D3.438 - Advances in Automated Pollen Monitoring in Mediterranean Regions: New Developments from the SYLVA Project in Córdoba (Spain)

SYLVA European project aims to enhance and standardize bioaerosol measurement techniques across Europe by testing novel equipment, algorithms, and training methodologies under demanding environmental conditions. Southern Spain’s Mediterranean region, featured by high biodiversity and extreme weather conditions, presents substantial challenges for automated pollen monitoring within this project.

In 2024, three commercially available devices—Swisens Poleno Jupiter, BAA500, and PollenSense APS-330 (this last out of SYLVA)—are being evaluated from their default untrainedconfiguration at Córdoba city (Southern Spain). At the same time, a collection of 44 reference datasets were created for all devices, including the key mediterranean pollen taxa (e.g. Olea, Morus).

Daily and three-hourly automatic pollen counts are compared here to those obtained by Hirst-type samplers during the year 2024. Results revealed that the average ratio of daily pollen concentrations relative to the Hirst sampler remained below 1.2 for APS-330 and BAA500, whereas it exceeded 3 for Swisens Poleno Jupiter. In most of the dominant pollen taxa high correlations between automatic and human counted data were revealed, as in the case of Cupressus with R values of 0.94 (APS-330), 0.87 (BAA500), and 0.82 (Swisens Poleno Jupiter). These variations highlight the necessity of scaling factors to align counts across devices when using out-of-the-box classification algorithms.

Based on these findings, new algorithmic developments have been introduced for the BAA500, now covering 40 pollen categories with an overall accuracy of 96% (Mean Precision: 94.46%, Mean Recall: 95.38%, Mean F1-Score: 94.77%). The APS-330 device has been upgraded to APS-400, and a revised classification model will be available by spring 2025, aiming to further improve detection reliability. Similarly, the Swisens Poleno Jupiter system is operating its own enhanced algorithm for local Mediterranean taxa. These enhancements are expected to reduce device-specific biases, improve correlation with standard reference methods, and contribute to a more robust automated pollen monitoring infrastructure across Europe.