EGU25-17983, updated on 15 Mar 2025
https://doi.org/10.5194/egusphere-egu25-17983
EGU General Assembly 2025
© Author(s) 2025. This work is distributed under
the Creative Commons Attribution 4.0 License.
Bridging Automatic and Manual Pollen Monitoring: A Path Towards Homogenized Long-Term Time Series
Mária Lbadaoui-Darvas1, Regula Gehrig1, Ingrida Sauliene2, Laura Sukiene2, and Jose Oteros3
Mária Lbadaoui-Darvas et al.
  • 1Federal Institute of Meteorology and Climatology, MeteoSwiss, Zurich, Switzerland
  • 2Vilnius University, Vilnius, Lithuania
  • 3University of Cordoba, Cordoba, Spain

The measurement of pollen concentrations began in the 1960s, initiated by medical doctors seeking to address allergenic concerns. Historically, pollen monitoring networks relied on the manual identification of daily samples collected on tapes in Hirst-type traps using optical microscopy. This approach persisted until a recent paradigm shift towards automatic, in situ monitoring solutions.

The new generation of automatic measurement systems employs advanced techniques, such as automated microscopy (e.g., BAA 500) or digital holography combined with fluorescence measurements (Swisens Poleno). These are augmented by AI-based identification algorithms, enabling real-time pollen monitoring with temporal resolutions of at least one hour. However, manual and automatic measurement systems exhibit different sampling efficiencies for various pollen species, stemming from disparities in instrumentation characteristics such as flow rates, identification and data processing methods, and temporal resolution. This technological transition has introduced a discontinuity in the historical pollen concentration time series, which are crucial for forecasting models.

In this study, we analyze and compare manual and automatic pollen concentration time series from the Swiss national pollen monitoring network for four major allergenic pollen types: alder, birch, grasses, and oak—species significant across different regions of Europe. Data from 2022 to 2024, collected simultaneously using both methods in rural and urban settings in Switzerland, are evaluated. Machine learning regression algorithms (Random Forest, GRNN) are leveraged to establish a transfer function that relates automatic and manual pollen data. The model incorporates environmental variables likely to influence pollen concentrations, including temperature, wind velocity, elevation, and particulate matter (PM) concentrations.

How to cite: Lbadaoui-Darvas, M., Gehrig, R., Sauliene, I., Sukiene, L., and Oteros, J.: Bridging Automatic and Manual Pollen Monitoring: A Path Towards Homogenized Long-Term Time Series, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17983, https://doi.org/10.5194/egusphere-egu25-17983, 2025.

Comments on the supplementary material

AC: Author Comment | CC: Community Comment | Report abuse

supplementary materials version 1 – uploaded on 01 May 2025, no comments