Development and verification of a taxa-specific gridded pollen modelling system for the UK

doi:10.21203/rs.3.rs-4629279/v1

Download PDF

Research Article

Development and verification of a taxa-specific gridded pollen modelling system for the UK

https://doi.org/10.21203/rs.3.rs-4629279/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 26 Apr, 2025

Read the published version in Aerobiologia →

You are reading this latest preprint version

Allergic rhinitis, often caused by allergies from grass, tree or weed pollen, affects a large proportion of the UK population, and leads to significant costs to the National Health Service. The existing UK pollen forecast, produced manually, provides a single daily level for each of 16 regions. We present here an implementation of a pollen modelling capability within the Met Office Numerical Atmospheric-dispersion Modelling Environment (NAME) dispersion model. This will provide taxa-specific outputs at high temporal (hourly) and spatial (5 km) resolutions, which will eventually transform the level of detail in a future forecast system and therefore be of significantly greater use to the public and health professionals for managing pollen risks. Initial developments are for the three taxa which are the most allergenic across the UK population: birch, oak and grass. Pollen grain emission maps have been estimated using species distribution modelling methods. The timing of the pollen season is controlled within NAME by an accumulated temperature sum parametrisation, while pollen release is estimated with short term meteorological dependencies based on precipitation, wind speed and the vapour pressure deficit, along with a diurnal cycle. We show that, when run in hindcast mode, NAME performance (verified against pollen observations) is comparable with the Copernicus Atmosphere Monitoring Service ensemble median prediction for birch and grass. Evaluation of NAME for simulating the UK Daily Pollen Index shows an improved correlation coefficient compared to the existing manual forecast.

In the UK around 26% of adults are affected by allergic rhinitis, commonly known as hay fever (Bauchau & Durham, 2004). This leads to significant costs, for example allergic diseases (including allergic rhinitis, asthma, and eczema) in the UK result in a direct cost of over £1 billion per annum to the NHS (Gupta et al., 2004). A common cause of these allergies is from pollen released from grasses, trees and weeds. Advice for treatment of allergic symptoms include limiting exposure to pollen, alongside medication which can be more effective if given in advance (Hossenbaccus et al., 2020). Therefore, knowledge of forecasted levels of pollen can be key to reducing allergic symptoms.

Pollen forecasting is dependent on multiple components, each with different temporal and spatial scales involved. The first of these is the location of the pollen producing taxa, which is dependent on quantities such as climate and land use (Kurganskiy et al., 2020). Secondly, estimates of the timing of pollen season phenology, including the start date, peak and length of season. Within the season, short-term daily variations in pollen release can be attributed to current meteorology, while after release dispersion through the atmosphere is important for carrying the pollen away from the flowers to people impacted by allergies (Scheifinger et al., 2013). Finally, the magnitude of the pollen season is also an important component which is dependent on the above factors but is also dependent on pre-season conditions (Stach et al., 2008).

Pollen forecast models can be split into three categories: the first is observation-based or receptor-oriented models, process-based models constitute the second category, while a new third category of machine learning methods are now also becoming more popular (Maya-Manzano et al., 2020). Traditionally it is the observation-based methods which have provided the majority of national pollen forecasts, including for the UK. These require the use of site-specific observations, potentially combined with other variables such as meteorology, to predict pollen levels at those sites. Many of these are based on a ‘pollen calendar’ which derives the shape and duration of the season and is often provided graphically; for example this was produced for the UK by Adams-Groom et al. (2020). These are useful for medical professionals and allergy sufferers to understand the likely presence of pollen through the year. More advanced methods for creating calendars have also been developed by Šikoparija et al. (2018) and Gehrig et al. (2018). Regression modelling has also been used to predict the timing of the season, for example by Grewling et al. (1996) and García-Mozo et al. (2002). For the daily variation within a season, Norris-Hill (1995) used multiple regression modelling to link meteorological variables to observed pollen concentrations and in turn enable predictions of the severity and daily variation of the grass pollen season in North London.

The Met Office currently uses an observation-based approach to provide a pollen forecast in collaboration with the National Pollen and Aerobiology Research Unit (NPARU), referred to henceforth as NPARU-MO. This is used to alert both the public and health professionals to allow the prevention and minimisation of allergies through the use of medication and behaviour modifications. This five-day forecast is for the overall daily pollen level (for a single prevailing taxa) for 16 regions across the UK. It is generated using expert judgement from NPARU based on current observations, coupled with reductions in the forecast due to meteorological variables by Met Office forecasters. The pollen risk scenarios under different weather conditions are built on historical data analysis and research articles such as Adams-Groom et al. (2020, 2022). The meteorological adjustment is based on simple threshold values for rainfall, temperature, humidity and wind speed. The severity of the season is predicted through observations, analysis of weather in preceding key pollen production months and comparison to previous similar seasons (personal communication, NPARU).

The current observation-based approach has a number of limitations. For example, the forecast is for a single daily pollen level, based on the prevailing taxa. This means that once the grass season begins, the forecast level is based on grass only, even if higher levels of other taxa, such as oak, are still being observed, which may harm allergy sufferers for these other taxa. In addition, the forecast is for a daily value, however the pollen concentration varies considerably during the day: it would be beneficial to the public to produce additional information regarding the hourly variation to enable them to better plan when to go outside. Pollen varies considerably at small spatial scales, due to the location of pollen emitting plants, or meteorological effects such as sea breezes: this is not captured by the current large scale regional forecasts. The current forecast also does not take into account long range transport of pollen, for example from the Republic of Ireland or France into the UK. A process-based dispersion model would enable both higher temporal and spatial resolution and taxa-specific pollen forecasts and should therefore provide a significant improvement on the Met Office’s current forecast capabilities.

The use of process-based numerical dispersion models, traditionally developed to simulate the transport of air pollutants(Leelőssy et al., 2014) for air quality modelling (Kukkonen et al., 2012), is well documented. However dispersion models are increasingly being used to simulate the atmospheric spread of pollen, which can travel large distances, up to 1000km (Rousseau et al., 2003; Sofiev et al., 2006). These models employ full three-dimensional simulations of the atmosphere, driven by meteorological fields and including pollen-specific parameters to estimate the transport of different pollen grains. These models can take in a gridded spatial field describing the location of pollen-producing taxa allowing grains to be transported around in the atmosphere. In more recent years these dispersion models have been coupled with plant physiological schemes to estimate spatio-temporal variations in pollen emissions, particularly for taxa with allergenic pollen such as birch (Sofiev et al., 2013; Zink et al., 2013). Plant physiological schemes that are coupled to dispersion models generally include two components, i) the seasonal timing of taxa-specific pollen seasons, and ii) the daily variation within the season, both of which are dependent on meteorology.

Knowledge of the location of pollen sources is an integral part of a dispersion-based pollen forecast. Methods for generating these are often limited by access to pollen observations or land use data which are on a small fixed geographical domain (e.g. McInnes et al., 2017). Species distribution modelling (SDM) techniques are one method which is now being used to overcome this. These make use of observed locations of species, which are combined with environmental features to determine habitat suitability. For example, Bullock et al. (2012) and Prank et al. (2013) used this method for ragweed mapping, while Stas et al. (2020) and Dujardin et al. (2022) developed tree pollen maps over Belgium. An advantage of this technique is that once the species distribution models have been developed, they can be projected onto future climatic environments (e.g. Mauri et al., 2022) or nearby regions such as France where we do not have access to consistent species distribution datasets. Compared to many of the alternative mapping methods, predictions of habitat suitability are not limited to strict land use categories, for example trees in forested areas, or grasses in ‘grassland’ categories, which is important when these taxa are present extensively outside these regions.

The first key component of plant physiological schemes is the prediction of the seasonal cycle. This is typically either done through fixed start and end dates of the year (Wozniak & Steiner, 2017), with varying lengths of season depending on temperature (Zink et al., 2013), or with an accumulated temperature sum based method (Linkosalo et al., 2010; Sofiev et al., 2006; Zhang et al., 2013). Secondly, the daily variation within each season is dependent on meteorology. Some of the common meteorological variables that have been used by models for this include wind speed, relative humidity and precipitation (Sofiev et al., 2013), although others also include friction velocity (Zhang et al., 2013), turbulent kinetic energy (Zink et al., 2013) and leaf area index (Helbig et al., 2004).

The process-based approach used in dispersion models would provide a major improvement in the Met Office pollen forecast. It would allow a gridded taxa-specific output, which would be more useful to the public due to (i) its localised nature with a greater level of spatial detail and (ii) distinction between taxa. In this paper, we discuss the implementation of a new pollen emission scheme within the Met Office’s Numerical Atmospheric-dispersion Modelling Environment (NAME) dispersion model (Jones et al., 2007). Three taxa were selected for inclusion in the initial version of the model. According to Bauchau & Durham (2004), 52% of subjects with a clinical diagnosis of allergic rhinitis in Europe are allergic to grass pollen, which peaks in June and July in the UK, so this is a key taxon to include. In addition, the two tree genera of birch and oak were included which represent the most important allergenic taxa in the UK in April and May respectively (Adams-Groom et al., 2020). A comparison between NAME model outputs and observations is given in the results section, along with a comparison against the existing NPARU-Met Office pollen forecast (‘NPARU-MO’) and the Copernicus Atmosphere Monitoring Service (CAMS) ensemble (Sofiev et al., 2015) which produces a pollen forecast for grass and birch. A discussion of the current capability of NAME is included in the discussion section, with a summary concluding this paper.

2.1. Pollen Observations

Pollen observations in the UK are compiled from a network of monitoring stations, which has been run by the Met Office since 2011. Up to 20 stations observed pollen between 2011 and 2023 (Fig. 1). However, the number of stations gradually reduced from a peak of 19 in 2011 to just 11 sites by 2023 and was severely affected by COVID-19 lockdown restrictions in 2020, when only 8 stations were active.

All the pollen monitoring stations are run by expert volunteers, following the procedures set out by Caulton et al. (2011). Pollen is collected by Burkard volumetric spore traps designed by Hirst (Hirst, 1952). From late March until early September, the traps are changed at 9am local time; daily in the grass season (mid-May to early August) or weekly outside of this period. Pollen, and other airborne particles, are drawn in through a slot on the Burkhard trap and stick to an adhesive film that is attached to a slowly rotating drum. The speed of rotation of the drum is set to enable the length of film per hour to be determined. Every day or week the sample drum is collected and replaced with one loaded with clean film. In a laboratory, the sample film is removed from the drum and sectioned into 24-hour lengths, which are mounted onto microscope slides. Under a microscope, pollen grains are identified and systematically counted in two-hour traverses. Both bihourly and daily total pollen counts are converted into concentrations in air (grains m^− 3), with the daily pollen count defined and recorded as the average pollen concentration in the 24-hour period from 9am local time (which is different to the pollen count defined by Galán et al. (2017)). Bihourly pollen concentrations are likely to exhibit higher variability and uncertainty than daily concentrations due to the lower numbers of grains counted over the bihourly traverses. However, the sub-daily concentrations can still be informative when considering a long time period or multiple sites. The bihourly concentrations are also used to calculate daily mean concentrations from 00Z (as opposed to 9am used in the pollen count) which are used in this paper to allow direct comparison to the NPARU-MO forecast.

Pollen from 12 plant taxa (species, genera or family, where species cannot be easily identified by microscopy) are routinely observed - grass (family), 8 trees (including birch and oak) and 3 weeds (Table S2). For development of model parametrisations, observations from 2011–2021 were used, while 2022 and 2023 were used to examine the model performance.

For the development of emission parametrisations, quality control steps were carried out on the observations. These included only keeping sites that have a) at least 75% data coverage in the peak season (Table S1), b) an Annual Pollen Integral (APIn, the total sum of daily average pollen concentrations for the year) of ≥ 10% mean APIn across all sites for the given year, c) an APIn of ≥ (mean – standard deviation) across all sites for the given year; these last two remove any erroneously low count sites where the quantity of data may not be sufficient to derive accurate trends and therefore parametrisations from.

2.2. Daily Pollen Level / Index

Observed daily mean concentrations for each taxon were combined and converted into the categorical Daily Pollen Level (DPL) used in the UK. Four different levels were defined - Low, Moderate, High (all hay fever sufferers are likely to be affected, (Adams-Groom et al., 2020) and Very High (sufferers are likely to have extensive symptoms). To facilitate quantitative verification, these categories are also referred to numerically as the Daily Pollen Index (DPI), where 1 represents Low, through to 4 which represents a Very High Daily Pollen Level. Each taxon was individually converted to a DPI value (Table 1), with the maximum across all taxa then taken as the overall DPI. Conversion thresholds for additional pollen taxa, which can also be included in the DPI are given in Table S2.

Table 1

Daily Pollen Level (DPI) for each taxon (latin and common names) used in this study, and their associated pollen concentration (c) thresholds in pollen grains per m³ of air.
Latin	Common name	Low (1)	Moderate (2)	High (3)	Very High (4)
Betula	Birch	c < 40	40 ≤ c < 80	80 ≤ c < 200	c ≥ 200
Quercus	Oak	c < 30	30 ≤ c < 50	50 ≤ c < 200	c ≥ 200
Poaceae	Grass	c < 30	30 ≤ c < 50	50 ≤ c < 150	c ≥ 150

2.3. NAME atmospheric dispersion model

The Numerical Atmospheric-dispersion Modelling Environment (NAME, Jones et al., 2007) is an atmospheric dispersion model that has been used for many different applications. For example, NAME has been used for modelling the dispersion of volcanic ash (Beckett et al., 2010), nuclear accidents (Draxler et al., 2015), biomass burning (Buus Hansen et al., 2019) and for biological agents such as bluetongue (Burgin et al., 2013) and wheat stem rust (Meyer et al., 2017).

NAME is an offline dispersion model which uses meteorological data from the UK Met Office’s operational numerical prediction model, the Unified Model (UM, Davies et al., 2005). For this work, we have used NAME version 8.4 which has been driven by high resolution 1.5km UM UKV (Tang et al., 2013; Bush et al., 2020) meteorology over the UK and Global UM (Walters et al., 2017) meteorology at approximately 10km resolution outside the UKV domain. NAME can be run either in Lagrangian mode, where ‘particles’ are released into the atmosphere and transported individually by the meteorology, or in Eulerian mode, where a gridded scalar field is evolved, and concentrations are therefore given directly instead of averaging over a grid box of particles as in the Lagrangian scheme. For pollen modelling we use a hybrid version where pollen is emitted as particles into the Lagrangian framework and then transferred to the Eulerian grid after five minutes. The modelling domain used covers the UK, Republic of Ireland and Northern France at a resolution of 0.05° (approximately 5km, illustrated in Fig. 6), with 38 vertical model levels up to 5 km. Skjøth et al. (2009) showed the potential of pollen monitored in London to have been transported from France, which illustrates the importance of including nearby parts of Europe.

Our model resolution of 0.05° is much higher than those used in Sofiev et al. (2015) (ranging from 0.15° − 0.25°), which is achieved through our smaller geographical domain and the need to estimate local variations. Pollen grains from each of the three taxa (birch, oak, grass) are treated as aerosols with diameters given in Table 3, which are allowed to both dry and wet deposit, including by sedimentation. Removal of pollen grains from the atmosphere by dry deposition is parameterised using a resistance-based deposition velocity (Webster & Thomson, 2011) and the treatment of wet deposition is based on the depletion equation using parametrised scavenging coefficients which vary depending on precipitation type (Webster & Thomson, 2014). NAME has been developed to include new parametrisations which allow taxa-level variation in pollen emissions depending on meteorology, which will be described in detail below.

NAME can also output meteorological variables at hourly resolution at specific locations such as at pollen measurement sites. This method has been used to extract meteorology at these sites from the Global UM for January 2011 – July 2013 and the UKV for dates since August 2013 which is used in the development of parametrisations discussed below.

2.4. Source emission maps

Pollen emission maps were developed using species distribution modelling (SDM) techniques. This statistical method allows explanatory data (such as land use and meteorology) to be fitted to known locations of species, given by ‘presence’ data. The fitted model can then also be used for projection onto different geographical domains (as required here for Northern France), or for different explanatory datasets, for example under future climate conditions.

Presence data were provided by the Botanical Society for Britain and Ireland (BSBI) for 11 common species of grass (Table 2), and 2 common native species of both birch (Betula pubescens and Betula pendula) and oak (Quercus robur and Quercus petraea). Additional species of Alnus glutinosa, Corylus avellana, Fraxinus excelsior, Platanus x acerifolia, six species of Salix, Ulmus glabra, Artemisia vulgaris, Urtica dioica and Ambrosia artemisifolia were also provided. Data were provided as observation counts in a 2 km grid box for the period of 1970 to the present day for the United Kingdom, the Republic of Ireland, the Channel Islands and the Isle of Man. To ignore any grid points where there were no relevant observations which would bias the results, the BSBI observations were converted to binary presence or absence data, similar to the method used in Hill et al. (2017): A grid point is set to absent only if another species provided by BSBI was recorded there; however, if no species were recorded, that point is not included in the model.

Three types of explanatory data were used. The first was land use from the CORINE 2018 dataset (CORINE, 2018), which is available at 100m resolution, but re-gridded to 0.05 degrees as a percentage covered by each land use type. The major CORINE land use types were grouped for birch and oak into “broadleaved forest”, “mixed forest”, “coniferous forest” and “urban”, while for grasses these were grouped as “forest”, “urban”, “urban grass”, “rural” and “rural grass” (Table S3Error! Reference source not found.). Soil types were also used as explanatory data, with topsoil properties used from the European Soil Database Derived data (ESDAC, 2013; Hiederer, 2013). Finally, 30 arc-second resolution bioclimatic variables from WorldClim (Fick & Hijmans, 2017) were also included, limited to the least-correlated fields - isothermality, mean temperature of warmest quarter, mean temperature of coldest quarter, annual precipitation and elevation.

The R package biomod2 (Thuiller, 2003) was used to fit an ensemble of SDM models and the probability of occurrence was projected onto the required 0.05° grid. Probability of occurrence was calculated as the mean value predicted by each model in the ensemble, weighted by the true skill statistic of each model, calculated using 5 cross-validation runs using 70% training data and 30% validation data. This process identifies grid points where the species could occur and the probability of this.

The number of pollen grains assumed to be available for emission per mature tree was taken to be 1x10¹⁰ grains per year for both birch and oak, based on very limited information in the literature for species of oak not common in the UK - ranging from 1x10⁸ for Quercus rotundifolia, (Tormo Molina et al., 1996) to 5.5x10¹⁰ grains for Quercus suber (Gómez-Casero et al., 2004). For each tree species, the emitted grains per km² was then calculated as the product of the probability of occurrence from the SDM modelling, the number of grains per tree and the estimated number of trees per km², which is dependent on land use (Table S4). For grasses, the number of grains released per inflorescence (flower head) varies depending on species (Table 2). There are limited data on grass inflorescence density, with most being from more arid habitats than the UK - such as 517 inflorescences m^− 2 in Morocco (Aboulaich et al., 2009) and 410 inflorescences m^− 2 in Southern Spain (Prieto-Baena et al., 2011). However, a grassland field experiment carried out by Bennie in 2017 (personal communication) gave a mean inflorescence density of 2200 inflorescences m^− 2 in an ungrazed meadow in Cornwall: given this is more characteristic of the modelling region this value is therefore used. The emission map in grains km^− 2 is derived for each grid point as the product of i) maximum probability of occurrence across all grass species, ii) probability of occurrence weighted average of all the individual species grains per inflorescence values, iii) inflorescence density and iv) estimated percentage of grass in each land use category (Table S4).

Table 2

*Grains per inflorescence for the grass species used in the emission estimates.* ¹Ali et al. (2022), ²Prieto-Baena et al. (2011), ³Severova et al. (2022). ⁴ *Agrostis capillaris* is assumed to have the same value as *Agrostis stolonifera*.
Latin name	Common name	Pollen grains / inflorescence (x10⁶)
Agrostis capillaris	Common bent grass	2.400⁴
Agrostis stolonifera	Creeping bent grass	2.400¹
Anthoxanthum odoratum	Sweet verbal grass	1.300¹
Bromus hordeaceus	Soft brome	2.450²
Dactylis glomerata	Cock’s-foot	4.600¹
Festuca arundinacea	Tall fescue	11.700²
Holcus lanatus	Yorkshire fog	5.300¹
Lolium perenne	Perennial rye grass	2.000¹
Phleum pratense	Timothy grass	5.849³
Poa annua	Annual meadow grass	0.116²
Poa pratensis	Smooth meadow grass	3.394³

Finally, emissions ready for use in the model were converted into units of grams m^− 2, using pollen diameter and density information (Table 3). Resulting emissions are shown in Fig. 2, representing the maximum number of grains to be released in a year per unit area.

Table 3

*Pollen grain parameters of diameter and density.* ¹Emmerson et al. (2019). ²Sofiev et al. (2006), ³Panahi et al. (2012), ⁴Assume same density as for birch.
	Pollen diameter (µm)	Pollen density (kg m^− 3)
Birch	22²	800²
Oak	29³	800⁴
Grass	35¹	1000¹

2.5. Seasonal cycle

Seasonal variations in meteorological conditions provide a major influence on the timing and magnitude of events in the seasonal cycle of pollen, for example, the start, peak and end dates for pollen emissions. This is estimated using an accumulated temperature sum, or ‘heat sum’ (HS) above a base threshold temperature (T_base) from an annual start date (DOY_base). This is similar to the method employed in SILAM (Sofiev et al., 2013) and COSMO-ART (Zink et al., 2013). Eq. (1) shows the heat sum equation at a given time t, with a current temperature T_t.

$$\begin{array}{c}HS= \sum _{t>DOYbase}\text{max}\left({T}_{t}-{T}_{base}, 0\right) \#\end{array}$$

(1)

The parametrisation of the pollen seasonal cycle is derived in two stages: Firstly, the optimum threshold temperature (T_base) and start date (DOY_base) are estimated by comparing the predicted to observed day of the year for varying thresholds; secondly, the heat sum is fitted to varying percentiles of APIn, with the area under its derivative representing the proportion of the total estimated emissions to be released in each timestep.

The first part involves calculating the threshold temperature (T_base) and starting day of year (DOY_base) for accumulations. For this, we use the quality-controlled observations for 2011–2021 and calculate the day of the year on which the 5th, 50th and 95th percentiles of APIn are exceeded. For each of these dates we then calculate the heat sum using all possible combinations of T_base in 1°C increments between 0–10°C and DOY_base in 15-day increments from 0 to the day of the year one week prior to the first observed start date. The median heat sum across all sites and years for each (T_base, DOY_base) pair can be used to calculate the predicted day of the year on which that heat sum is exceeded. Statistical metrics comparing the predicted to observed day such as Root Mean Square Error (RMSE), Pearson correlation coefficient and the percentage of points correct to within ≤ 4 and ≤ 7 days can therefore be calculated, similar to the method used by Grundström et al. (2019). Finally, a judgement was taken by examining all metrics at the three different API levels as to which (T_base, DOY_base) is best, as concluded in Table 4.

Table 4

Heat sum and heat sum factor (HFactor) parameters for each pollen group.
	Birch	Oak	Grass
T_base (°C)	0	4	0
DOY_base	60 (~ 1st March)	75 (~ 15th March)	90 (~ 1st April)
First linear region (ramp up at beginning of season)
HS_l1start	149.120	81.441	113.025
HS_l1end	263.150	175.238	680.579
HS_l1c	-1.293e-3	-9.567e-4	-3.614e-5
HS_l1m	8.671e-6	1.175e-5	3.197e-7
Derivative curve region
HS_qd0	5.621e-1	-1.691e-1	-1.327e-1
HS_qd1	-2*3.351e-3	2*1.008e-3	2*2.531e-4
HS_qd2	3*9.757e-6	-3*2.690e-6	-3*2.344e-7
HS_qd3	-4*1.379e-8	4*3.261e-9	4*1.070e-10
HS_qd4	5*7.583e-12	-5*1.415e-12	-5*1.931e-14
Second linear region (ramp down at end of season)
HS_l2start	474.920	377.264	1443.842
HS_l2end	955.475	795.745	2050.539
HS_l2c	7.324e-4	8.233e-4	5.549e-4
HS_l2m	-7.666e-7	-1.035e-6	-2.706e-7

Once the taxa-dependent heat sum was calculated, this was used to derive the seasonal cycle to be input to the model through the heat sum factor (HFactor). For each five-percentile increment of the API, the median heat sum was calculated across all sites and years, with a quintic curve then fitted to this data. The derivative was calculated, and the curve start and end points limited to remove any inflection points. A straight line was fitted from these end points to the 0th percentile heat sum (start of season) and to the 100th percentile (end of season). These lines were adjusted such that the area under them match their known APIn percentile at the ends of their line (e.g. 5%, or more if the inflection points were removed). The derivative curve was then adjusted to ensure the total area under the entire season curve was unity. This ensures that over the entire season the area represents the emission for the whole year.

The combined line, showing the seasonal cycle, is illustrated in Fig. 3, with points on this line representing the fraction emitted at a given time t, FE_t; this is shown mathematically in Eq. (2). Here qd refers to parameters of the quintic derivative curve, l1 the first linear region at the start of the season and l2 the second linear line at the end of the season. HS_l1start and HS_l1end are the heat sums at the start and end of the first linear region, while HS_l2start and HS_l2end represent the heat sum at the start and end of the second linear region at the end of the season: these parameters are all shown in Table 4. Heat sum and heat sum factor (HFactor) parameters for each pollen group.

The area under the curve between the fraction emitted at time t and at the previous timestep t-Δt represents the proportion of the annual total emission that is released during that timestep and is given by a trapezium-rule calculation of the area. The HFactor_t is therefore given in Eq. (3) where HS_t is the heat sum at time t.

$$\begin{array}{c}{FE}_{t}=\left\{\begin{array}{ll}{HS}_{l{1}_{c}}+\left({HS}_{l{1}_{m}}.{HS}_{t}\right),& {HS}_{l{1}_{start}}\le {HS}_{t}<{HS}_{l{1}_{end}}\\ \begin{array}{c}{HS}_{qd0}+\left({HS}_{qd1}.{HS}_{t}\right)+\left({HS}_{qd2}.{HS}_{t}^{2}\right)\\ +\left({HS}_{qd3}.{HS}_{t}^{3}\right)+\left({HS}_{qd4}.{HS}_{t}^{4}\right),\end{array}& {HS}_{l{1}_{end}}\le {HS}_{t}\le {HS}_{l{2}_{start}}\\ {HS}_{l{2}_{c}}+\left({HS}_{l2m}.{HS}_{t}\right),& {HS}_{l{2}_{start}}<{HS}_{t}\le {HS}_{l{2}_{end}}\\ 0,& Otherwise\end{array}\right.\#\end{array}$$

(2)

$${HFactor}_{t}=0.5({HS}_{t}-{HS}_{t-\varDelta t})({FE}_{t}+{FE}_{t-\varDelta t})$$

(3)

2.6. Short term meteorological dependencies

In any pollen season, the modelled pollen emissions will start, peak and end based on their general season cycle as parametrised by the heat sum described above. However, within a season there will be short-term variability, driven by the changing meteorological conditions, which can prohibit emissions. We will therefore now discuss some of the key parameters which are used in NAME to model these effects.

2.6.1. Precipitation

Rainfall can suppress pollen emissions so a precipitation factor PFactor is used in the model. We use the same parameterisation as given by Sofiev et al. (2013) and shown in Eq. (4), where P_t is the precipitation rate in mm hr^− 1. This results in emissions decreasing linearly until precipitation reaches 0.5 mm hr^− 1 beyond which point there are no emissions. Although other options for a rainfall scheme were tested, the modelled concentrations were found to be not particularly sensitive to the scheme used, but that used by Sofiev et al. (2013) matched the observed concentration variations best during the 2018 and 2019 seasons.

$$\begin{array}{c}{PFactor}_{t}= \left\{ \begin{array}{cc}1,& {P}_{t}=0\\ 1-2{P}_{t},& {P}_{t}<0.5\\ 0,& {P}_{t}\ge 0.5\end{array}\right.\#\end{array}$$

(4)

2.6.2. Wind

Some pollen can be emitted at low wind speeds when there is thermal convection. As wind speed increases, pollen emissions will increase to a certain point whereby they will be limited by availability of pollen grains. We use the wind factor Wfactor parametrisation fromSofiev et al. (2013) as given in Eq. (5), where U_t is the wind speed (m s^− 1) at 10 m, w*_t is the convective velocity scale (m s^− 1) and the constants are f_stagnant = 0.5, f_promote = 1.0 and U_satur = 5 m s^− 1.

$$\begin{array}{c}{WFactor}_{t}={f}_{stagnant}+ {f}_{promote}+\left( 1-\text{exp}\left(-\frac{\left({U}_{t} +{{w}^{*}}_{t}\right)}{{U}_{satur}}\right)\right) \end{array}$$

(5)

2.6.3. Vapour pressure deficit

Vapour pressure deficit (VPD) is the difference, or deficit, between the amount of moisture in the air and how much moisture the air can hold when it is saturated. It is an important variable for vegetation as transpiration, and therefore plant growth is dependent on it. In the context of pollen emissions, a dry atmosphere promotes pollen release. Although only a few studies consider VPD as a parameter for controlling pollen emissions (Schueler & Schlünzen, 2006; van Hout et al., 2008), by combining temperature and relative humidity to calculate VPD (Pa) should lead to improved emission predictions. VPD is calculated via the saturated vapour pressure (SVP), as shown in Equations (6) and (7), where T is temperature (°C) and RH is relative humidity (%).

$$SVP=610.7*{10}^{\frac{7.5T}{237.3+T}}$$

(6)

$$VPD= \left(1-\frac{RH}{100}\right).SVP$$

(7)

To develop a parametrisation based on VPD, hourly temperature and relative humidity values were extracted at pollen observation sites using NAME and then VPD calculated. Hourly pollen concentrations were matched to hourly VPD. 25 equal size VPD bins were then created, with mean pollen concentrations for each bin calculated. For some bins, the data were classified as less reliable when there is less than 5% of the expected number of data points (given the total number of data points and bins) – these points were therefore given a lower weighting, when a sine curve was then fitted through the mean binned data.

The VPDfactor was therefore modelled as a sine curve with maximum at one, shown by Eq. (8), where VPDmax parameter is given in Table 5 and varies by taxa.

$${VPDFactor}_{t}= \left\{\begin{array}{cc}\text{sin}\left(\frac{\pi }{VPDmax}.\left({VPD}_{t}+\pi \right)\right),& VPD<VPDmax\\ 0,& VPD\ge VPDmax\end{array}\right.$$

(8)

Table 5

Taxa-specific parameters for Vapour Pressure Deficit (VPD) and diurnal cycle parameterisations.
	VPDmax (Pa)	Tshift (hours)
Birch	2725	9
Oak	2619	9
Grass	4135	12

2.6.4. Diurnal cycle

Pollen concentrations vary significantly during the day. Application of the meteorological dependant parameterisations described above in NAME results in some diurnal variation, however these were found to poorly match the variation shown in the observations. Consequently, a simple diurnal profile was developed using a cosine curve as shown in Eq. (9), where Hr_t is the current hour in the day and tshift is given in Table 5. The value of tshift was found by running NAME for 2018 with varying values of tshift to find the best match in the peak season. NAME was used here to account for the difference between when plants emit pollen and the time taken to advect downwind to the monitoring sites. The importance of including this factor is shown in Fig. 4, which illustrates moving the peak from around 8pm to mid-afternoon for the trees and late afternoon for grass as indicated by the bihourly observations.

$$Dfactor=1+\text{cos}\left(\left({Hr}_{t}+24-tshift\right)* \frac{\pi }{12}\right)$$

(9)

2.7. Implementation in NAME

Within the NAME model, the heat sum is incremented every time step (5 minutes) and can be output and read in again should the model need to be stopped and restarted (which is done daily). A new pollen emission scheme has been added to the code which then applies the seasonal and daily parameterisations described above. This is shown by Eq. (10) for S which is the source emission (g gridbox^− 1) for the current timestep, where the factors are described above, refSS is the reference source strength as read from the emissions (Section 2.4) and converted within NAME to g gridbox^− 1 s^− 1; sPerYear converts emission strength to g gridbox^− 1 yr^− 1 and deltaTime accounts for the time step duration of the current release. Once emitted within NAME it is advected and undergoes deposition before hourly concentrations are calculated.

$$S=refSS*HFactor*PFactor*WFactor*VPDFactor*DFactor* \frac{sPerYear}{deltaTime}$$

(10)

2.8. Model evaluation set-up

The NAME pollen model has been run for two years, 2022 and 2023, which were both different in terms of meteorology and the resulting pollen levels. Spring 2022 was warmer than average in March and May and drier than average in April (UK Met Office, 2023). In contrast Spring 2023 had generally more average rainfall and temperatures, although March was very wet, especially in the south where there was double the average rainfall and May was warmer than average. June was warm and dry in both years, but particularly hot in 2023, which was the warmest June recorded in the UK. July 2022 was very hot and dry, with a new UK record of 40.3°C recorded in Lincolnshire on the 19th, but in contrast, July 2023 was cooler than average and the wettest since July 2009.

The CAMS ensemble (METEO FRANCE et al., 2022; Sofiev et al., 2015) produces a daily pollen forecast for Europe, comprised of 11 process-based air quality forecasting models, adapted to include pollen, which are combined to produce an ensemble median. Their modelling domain is at 0.1° resolution over Europe (compared to 0.05° for NAME). Six pollen taxa are incorporated in their models which include both birch and grass as well as alder, mugwort, olive and ragweed although oak pollen is not currently modelled. This therefore provides a good comparator to the Met Office models and is included in verification metrics where appropriate.

There is a significant interannual variation in observed pollen levels, particularly in trees such as birch and oak. This is dependent on factors outside of the current season, for example weather conditions in the preceding summer (Adams-Groom et al., 2022) and is not currently captured by the NAME model. However, for examining the behaviour of the model for historical seasons, this can be overcome by scaling the model forecasts by a taxa and year-dependent factor. Here we have defined a scaling factor as the ratio of the observed to NAME-predicted annual pollen integral (APIn). This varies by site, but we take the median scaling factor across all sites to apply a consistent value suitable for use with gridded data. This factor is only applied to the NAME forecast on the assumption that the NPARU-MO and CAMS forecasts are able to predict this variation given they are used operationally. This may give NAME an advantage for some statistics, but the correlation coefficient will remain a fair metric. The use of the scaling factor also allows for correction of uncertainties in the pollen production values used in the emission maps. In theory a mean scaling factor, calculated from multiple years, could be applied to the emissions prior to running NAME, but this should give the same results as scaling the final concentrations in post-processing. Operationally, a mean scaling factor value could be used, but this would not account for the interannual variability which is still an active area of research.

The skill of the NAME model has been verified against daily mean observed pollen concentrations from the UK Pollen Network, which are available routinely from typically the 14th March until around the 5th September, with some variation between sites. Although NAME produces hourly concentrations, we have focussed the verification in this paper on daily mean values as this allows easier comparison to the NPARU-Met Office forecast and is the clearest way to illustrate the skill of the model at predicting the key seasonal variations. The observed DPI included in the verification includes all nine taxa (birch, oak, grass, hazel, alder, ash, plane, nettle and mugwort) which are both regularly monitored and can be converted to DPI, using the thresholds in Table S2. NAME has also been compared to the existing NPARU-Met Office Daily Pollen Index forecast (‘NPARU-MO’) and the CAMS ensemble median (‘CAMS’) individual birch and grass concentrations and the calculated Daily Pollen Index (noting that CAMS does not include oak which is an important component of the Daily Pollen Index). We also provide statistical metrics for a DPI calculated using only the three taxa of birch, oak, and grass, which gives a greater insight into the usability of the NAME model.

This section presents the results of the NAME model runs for 2022 and 2023 and their comparison to the CAMS ensemble median and the NPARU-MO forecast as well as verification against observations. Daily mean concentrations for each taxon are examined individually, concluding with the combined DPI verification.

3.1. Birch

The birch season runs from late March through to early May. The two years that were modelled show the extremes of birch pollen levels, with 2022 being a particularly low year, while 2023 was a year with very high concentrations of birch pollen: this is reflected in much higher APIn observed in Fig. 6 and the subsequent scaling factors used for NAME as shown in Table 6. Correlation coefficients between NAME and CAMS were comparable for 2023 but show slightly poorer performance for NAME in 2022 (Table 6). Timeseries plots comparing NAME and CAMS to daily mean observations are shown in Fig. 5, illustrating the ability to capture the daily variation fairly well, however a peak at the end of April 2022 is missed resulting in the lower correlation coefficients. This is probably due to the particularly warm March which increased the heat sum to the extent that the model was only emitting low, end-of-season, levels of birch pollen by this period. In 2023 by contrast, the NAME modelled season continues at low levels for too long, possibly driven by higher values of vapour pressure deficit in the second half of May during warmer and less humid conditions. Figure 8a captures the skill of the NAME model at predicting the season start and end dates, as defined by the 5th and 95th percentile API. These are generally forecast to be within 7 days of observed dates, however in 2023, the 95th percentile is reached an average of 20 days too late due to the extended model season. An example gridded field showing NAME daily mean concentrations of birch on one of the peak days in 2023 can be seen in Fig. 7a with the spatial pattern and magnitudes forecast matching well against the observations (overplotted in circles), with higher values in the south-west. This also shows the benefit of high-resolution gridded output from NAME, with the impacts of easterly winds clearly seen with lower concentrations found near the eastern coastline.

Table 6

Statistical metrics (Pearson Correlation Coefficient, Bias, Root Mean Squared Error – RMSE, Scaling factor) for 2022 and 2023, comparing daily mean pollen concentrations and Daily Pollen Index (DPI) between models and those calculated from bihourly observations.
	2022			2023
	NAME	CAMS ensemble	NPARU-MO	NAME	CAMS ensemble	NPARU-MO
Daily Mean Birch (15th March – 31st July)
Correlation	0.49	0.58		0.63	0.62
Bias (pollen grains m^− 3)	-2.17	-0.23		-2.88	-16.76
RMSE (pollen grains m^− 3)	19.33	18.01		57.22	69.02
Scaling factor	0.848	1.0		6.98	1.0
Daily Mean Oak (15th March – 31st July)
Correlation	0.58			0.68
Bias (pollen grains m^− 3)	-1.35			0.80
RMSE (pollen grains m^− 3)	16.34			12.78
Scaling factor	0.931			1.74
Daily Mean Grass (15th March – 31st August)
Correlation	0.74	0.80		0.78	0.82
Bias (pollen grains m^− 3)	3.52	-9.82		2.10	-11.01
RMSE (pollen grains m^− 3)	32.71	33.62		29.26	35.22
Scaling factor	0.0153	1.0		0.0155	1.0
Daily Pollen Index (17th March – 31st July) – using only grass, oak, birch
Correlation	0.62	0.61	0.56	0.62	0.57	0.59
Bias	0.07	-0.27	0.76	0.12	-0.43	0.58
RMSE	0.81	0.78	1.16	0.86	0.92	1.07
Daily Pollen Index (17th March – 31st July) - using all nine DPI taxa
Correlation	0.58	0.42	0.53	0.61	0.44	0.59
Bias	0.001	-0.24	0.69	0.08	-0.42	0.54
RMSE	0.85	0.94	1.15	0.87	1.01	1.06

3.2. Oak

The oak pollen season generally peaks in late April and early May. Time series plots shown in Fig. 5 and correlation coefficients in Table 6 indicate that NAME has captured the seasonal variability well, although struggles more with the timing of the end of the season, as shown by Fig. 8b. 2023 had generally higher levels of oak pollen concentrations, with a nearly a factor of two increase in the required scaling factor (Table 6), of which the impact of this can be clearly seen in Fig. 5. This may be due to cooler spring/summer temperatures in 2021 reducing the 2022 APIn (important variables according to Adams-Groom et al., 2022). The missed peaks in the mean values in the 2022 timeseries are driven by extreme peaks observed at single sites which NAME was not able to capture the magnitude of (for example Ipswich peaked at 207 pollen grains m^− 3 on the 1st May 2022). The spatial variation shown in Fig. 3b shows a good correlation between NAME and the observations on this day in the middle of the oak season. Lower values of oak pollen are generally observed at the Kings College London site, as shown in the gridded maps (Fig. 7); these are believed to be non-representative of the general Greater London area due to the urban location, high trap height and next to the river. It is also evident in this figure that there is transport of some pollen from France, highlighting the need to include this region in the NAME dispersion domain.

3.3. Grass

The grass season generally runs through both June and July in the UK. NAME shows good correlations with the observations (Table 6) and is comparable to the CAMS ensemble. In 2022, there are two distinct peaks in the timeseries in Fig. 5, split by a period of wetter weather. There is an over-prediction by NAME in July, which is driven by a lack of a second peak in some more Eastern observation sites. This is potentially caused by the absence of later flowering species of grasses near these sites. However, the individual species cannot be distinguished under a microscope as used in the observation network. Therefore, NAME currently models all the grass species in a single group, limiting the ability to capture the full complexity of the phenology of earlier and later flowering species. The rapid drop of pollen concentrations in 2023 is well modelled and is due to a very wet July. The timing of the season as shown in Fig. 8c is well predicted with most sites forecast with 7 days accuracy and many within 4 days. An example gridded field for June 2022 is shown in Fig. 7c, which shows a good match between NAME forecasts and the spatial variation of observations. As mentioned for oak, the London site generally captures much lower concentrations of pollen. There is much less interannual variability in the grass concentrations, consequently scaling factors are similar between years and an average over multiple years could be used for forecast purposes. It is worth noting however that given the average scaling factor of 0.0154 currently used in NAME, it is predicting concentrations around 65 times too high. This is likely to be due to errors in the magnitude of emissions where there are many uncertain factors.

3.4. Daily Pollen Index

The Daily Pollen Index (DPI), when translated to Daily Pollen Levels (Low, Moderate, High or Very High) is the basis for the UK pollen forecast: it is therefore a critical variable to be able to model. Table 6 shows verification metrics for two versions of the DPI: the first is calculated from birch, oak and grass only, in order to examine the current capability of NAME; the second uses all nine taxa which are regularly monitored at observation sites and have thresholds for conversion to DPI. This latter version gives a true representation of the ability of the models to reproduce the UK Daily Pollen Index, however for both versions NAME has a higher correlation coefficient than both CAMS and NPARU-MO. The nine-taxa version of the DPI is also shown as a timeseries plot in Fig. 5 and used in the contour plots in Fig. 9.

The manual NPARU-MO forecast is for the overall DPI only, with a single value representing a large geographical area. This forecast includes all nine taxa of the DPI, however in practice levels of other taxa are usually lower than the three examined in this paper, or in a different season, for example nettle in August. We therefore show model performance statistics from the 17th March (when the NPARU-MO forecast is first issued), to the end of July, after which nettle becomes more dominant. It is noticeable that there is only a slight difference in the correlation coefficient between the two versions of the DPI (Table 6), suggesting that the three taxa of birch, oak and grass constitute the majority of the taxa driving the DPI in this period. The DPI is consistently over-predicted by the NPARU-MO as indicated by the bias in Table 6 and the timeseries plots in Fig. 5. This may be due in part to a higher forecast expected in one location in the forecast region, but the single observation only occurring in another lower-DPI location.

The CAMS ensemble does not include oak in its models and subsequently the DPI is 1 (Low) for the whole of May. Given the prominence of oak pollen allergies in the UK (Ross et al., 1996) and the contribution of oak to the Daily Pollen Level, CAMS is therefore not currently a suitable forecasting model for use in the UK. In addition, CAMS predicts higher values of DPI in late July due to mugwort: this has limited spread in the UK and consequently is only observed on occasion and not a large contributor to the DPI – the CAMS forecast shows a high bias and a correlation coefficient close to zero for daily mean mugwort in the UK (results not shown). The poor performance of mugwort explains the lower statistical scores with the all-taxa DPI.

The NAME model has the highest correlation coefficient of the three models, for both versions of the DPI. The low bias is expected here due to the scaling factors used to account for the interannual variation in APIn. There is some degradation in skill when verified against all taxa: this is due to the DPI being driven by taxa such as ash in early April and nettle in late July, which are not yet included in the model.

Example fields of the Daily Pollen Index are shown in Fig. 9 for 14th June 2023, in the middle of the grass season. The low resolution of the regional NPARU-MO forecast is clear in Fig. 9b. By contrast, the NAME and CAMS ensemble forecasts give much more spatial variation; for example, it is evident that easterly winds are reducing pollen levels down the East coast. In a forecast system this higher resolution will provide valuable additional information to the public and health professionals.

4.1. Comparison to other pollen forecasts

The results in Section 3 show that the NAME model is performing well when run in hindcast mode (to allow appropriate scaling factors to be calculated). For birch, results are generally comparable to the CAMS ensemble, although NAME results were poorer in 2022. A substantial benefit of NAME over the CAMS ensemble is the inclusion of oak pollen which is a major contributor to the Daily Pollen Level in the UK. For grasses, the skill of the NAME model is better than for other taxa, although slightly worse than for the CAMS ensemble. The key variable to predict however is that of the Daily Pollen Index (DPI) as this is the variable currently provided to the public due to its links to health impacts. From mid-March to the end of July, we have shown that the nine-taxa DPI is driven mainly by the three taxa we have initially included in NAME (birch, oak and grass). The NAME model (in hindcast mode) is out-performing the existing NPARU-MO forecast for both years and both versions of the DPI. NAME also performs better than the CAMS ensemble for the DPI given the lack of oak in the CAMS models.

4.2. Emissions

The source maps we have developed are generated using species distribution modelling (SDM). For comparison, the CAMS ensemble source map for birch is based on only the locations of forests in Europe (Sofiev et al., 2006) The benefit of the SDM approach is that it enables us to account for trees outside of forest areas, for example those in hedgerows, as well as grasses growing outside of ‘grassland’ categories in land use datasets, for example in urban gardens and parklands which can be critical for the health of urban populations.

One of the biggest weaknesses of the NAME model is the current inability to predict the significant interannual variability in the magnitude of the pollen season which is seen in the observations. In this paper we have accounted for this by scaling the model such that the median APIn at the observation sites calculated by NAME is matched to the same value calculated from the measurements. This is suitable for examining the skill of the model for predicting seasonal metrics such as season length and daily variations. However further research is required to predict the interannual variability in advance of the season such that NAME can be used for forecasting. This is particularly important for trees such as birch which varies by a factor of 8 between the two years considered here. The variability between years for grass is much smaller so NAME could now potentially be used with a constant mean scaling factor for forecasting purposes. An alternative option would be for in-season bias correction methods to be developed to modify the pollen forecast using recent observations, similar to Neal et al. (2014), However, the weekly frequency of observation counting during the birch season may prove problematic as this could mean correcting the model in the peak of the season to the lower observed levels from a week earlier, potentially before the season had started.

4.3. Benefits of the process-based model

The existing NPARU-MO forecast is generated using expert judgment with observation-based methods. This is only able to provide a single daily value across all taxa for each of just 16 regions. The process-based dispersion model NAME however has been shown to have much higher (0.05°) resolution spatial fields. This is able to capture short range meteorological features such as sea breezes, as evidenced by the lower birch concentrations near the eastern coastline, as well as longer range transport where pollen grains are carried out to sea and from France across to Southern England (e.g. Figure 7b). Although daily mean concentrations have been discussed here, NAME is also able to output at hourly resolution and is able to correctly model the diurnal profile (as shown in Fig. 4) which could be of significant benefit to the public to allow them to avoid hours in the day with higher levels of pollen concentrations. Finally, the NAME model can react to changing weather conditions, potentially increasing pollen levels over the following days as the meteorology becomes more conducive to pollen release. This is an improvement compared to the existing manual forecast which can currently only be moderated to lower levels.

4.4. Future developments

We have introduced birch, oak and grass pollen into the NAME model, which we have shown accounts for the majority of the variation in the DPI during the mid-March to end of July period we have evaluated. However outside of these periods other taxa become more significant, for example nettle, which is used to represent weeds, is a major allergen for August and September, while alder and hazel are important for earlier in the year, covering January to March – these two are particularly relevant as they are in the same homologous group as birch and therefore affect the same groups of people (Biedermann et al., 2019). These three additional taxa are currently being actively developed for inclusion in NAME, while ash (measurements of which caused the raised DPI in March in Fig. 5g) and London plane (important in cities) are then the two remaining taxa that will need to be added in order to capture almost all occurrences of Moderate or higher Daily Pollen Levels in the UK.

Another area of active research is currently understanding features of the grass season. This often appears in two distinct peaks of differing magnitude, such as seen in Fig. 5e, however NAME will predict a similar magnitude for both periods. It is thought that this may be caused by different species of grass which flower at different times and with different phenology.

Finally, the ability of NAME to predict concentrations in hindcast mode leads to the potential to produce a high-resolution pollen reanalysis over the UK, driven by a consistent meteorology. This could be bias corrected to accurately match observations. This would have significant potential benefits for research purposes such as health impact studies.

We have presented an initial implementation of the emission, transport and dispersion of airborne pollen within the NAME model. New emissions, derived via species distribution modelling, have been developed for three taxa: birch, oak and grass. Emissions within NAME are dependent on a heat sum parametrisation to control the pollen season, with short term meteorological dependencies implemented based on precipitation, wind speed and the vapour pressure deficit, along with a diurnal cycle. We have shown that the performance of the NAME model for pollen predictions is comparable with that of the CAMS ensemble for birch and grass and is better than the manual NPARU-MO forecast for the Daily Pollen Index, when run in hindcast mode to allow appropriate scaling factors to be calculated. Further work is required to predict either in advance, or within season, the scaling factor that should be applied to the pollen emissions to represent the interannual variation, particularly for birch. In addition, work is currently ongoing to extend the NAME scheme to additional taxa. The system described here has the potential to be used to provide a new pollen forecast for the UK, bringing the significant benefits of higher spatial and temporal resolution, improved physical basis for the transport of pollen and explicit taxa-specific forecasting.

Author Contribution

L.S.N. wrote the main manuscript text. All authors reviewed the manuscript.

Acknowledgements

We would like to thank the Met Office Public Weather Service for funding to allow us to develop the NAME pollen model. We would also like to thank the Botanical Society for Britain and Ireland (BSBI) for their plant distribution maps and the Copernicus Atmosphere Monitoring Service for the CAMS ensemble data downloaded from their Atmospheric Data Store.

Data Availability

Data sets generated during the current study are available from the corresponding author on reasonable request.

Aboulaich, N., Bouziane, H., Kadiri, M., Del Mar Trigo, M., Riadi, H., Kazzaz, M., & Merzouki, A. (2009). Pollen production in anemophilous species of the Poaceae family in Tetouan (NW Morocco). Aerobiologia, 25(1), 27–38. https://doi.org/10.1007/S10453-008-9106-2/FIGURES/2
Adams-Groom, B., Ambelas Skjøth, C., Selby, K., Pashley, C., Satchwell, J., Head, K., & Ramsay, G. (2020). Regional calendars and seasonal statistics for the United Kingdom’s main pollen allergens. Allergy, 75(6), 1492–1494. https://doi.org/10.1111/ALL.14168
Adams-Groom, B., Selby, K., Derrett, S., Frisk, C. A., Pashley, C. H., Satchwell, J., King, D., McKenzie, G., & Neilson, R. (2022). Pollen season trends as markers of climate change impact: Betula, Quercus and Poaceae. Science of The Total Environment, 831, 154882. https://doi.org/10.1016/j.scitotenv.2022.154882
Ali, A. M. S., Rooney, P., & Hawkins, J. A. (2022). Automatically counting pollen and measuring pollen production in some common grasses. Aerobiologia, 38(4), 433–455. https://doi.org/10.1007/S10453-022-09758-3/FIGURES/6
Bauchau, V., & Durham, S. R. (2004). Prevalence and rate of diagnosis of allergic rhinitis in Europe. European Respiratory Journal, 24(5), 758–764. https://doi.org/10.1183/09031936.04.00013904
Beckett, F. M., Witham, C. S., Leadbetter, S. J., Crocker, R., Webster, H. N., Hort, M. C., Jones, A. R., Devenish, B. J., & Thomson, D. J. (2010). Atmospheric Dispersion Modelling at the London VAAC: A Review of Developments since the 2010 Eyjafjallajökull Volcano Ash Cloud. https://doi.org/10.3390/atmos11040352
Biedermann, T., Winther, L., Till, S. J., Panzner, P., Knulst, A., & Valovirta, E. (2019). Birch pollen allergy in Europe. In Allergy: European Journal of Allergy and Clinical Immunology (Vol. 74, Issue 7, pp. 1237–1248). Blackwell Publishing Ltd. https://doi.org/10.1111/all.13758
Brewer, A. (2012). NFI preliminary estimates of quantities of broadleaved species in British woodlands, with special focus on ash. https://cdn.forestresearch.gov.uk/2022/02/nfi_prelim_bl_ash_estimates.pdf
Britt, C., & Johnston, M. (2012). Trees in Towns II: A new survey of urban trees in England and their condition and management. Arboricultural Journal, 34(2), 119–119. https://doi.org/10.1080/03071375.2012.708498
Bullock, J. M., Chapman, D., Schafer, S., Roy, D., Girardello, M., Haynes, T., Beal, S., Wheeler, B., Dickie, I., Phang, Z., Tinch, R., Čivić, K., Delbaere, B., Jones-Walters, L., Hilbert, A., Schrauwen, A., Prank, M., Sofiev, M., Niemelä, S., … Brough, C. (2012). Assessing and controlling the spread and the effects of common ragweed in Europe.
Burgin, L. E., Gloster, J., Sanders, C., Mellor, P. S., Gubbins, S., & Carpenter, S. (2013). Investigating Incursions of Bluetongue Virus Using a Model of Long-Distance Culicoides Biting Midge Dispersal. Transboundary and Emerging Diseases, 60(3), 263–272. https://doi.org/10.1111/J.1865-1682.2012.01345.X
Bush, M., Allen, T., Bain, C., Boutle, I., Edwards, J., Finnenkoetter, A., Franklin, C., Hanley, K., Lean, H., Lock, A., Manners, J., Mittermaier, M., Morcrette, C., North, R., Petch, J., Short, C., Vosper, S., Walters, D., Webster, S., … Zerroukat, M. (2020). The first Met Office Unified Model-JULES Regional Atmosphere and Land configuration, RAL1. Geoscientific Model Development, 13(4), 1999–2029. https://doi.org/10.5194/GMD-13-1999-2020
Buus Hansen, A., Suzanne Witham, C., Ming Chong, W., Kendall, E., Ning Chew, B., Gan, C., Craig Hort, M., & Lee, S. Y. (2019). Haze in Singapore-source attribution of biomass burning PM10 from Southeast Asia. Atmospheric Chemistry and Physics, 19(8), 5363–5385. https://doi.org/10.5194/ACP-19-5363-2019
Caulton, E., Lacey, M. E., Allitt, U., Crosby, R., Emberlin, J., & Hirst, J. M. (2011). BAF GUIDE Airborne Pollens Spores - A Guide to Trapping and Counting.
CORINE. (2018). CLC 2018 — Copernicus Land Monitoring Service. European Environment Agency. https://land.copernicus.eu/pan-european/corine-land-cover/clc2018
Crowther, T. W., Glick, H. B., Covey, K. R., Bettigole, C., Maynard, D. S., Thomas, S. M., Smith, J. R., Hintler, G., Duguid, M. C., Amatulli, G., Tuanmu, M. N., Jetz, W., Salas, C., Stam, C., Piotto, D., Tavani, R., Green, S., Bruce, G., Williams, S. J., … Bradford, M. A. (2015). Mapping tree density at a global scale. Nature 2015 525:7568, 525(7568), 201–205. https://doi.org/10.1038/nature14967
Davies, T., Cullen, M. J. P., Malcolm, A. J., Mawson, M. H., Staniforth, A., White, A. A., & Wood, N. (2005). A new dynamical core for the Met Office’s global and regional modelling of the atmosphere. Quarterly Journal of the Royal Meteorological Society, 131(608), 1759–1782. https://doi.org/10.1256/QJ.04.101
Draxler, R., Arnold, D., Chino, M., Galmarini, S., Hort, M., Jones, A., Leadbetter, S., Malo, A., Maurer, C., Rolph, G., Saito, K., Servranckx, R., Shimbori, T., Solazzo, E., & Wotawa, G. (2015). World Meteorological Organization’s model simulations of the radionuclide dispersion and deposition from the Fukushima Daiichi nuclear power plant accident. Journal of Environmental Radioactivity, 139, 172–184. https://doi.org/10.1016/J.JENVRAD.2013.09.014
Dujardin, S., Stas, M., Van Eupen, C., Aerts, R., Hendrickx, M., Delcloo, A. W., Duchêne, F., Hamdi, R., Nawrot, T. S., Van Nieuwenhuyse, A., Aerts, J. M., Van Orshoven, J., Somers, B., Linard, C., & Dendoncker, N. (2022). Mapping abundance distributions of allergenic tree species in urbanized landscapes: A nation-wide study for Belgium using forest inventory and citizen science data. Landscape and Urban Planning, 218, 104286. https://doi.org/10.1016/J.LANDURBPLAN.2021.104286
Emmerson, K. M., Silver, J. D., Newbigin, E., Lampugnani, E. R., Suphioglu, C., Wain, A., & Ebert, E. (2019). Development and evaluation of pollen source methodologies for the Victorian Grass Pollen Emissions Module VGPEM1.0. Geosci. Model Dev, 12, 2195–2214. https://doi.org/10.5194/gmd-12-2195-2019
ESDAC. (2013). European Soil Database Derived data - ESDAC - European Commission. European Commission Joint Research Centre. https://esdac.jrc.ec.europa.eu/content/european-soil-database-derived-data
Fick, S. E., & Hijmans, R. J. (2017). WorldClim 2: new 1-km spatial resolution climate surfaces for global land areas. International Journal of Climatology, 37(12), 4302–4315. https://doi.org/10.1002/JOC.5086
Galán, C., Ariatti, A., Bonini, M., Clot, B., Crouzy, B., Dahl, A., Fernandez-González, D., Frenguelli, G., Gehrig, R., Isard, S., Levetin, E., Li, D. W., Mandrioli, P., Rogers, C. A., Thibaudon, M., Sauliene, I., Skjoth, C., Smith, M., & Sofiev, M. (2017). Recommended terminology for aerobiological studies. Aerobiologia, 33(3), 293–295. https://doi.org/10.1007/s10453-017-9496-0
García-Mozo, H., Galán, C., Aira, M. J., Belmonte, J., Díaz De La Guardia, C., Fernández, D., Gutierrez, A. M., Rodriguez, F. J., Trigo, M. M., & Dominguez-Vilches, E. (2002). Modelling start of oak pollen season in different climatic zones in Spain. Agricultural and Forest Meteorology, 110(4), 247–257. https://doi.org/10.1016/S0168-1923(02)00003-5
Gehrig, R., Maurer, F., & Schwierz, C. (2018). Designing new automatically generated pollen calendars for the public in Switzerland. Aerobiologia, 34(3), 349–362. https://doi.org/10.1007/S10453-018-9518-6/FIGURES/7
Gómez-Casero, M. T., Hidalgo, P. J., García-Mozo, H., Domínguez, E., & Galán, C. (2004). Pollen biology in four Mediterranean Quercus species. Grana, 43(1), 22–30. https://doi.org/10.1080/00173130410018957
Grewling, Ł., Jackowiak, B., & Smith, M. (1996). Quercus sp. pollen seasons. Poznan´. https://doi.org/10.1007/s10453-013-9313-3
Grundström, M., Adams-Groom, B., Pashley, C. H., Dahl, Å., Rasmussen, K., de Weger, L. A., Thibaudon, M., Fernández-Rodríguez, S., Silva-Palacios, I., & Skjøth, C. A. (2019). Oak pollen seasonality and severity across Europe and modelling the season start using a generalized phenological model. Science of the Total Environment, 663, 527–536. https://doi.org/10.1016/j.scitotenv.2019.01.212
Gupta, R., Sheikh, A., Strachan, D. P., & Anderson, H. R. (2004). Burden of allergic disease in the UK: secondary analyses of national databases. Clinical & Experimental Allergy, 34(4), 520–526. https://doi.org/10.1111/J.1365-2222.2004.1935.X
Helbig, N., Vogel, B., Vogel, H., & Fiedler, F. (2004). Numerical modelling of pollen dispersion on the regional scale. Aerobiologia, 3, 3–19.
Hiederer, R. (2013). Mapping Soil Properties for Europe-Spatial Representation of Soil Database Attributes. https://doi.org/10.2788/94128
Hill, L., Hector, A., Hemery, G., Smart, S., Tanadini, M., & Brown, N. (2017). Abundance distributions for tree species in Great Britain: A two-stage approach to modeling abundance using species distribution modeling and random forest. Ecology and Evolution, 7(4), 1043–1056. https://doi.org/10.1002/ECE3.2661
Hirst, J. M. (1952). AN AUTOMATIC VOLUMETRIC SPORE TRAP. Annals of Applied Biology, 39(2), 257–265. https://doi.org/10.1111/J.1744-7348.1952.TB00904.X
Hossenbaccus, L., Linton, S., Garvey, S., & Ellis, A. K. (2020). Towards definitive management of allergic rhinitis: best use of new and established therapies. Allergy Asthma Clin Immunol, 16, 39. https://doi.org/10.1186/s13223-020-00436-y
Jones, A., Thomson, D., Hort, M., & Devenish, B. (2007). The U.K. Met Office’s Next-Generation Atmospheric Dispersion Model, NAME III. Air Pollution Modeling and Its Application XVII, 580–589. https://doi.org/10.1007/978-0-387-68854-1_62
Kosztra, B., Büttner, G., Hazeu, S., & Arnold, G. (2019). Updated CLC illustrated nomenclature guidelines.
Kukkonen, J., Olsson, T., Schultz, D. M., Baklanov, A., Klein, T., Miranda, A. I., Monteiro, A., Hirtl, M., Tarvainen, V., Boy, M., Peuch, V. H., Poupkou, A., Kioutsioukis, I., Finardi, S., Sofiev, M., Sokhi, R., Lehtinen, K. E. J., Karatzas, K., San José, R., … Eben, K. (2012). A review of operational, regional-scale, chemical weather forecasting models in Europe. Atmospheric Chemistry and Physics, 12(1), 1–87. https://doi.org/10.5194/ACP-12-1-2012
Kurganskiy, A., Ambelas Skjøth, C., Baklanov, A., Sofiev, M., Saarto, A., Severova, E., Smyshlyaev, S., & Kaas, E. (2020). Incorporation of pollen data in source maps is vital for pollen dispersion models. Atmospheric Chemistry and Physics, 20(4), 2099–2121. https://doi.org/10.5194/ACP-20-2099-2020
Leelőssy, Á., Molnár, F., Izsák, F., Havasi, Á., Lagzi, I., & Mészáros, R. (2014). Dispersion modeling of air pollutants in the atmosphere: a review. Central European Journal of Geosciences, 6(3), 257–278. https://doi.org/10.2478/S13533-012-0188-6/XML
Linkosalo, T., Ranta, H., Oksanen, A., Siljamo, P., Luomajoki, A., Kukkonen, J., & Sofiev, M. (2010). A double-threshold temperature sum model for predicting the flowering duration and relative intensity of Betula pendula and B. pubescens. Agricultural and Forest Meteorology, 150(12), 1579–1584. https://doi.org/10.1016/j.agrformet.2010.08.007
Mauri, A., Girardello, M., Strona, G., Beck, P. S. A., Forzieri, G., Caudullo, G., Manca, F., & Cescatti, A. (2022). EU-Trees4F, a dataset on the future distribution of European tree species. Scientific Data 2022 9:1, 9(1), 1–12. https://doi.org/10.1038/s41597-022-01128-5
Maya-Manzano, J. M., Smith, M., Markey, E., Hourihane Clancy, J., Sodeau, J., & O´Connor, D. J. (2020). Recent developments in monitoring and modelling airborne pollen, a review. Grana, 60(1), 1–19. https://doi.org/10.1080/00173134.2020.1769176
McInnes, R. N., Hemming, D., Burgess, P., Lyndsay, D., Osborne, N. J., Skjøth, C. A., Thomas, S., & Vardoulakis, S. (2017). Mapping allergenic pollen vegetation in UK to study environmental exposure and human health. Science of the Total Environment, 599–600, 483–499. https://doi.org/10.1016/j.scitotenv.2017.04.136
METEO FRANCE, Institut national de l’environnement industriel et des risques (Ineris), Aarhus University, Norwegian Meteorological Institute (MET Norway), Jülich Institut für Energie- und Klimaforschung (IEK), Institute of Environmental Protection – National Research Institute (IEP-NRI), Koninklijk Nederlands Meteorologisch Instituut (KNMI), Nederlandse Organisatie voor toegepast-natuurwetenschappelijk onderzoek (TNO), Swedish Meteorological and Hydrological Institute (SMHI), Finnish Meteorological Institute (FMI), Italian National Agency for New Technologies Energy and Sustainable Economic Development (ENEA), & Barcelona Supercomputing Center (BSC). (2022). CAMS European air quality forecasts, ENSEMBLE data. Copernicus Atmosphere Monitoring Service (CAMS) Atmosphere Data Store (ADS). https://ads.atmosphere.copernicus.eu/cdsapp#!/dataset/cams-europe-air-quality-forecasts?tab=overview
Meyer, M., Burgin, L., Hort, M. C., Hodson, D. P., & Gilligan, C. A. (2017). Large-scale atmospheric dispersal simulations identify likely airborne incursion routes of wheat stem rust into Ethiopia. Phytopathology, 107(10), 1175–1186. https://doi.org/10.1094/PHYTO-01-17-0035-FI/ASSET/IMAGES/LARGE/PHYTO-01-17-0035-FI_T2.JPEG
Neal, L. S., Agnew, P., Moseley, S., Ordóñez, C., Savage, N. H., & Tilbee, M. (2014). Application of a statistical post-processing technique to a gridded, operational, air quality forecast. Atmospheric Environment, 98, 385–393. https://doi.org/10.1016/j.atmosenv.2014.09.004
Norris-Hill, J. (1995). The modelling of daily Poaceae pollen concentrations. Grana, 34(3), 182–188. https://doi.org/10.1080/00173139509429041
Panahi, P., Pourmajidian, M. R., Fallah, A., & Pourhashemi, M. (2012). Pollen morphology of Quercus (subgenus Quercus, section Quercus) in Iran and its systematic implication. Acta Societatis Botanicorum Poloniae, 81(1), 33–41. https://doi.org/10.5586/asbp.2012.005
Prank, M., Chapman, D. S., Bullock, J. M., Belmonte, J., Berger, U., Dahl, A., Jäger, S., Kovtunenko, I., Magyar, D., Niemelä, S., Rantio-Lehtimäki, A., Rodinkova, V., Sauliene, I., Severova, E., Sikoparija, B., & Sofiev, M. (2013). An operational model for forecasting ragweed pollen release and dispersion in Europe. Agricultural and Forest Meteorology, 182–183, 43–53. https://doi.org/10.1016/J.AGRFORMET.2013.08.003
Prieto-Baena, J. C., Hidalgo, P. J., Domínguez, E., & Galán, C. (2011). Pollen production in the Poaceae family. Http://Dx.Doi.Org/10.1080/00173130310011810, 42(3), 153–159. https://doi.org/10.1080/00173130310011810
Reid, C., Hornigold, K., McHenry, E., Nichols, C., Townsend, M., Lewthwaite, K., Elliot, M., Pullinger, R., Hotchkiss, A., Gilmartin, E., White, I., Chesshire, H., Whittle, L., Garforth, J., Gosling, R., Reed, T., & Hugi, M. (2021). State of the UK’s Woods and Trees 2021. In Woodland Trust.
Ross, A. M., Corden, J. M., & Fleming, D. M. (1996). The role of oak pollen in hay fever consultations in general practice and the factors influencing patients’ decisions to consult. The British Journal of General Practice, 46(409), 451. /pmc/articles/PMC1239713/?report=abstract
Rousseau, D.-D., Duzer, D., Ve Cambon, G., Jolly, D., Poulsen, U., Ferrier, J., Schevin, P., Gros, R., Rousseau, D.-D., Duzer, D., Cambon, G., Jolly, D., Poulsen, U., Ferrier, J., Schevin, P., & Gros, R. (2003). Long distance transport of pollen to Greenland. Geophysical Research Letters, 30(14), 1765. https://doi.org/10.1029/2003GL017539
Scheifinger, H., Belmonte, J., Buters, J., Celenk, S., Damialis, A., Dechamp, C., García-Mozo, H., Gehrig, R., Grewling, L., Halley, J. M., Hogda, K. A., Jäger, S., Karatzas, K., Karlsen, S. R., Koch, E., Pauling, A., Peel, R., Sikoparija, B., Smith, M., … De Weger, L. A. (2013). Monitoring, modelling and forecasting of the pollen season. In Allergenic Pollen: A Review of the Production, Release, Distribution and Health Impacts (Vol. 9789400748811, pp. 71–126). Springer Netherlands. https://doi.org/10.1007/978-94-007-4881-1_4
Schueler, S., & Schlünzen, K. H. (2006). Modeling of oak pollen dispersal on the landscape level with a mesoscale atmospheric model. Environmental Modeling and Assessment, 11(3), 179–194. https://doi.org/10.1007/s10666-006-9044-8
Severova, E., Kopylov-Guskov, Y., Selezneva, Y., Karaseva, V., Yadav, S. R., & Sokoloff, D. (2022). Pollen Production of Selected Grass Species in Russia and India at the Levels of Anther, Flower and Inflorescence. Plants 2022, Vol. 11, Page 285, 11(3), 285. https://doi.org/10.3390/PLANTS11030285
Šikoparija, B., Marko, O., Panić, M., Jakovetić, D., & Radišić, P. (2018). How to prepare a pollen calendar for forecasting daily pollen concentrations of Ambrosia, Betula and Poaceae? Aerobiologia, 34(2), 203–217. https://doi.org/10.1007/S10453-018-9507-9/TABLES/4
Skjøth, C. A., Smith, M., Brandt, J., & Emberlin, J. (2009). Are the birch trees in Southern England a source of Betula pollen for North London? International Journal of Biometeorology, 53(1), 75–86. https://doi.org/10.1007/S00484-008-0192-1/FIGURES/9
Sofiev, M., Berger, U., Prank, M., Vira, J., Arteta, J., Belmonte, J., Bergmann, K.-C., Chéroux, F., Elbern, H., Friese, E., Galan, C., Gehrig, R., Khvorostyanov, D., Kranenburg, R., Kumar, U., Marécal, V., Meleux, F., Menut, L., Pessi, A.-M., … Peuch, V.-H. (2015). MACC regional multi-model ensemble simulations of birch pollen dispersion in Europe. Atmospheric Chemistry and Physics, 15(14), 8115–8130. https://doi.org/10.5194/acp-15-8115-2015
Sofiev, M., Siljamo, P., Ranta, H., Linkosalo, T., Jaeger, S., Rasmussen, A., Rantio-Lehtimaki, A., Severova, E., & Kukkonen, J. (2013). A numerical model of birch pollen emission and dispersion in the atmosphere. Description of the emission module. International Journal of Biometeorology, 57(1), 45–58. https://doi.org/10.1007/s00484-012-0532-z
Sofiev, M., Siljamo, P., Ranta, H., & Rantio-Lehtimäki, A. (2006). Towards numerical forecasting of long-range air transport of birch pollen: Theoretical considerations and a feasibility study. International Journal of Biometeorology, 50(6), 392–402. https://doi.org/10.1007/S00484-006-0027-X/FIGURES/5
Stach, A., Emberlin, J., Smith, M., Adams-Groom, B., & Myszkowska, D. (2008). Factors that determine the severity of Betula spp. pollen seasons in Poland (Poznań and Krakow) and the United Kingdom (Worcester and London). International Journal of Biometeorology, 52(4), 311–321. https://doi.org/10.1007/s00484-007-0127-2
Stas, M., Aerts, R., Hendrickx, M., Dendoncker, N., Dujardin, S., Linard, C., Nawrot, T., Van Nieuwenhuyse, A., Aerts, J. M., Van Orshoven, J., & Somers, B. (2020). An evaluation of species distribution models to estimate tree diversity at genus level in a heterogeneous urban-rural landscape. Landscape and Urban Planning, 198, 103770. https://doi.org/10.1016/J.LANDURBPLAN.2020.103770
Tang, Y., Lean, H. W., & Bornemann, J. (2013). The benefits of the Met Office variable resolution NWP model for forecasting convection. Meteorological Applications, 20(4), 417–426. https://doi.org/10.1002/MET.1300
Thuiller, W. (2003). BIOMOD – optimizing predictions of species distributions and projecting potential future shifts under global change. Global Change Biology, 9(10), 1353–1362. https://doi.org/10.1046/J.1365-2486.2003.00666.X
Tormo Molina, R., Muñoz Rodríguez, A., Silva Palaciso, I., & Gallardo López, F. (1996). Pollen production in anemophilous trees. Grana, 35(1), 38–46. https://doi.org/10.1080/00173139609430499
UK Met Office. (2023). Climate summaries - Met Office. https://www.metoffice.gov.uk/research/climate/maps-and-data/summaries/index
van Hout, R., Chamecki, M., Brush, G., Katz, J., & Parlange, M. B. (2008). The influence of local meteorological conditions on the circadian rhythm of corn (Zea mays L.) pollen emission. Agricultural and Forest Meteorology, 148(6–7), 1078–1092. https://doi.org/10.1016/J.AGRFORMET.2008.02.009
Walters, D., Boutle, I., Brooks, M., Melvin, T., Stratton, R., Vosper, S., Wells, H., Williams, K., Wood, N., Allen, T., Bushell, A., Copsey, D., Earnshaw, P., Edwards, J., Gross, M., Hardiman, S., Harris, C., Heming, J., Klingaman, N., … Xavier, P. (2017). The Met Office Unified Model Global Atmosphere 6.0/6.1 and JULES Global Land 6.0/6.1 configurations. Geoscientific Model Development, 10(4), 1487–1520. https://doi.org/10.5194/GMD-10-1487-2017
Webster, H. N., & Thomson, D. J. (2011). Dry deposition modelling in a Lagrangian dispersion model. Int. J. Environment and Pollution, 47, 2011.
Webster, H. N., & Thomson, D. J. (2014). The NAME wet deposition scheme. In Met Office Forecasting Research Technical Report (Issue 584). https://digital.nmla.metoffice.gov.uk/IO_badb003d-26ec-4aab-ab12-84dd6b879656/
Wozniak, M. C., & Steiner, A. L. (2017). A prognostic pollen emissions model for climate models (PECM1.0). Geoscientific Model Development, 10(11), 4105–4127. https://doi.org/10.5194/GMD-10-4105-2017
Zhang, R., Duhl, T., Salam, M. T., House, J. M., Flagan, R. C., Avol, E. L., Gilliland, F. D., Guenther, A., Chung, S. H., Lamb, B. K., & VanReken, T. M. (2013). Development of a regional-scale pollen emission and transport modeling framework for investigating the impact of climate change on allergic airway disease. Biogeosciences (Online), 10(3), 3977. https://doi.org/10.5194/BGD-10-3977-2013
Zink, K., Pauling, A., Rotach, M. W., Vogel, H., Kaufmann, P., & Clot, B. (2013). EMPOL 1.0: A new parameterization of pollen emission in numerical weather prediction models. Geoscientific Model Development, 6(6), 1961–1975. https://doi.org/10.5194/gmd-6-1961-2013

No competing interests reported.

LucyNealSupplementaryMaterialsForSubmission.docx

Download PDF

Journal Publication

published 26 Apr, 2025

Read the published version in Aerobiologia →

Editorial decision: Revision requested
02 Dec, 2024
Reviews received at journal
01 Dec, 2024
Reviews received at journal
16 Sep, 2024
Reviewers agreed at journal
22 Jul, 2024
Reviewers agreed at journal
19 Jul, 2024
Reviewers invited by journal
17 Jul, 2024
Editor assigned by journal
11 Jul, 2024
Submission checks completed at journal
29 Jun, 2024
First submitted to journal
24 Jun, 2024

You are reading this latest preprint version

Development and verification of a taxa-specific gridded pollen modelling system for the UK

Status:

Journal Publication

Version 1

Abstract

Figures

1. Introduction

2. Model Description

2.1. Pollen Observations

2.2. Daily Pollen Level / Index

2.3. NAME atmospheric dispersion model

2.4. Source emission maps

2.5. Seasonal cycle

2.6. Short term meteorological dependencies

2.6.1. Precipitation

2.6.2. Wind

2.6.3. Vapour pressure deficit

2.6.4. Diurnal cycle

2.7. Implementation in NAME

2.8. Model evaluation set-up

3. Results

3.1. Birch

3.2. Oak

3.3. Grass

3.4. Daily Pollen Index

4. Discussion

4.1. Comparison to other pollen forecasts

4.2. Emissions

4.3. Benefits of the process-based model

4.4. Future developments

5. Summary

Declarations

Author Contribution

Acknowledgements

Data Availability

References

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1