2.1 Atmospheric chemistry transport modelling

The Fine Resolution Multi-pollutant Exchange (FRAME) is an atmospheric chemistry transport model (ACTM) used to calculate annual deposition of reduced and oxidised nitrogen (N) over the United Kingdom. The model is fully described elsewhere (Aleksankina et al., 2018; Dore et al., 2012, 2016; Vieno et al., 2010; Singles et al., 1998), and only the relevant information for this work is reported here. The domain of the model covers Europe at 50 km × 50 km resolution to provide the boundary conditions for the UK model domain with a grid resolution of 1 km × 1 km. The UK model domain is represented by the British National Grid (EPSG:27700) projected coordinate system. A column of air with a depth of 2500 m is used to represent the relevant atmospheric processes. The column of air is advected across the model domain from all edge grid points and all wind directions with an angular resolution of 1^∘. Figure 1 shows the 1 km × 1 km UK model domain – which captures both the UK and the Republic of Ireland to allow for high-resolution modelling of the closest neighbouring territory – in the European context. Further figures in this work did not show lines of latitude or longitude.

Figure 1The UK FRAME modelling domain, shown by a red rectangle, within which 1 km × 1 km estimates of N deposition are made. The inset shows the context within Europe and lines of latitude and longitude are also shown, while the inputs and outputs of the model are in the British National Grid projected coordinate system.

Emission of gaseous pollutants, vertical diffusion, chemical transformation, and wet and dry removal processes take place within the air column. The model has 33 vertical layers with thickness varying from 1 m at the surface to 100 m in the upper layers. The model requires input data of both diffuse and point source emissions of ammonia (NH₃), oxides of nitrogen (NO_x) and sulfur dioxide (SO₂) (Vieno et al., 2010).

FRAME uses land-cover-specific deposition velocities to generate dry deposition for up to five land cover categories: woodland, low-growing semi-natural vegetation, improved grassland, arable and urban (Land Cover Map 2015; Rowland et al., 2017). The model uses different scavenging coefficients for soluble gases and particles and assumes constant drizzle for calculation of wet deposition. An annual precipitation map (Tanguy et al., 2019; Walsh, 2012) is used to drive the spatial variation in wet removal rate.

The FRAME model used for this work uses long-term radiosonde mean wind speed (Dore et al., 2006) for all the years included here (1990–2017). The wind frequency is derived from modelled data from the Weather and Research Forecast model (Skamarock et al., 2019). The wind frequency used here is kept constant to a 2001–2012 mean for the years 1990–2001, and the specific years afterwards (2001–2017).

The FRAME model, for both the European and British Isles domains, was run for each year from 1990 to 2017, using the corresponding emission and wind–rainfall data. The land cover was kept constant throughout. The FRAME model version used was 9.15.0.

2.2 Emissions data

2.2.1 Data sources

Input data were extracted and processed from the most recently available national emission inventory submissions from both the UK and the Republic of Ireland (EMEP, 2019; E-PRTR, 2019; NAEI, 2019). Emissions for the European domain were taken from Convention on Long-Range Transboundary Air Pollution (CLRTAP) submissions (EMEP, 2019). For agricultural NH₃ emissions, the latest set of annual emission maps from 1990–2017 was used, as derived for the UK's national atmospheric emission inventory. This inventory work utilises annual activity data at the holding level from the devolved authorities in the UK, i.e. Defra (England), the Scottish Government (Scotland), Welsh Assembly (Wales) and Daera (Northern Ireland) (see Carnell et al., 2019a, for details).

Emissions data are routinely made available via sectors (e.g. energy production) and to create a consistent structure for all data sources. NO_x and SO₂ emissions were restructured into the 11 Selected Nomenclature for sources of Air Pollution (SNAP) sectors (Table 1), developed by the European Topic Centre on Air Emissions (ETC/AE). Given the dominance of agriculture in NH₃ emissions, the FRAME model requires agricultural data to be split into livestock fertiliser emissions, with all non-agricultural sources as one sector (see Sect. 2.1.3).

The SNAP system is used in the UK for the annual updates to the National Atmospheric Emissions Inventory (NAEI, 2019). This corresponds to the main area of interest for the deposition outputs, and the Irish and wider European emissions were reformatted to match that reporting system. Whilst the UK, Ireland and the collated European data all use the Nomenclature for Reporting system (NFR, ca. 240 sectors – EEA, 2019), the UK collate the fine-resolution categories into SNAP sectors whereas the latter two report via the aggregated Generalised/Gridded Nomenclature for Reporting (GNFR). Table 1 also shows how these two aggregated reporting systems broadly relate to each other.

Table 1Selected Nomenclature for sources of Air Pollution (SNAP) sectors for emissions inventory reporting as outlined by CORINAIR, alongside the Generalised/Gridded Nomenclature for Reporting (GNFR) sectors (broadly matched).

Download Print Version | Download XLSX

It is worth noting that emissions data for “international shipping” and “aviation cruise” do not count within a specific national inventory but are reported into a “pooled” total by all countries. Separate totals for national shipping, airports, and the take-off and landing of aircraft are reported on a country basis. Finally, emissions data should ideally be translated between the aggregated classification systems using the NFR codes upon which they are built (which still has some one-to-many relationships) but spatial data are not available at this level, and therefore the aggregated spatial data should not be broken down in an attempt to make the NFR level data.

2.2.2 Point and diffuse emissions of NO_x, SO₂ and NH₃

NH₃, NO_x and SO₂ emission inputs were produced for the years 1990 to 2017, for both diffuse and point source emissions. Diffuse sources are those deemed to be areal, non-exact locations such as agriculture, vehicles, population-related sources, etc. Point sources can be located by exact coordinates, for example the actual chimney/exhaust stacks of power stations and industry (Vieno et al., 2010). Point source information in the UK is nearly (but not totally) exclusive to energy generation and industry.

Figure 2 shows an overview of the processes to combine the various spatial and tabulated emissions data that are required for the 28 annual model runs. There are some important methodological details, for both diffuse and point emissions, worth noting. In the UK, diffuse data are produced and published for 11 SNAP sectors for the latest emissions inventory year, superseding any previous data. This is principally due to the fact that every year in the inventory compilation, minor to major changes are made to the way the data are compiled – this could be changes to emission factors with the latest research being incorporated or how underlying spatial methods and datasets are updated. While the non-spatial data are “back-cast” to 1990 (or earlier, depending on the pollutant), the maps are not currently updated as a time series. Consequently, it is unwise to compare previous years' gridded emissions surfaces to the latest available. For this reason, at the time of publication, only the latest 2017 emissions maps were used in the UK for the entire time series and were scaled back through the time series using the annual tabulated NFR totals, for SO₂, NO_x and non-agricultural NH₃. For agricultural NH₃, the latest mapped time series (using annual livestock and crop data) was used (Carnell et al., 2019a). For point sources – which in the more recent data number in the thousands – some earlier data were obtained back to 1990 but only for a subset of major polluters and not for all years (missing years were linearly interpolated). For the very largest emitters, information (when known) regarding the stack/chimney height, stack/chimney diameter and emission exit velocities is also used by the model to create plume characteristics. It is the non-coordinate parameters that are important in determining to what height into the atmosphere the emissions travel, and therefore what subsequent chemical interactions occur, which is important for the deposition modelling.

Table 2Deposition outputs as provided in this dataset from the Fine Resolution Multi-pollutant Exchange (FRAME) atmospheric chemistry transport model.

Download Print Version | Download XLSX

Figure 2Visualised methodology of steps to create inputs for the Fine Resolution Multi-pollutant Exchange (FRAME) atmospheric chemistry transport model: rectangle with corners missing (solid border) denotes spatial data, rectangle with corners missing (dashed border) denotes tabulated data, rectangle with rounded corners denotes process and oval denotes model.

Download

Emissions from the Republic of Ireland influence the deposition of N species in the UK. To allow for similarly high-resolution emissions inputs, the outputs from the National Mapping of Greenhouse Gas and Non-greenhouse Gas Emissions Sources project (MapEire, 2019; Pjeldrup et al., 2018) were used in a similar manner to the latest emissions surfaces produced for the UK in the NAEI. The MapEire project produced 1 km × 1 km resolution gridded emissions for all GNFR sectors for the year 2016, which were scaled to other years by the totals reported to the CLRTAP by the Republic of Ireland. These surfaces were then transformed to SNAP sectors (see Table 2) to be joined to the UK data. One important difference to note is that the MapEire gridded data include all sources of emissions, including point sources (the UK data do not). Therefore, the major emitting point sources, as reported to the European Pollutant Release and Transfer Register (E-PRTR, 2019), were extracted for NO_x and SO₂ for all available years back to 1990 (gaps were linearly interpolated). To conserve totals, Irish point values were removed from the Irish total gridded surface by subtracting the point value from the grid cell in which it was located, with any surplus emissions removed from the surrounding eight cells on an equal-share basis (if required). This created a diffuse surface and a point source input, consistent with the UK data.

Table 3Four measurement networks used within the UK Acidifying and Eutrophying Atmospheric Pollutants (UKEAP) network, along with the 10 compounds used to evaluate the atmospheric modelling.

Download Print Version | Download XLSX

A consistent time series of UK agricultural NH₃ emission estimates was created at a 1 km × 1 km grid resolution for the years 1990–2017. These high-resolution agricultural NH₃ emission maps are produced annually for the NAEI, using an agricultural emission model jointly developed by the UK Centre for Ecology & Hydrology, Rothamsted Research, ADAS and Cranfield University. The emission model uses annual agricultural census data (e.g. livestock numbers and crop areas – see Carnell et al., 2019a) at the holding level, agricultural practice information (e.g. fertiliser application rates, stocking densities) and emission source strength data from the UK emissions inventories for agriculture (e.g. Brown et al., 2019; Richmond et al., 2019). Emission estimates are output for each individual emission source at a 10 km × 10 km grid resolution, which are spatially disaggregated to a 1 km × 1 km grid resolution using land cover data (Rowland et al., 2017) and methods outlined in Dragosits et al. (1998), Hellsten et al. (2008) and Carnell et al. (2019a). Emissions sources are numerous and include grazing, storage, spreading and housing for cattle, pigs, poultry, sheep and minor livestock (plus all sub-types) as well as differing fertiliser applications for varying crop and grass types.

Figure 3Locations of sites in four measurement networks, across four periods of the time series in this study: Acid Gases & Aerosol Network (AGANET), National Ammonia Monitoring Network (NAMN), Rural Background NO₂ (NO₂NET) and Precipitation Network (PrecipNet). Some sites from different networks are co-located, and therefore not all dots represented in the table are visible in the maps.

2.3 Outputs

Outputs from the model as presented in this dataset are the annual values of wet and dry deposition of reduced nitrogen (“NH_x”) and wet and dry deposition of oxidised nitrogen (“NO_y”) as a weighted mean of all land cover types within a given cell, as well as vegetation-specific values for both forest and moorland – Table 2 provides more detail.

Deposition data are provided on a 1 km × 1 km resolution surface, using the British National Grid projection (same domain as the emission files) for UK terrestrial cells (no. of cells = 259 436). Other land cover types used in the calculations (but not output) are arable, urban and improved grassland.

2.4 Evaluation

2.4.2 Evaluation metrics

It is unlikely for an ACTM to perfectly reproduce reality due to errors in, but not limited to, input data, model physics and chemistry schema, uncertainty in meteorological data, and the random effects of the real world. However, using methods outlined in Chang and Hanna (2004), several statistical metrics may be used to evaluate the agreement between the modelled predictions and the real-world observations: fraction of predictions within a factor of 2 of observations (FAC2), the fractional bias (FB), the normalised mean square error (NMSE) and the geometric mean bias (MG). These metrics are defined in the following way:

$$\begin{array}{}\text{(1)}& {\displaystyle }& {\displaystyle }\mathrm{FAC}\mathrm{2}=\mathrm{fraction}\mathrm{of}\mathrm{data}\mathrm{that}\mathrm{satisfy}\mathrm{0.5}\le {\displaystyle \frac{C_{\mathrm{p}}}{C_{\mathrm{o}}}}\le \mathrm{2.0},\text{(2)}& {\displaystyle }& {\displaystyle }\mathrm{FB}={\displaystyle \frac{\left(\stackrel{\mathrm{‾}}{C_{\mathrm{o}}}-\stackrel{\mathrm{‾}}{C_{\mathrm{p}}}\right)}{\mathrm{0.5}\left(\stackrel{\mathrm{‾}}{C_{\mathrm{o}}}+\stackrel{\mathrm{‾}}{C_{\mathrm{p}}}\right)}},\text{(3)}& {\displaystyle }& {\displaystyle }\mathrm{NMSE}={\displaystyle \frac{{\stackrel{\mathrm{‾}}{\left(C_{\mathrm{o}}-C_{\mathrm{p}}\right)}}^{\mathrm{2}}}{\stackrel{\mathrm{‾}}{C_{\mathrm{o}}}\cdot \stackrel{\mathrm{‾}}{C_{\mathrm{p}}}}},\text{(4)}& {\displaystyle }& {\displaystyle }\mathrm{MG}=\exp \left(\stackrel{\mathrm{‾}}{\ln C_{\mathrm{o}}}-\stackrel{\mathrm{‾}}{\ln C_{\mathrm{p}}}\right),\end{array}$$

where C_o represents measured observations and C_p represents model predictions, the former being paired with the latter spatially. A perfect reproduction of measurement data would have FAC2 = 1, FB = 0, NMSE = 0 and MG = 1.

FAC2 is a robust measure of performance, not overly influenced by outliers, indicating the proportion of modelled–measured pairs falling within a factor of 2 of each other. FB is a linear metric that measures the mean systematic bias of the model and may have predictions out of phase with measurements but still return a value of 0 due to cancelling errors. NMSE is a measure of mean relative scatter and reflects both systematic and random errors. Finally MG is also a measure of mean systematic bias, but is less influenced by extreme values as it is a logarithmic metric (see Chang and Hanna, 2004, for more detail). Hanna and Chang (2012) suggest that a model should satisfy at least 50 % of the criteria used (two of four in this study), while the acceptability criteria for each metric are as defined in Theobald et al. (2016): FAC2 > 0.5, |FB| < 0.3, NMSE < 1.5 and 0.7 < MG < 1.3.

Table 4Evaluation metrics of modelled concentrations of six nitrogen compounds in gas, aerosol and precipitation in the UK for 2017 (see Table 3 for definitions). Bold numbers represent where that metric has been satisfied (see Sect. 2.4.2 for metric definitions).

n/a stands for not applicable.

Download Print Version | Download XLSX

3.1 Emissions

In the UK, stricter air pollution policies, improving technology and changes in fuel use, have all contributed to the reduction of emissions. Initially, mitigation strategies concentrated on SO₂ emissions, but the focus was extended to nitrogen compounds such as NO_x (as well as VOCs) in an attempt to abate acidification and, latterly, to NH₃ (Grennfelt and Hov, 2005; Carnell et al., 2019b). Within the model domain, emissions of NH₃ and NO_x have decreased by ∼12 % and ∼64 % respectively from 1990 to 2017 (Fig. 4).

Figure 4Emissions (in kt) of ammonia (NH₃), nitrogen oxides (NO_x) and sulfur dioxide (SO₂) in the model domain, covering the UK and Ireland, from 1990 to 2017, split into the main broad reporting sectors.

Download

Much of the decrease in emissions of NO_x in the UK has been driven by the decline of coal use in power stations (95% decrease in emissions over the time series) and the improvement and modernisation of petrol combustion in road transport (98 % decrease in emissions over the time series). Decreases in NO_x have been offset by increases in emissions from DERV (diesel fuels) and aviation fuels. With regard to NH₃ emissions, which are dominated by agriculture, changes in farm practices have seen a patchwork of decreases and increases to various emissions sources, with a generally decreasing trend that has plateaued from ca. 2001. It is the non-agricultural sources, however, that have shown marked increases from 1990 to 2017, including those activities associated with the circular economy: anaerobic digestion, composting of organic materials, application of sewage sludge to land and the combustion of biomass for industry (total increase: ∼5 to ∼26 kt). Finally, SO₂ emissions have reduced by ∼94 % in the same time period (mean of ∼5 % yr⁻¹), which is a direct result of the decline of coal use, especially in power stations, and restrictions being placed on the sulfur content of various fuels.

As all three pollutants are reactive in the atmosphere, differing rates of emissions reductions have varying effects on chemical reactions and subsequent deposition. Changes to emissions over time vary in space and so does, therefore, N deposition (Fowler et al., 2007).

3.2 Model evaluation

Scatter plots of the modelled predictions vs. measurements in 2017, for data collected in Table 3, are shown in Fig. 5. The associated performance metrics are given in Table 4.

Figure 5Evaluation of modelled (x axis) and measured (y axis) concentrations of six nitrogen compounds in the UK for 2017 (see Table 3 for definitions). The solid black line represents a 1:1 relationship, and the dotted lines represent a factor of 2 (FAC2) relationship; the blue dashed lines are linear regressions.

Download

For the latest year included in this study, all six N forms in Table 4 comply with the FAC2 metric and all six comply with the recommended NMSE limit of 1.5. FB and MG are met with less success, though all are close to the recommended thresholds, aside from NH₄ in aerosol (which contributes to dry deposition). FB and MG measure the systematic bias of the model, and for both NH₄ and NO₃, the model slightly under-predicts the aerosol phase and over-predicts the aqueous phase. Not shown in Fig. 5 and Table 4 is the evaluation of HNO₃ in gas, which similarly fulfils recommendations for FAC2 (0.54) and NMSE (0.48), but not for |FB| (0.48) or MG (0.56). Modelled predictions were also evaluated for 2016, with all seven compounds achieving 50 % compliance, with NH₃ in gas, NO₂ in gas and HNO₃ in gas satisfying all four. It is not fully known why 2016 achieves better evaluation scores; it may be random variations in real-world conditions, but one reason may be that 2017 was a relatively warm year by annual mean temperature standards (and fourth warmest on record for England only). It is known that NH₃ emissions are affected by temperature (e.g. Hempel et al., 2016; Sutton et al., 2013; Riddick et al., 2018), and, as temperature fluctuations are not factored into the model or into the underlying emission inventories, this may have driven higher spring–summer emissions of NH₃ and therefore higher dry deposition episodes.

For context, Carslaw (2011) undertook a model inter-comparison exercise for the UK Department for Environment, Food & Rural Affairs (Defra), with a specific focus on deposition from the CMAQ, EMEP4UK, FRAME, HARM and NAME models. Respectively, those models (at the time) were run at resolutions of 12, 5, 5, 10 and 12 km. In Carslaw (2011), the models performed with a similar correlation coefficient (“r”) for all N compounds, aside from NH₄ and NO₃ in precipitation, for which the 2017 model run in this study had a weaker correlation (0.51–0.61 compared to 0.7–0.88).

The evaluation for 2017 would indicate that total wet deposition was over-predicted and total dry deposition was under-predicted. To provide further context and evaluation, measurement data were obtained for three previous years spanning the time series at equal intervals; 1990, 1999 and 2008. Data for historic years, especially prior to ∼1998, are limited, and so scatter plots in Fig. 6 show the relationship between modelled predictions and measured data for four N compounds while Table 5 shows the associated performance metrics.

Figure 6Evaluation of modelled (x axis) and measured (y axis) concentrations of four nitrogen compounds in the UK for 1990, 1999 and 2008 (see Table 3 for definitions; no NH₃ gas data exist for 1990). The solid black line represents a 1:1 relationship, and the dotted lines represent a factor of 2 (FAC2) relationship. The blue, green and red dashed lines are linear regressions.

Download

Table 5Evaluation metrics of modelled concentrations of six nitrogen compounds in the UK for 1990, 1999 and 2008 (see Table 3 for definitions). Dashed lines represent no available data. Bold numbers represent where that metric has been satisfied (see Sect. 2.4.2 for metric definitions).

Download Print Version | Download XLSX

All N forms for which data were available in 1990, 1999 and 2008 satisfy at least two of the four evaluation metrics, with four gas and aerosol N compounds fulfilling all metrics in 2008. An example of the benefit of multiple evaluation metrics is shown in Fig. 6 when looking at NO₂ and NH₃ in gas in 2008. Both have very low FB values (indicating very low mean bias) due to the cancelling effect around the 1:1 line, but the scatter of predictions to measurements of NH₃ is clearly much larger than for NO₂. Information of the NMSE and the FAC2, plus visual inspection of the plots, helps to illustrate that NH₃ has a larger error than NO₂.

From the perspective of model sensitivity and/or uncertainty, there were no further runs made with adjusted emissions inputs or adjusted deposition parameters within this study. However, Aleksankina et al. (2018) employed statistical techniques to obtain uncertainty estimates of the FRAME model, representing model runs with a ±40 % variation range for the UK emissions of SO₂, NO_x and NH₃. They found that the sensitivity of concentrations of primary precursors NO_x and NH₃, plus the deposition of N, was dominated by emissions. However, concentrations of secondary species such as particulate ${\mathrm{NO}}_{\mathrm{3}}^{-}$ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ were more geographically dependent.

3.3 Nitrogen deposition

Grid average N deposition – NH_x wet and dry, NO_y wet and dry – is plotted in Fig. 7 at a 1 km × 1 km resolution over the UK terrestrial surface, for 2017. The total N deposition over the UK is 278.3 kt N ($\stackrel{\mathrm{‾}}{x}$ = 10.7 kg N ha⁻¹ yr⁻¹, SD = 4.5 kg N ha⁻¹ yr⁻¹), with a maximum of 74.3 kg N ha⁻¹ yr⁻¹. Such high deposition values are reasonably rare (no. of cells > 30 kg N ha⁻¹ yr${}^{-\mathrm{1}}=\mathrm{118}$; no. of cells > 50 kg N ha⁻¹ yr${}^{-\mathrm{1}}=\mathrm{8}$) and are a direct result of the increased resolution of the model, when compared to the maximum deposition of 5 km × 5 km resolution N deposition.

Figure 7Four forms of nitrogen (N) deposition over the UK terrestrial surface in 2017 at 1 km × 1 km resolution, for grid average land cover: wet and dry deposition of reduced N (NH_x) and wet and dry deposition of oxidised N (NO_y) (kg N ha⁻¹ yr⁻¹).

Figure 8Four forms of total nitrogen (N) deposition over the UK terrestrial surface from 1990 to 2017, for grid average land cover: total wet and dry deposition of reduced N (NH_x) and wet and dry deposition of oxidised N (NO_y) (kt N yr⁻¹).

Download

The two wet deposition surfaces in Fig. 7 exhibit smoother spatial distributions and less heterogeneity (compared to dry deposition), a reflection of the precipitation surface across the UK, and they constitute ∼67 % of the total deposition. Wet deposition is nearly always of a longer-range than dry deposition, due to the transport in more elevated atmospheric layers, but some enhanced local washout around strong sources is also represented. This longer range transport acts as a smoothing effect on the deposition field due to the increased distance from the emission source. It should be noted that, as shown in Figs. 5 and 6, deposition in precipitation of both NH₄ and NO₃ is consistently overpredicted by the model throughout the time series. Upland areas are subject to the highest values of wet deposition, and most of the highest value cells between 25 and 50 kg total N ha⁻¹ yr⁻¹ are dominated by wet deposition. Dry deposition of NO_y, as modelled in this study, is the smallest contributor to total N deposition (∼14 %) and is dominated by NO₂ and HNO₃, which both follow their respective concentration fields closely (RoTAP, 2012). Dry deposition of NO₂, therefore, is largest in urban areas and close to road networks such as motorways. Dry deposition of NH_x, ∼20 % of total N deposition, is a highly heterogeneous surface with the highest values associated with areas of intensive livestock farming (including beef, dairy, pigs and poultry). Gaseous NH₃ has a short atmospheric lifetime and so is deposited close to the sources. The very highest values of total N deposition (>50 kg N ha⁻¹ yr⁻¹) are all dominated by dry deposition of NH_x and are located near high agricultural emissions. An important factor in the deposition of NH_x is the presence of oxidised SO₂, sulfuric acid (H₂SO₄), to form the aerosol (NH₄)₂SO₄. With decreasing SO₂ available to create H₂SO₄, more NH₃ is deposited within short distances as dry deposition. This effect is further enhanced by the increased rate of dry deposition of the available SO₂, a result of the increase in the concentration ratio of NH₃ : SO₂, which increases surface water pH, which further limits the available SO₂ to oxidise to H₂SO₄ (Baek and Aneja, 2004; Fowler et al., 2007; RoTAP, 2012; Tan et al., 2020).

Looking at the pattern of modelled N deposition from 1990 to 2017, Fig. 8 shows a steady decrease in wet and dry NO_y deposition, a slow decrease in wet NH_x deposition, and no apparent decrease in dry NH_x deposition. The latter is due to the change in atmospheric chemistry with declining sulfur emissions due to successful policy implementation. Total N deposition over the UK has decreased from 465 to 278 kt N, though no significant reductions in the total have occurred since around 2011.

Total oxidised N deposition has decreased by ∼56 % from 1990 to 2017, while reduced N deposition has decreased by ∼19 %. This reflects the larger emissions reductions achieved for NO_x than for NH₃ from 1990. Mean deposition values for all four N forms have changed in a similar fashion to their respective totals from 1990, but the standard deviation across all 5 km × 5 km cells for oxidised N (both wet and dry) has decreased over time, possibly due to the heavy reductions in emissions sources such as road traffic and power stations, which previously created very high localised dry deposition. Figure 9 shows every year of total N deposition from 1990 to 2017 and highlights the non-linear relationship between decreasing emissions and deposition.

Figure 9Spatial distribution of total nitrogen (N) deposition over the UK terrestrial surface, 1 km × 1 km resolution, from 1990 to 2017, for grid average land cover (kt N yr⁻¹).

Some of the areas with the highest N deposition in later years are remote upland areas, which are principally affected by longer-range wet deposition (and transboundary deposition) and have seen much lower relative decreases in N deposition than lowland areas such as southeast England. NO_x emissions have decreased by ∼64 % across the time series and resulted in wet and dry NO_y deposition decreases of ∼48 % and ∼66 %, respectively. This illustrates the non-linear processes involved with the chemical processing of NO_x emissions, in particular the resulting concentrations of NO₃ in precipitation which are not decreasing at the same rate as gas and/or aerosol forms of oxidised N (see Fowler et al., 2007; Sickles and Shadwick, 2015; Feng et al., 2020). It must be recognised again, however, that the model over-estimates wet deposition of N to a degree.

As a result of emissions changes and non-linear chemistry, estimates of modelled dry deposition have decreased as a percentage of the total N deposition ($\mathrm{1990}=\sim \mathrm{38}$ %, $\mathrm{2017}=\sim \mathrm{33}$ %) (see Fig. 10). This dataset models wet deposition as the dominant source of total N deposition.

Figure 10Fraction of the total nitrogen (N) deposition over the UK terrestrial surface for four forms of nitrogen (N) deposition, for grid average land cover, from 1990 to 2017: total wet and dry deposition of reduced N (NH_x) and wet and dry deposition of oxidised N (NO_y).

Download

As a result of the large decreases in NO_x emissions, and fewer regulations on most NH₃ emission sources in the UK compared to NO_x, reduced N is now the major component of N deposition. In this dataset, the proportion of dry deposition has moved from being dominated by oxidised N in 1990 (∼65 %) to reduced N in 2017 (∼59 %). This has resulted in a highly heterogeneous spatial distribution of N deposition that is more reflective of both agricultural practice and rainfall patterns.