Impact of Long Range Pollen Transport to Allergenic Episodes Forecasting Possibilities

The above-outlined main mechanisms that control the plant behavior at a specific place lead to significantly different regional phenological calendars dependent on specifics of regional climate and vegetation. Some key features of these processes can be simulated using semi-empirical models, such as Thermal-Time, thus allowing the forecasting of pollen seasons using meteorological forecasts and local plant observations. The current section will discuss the process that challenges such regionalization and provides large-scale links between different parts of continents: a long-range transport of genetic and allergenic material released during the flowering.

A typical shape of observed pollen concentrations in air during the spring season is shown in Fig. 5.4. One can distinguish the main peak representing the local flowering season with well-seen start and end times, which is surrounded by smaller-scale increases of counts a few days before and after the main rise of concentrations. In some places these "tails" turn into two peaks comparable with the main one. These early and late blows of pollen often originate from longdistance transport. The evidence of such phenomenon is quickly growing and obtained at nearly all climatic zones (Corden et al. 2002; Latalowa et al. 2002; Hjelmroos 1992; Damialis et al. 2004; Rantio-Lehtimaki 1994). However, from the point of view of atmospheric science and air pollution, pollen could not be expected to be dispersed farther than a few tens of kilometres in the atmosphere: the grains are too big (tens of micrometres in diameter) and thus ought to be deposited much too fast to be a large-scale pollutant. According to atmospheric science, a so-called Junge size spectrum of aerosol with long air lifetime is confined between 0.1 and 1 |im.

A more detailed theoretical consideration (Sofiev et al. 2006) showed, however, that at least some types of pollen are susceptible for distribution with air masses for a distance of several hundreds or thousands kilometers, which is well sufficient for transport between climatic zones. These conclusions can be extended by suggesting that the allergenic material can be emitted from the pollen grain during its active period. There are also suggestions that anthropogenic pollution can escalate such process (Majd et al. 2004; Motta et al. 2006). As stated in the previous section, this material consists of organic particles with characteristic size close to the "classical" fine-particle range 0.10-1 |im, which are known to travel over large scales.

The four most-evident consequences of the long-range pollen transport are: (i) earlier start of pollen seasons in northern regions where the local flowering is late in comparison with (remote) neighbors (Fig. 5.5); (ii) a delay of the end of the season in central and southern regions where the local flowering ends early (Fig. 5.6); (iii) an increase of the actual concentrations during the main season, especially in regions with weak or moderate flowering intensity (this is not observable directly, and requires detailed modelling to highlight); (iv) a fast (in a scale of days) transport of genetic material across the continents, climate and vegetation zones.

Fig. 5.5 An example of a strong early-spring pollen episode in central Finland caused by the long-range transport from central Europe in 1999 (Sofiev et al. 2006)


3500 n 300025002000150010005000 14

Leaf bud break Turku, 4. May

Leaf bud break Oulu, 20. May

Leaf bud break Turku, 4. May

Leaf bud break Oulu, 20. May m I I I I h 8.5. 16.5.

Fig. 5.6 An example of north-to-south transport of pollen from Finnish Lapland to Denmark after the end of the season in the south. Left panel: observed by European Aeroallergen Network sites (courtesy of S. Jaeger), right-hand panel: probability (Area Of Risk) and pollen concentrations modelled by SILAM model,

Fig. 5.6 An example of north-to-south transport of pollen from Finnish Lapland to Denmark after the end of the season in the south. Left panel: observed by European Aeroallergen Network sites (courtesy of S. Jaeger), right-hand panel: probability (Area Of Risk) and pollen concentrations modelled by SILAM model,

An analysis of the phenomenon and its forecasting requires the involvement of atmospheric dispersion models in combination with tools for the prediction of pollen emissions over large territories, such as continents. A principal scheme of such a system is shown in Fig. 5.7.

From the point of view of atmospheric science, pollen is a specific aerosol and simulation of its transport does not pose principal difficulties, providing that existing dispersion models are applicable. These models assume that the transported species follow the path of surrounding air, including small turbulent eddies, and do not pose any feedback to the atmospheric flows. As shown by Sofiev et al. (2006), these assumptions are fulfilled for birch pollen (a comparatively small grain), but are not necessarily correct for larger particles because the neglected terms"/>
Fig. 5.7 A scheme of analysis and forecasting system for evaluation of pollen long-range transport (being developed in Finland,

in the transport and deposition equations might no longer be small. The inertial penetration of such grains through low-turbulence layer can be taken into account via a classical parameterization described by Seinfeld and Pandis (2006), chapter 19.4.2. However, it would improve only the estimation of dry deposition, still assuming that the grain follows the main air flows far from the surface. Additional complexity can be introduced by including the aerodynamic shape of some grains. The presence of wings or strong asymmetry may lead to detachment of the grain from surrounding air. For such pollen, the existing dispersion models will manifest large errors in deposition intensity and, possibly, miscompute the diffusion. Therefore, the usage of existing models for each specific taxon requires the explicit applicability checking.

It is known that pollen grains loose and gain water depending on air humidity and temperature, but the corresponding processes are quite poorly known and so far none of working systems takes these into account.

The most important and the most difficult challenge, however, is the construction of the pollen emission model. As seen from Fig. 5.5 and also pointed by e.g. Damialis (2004), Rantio-Lehtimaki (1994), Sofiev et al. (2006) and Siljamo et al. (2004), the long-range transport episode usually lasts for a couple of days and entirely depends on meteorological conditions. Flowering start time varies between the neighboring climatic zones for just a few days. Therefore, in order to catch the long-range transport episode the emission model has to forecast the emission timing with accuracy of 1-2 days - homogeneous over large areas. Unfortunately, even local-scale phenological models have the standard deviations of predicted flowering time about 4-5 days and therefore cannot serve as a sole source of emission information for deterministic pollen forecasts.

There can be two ways to cope with the problem of high uncertainty of pollen emission: (i) to use probabilistic forecasts via ensemble simulations or straightforwardly computed probability distributions, and (ii) to utilize additional information via data assimilation mechanisms to adjust the phenological models "on-the-fly".

Three most-evident data sources for the assimilation are the near-real-time aerobiological and phenological observations and satellite images.

Probabilistic large-scale simulations and forecasting have been used in Finland for several years with positive outcome (see, Sofiev et al. 2006). The model estimated both absolute pollen concentrations over Europe and the territories affected by each of six major birch forest areas (probability distributions). The concentration estimate relies on an European-scale phenological emission model, while the second one does not use it at all - just a map of birch forests. The combination of these two characteristics appeared to be very useful for the allergy forecasting as it (i) shows a risk of getting pollen from major remote sources, (ii) indicates the forests affecting the receptor region, and (iii) estimates an absolute concentration of pollen grains in air using some emission model with known formulations and accuracy. The next level of sophistication of this methodology is to build a probabilistic emission model that would include the above-mentioned uncertainty as a feature of the flux probability distribution.

Adjustment of the emission module using independent observations are proved to be useful but also has certain limitations. In particular, near-real-time phenological observations currently do not exist. Many countries have very dense networks (Fig. 5.8) but the data collection is manual, as well as the processing and compilation

Fig. 5.8 Combined map of 15 national phenological networks in Europe. Colour denotes the number of observed years by each station

of national databases. There is no European-wide regular phenological database; the one presented in Fig. 5.8 was compiled for birch at Finnish Meteorological Institute and is owed to courtesy of the national phenological networks listed in Table 5.1.

Pollen counts are the only data partly available with a delay of less than 2 days ( However, such observations do not provide information about the pollen source and thus cannot distinguish between the local and long-range transported pollen. Therefore, this information can be used for two purposes: now-casting the pollen concentrations in the area surrounding the trap and forecasting the counts downwind. Size of the area where a specific pollen trap is representative varies strongly depending on local conditions and microclimate. Usually, however, it extends over a few tens of km - for daily pollen concentrations.

A promising set of information emerges from the growing satellite fleet. Several modern methodologies for evaluating the state of vegetation with very high spatial and temporal resolutions (250 m and 15 min, respectively) are expected or already

Table 5.1 National phenological networks contributed to the database presented in Fig. 5.8


No. of stations

Years available at least for some stations


Data handler/ provider





Minsk aerobiology



Czech Republic


1955-2004, few

Betula pendula

Czech Hydrometeoro-


logical Inst.









Betula pen

Finnish Forest Research


dula, Betula

Institute METLA





Betula pendula






University of Tartu





University of Tartu




Betula pubescens

Biofork Nord




Betula pendula

Institute of Meteorology


and water manage-






University of Tartu,

Moscow State





Betula pendula

Slovak Hydromet. Inst.




Betula pendula,

University of Vigo

Betula alba




Betula pendula






Moscow State


United Kingdom



Betula pendula

UK Phenological




available for users (,, These products allow various approaches for evaluation of phenological stages, such as NDVI (Normalized-Difference Vegetation Index), micro-channel based evaluation of leaf area indices, etc. (Manninen et al. 2006; H0gda et al. 2002). However, all these tools have a significant limitation: the flowering itself is not visible from the space. The observable set of parameters describes the status of leaves, their size, area, color, etc. For some species the moment of e.g. bud burst correlates well with the flowering (with a possible shift for a few days), while for the others this dependence is either weak or overshadowed by other species opening their leaves earlier (the satellite monitors evidently see the unfolding of leaves of the earliest taxa). Therefore, the satellite-based correction of predicted flowering timing, being otherwise highly accurate and timely, is available not for all plants and not at all for herbaceous plants.

To conclude the list of challenges on the way of pollen large-scale forecasting, it is worth recalling that absolute amount of pollen released during flowering is largely determined by the previous-year conditions during the active vegetation period and, to a less extent, by winter conditions. This is an entirely new feature for atmospheric dispersion modelling: usually such models have to "remember" the situation for just a few hours to the past. Pollen forecasting model has to keep (albeit, in an aggregated form) the information over the past year.

The above-outlined complexity resulted in existence of only few modelling systems approaching the pollen forecast problem. Two of them working at local scale are mentioned above - the A.S.T.H.M.A system in S. Europe and University of Tulsa in the USA. To our knowledge, the only system approaching the European scale exists in Finland (see example of its results in Fig. 5.6). Similar models are being developed in Denmark, France, the US and other countries.

The usefulness of such systems for the purpose of the human adaptation to the pollen seasons strongly depends on the forecast horizon: a warning made few hours before the episode may be too late for preparatory actions. Based on existing statistics of application meteorological models, the high-quality detailed forecast is usually possible for a period of about 2 days, while the main trends can be computed almost a week in advance. Providing that the pollen emission is predicted accurately enough, the allergenic forecasting can be made for a similar period.

Was this article helpful?

0 0
Allergy Relief

Allergy Relief

Have you ever wondered how to fight allergies? Here are some useful information on allergies and how to relief its effects. This is the most comprehensive report on allergy relief you will ever read.

Get My Free Ebook

Post a comment