| 30 Mar 2026
Real-Time Pollen Dynamics and Automated Detection: Novel insights from Wrocław (Poland) 2024–2025
Abstract. Accurate monitoring and forecasting of airborne pollen are essential for public health and allergy management. This study evaluated a neural network model for real-time pollen monitoring using locally collected data in Wrocław, Poland (2024–2025) with the Swisens Poleno Jupiter detector. The retrained model, based on local data, outperformed the reference model trained on Swiss datasets and validated against a Hirst-type pollen trap. The coefficient of determination (R²) remained high (~0.8) especially for Alnus, Betula and Quercus, while the root mean square error (RMSE) was lower, particularly at low and medium concentrations, showing improved sensitivity, real-time detection and better representation of seasonal and diurnal dynamics. Hourly analyses revealed distinct taxon-specific diurnal patterns in pollen release. Temperature and relative humidity were the main drivers of variability, while wind speed influenced all taxa except Pinus. Hourly pollen concentrations were positively correlated with planetary boundary layer height, especially for Betula and Alnus, highlighting the role of atmospheric mixing in pollen dispersion. Wind direction, particularly from southern and southeastern sectors, modulated local transport, reflecting land cover effects. Correlations with meteorological variables varied by month and flowering stage. Validation against Hirst-type data confirmed that the locally retrained model accurately captures taxon-specific pollen dynamics, demonstrating its effectiveness for real-time allergen monitoring and improving the reliability of allergy risk assessments.
The manuscript describes the retraining of an automatic pollen identification algorithm for use with an airflow cytometer in a new environment. This allows in turn the authors to present an interesting case study of real-time pollen dynamics in correlation with weather parameters at the site in question.
While the study is of interest to the community, two critical points need to be addressed before it can be considered for publication:
1) No code or training dataset is made directly available, which significantly hinders the reproducibility of the study and the reuse of the model for operational or research purposes. This is particularly regrettable given that dedicated portals would allow this with minimal effort. For example, the Sylva/AutoPollen data portal or the Zenodo repository could be used for the training datasets, and GitHub could be used for the code. The issue is exacerbated by the lack of information on model training and architecture.
2) The benchmark model (“old model”) shows reasonable correlations with reference measurements, but there are massive scaling issues (Figure 3). This could be related to issues when applying scaling factors used for converting raw events to concentrations, as different versions of the Swisens Poleno software have been known to circulate. I strongly recommend that the authors double-check how scaling is applied when using the benchmark model, as this has a dramatic effect on the metrics for exposure class determination (leading to the old model systematically overestimating classes), the RMSE, and consequently the conclusions of the paper.
In addition to those main points, here are my other comments and suggestions in the order as they appear in the manuscript:
Line 71: “advanced instrument” I suggest removing “advanced” as this adds no information per se.
Line 75: “detections” remove “s”
Line 76: remove “advanced”
Line 81-82: Which “reference pollen database” are the authors referring to? How can this database be accessed?
Line 92: “These tree species” If I understand correctly you work at genus level throughout the paper. I thus suggest to use “genera” instead of “species”.
Line 109-111: Maybe compare hourly correlations you obtain with the results from Chappuis et al. 2020, Aerobiologia for a different automatic pollen monitor.
Line 119: “20 meters” what about surrounding buildings height /height of the urban layer if any.
Line 132: “airborne cytometer” to be replaced by “airflow cytometer”
Line 157-158: Why were Poaceae, particularly important for respiratory allergies disregarded in the study?
Line 221: How was the size of the buffer determined ?
Line 265: “Cruozy” -> “Crouzy”
Line 324: “Polleno” takes one “l”, check throughout the manuscript as this appears one more time on line 413.
Figure 3: I suggest to use a capital letter in front of each genus name
Line 340, 348 and 352: the massive RMSE differences might be related to improper scaling.
Figure 5, 7 and the S1 to S5 should be made self-contained by the caption for the reader to understand all the symbols used and different metrics represented.
Line 501-510: the case of rainfall could be discussed in more details as the result is a bit counter intuitive (rainfall is expected to scavenge airborne pollen).
Table 2: the results on PBL are interesting but again counter intuitive, how could that be explained? One might expect a compression effect.
Line 544 “Dissucion” typo
Whole discussion: correlation vs. causality issues, for example “crucial” Line 642, “stronger effect” Line 658, “influential” Line 660, “responsive”
Line 674 or “shaping “ Line 690”.
Line 717: Pine pollen might not be relevant for allergies.