pollen data for adapting to exposure risks in a changing climate

Increasing trends in allergic respiratory diseases may be linked to climate change affecting pollen levels and seasonality. Pollen data from monitoring stations are usually incomplete, and limited research has assessed the accuracy of imputation methods in aerobiological data. Moreover, monitoring networks have limited spatial coverage. Increasing the temporal and spatial resolution of the information on pollen levels will be useful from a public health perspective. In particular, implementing accurate pollen forecasting models can help patients with respiratory allergies manage their symptoms more effectively. The aim of the thesis would cover these gaps of knowledge, (i) testing a new data-driven method to impute missing data in aerobiological datasets, (ii) assessing temporal trends of pollen exposure indicators in the Veneto Region (northern Italy) in the last two decades, and (iii) implementing a novel pollen modelling system that can be used to provide high-resolution forecasts of pollen concentrations in the Veneto Region. Pollen data for the Veneto Region (and the surrounding areas) were downloaded freely from the POLLnet database or directly provided by ARPAV. The “AeRobiology” R package was used to manage pollen data and calculate pollen seasons and seasonal indexes. Different methodologies were applied to achieve the three objectives. For (i), a simulation study was conducted to compare a new data-driven method (Gappy Singular Value Decomposition, GSVD) to a statistical one (moving mean). The gaps were randomly generated in an annual pollen dataset for 2 pollen types and 2 monitoring stations, imputed with the two methods, and lastly, the imputation accuracy was assessed. For (ii), a trend analysis was conducted to estimate temporal changes in seasonal pollen indexes between 2001 and 2022 (9 pollen types, 20 monitoring stations). The non-parametric Theil-Sen median slope method was used to examine the trends in the whole region and by climatic areas. For (iii), a novel pollen modelling system was validated to simulate the dispersion, diffusion, and deposition processes of five pollen types over the Veneto Region, investigating the impact of vegetation cover maps resolution on predicted concentrations. The results of this thesis showed: (i) the GSVD method and the moving mean approach showed a similar effectiveness in imputing pollen data, and that the imputation accuracy was affected by temporal variability of pollen, (ii) for several pollen taxa, concentrations and the duration of pollen seasons have increased in the Veneto Region over the last twenty years, particularly in subcontinental areas, and (iii) the use of detailed vegetation maps as input to pollen modelling systems can improve the estimation of arboreal pollen concentrations, especially on complex surfaces (such as alpine zones), potentially increasing the quality of daily and seasonal forecasts. In conclusion, our findings contribute to knowledge in terms of monitoring, analysis and prediction of aerobiological data. They also highlight the need for further studies aimed at improving pollen data imputation techniques and to better understand and predict temporal and spatial variations in pollen. Our analyses have confirmed variations in pollen concentrations in the atmosphere over the past two decades, suggesting a substantial impact of climate change. Research in this field can lead to the development of new measures of prevention and adaptation to address future variations in pollen exposure due to climate change.