This work used a novel, globally nested setup of the EMEP4UK atmospheric chemistry transport model (ACTM), consisting of the three domains shown in Fig. . In its standard setup , EMEP4UK operates over the two domains labelled B and C in the figure, with mostly prescribed boundary concentrations for domain B, and is a UK application of the European Monitoring and Evaluation Programme Meteorological Synthesizing Centre – West (EMEP MSC-W) Eulerian ACTM . As the standard setup of EMEP4UK cannot accurately account for the transient influences of pollutant transport into domain B, the model was extended to full global coverage. The global domain A provides hourly boundary conditions for the intermediate European domain B, which provides hourly boundary conditions for the inner domain C covering the UK and Republic of Ireland (ROI). Simulations were carried out with EMEP MSC-W model version 5.0.

Meteorology to drive the ACTM was calculated with the Weather Research and Forecast (WRF) model v4.2.2 at spatial resolutions of 1° × 1°, 27 km × 27 km and 3 km × 3 km for domains A–C respectively. There are 21 vertical layers extending up to 100 hPa. Reanalysis data from the US National Centers for Environmental Prediction (NCEP)/National Center for Atmospheric Research (NCAR) Global Forecast System (GFS) and Newtonian nudging of wind vectors and temperature every 6 h at 1° resolution were used. WRF parameterisations are as described by .

Model runs were conducted for the year 2019 and included a full year of spin-up (2018). Global domain simulations used anthropogenic emissions from the Task Force on Hemispheric Transport of Air Pollution (HTAP) v3 inventory for 2018 . The agricultural waste burning sector was not included to avoid the double counting of this source of BB emissions. European domain simulations used 2019 anthropogenic emissions from the Centre for Emission Inventories and Projections (CEIP) . In the innermost domain, simulations used 2019 anthropogenic emissions from the for the UK and MapEIre from the for ROI. Emissions of isoprene and other biogenic VOCs from vegetation, NO_x from lightning and soils, marine dimethyl sulfide (DMS), and wind-derived dust and sea salt are linked to the meteorological year and simulated as reported in and model update reports .

Gas-phase chemistry and inorganic aerosol thermodynamics are simulated with the EmChem19 chemical scheme and the Model for an Aerosol Reacting System (MARS) , respectively. Secondary organic aerosol (SOA) formation, ageing and phase partitioning are parameterised using a 5-bin 1-D volatility basis set with effective saturation concentration C* mid-points of 0.1, 1, 10, 100, 1000 µgm-3 . Primary organic aerosol (POA) is treated as non-volatile and chemically inert, as is assumed by emissions inventories . The model quantifies dry and wet removal processes as described by , , and .

Model output includes hourly gaseous and aerosol concentrations for all vertical model layers. The lowest model layer has a thickness of ∼ 48 m, and modelled air pollutant concentrations described here as surface concentrations have been adjusted to correspond to 3 m above the surface . PM_2.5 is calculated as the sum of the fine (< 2.5 µm diameter) fractions of sulfate (SO42-), nitrate (NO3-), ammonium (NH4+), organic matter (OM), sea salt, windblown dust, road dust, black carbon (BC), ash and a remaining primary component. A water component is not included to avoid ambiguity about how much water is associated with each PM_2.5 constituent.

BB emissions for 2018 and 2019 were obtained from the Fire INventory from NCAR (FINN) v2.5 dataset, which uses fire detections from both Moderate Resolution Imaging Spectroradiometer (MODIS) and Visible Infrared Imaging Radiometer Suite (VIIRS) . The latter yields fire detection down to 375 m resolution. FINNv2.5 provides daily estimates of aerosol and trace gas emissions from BB globally at 0.1° × 0.1° resolution, calculated using burned area from active fire detections. BB emissions are regridded to model resolution and evenly distributed from the surface up to 800 hPa .

The following model experiments were carried out: 1.

“BASE”: the base run with all emissions included.

The regions 1 to 8 are based on those proposed for perturbation experiments under the “HTAP3 Fires” model intercomparison project , which were derived from the 14 Global Fire Emissions Database (GFED) regions frequently used in fire emissions datasets . These regions were chosen to allow comparison to experiments carried out under the “HTAP3 Fires” project. Some minor changes were made to increase the relevance for the UK.

Concentrations conditional on BB emissions globally were calculated by subtracting the NBB run from the BASE run. Concentrations conditional on BB in the European domain illustrated in Fig.  were calculated by subtracting the NEBB model run from the BASE model run. Concentrations conditional on BB in the UK were calculated by subtracting the NUBB model run from the BASE model run. Concentrations conditional on BB emissions in each Region x defined in Fig.  were calculated by subtracting each NRxBB run from the BASE run. Population-weighted concentration means for the UK were calculated following the methodology described by . Gridded 2021 UK population data were obtained from (Fig.  of Appendix ), which uses data from the 2022 (Scotland) and 2021 (rest of the UK) Censuses and a 2021 Land Cover Map.

The EMEP4UK model in its standard configuration is regularly evaluated against measurements and is widely used for air quality studies . To evaluate the globally nested configuration of the model, the BASE model run was repeated using the standard configuration of EMEP4UK, the setup of which is described in and Appendix . The globally nested version of EMEP4UK is compared in Appendix  both to the standard configuration of EMEP4UK and to UK supersite measurements, focussing on annual mean PM_2.5 components. Secondary inorganic and total organic aerosol components of PM_2.5 are generally well represented in the globally nested version, particularly at sites most influenced by PM_2.5(BB). The major instance of global model overestimation is sea salt, which is not relevant to this study and is linked to changes made in recent versions of the EMEP MSC-W model code, where a larger percentage of the sea salt uplift is attributed to the fine fraction to improve model performance for Continental Europe .

Data in Table c reveal that a majority (73 %) of the 2019 UK annual mean PM_2.5(BB) derives from BB emissions outside the EMEP4UK model's European domain. This clearly demonstrates that continental-scale modelling is insufficient to capture the full contribution of BB and that a global nesting approach is needed to provide realistic and spatially and temporally resolved boundary conditions to regional ACTMs in order to accurately capture the very long-range impacts of BB emissions. These long-range contributions from episodic emissions, such as from BB, would not be accurately captured through the prescribed boundary conditions of the standard configuration of EMEP4UK.

In 2019, the annual mean PM_2.5(BB) associated with all BB emissions globally is 0.99 µgm-3 (averaged over the UK), which is a significant proportion (10 %) of total annual mean UK PM_2.5 from all sources (Table ). To the authors' knowledge, no other studies quantify the long-term contributions of the long-range transport of BB globally on UK PM_2.5, to allow comparison here. The equivalent population-weighted PM_2.5(BB) value associated with all BB emissions globally is 1.1 µgm-3, or 22 % of the WHO PM_2.5 annual mean guideline concentration . Averaged over the UK and the full year, the PM_2.5(BB) comprises of more secondary aerosol than primary aerosol (58 % and 42 %, respectively) (Table d). The dominant primary component is primary OM, constituting a proportion of 31/42, or 74 %, of the primary PM_2.5(BB). BC from BB emissions comprises just 3 % of PM_2.5(BB) (7 % of the primary PM_2.5(BB)). Within the secondary component of UK annual mean PM_2.5(BB), 55 % is SOA and 45 % is SIA, the latter dominated by NH₄NO₃ (Table e). The enhanced NH₄NO₃ formation is a subtle but important indirect consequence of BB emissions: the BB emissions change global oxidant concentrations which react with the large anthropogenic emissions of NO_x and NH₃ in the UK (and elsewhere) . The extent to which this component is reproduced by a standard regional implementation of a model such as the EMEP MSC-W model depends on the concentrations of oxidant drivers, such as CO, used as boundary concentrations. The standard setup of the EMEP4UK model uses prescribed boundary concentrations for CO (with a latitudinal gradient); these long-range chemical influences are another reason why a global version of the model is required. Other models such as GEOS-Chem and CHIMERE often pick up their boundary concentrations from global model outputs, and if these include BB emissions should take account of this contribution. In contrast to NH₄NO₃, the model output indicates that the UK SOA conditional on BB is formed through the oxidation of pyrogenic VOC emissions, rather than through oxidation of locally emitted VOCs via a mechanism similar to that underpinning the BB-induced NH₄NO₃ formation .

The above discussion is based on UK averages for the whole year. The PM_2.5(BB) concentrations vary spatially across the UK (Fig. ) and temporally during the year (Fig. ). Largest values of PM_2.5(BB) occur in the southeast (maximum model grid cell annual mean value of 1.3 µgm-3), and lowest values in the northwest (miniumum grid cell annual mean value of 0.66 µgm-3). Meteorology plays a major role in explaining this southeast-northwest gradient, with the majority of contributions to PM_2.5(BB) arising from BB emissions in Regions 1, 2 and 3 to the east of the UK (Fig. ). South, east and central England also have large anthropogenic emissions of NO_x and NH₃, as do other densely populated areas of the UK such as central Scotland (see Figs.  and  in Appendix ). As a result, SIA formation conditional on BB is particularly enhanced in these areas , which contributes to the southeast-northwest gradient and to the superposition of spatial patterns of UK anthropogenic emissions on this gradient.

The greater population-weighted 2019 annual UK mean PM_2.5(BB) of 1.1 µgm-3, compared with the annual mean PM_2.5(BB) of 0.99 µgm-3, shows that larger absolute PM_2.5(BB) exposures coincide with more densely populated areas of the UK (see Fig.  in Appendix ). This is a consequence of both meteorology and higher anthropogenic emissions of NH₃ and particularly of NO_x in highly populated regions, resulting in enhanced NH₄NO₃ formation conditional on BB in the higher populated regions .

With respect to the temporal variabilities in BB contributions to UK PM_2.5, Fig.  shows that the colder months of October to March are generally characterised by low concentrations of PM_2.5(BB) (daily mean values less than 1 µgm-3). For the year of study here – 2019 – this period of low PM_2.5(BB) concentrations is interspersed with episodes of higher concentrations in February, March and April, which can be attributed to sources closer to the UK. Highest PM_2.5(BB) concentrations occur during a prolonged episode in April when daily mean PM_2.5(BB) exceeds 5 µgm-3 for 6 d and contributes between 13 % and 42 % of daily mean PM_2.5 from all sources. The PM_2.5(BB) contribution is superimposed on an already elevated episode of PM_2.5 pollution caused by the easterly air flow conditions that increase long-range transport of PM_2.5 and its precursors from continental Europe into the UK, together with reducing dispersion, as per analyses of previous spring-time episodes . Particulate nitrate in particular is often found to peak in early spring in the UK due to easterly air flow and increased agricultural emissions of NH₃ at this time . The majority of the PM_2.5(BB) component during this episode is associated with BB in the model's European domain (predominantly from eastern Europe and the western areas of Russia also included in that domain). Similar episodic peaks in PM_2.5(BB) occur at the end of February and March 2019, with values exceeding 1 µgm-3 on 13 d, and 2 µgm-3 on 5 d. Notably, there is a larger contribution from BB in the UK here, as revealed in Fig. c, as well as contributions from southern Europe in the February episode. This is consistent with Copernicus Atmosphere Monitoring Service (CAMS) reports of notable BB activity in the UK, northern Spain, southern France, Portugal and southeastern Europe in February 2019 .

Although there are no large variations in the proportions of primary and secondary PM_2.5(BB) during the year (Fig. d), there is a notable trend for the secondary component to consist more of SIA in winter and more of SOA in summer (Fig. e); the lower temperatures in winter shift the NH₄NO₃ equilibrium to the particle phase . Figure b and d show a tendency for the lowest concentrations of PM_2.5(BB) to have a larger proportion of secondary aerosol (confirmed by a scatter plot of daily percentage secondary contribution vs. daily mean PM_2.5(BB), not shown). This is because PM_2.5(BB) concentrations are lowest when the associated BB sources are further removed from the UK receptor region, but the longer transport distances provide more time for secondary chemical transformations.

The warmer months of May to September are characterised by a continuous period of moderately elevated PM_2.5(BB), with daily mean values ranging between 0.4 and 4.1 µgm-3. The mean (±1 standard deviation) daily value over this period is 1.4 ± 0.7 µgm-3. This occurs at a time of lower total PM_2.5 concentrations (all sources), resulting in a contribution of PM_2.5(BB) ranging between 6 and 36 %. In this period, the contribution from SIA is lower due to the increased NH₄NO₃ dissociation constant at warmer temperatures . There is a higher contribution of SOA relative to the primary component due to increased oxidation and emission of VOCs at higher temperatures and greater sunlight. The majority of this PM_2.5(BB) is attributed to BB outside the model's European domain, with larger contributions from Regions 1 (2019 Siberian wildfires; ), 3 and 4 of Fig. .

The extended period of elevated PM_2.5(BB) during the warmer months dominates the annual mean values in Table . The largest contributions to PM_2.5(BB) are ascribed to BB in Regions 1 (Russia), 2 (Europe), 3 (Asia excluding Russia) and 4 (boreal North America), with respective contributions of 43 %, 19 %, 15 % and 11 % (Fig. ). Only 3 % is attributed to BB within the UK. Southern hemispheric BB makes negligible contribution to UK PM_2.5(BB) (< 2 %). Approximately 5 % of PM_2.5(BB) is attributed to BB in the previous year (2018), for which the NRxBB experiments were not performed due to the computational expense involved. However, whilst its geographic origin has not been identified, it provides useful information about the spin-up time of BB-related species, which Fig. f shows is approximately 3 months. Although only global model runs were used for the NRxBB experiments – and therefore the exact percentages shown in Fig.  would likely differ slightly if these had been followed by additional European and UK nesting – these experiments nevertheless provide a good indication of the relative contribution of BB emissions in Regions 1–8 to UK PM_2.5(BB) values.

Figure a compares the annual total global BB emissions of PM_2.5 for 2019, split by Regions 1–8 (Fig. ), with the emissions from all other years from 2012 to 2023 according to the FINNv2.5 dataset. The spatial distribution of BB emissions for 2019 is shown in Fig.  in Appendix . Although 2019 has the highest global emissions in the 2012–2023 period, this is driven by anomalously high emissions in the southern hemisphere (especially Region 8 of Fig.  – the Australian “Black Summer” of 2019–2020; ), which our modelling shows do not impact the UK.

Figure b provides an estimate of how much PM_2.5 in the UK would likewise have been conditional on BB emissions from each Region 1–8 in each of the other years from 2012 to 2023, based on the assumption that the source-receptor relationships calculated for 2019 are applicable also to the other years. Annual BB emissions of PM_2.5 from each region were weighted according to the impact of the 2019 BB emissions from that region on UK PM_2.5(BB) in 2019, using the following method. First, the multiplication factor required to convert the 2019 BB emissions of PM_2.5 (Fig. a) into the UK 2019 annual mean PM_2.5(BB) for Regions 1–8 was calculated. This was applied to the annual BB emissions of PM_2.5 for 2012 to 2023 to give the corresponding values in Fig. b. The component attributed to BB in the previous year (grey stack in Fig. b) was obtained by calculating a separate multiplication factor, relating the component of 2019 UK annual mean PM_2.5(BB) attributed to BB in 2018 (grey stack in the 2019 bar in Fig. b) with the estimated UK annual mean PM_2.5(BB) for 2018 (excluding the remaining component from 2017) (non-grey stacks in the 2018 bar in Fig. b). This factor was then applied to each year, n - 1, between 2012 and 2022, to give the component of UK annual mean PM_2.5(BB) in year n attributed to BB in year n - 1. The value for the year 2012 is an average of the years 2013–2023, as this is the earliest year for which the BB emissions data used here were available.

This methodology makes the assumptions that: (i) most importantly, annual source-receptor relationships hold across each year, i.e. the combination of the locations and times of the BB emissions and the long-range meteorological transport in other years is similar to that in 2019; (ii) emissions of other species from BB, for example CO, are proportional to the trends in PM_2.5 emissions; (iii) anthropogenic emissions remain sufficiently similar across the time period that variations in oxidant fields and secondary aerosol formation depend principally on changes in magnitudes of BB emissions.

Whilst the values in Fig. b include these assumptions, the figure provides an indication of the contributions of BB emissions globally to UK PM_2.5 in all these other years without running an unfeasibly large number of sensitivity experiments. The 2023 estimate for Region 4 (0.5 µgm-3) can be validated by comparison to literature values of European PM_2.5 exposure from the 2023 Canadian wildfires . It is within the 95 % confidence interval of 0.32–0.50 µgm-3, providing confidence that the assumptions made here are not unreasonable.

The mean contribution calculated across these years is 0.87 ± 0.17 µgm-3, where the uncertainty range is the associated standard deviation of the annual values. This weighting method suggests that 2019 is not an exceptional year for UK PM_2.5(BB) (within 1 standard deviation of the mean), despite this year having high BB emissions globally. This is because inter-annual variability of UK PM_2.5(BB) is dominated by variability in northern hemispheric BB emissions, particularly Regions 1, 2 and 4 (Fig. ), which are not exceptionally high in 2019. The contribution of BB in Region 4 (boreal North America), in particular, is expected to be significantly larger in recent years, with intense wildfire activity in 2023 , 2024 and 2025 .

This study uses a single model – a novel, globally nested version of the EMEP4UK model. Model output will vary with the associated chemical and deposition schemes and meteorological model used, and the spatial resolution of the global model run. It will also vary with the choice of anthropogenic and BB emissions datasets (which may have different spatial patterns, timings and magnitudes of emissions). For example, FINNv2.5 has generally larger emissions than other BB datasets such as FINNv1.5 , GFED4 or Global Fire Assimilation System (GFAS)v1.2 . Choice of the BB emissions dispersion scheme is also an important factor, particularly on a regional scale, though this has been found to be less important when considering long-range transport and longer time-scales , as are being considered in this work. These caveats apply to any similar study using an atmospheric model.

The comparison between modelled and measured PM_2.5 components for the model setup used here is discussed in Appendix . Although, on an annual-mean scale the model overestimates total PM_2.5 compared to measurements, the predominant contributor to this is the overestimation of sea salt, which has no influence on this study. It is not possible directly to validate model results of PM_2.5(BB) and its components because measurement-based source apportionment approaches cannot distinguish between domestic wood burning and open BB, as their levoglucosan and potassium marker compounds are common to both sources. Measurements also cannot distinguish the portion of inorganic NH₄NO₃ that depends on the BB impact on atmospheric oxidants. This illustrates the advantages of using ACTMs to reveal the complex relationship between source and receptor regions which measurements alone cannot.

While the specific numerical values of model output presented here are inevitably subject to much uncertainty, the use of a long-standing and well-validated ACTM and internationally-accepted input datasets provides confidence that these findings are broadly correct.