SYLVA, A SYstem for reaL-time obserVation of Aeroallergens

SYLVA (A SYstem for reaL-time obserVation of Aeroallergens) started at the first of januari 2023 and will run until december 2026. It builds upon the experience from the AUTOPOLLEN project (2018-2022). A few questions for Mikhail Sovief, coördinator of SYLVA:

  1. One of the SYLVA goals is to increase frequency of the countings. Why is it important to increase the frequency?

For allergy, it is well known that reactions are much faster than one day. Peaks, even very short, are more relevant than lows. There is also a possibility that the current thresholds need to be reviewed: they are based on daily levels while provocation chambers make tests at much shorter times. The daily resolution was out of necessity, which may be no longer valid. All-in-all, we might be much more accurate in relating the pollen concentrations and allergy symptoms if dive into hourly resolution.

For forestry/agriculture, it is deposition of pathogens rather than concentration that important – and it needs accurate colocation of rains and spore plumes. This has to happen at high resolution, tens of minutes max.


  1. Another goal is to make te data available faster, why is it important to speed up the availability of the data?

The week-old data are predominantly of historical value, they do not help much in daily life. For decades, the struggle of aerobiologists was to get the data as soon as possible – including daily change of tapes with immediate counting during the season – and still the minimum was about 2 days of a delay. Such lag dramatically reduces the use of the data for daily procedures in hospitals, agriculture and forestry companies – the information is needed in real time or, in most cases, as a forecast. What concerns forecasts, model can assimilate data of at most one day old, preferably, not more than 3 hours old. Longer delay sharply reduces the effect of assimilation.


  1. Coverage, why is it important to have a higher resolution of measuring locations?

It depends. The topic is much more complicated than just coverage. Every specific user requires own distribution of observations. Allergologists would argue for hospital roofs, farmers – for fields, forestry – for their monitoring sites, climatologists – for long-term observation sites, etc. There will be never funding to put these devices to every hospital roof, farm field, and forest monitoring site. Even if we take only hospitals, the current de-facto target of EAN, the existing ~600 locations are nowhere close to having a trap at each hospital roof. In our just-submitted paper on the aerobiological network design we argue for a paradigm change: comparatively homogeneous distribution of limited number of automatic sites, whose data are combined with models thus getting realistic analysis and accurate forecasts for the whole Europe.


  1. Availability, What could be improved compared to the current availability of pollendata?

Biological composition of the atmosphere must be public information, just as chemical composition (a.c.a. air quality) is. Today’s situation originated from history of EAN but it is time to change. The automatic devices are too expensive for small groups, too demanding in terms of IT, storage and communications. They will have to be established and maintained by big institutions, such as meteorological services or environmental institutes, in cooperation with aerobiological teams. They will rely on public money just as AQ and weather observations do. And, what has been paid by the tax payers must be publicly available. This is an imperative in SYLVA.


  1. Are there, besides Hund and Swisens, other Europeaan companies, universities or start-ups developing automatic counting systems?

There are at least 20 different initiatives and even more talks about, in Europe and the world. The situation is extremely dynamic. Today, these two companies produced good machines, attracted users, engaged and actively supported them, proved open, reliable and agile in helping to overcome problems inevitable at the start of so complicated work. Please note that machine itself is not enough. Someone has to develop recognition algorithm to interpret the raw data, arrange storage for them, IT infrastructure to handle this machinery, and train the devices to recognize bioaerosols from all over Europe. Not any single company can do all that alone – but a company working together with its users can. Hund Wetzlar and Swisens succeeded this very task, which, together with good machines, secured their currently leading position. More about devices and their relative performance can be obtained here:

Towards European automatic bioaerosol monitoring: Comparison of 9 automatic pollen observational instruments with classic Hirst-type traps, Science of The Total Environment, 866, 161220, https://doi.org/10.1016/j.scitotenv.2022.161220, 2023.


  1. Open source, will the data be available for everyone worldwide and in what form?

The SYLVA infrastructure will be directly linked to the international EBAS database, so the standard follows that very system: AMES-1001, or netcdf. The data will be automatically uploaded to EBAS in real time. Open availability of the information is the must: what is funded by tax money must be publicly open.


  1. Can you give an example of how the open-source principle, applied by the European environmental observing systems, resulted in beter quality information for European citizens?

Let’s differentiate the open-source and open-data principles.

Open-source means that the methodology, algorithms, interfaces, etc, are publicly available, e.g., via github. An example is SILAM or WRF models – anyone can get their code by just going to the corresponding Web sites. The same will refer to the Hund and Swisens recognition algorithms: Swisens opened its algorithm for the users a while ago, and both companies pledged to maintain new developments open-source. Such approach is particularly valuable as it creates a community of specialists, who can improve the technology further. This is how several existing versions of the Swisens algorithm were created – and this is the guarantee that it will be getting better.

The open-data principle means that the results of the model or observations of the device are publicly available. The key added value is then for the data users and for developers of downstream services. For instance, SILAM follows open-data principle, and its forecasts can be freely downloaded from the model Web site. We know that several small teams and companies use these data for derived products, and helping them wherever needed. For monitoring devices, the open-data principle has to be decided by the device owners – and in SYLVA such a pledge is a cornerstone of the operations of its partners.


  1. Will the open source data exist of the specific pollencountings by location or is it already combined with weatherdata and vegetationmaps and turned into a complete validated pollenforecastmodel like Silam?

SYLVA aims at development of the baseline infrastructure for the future networks; it does not build the network. There is a lot of homework to be done before we can seriously consider automatic bioaerosol monitoring network in Europe. However, SYLVA partners also represent the major efforts currently undertaken in Europe towards construction of such network. Work goes on to make these data available.

SYLVA builds an open system suitable, in principle, for connecting various devices, not only those from Swisens and Hund. For instance, we also do some work with Plair Rapid-E+, albeit at a much smaller scale. Apart from automatic monitors, we are pushing forwards the ground-breaking technique based on DNA analysis and metagenomics of atmospheric aerosols.

The idea of the project is to construct an integrated system, not just a few unconnected pieces of hard- and software. This system includes automatic monitors, their IT infrastructure and processing algorithms, numerical models, and DNA metagenomics. These pieces are glued together by machine learning tools, which also connect them to the existing manual instruments, thus securing continuation of the existing time series all over Europe. This system, however, will not be closed: any its piece can be used and reused by others. We will only demonstrate the added value of the integrated approach.


  1. How can you validate these high-tech countingmachines which use AI? What is the reference?

Validation is multi-step. For the first time in aerobiology, detailed metrological examination and calibration of the devices is planned. SYLVA collaborates with several national projects to secure funding for this vital but expensive exercise. On the next level, recognition algorithms are confronted with “problems with known solutions”: Swisens Poleno can be supplied with pollen or spores of known types, and then requested to classify each particle it registered. For Hund POMO, its high-quality images can be used for explicit validation of the algorithm conclusion. And metagenomic analysis will give an ultimate answer on what is in the air. There will be, of course, comparison with Hirst traps, but we keep in mind that, providing the above homework is done properly, quality of automatic data may turn higher than those obtained from manual device, at least for some pollen types.


  1. What are the European hotspots related to automatic counting?

SYLVA will work in three climatic regions: nordic (Finnish Lapland), high-altitude (Alps), and Southern conditions (southern Spain, Serbia, and Italy). One of our tasks is to quantify performance of the devices in extreme conditions, thus making sure that they can be applied anywhere in Europe.


  1. Is a budget of 3 mln, spread over 4 years, sufficiënt for reaching the goals?

The goals are extremely ambitious but realistic. First of all, the project budget is 3.5 MEuro: Swiss government provided complementary funding for MeteoSwiss and Swisens. Secondly, we do not build the network but rather use devices already possessed by the partners, with some upgrades where necessary. Thirdly, we have several years of experience with these devices, first algorithms have been built, first adaptations to extremes done, first interfaces and databases existing, etc. Fourthly, we actively collaborate with other projects in the area, both national and European. All these factors help us to achieve the goals.


  1. Which sectors, besides hayfeverpatiënts, will benefit from this infrastructure?

SYLVA is now organizing a wide Stakeholder Forum (SF), which first meeting will take place in autumn 2023. It will include, first of all, aerobiologists, who will be the most-direct beneficiaries of the infrastructure. But not only them. We would like to uptake other bioaerosols, such as spores and, hopefully, some bacteria. Therefore, among the users of the future network data, SF will include allergologists, agriculture and forestry representatives, biodiversity and climate scientists, etc.


  1. How can you ensure the long-term sustainability of bioaerosol technology and infrastructure across Europe?

Infrastructure of such scale can run only on the basis of sustainable financial support, similar to that currently existing for AQ and climate monitoring. Our goal is to secure such funding. European Union has several infrastructure programs, such as ACTRIS, who have already been contacted.


Links:

SYLAM project website >>

Cordis website >>

Mikhail Sovief works at FMI, the finnisch meteorological institute >>

www.autopollen.net >>