Surface–atmosphere exchange of inorganic water-soluble gases and associated ions in bulk aerosol above agricultural grassland pre- and postfertilisation

. The increasing use of intensive agricultural prac-tices can lead to damaging consequences for the atmosphere through enhanced emissions of air pollutants. However, there are few direct measurements of the surface–atmosphere exchange of trace gases and water-soluble aerosols over agricultural grassland, particularly of reactive nitrogen compounds. In this study, we present measurements of the concentrations, ﬂuxes and deposition velocities of the trace gases HCl, HONO, HNO 3 , SO 2 and NH 3 as well as their associated water-soluble aerosol counterparts Cl − , NO − 2 , NO − 3 , Registrator


Introduction
As the demand for food production grows in line with an increasing global population, so too does the development of intensive agricultural practices. These can have deleterious impacts on the environment and human health (Godfray et al., 2010;Foley et al., 2011), particularly through the emission of trace gases and the formation of airborne particles generated by their reactive chemistry. It is therefore important that measurements be made of the surface-atmosphere exchange of trace gases and associated aerosol compounds to quantify the emissions from -and deposition to -land used for agriculture, in order to quantify the impact of agricultural activities on the atmosphere and environment. Such studies can inform the development of abatement strategies and legislation designed to control emissions from agricultural activities. This also provides important process understanding to represent better the dry deposition processes in chemistry and transport models used to predict air quality and climate change.
Of particular importance to the surface-atmosphere exchange over agricultural land is the formation of atmospheric reactive nitrogen (N r ) compounds. These include the gases ammonia (NH 3 ) and nitrous acid (HONO), the latter of which, together with nitric acid (HNO 3 ), also derives from the oxidation of nitrogen oxides (NO x ) emitted by combustion sources. NH 3 and HNO 3 , and their aerosol equivalents NH + 4 and NO − 3 , are the primary contributors to atmospheric reactive nitrogen (N r ) dry deposition (Andersen and Hovmand, 1999). The majority of NH 3 emissions originate from agricultural sources, either from direct point sources from the application of N-containing fertilisers or from long-term sources from livestock (Behera et al., 2013). The use of urea as a fertiliser is associated with particularly large losses of NH 3 after application, due to the action of the urease enzyme present in soil, which leads to NH 3 volatilisation (Suter et al., 2013). Ferm (1998) estimate that fertiliser losses as NH 3 average 14 % of the N applied. Nitrogen losses from animal waste present on grassland used for sheep grazing have also been observed (Cowan et al., 2015). While NH 3 is predominantly deposited close to the source, resulting NH + 4 aerosol can be transported over large distances.
HNO 3 is primarily formed from the oxidation of nitrogen oxides (NO x ), which are principally anthropogenic in origin but also have a soil biogenic origin (Pilegaard, 2013). HNO 3 is extremely water soluble and is rapidly removed from the atmosphere through deposition or by gas-particle interactions, leading to a high deposition velocity. The gas-phase equilibrium reaction of HNO 3 with NH 3 , which is dependent upon temperature and relative humidity (RH) (Mozurkewich, 1993;Robertson et al., 2013), gives rise to ammonium nitrate (Reaction R1). NH 3 + HNO 3 NH 4 NO 3 (R1) The associated condensed-phase components of ammonium (NH + 4 ) and nitrate (NO − 3 ) exist in equilibrium (as ammonium nitrate (NH 4 NO 3 )) with NH 3 and HNO 3 . Higher temperatures and humidity favour the decomposition of NH 4 NO 3 . The emission of these N r species and their subsequent deposition by washout (wet deposition) or uptake on the surface (dry deposition) have high spatial and temporal variability and can have critical impacts on terrestrial and aquatic ecosystems, especially those which are nitrogen limited (Galloway et al., 2003;Fowler et al., 2013).
The interaction of NH 3 with HNO 3 can also lead to overestimation of the HNO 3 deposition rate, as the additional sink for HNO 3 deposition provided for by the reaction violates the theoretical deposition rate modelled on a zero surface resistance model for HNO 3 . The dissociation of NH 4 NO 3 over vegetation can induce an opposite effect, with apparent emissions of HNO 3 occurring with associated high deposition rates for NO − 3 and NH + 4 . The sums of the total ammonium (tot-NH + 4 = NH 3 + NH + 4 ) and of total nitrate (tot-NO − 3 = HNO 3 +NO − 3 ), however, are conservative quantities (Kramm and Dlugi, 1994), and the use of them in the measurement of exchange fluxes can help to account for the NH 3 -HNO 3 -NH 4 NO 3 triad on overall deposition rates.
HONO is similar to HNO 3 in that it can derive from oxidation of NO x precursors. Although it can be formed homogeneously in the atmosphere by the reaction of the hydroxyl radical OH with NO (Reaction R2) (Pagsberg et al., 1997), the rate of this reaction is too slow to account for measured concentrations of HONO. Similarly, the heterogeneous reaction involving the reaction of NO 2 with H 2 O on terrestrial surfaces, while potentially a contributory source to atmospheric HONO, has also been found to be too slow to account for measured concentrations (Kleffman, 2007). HONO is photolysed during daytime, being a primary source of OH radicals depending on the source and sink mechanisms that govern its abundance (Sörgel et al., 2015). However, a growing number of field measurements of non-zero HONO concentrations during the day points to the presence of daytime sources (Acker et al., 2006), including the emissions of HONO from soils (Su et al., 2011;Oswald et al., 2013;Scharko et al., 2015).
As the primary basic gas in the atmosphere, NH 3 also reacts with trace acidic gases, such as hydrogen chloride (HCl) and sulfuric acid (H 2 SO 4 ). The products of these reactions give rise to the aerosols ammonium chloride (NH 4 Cl) and ammonium sulfate ((NH 4 ) 2 SO 4 ), which along with NH 4 NO 3 act as scattering aerosols that alter the Earth's total albedo and contribute significantly to regional and global climate (Fiore et al., 2015). The ammonium salts make a significant contribution to inhalable particulate matter (PM) associated with human health impacts, with NH 4 NO 3 often dominating PM pollution events in northern Europe (Vieno et al., 2014). Ammonium sulfate is particularly long lived, and its transport and subsequent deposition to surfaces such as agricultural soils can affect plant health (van der Eerden et al., 1992) and lower soil pH (Elliott et al., 2006). The dry deposition of the acidic gases themselves can also induce soil acidification, which on agricultural soils can limit the growth of crops through perturbation of the uptake of nutrients. HCl, like HNO 3 , is highly water soluble; is deposited quickly to the surface; and consequently has a high deposition velocity. It can be formed by the reaction of other acidic gases, such as HNO 3 and SO 2 , with sodium chloride found in sea spray (Pio and Harrison, 1987) SO 2 , which is the precursor for H 2 SO 4 in the atmosphere and is primarily anthropogenic in origin, being emitted via the burning of fossil fuels that contain sulfur.
Measurements of trace gases and associated aerosols are, however, restricted by the availability of appropriate instrumentation, complications in their measurement due to their reactivity and water solubility, and the potential interference of gas-particle interactions.
Techniques to measure concentrations and fluxes of these trace gas and associated aerosol components require multispecies quantification, low detection limits and fast temporal resolution. Eddy covariance, the most direct micrometeorological technique for the measurement of trace gas fluxes, requires fast-response sensors that are not available for some species (such as HNO 3 ) or are limited by the time-response and potential for chemical interferences of the inlet (Neuman et al., 1999). While eddy covariance has been used to measure NH 3 concentrations using laser absorption spectroscopy, such as through the use of quantum cascade lasers (QCLs) (Famulari et al., 2004;Zöll et al., 2016), inter-comparisons with more established techniques are still lacking.
The aerodynamic gradient method (AGM) derives fluxes of a tracer from its vertical concentration gradient, which can be obtained from concentration measurements at two or more heights, avoiding the requirement for fast response measurement. Developments in automated wet-chemistry instrumentation have in turn led to the development of the Gradient of Aerosols and Gases Online Registrator (GRAEGOR), a two-point gradient system that measures the concentrations of HCl, HONO, HNO 3 , SO 2 and NH 3 as well as their associated aerosol counterparts Cl − , NO − 2 , NO − 3 , SO 2− 4 and NH + 4 (Thomas et al., 2009). One of the advantages of the modified aerodynamic gradient method is the ability to determine the deposition velocities (V d ) of chemical tracers, provided the flux and concentration at a reference height have been calculated. With the use of the GRAEGOR, which takes measurements of tracers at two heights over 1 h, high-resolution timescale measurements of deposition velocities can be acquired.
Other wet-chemistry instruments have also been developed to measure individual species at one height, such as the Long Path Absorption Photometer (LOPAP), which measures concentrations of HONO with fewer artefacts than the GRAEGOR (Heland et al., 2001). A comparison study between LOPAP HONO measurements and the Gas and Aerosol Collector (GAC) -an instrument which uses similar measurement techniques to the GRAEGOR -was conducted by Dong et al. (2012), but there has not yet been a published comparison between the LOPAP and GRAEGOR in measurements of HONO. Similarly, measurements of trace gases and aerosols above agricultural grassland using the GRAE-GOR are limited, and previous studies above these land sys-tems have been restricted to measurements of a limited number of species within a limited particle size range.
The aim of this study was to use the GRAEGOR to measure concentrations and fluxes of the trace gases HCl, HONO, HNO 3 , SO 2 and NH 3 as well as their water-soluble aerosol counterparts Cl − , NO − 2 , NO − 3 , SO 2− 4 and NH + 4 over agricultural grassland in Scotland during a period in early summer (May-June 2016) that included a fertilisation event using urea pellets. The possible formation of NH 4 NO 3 postfertilisation, a link between aerosol deposition velocity and size (specifically, a proxy for size based on the PM 2.5 /PM 10 ratio from measurements nearby), and the potential ground source formation of HONO are discussed. A further aim of this study was to undertake inter-comparisons between the measurements of HONO by the GRAEGOR and two LOPAP instruments and between measurements of NH 3 recorded by a parallel quantum-cascade-laser eddy covariance system.

Easter Bush site description
The campaign was conducted during the late spring-summer 2016 (21 May-24 June) at the Easter Bush measurement site (3 • 12 W, 55 • 52 N; 190 m above sea level), located 10 km south of Edinburgh, UK. Measurements were made at a 3 m tower situated on the boundary of two intensively managed grassland fields (hereafter referred to as the north and south field) of 16 ha total area, composed principally of Lolium perenne (perennial ryegrass) (Fig. 1). Due to the presence of the Pentland Hills close by to the west, local wind direction is channelled such that SW winds -the predominant wind direction at the site -yield flux footprints over the south field, while NE winds produce flux footprints over the north field.
Both fields are used for year-round (although not continuous) sheep grazing, in rotation with adjacent fields, but the south field also typically has an annual cutting for silage. Mineral fertilisation is carried out twice a year on both fields. During this study, fertilisation of the two fields occurred between 08:00 and 09:00 on 13 June, using urea mineral fertiliser at a rate of 69.9 kg N ha −1 . In preparation for this application, sheep that had been present in the fields since April were removed from the south field on 2 June and removed from the north field on 9 June. Sheep were reintroduced to the north field on 21 June.
Over the years the Easter Bush field site has hosted several long-term measurements of CO 2 , CH 4 and NO 2 and has participated in a number of international projects, such as GRAMINAE (GRassland AMmonia INteractions Across Europe) (Sutton et al., 2009), Greengrass (Soussana et al., 2007) and NitroEurope . It has also supported several individual campaigns of trace gas measurements (Di Marco et al., 2004;Famulari et al., 2004;Jones et al., 2017). In particular, fluxes of NH 3 were measured over an 18-month period (Milford et al., 2001) and the GRAEGOR was operated during a period of manure application (Twigg et al., 2011).

Gradient of Aerosols and Gases Online Registrator
The GRAEGOR (Energy Research Centre of the Netherlands) is a wet-chemistry instrument that measures the concentrations of reactive trace gases (HCl, HONO, HNO 3 , SO 2 and NH 3 ) and water-soluble aerosols (Cl − , NO − 2 , NO − 3 , SO 2− 4 and NH + 4 ) continuously, semi-autonomously and with online analysis at hourly resolution (Thomas et al., 2009;Wolff et al., 2010b). The instrument consists of two sampling boxes placed at two heights (during this campaign, z 1 = 0.6 m, z 2 = 2.4 m), from which concentration gradients and hence fluxes can be derived.
Each sample box contains a horizontal wet rotating annular denuder (WRD) (Keuken et al., 1988) and a steam jet aerosol collector (SJAC) (Khlystov et al., 1995;Slanina et al., 2001) connected in series. Air is drawn through each sample box simultaneously by an air pump at a rate of 16.7 L min −1 , passing first through the WRD, which is continuously coated with a feeding solution of double-deionised water (DDI) of 18.2 M resistance. Trace gases within the laminar airflow are absorbed into the sorption solution which is then fed from the sample box to a detection unit located at ground level. The trace-gas-free air then passes through the SJAC, where particles within the airflow are mixed with steam generated from the DDI water feeding solution, precipitating a supersaturation event causing the water-soluble particles to grow into droplets. The enlarged droplets are separated out of the air stream by a cyclone and fed as a liquid sample to the detection unit. Liquid samples from the SJAC and WRD of each sample box are analysed for NH 3 /NH + 4 using flow injection analysis (FIA) (Wyers, 1993;Norman et al., 2009).
An ion chromatography (IC) unit equipped with a Dionex AS12 column quantifies the concentration of HONO/NO − 2 , HNO 3 /NO − 3 and SO 2 /SO 2− 4 based on the measured conductivity of the respective anions within the liquid sample compared to a reference standard of 50 ppb Br − added to the sample solution. Analysis by FIA and IC is carried out over 15 min, and using a flow control scheme a half-hourly averaged concentration of trace gases and water-soluble aerosols is generated for each height every hour.
A high-density polyethylene (HDPE) tube (0.3 m length, and 1/3 outer diameter) with a HDPE filter is placed at the inlet of the WRD in order to minimise the loss of HNO 3 and NH 3 and to ensure a particle diameter cutoff of 0.2 nm. A biocide of 0.6 mL of hydrogen peroxide (30 %) is added to every 1 L of the DDI water feeding solution to prevent biological contamination in the WRD of each sample box. Airflow is controlled using a critical orifice downstream of the SJAC.
Autonomous calibration of the FIA system was carried out 24 h after the beginning of the campaign and every 72 h thereafter, giving a total of five internal calibrations of this system. Calibration was conducted using three liquid NH + 4 standards of 0, 50 and 500 ppb concentration. The IC unit is continuously checked for analytical performance by the addition of a liquid Br − internal standard (50 ppb concentration) to each column injection. Calibration of the IC unit was conducted twice during the campaign (23 May and 28 June, prior to and after the campaign, respectively) using a mixed ionic liquid standard consisting of 25 ppb SO 2− 4 , 20 ppb NO − 3 and 20 ppb Cl − .
Measurements of the airflow into the sample boxes were conducted using an independent device (TSI Mass Flowmeter 4140) once every fortnight during the campaign. Additional checks of the field performance of the instrument included daily checks of the WRD tubes and sample box air inlets for signs of visible contamination.
The GRAEGOR sampling boxes have very short inlets with no size selection. Consequently, the aerosol concentration reflects water-soluble total suspended particles (TSP). It detects any compound that dissociates to form the measured ions and therefore has a number of artefacts. These include interferences in HONO measurements through NO 2 , particularly during periods of high SO 2 concentrations (Spindler et al., 2003); the inclusion of dinitrogen pentoxide (N 2 O 5 ) concentrations in measurements of HNO 3 during the nighttime measurement periods, though the magnitude of this unclear in rural environments (Phillips et al., 2013); and the potential for organic chloride compounds to be included in measurements of overall Cl − aerosol (Nemitz et al., 2000b).
The GRAEGOR has been demonstrated to be capable of measuring fluxes in a number of studies both in identical form to the one used here (Wolff et al., 2010a, b;Twigg et al., 2011) and in related variants Rumsey and Walker, 2016). Ammonia-specific instruments based on the same technology (AMANDA, GRA-HAM, ECN, Petten, the Netherlands; Wyers et al., 1993) represent the most commonly used instrument for the automated measurement of ammonia fluxes.

Supplementary measurements
Vertical profiles of temperature were measured at the tower using fine-thread, custom-made thermocouples set at the same heights as the GRAEGOR sample boxes. Located 0.4 m from the tower, an eddy covariance system (Gill Anemometer R01012 with LI-COR-7000) at a height of 2.6 m measured three-dimensional wind speed, sensible heat flux (H ), frictional velocity (u * ) and wind direction. Ongoing, longterm measurements of relative humidity (Vaisala 50/Y Humitter), global radiation (Skye Instruments SKS 110 pyranometer) and total rainfall (Campbell Scientific ARG110 tipping bucket rain gauge) were also available at the site for the campaign period. Measurements of HONO taken by a LOPAP (QUMA Elektronik & Analytik, Wuppertal, Germany) and NH 3 measurements taken by a quantum cascade laser (Aerodyne Research Inc., Billerica, USA) during the campaign period were used for comparison studies with GRAEGOR measurements. Measurements of NO 2 concentration, used in Sect. 4.4.1 to quantify an artefact in GRAE-GOR HONO measurements, were recorded by a chemiluminescence NO 2 detector (200E, Teledyne API, San Diego, California, USA) located 300 m south-east of the Easter Bush site.

Aerodynamic gradient method
The aerodynamic gradient method (AGM), based upon fluxgradient similarity theory, calculates the flux of a tracer (χ, such as a gas or aerosol species) based on its vertical concen-tration gradient coupled with turbulence parameters (Foken, 2008). In this paper a hybrid version of the AGM is used, in which the flux is calculated as (Flechard, 1998) where the friction velocity (u * ) is derived from eddy covariance measurements with a sonic anemometer; κ is the von Karman constant (κ = 0.41); z 2 and z 1 are the heights of the sample boxes; d is the displacement height; and ζ is a dimensionless stability parameter expressing the ratio (z − d)/L, where L is the Obukhov length, a measure for atmospheric stability. The parameter H , an integrated form of the heat stability correction term, accounts for deviations from the log-linear profile under non-neutral stratification. By convention, negative and positive flux values denote deposition and emission, respectively.

Choice of displacement height, d, value
A temperature gradient profile for the campaign was derived from measurements of air temperature at the two heights at which concentrations were measured (0.6 and 2.4 m). Sensible heat flux (H ) was calculated from the temperature gradient as per Wang and Bras (1998): where c p is the heat capacity of air, ρ air is the density of air, and K H is the eddy diffusivity constant for heat. K H can be calculated as where z is the absolute height above ground, d is the displacement height, u * is the friction velocity, and H is the stability correction for sensible heat. Sensible heat flux and, by extension, the flux of the trace gas and aerosol species, are dependent upon the value of d. In order to ensure that the correct displacement height was chosen, the sensible heat flux based upon the temperature gradient developed from thermocouple measurements was calculated using a variety of different values for displacement height. The resulting values for the sensible heat flux were then compared through linear regression to the value for the sensible heat flux recorded by the eddy covariance system also present. A displacement height value of 0.14 m gave the closest agreement between the sensible heat fluxes derived by the aerodynamic gradient approach and eddy covariance, with a linear regression slope of 0.997 and R 2 = 0.945.

Determination of dry deposition velocities
The dry deposition velocity (V d ) of a tracer is the negative ratio of its flux to its concentration (χ) at height z − d: . (4) The V d for gas species may also be expressed as the reciprocal of the total resistance for deposition, which is composed of R a (the aerodynamic resistance), R b (the quasi-laminar boundary layer resistance) and R c (the canopy resistance) as per the resistance analogy for dry deposition (Fowler and Unsworth, 1979;Wesely et al., 1985). R a and R b were calculated from Eqs. (5) and (6) using meteorological measurements taken at the site using (Garland, 1977) where B −1 , B being the Stanton number, is parameterised by the turbulent Reynolds number, Re * (the ratio of the frictional force to the kinematic velocity of air) and the Schmidt number, Sc (the ratio of kinematic velocity of air to the molecular diffusivity coefficient of the gas species): If R a and R b are calculated from measurements, R c can be inferred via For gases, a theoretical maximum deposition velocity can be calculated when it is assumed that the gas is completely absorbed by the canopy (i.e. for R c = 0): The canopy resistance approach can only describe deposition and fails when the exchange of a gas is bidirectional, such as is often the case with NH 3 . In this case, the canopy compensation point model can be adopted, which considers the surface interaction of NH 3 in terms of parallel resistance pathways, composed of individual resistances such as stomatal resistance and cuticular resistance (Nemitz et al., 2000a;Flechard et al., 2007). The gradient technique is only applicable for inert species whose flux is constant with height. Most studies of surface exchange fluxes of reactive compounds do not have the information to assess whether chemical reactions might interfere with the flux measurement, but in this study the behaviour of HNO 3 and HCl allows us to draw conclusions on flux divergence (Sect. 4.2.1). Following precedence in the literature (e.g. Nemitz and Sutton, 2004) we initially evaluate fluxes assuming that chemistry can be ignored, and then discuss the validity of this discussion based on the results.

Limits of detection and estimation of uncertainties in concentration measurements and flux calculations
The concentration limit of detection (LOD) of the instrument for each of the species measured was quantified from a field blank test. The field blank test was carried out prior to the campaign on the 20 March over 24 h by switching off the sample box air pump and sealing the air inlets, but leaving the rest of the system unaltered, as per Thomas et al. (2009). Limits of detection were then calculated as 3 standard deviations from the average background signal. Results from this test are presented in Table 1, expressed as LOD values for each trace gas and corresponding water-soluble aerosol species.
The minimum detectable flux for each aerosol and gas species measured by the GRAEGOR is dependent upon atmospheric stability and the ambient concentration of the given trace gas or aerosol species. Based on the method described by Thomas et al. (2009), median minimum detectable fluxes (F LOD ) were calculated for each trace gas and aerosol species measured and are detailed in Tables 2 and 3, respectively.
When calculating the flux of a species using the aerodynamic gradient method, it is apparent that errors in individual concentration measurements propagate into an error in the concentration differences and, subsequently, affect the accuracy of the calculated vertical concentration gradient. Some errors systematically affect both heights and therefore affect the gradient to a lesser extent than systemic errors in sampling efficiency at a single height, such as the difference in capture efficiency of the WRD tubes or slight differences in airflow caused by differences in the critical orifices, which may impact the accuracy of concentration measurements and resultantly affect the precision in the error of the concentration difference.
The overall random error in the measurements of the trace gas and water-soluble aerosol concentrations (σ m ) can be determined using a Gaussian error propagation approach, in which the concentration error is expressed as a product of several individual measurement errors with the mixing ratio, m (Trebs et al., 2004) (Eq. 9): Here, m liq is the mixing ratio of the compounds found in the analysed liquid sample in ppb, Br (std) the stated mixing ratio of the internal Br − standard, Q Br the flow rate of the internal Br − standard, m Br the analysed Br − mixing ratio and Q air the air mass flow through the system. All values have an associated standard deviation, σ x . This formulation holds strictly for the species measured by ion chromatography; for NH 3 and NH + 4 , the equation is altered by omitting the factor relating to Br − addition and substituting the factor for Q Br and its associated standard deviation with the term Q S , the flow of the analysed liquid sample of NH 3 or NH + 4 . Uncertainties for the trace gases and water-soluble aerosols measured calculated by error propagation ranged from 8 % to 18 % (3σ ) throughout the campaign, varying primarily due to fluctuations in the measured flow rate and analysed concentration of the internal Br − standard.
The error in the concentration difference (σ c ) can be characterised experimentally, by placing both sample boxes at one height or -provided that the absolute difference between sample heights is small -by using one common air inlet at a specified height, with the instrument operated normally. From this side-by-side measurement, linear regression analysis accompanied by orthogonal best fit between the concentrations measured by each sample box can be conducted, with deviation from a 1 : 1 fit between sample heights defined as a systemic error. Using the calculated orthogonal fit equation, corrections in the concentrations can then be applied, accounting for the systemic bias (Wolff et al., 2010b). After correction using the orthogonal fit, the remaining scat- Table 3. Mean (µ A ), median (µ M ), minimum and maximum values for flux and deposition velocity (V d ) for water-soluble aerosols measured during the Easter Bush campaign. Also shown are the median relative standard error (σ F ), the flux limit of detection (F LOD ) evaluated for typical conditions (median u * and median concentration) and the fraction of the hourly flux value that exceeds the flux detection limit evaluated for that hour (f LOD ). ter -termed the residuals -was used to determine the error in the concentration difference. During this campaign, one side-by-side measurement was conducted on 8 June for 16 h by connecting a common air inlet set at z = 1.2 m between each sample box. From the results obtained, it was found that for the gases NH 3 , HCl, HONO, HNO 3 and SO 2 that deviation from the 1 : 1 fit resulted in a precision of measurements < 4 % (3σ ). For the aerosol species Cl − , NO − 3 and SO 2− 4 , precision was calculated as < 8 % (3σ ), while for NH + 4 it was calculated as < 9 % (3σ ).
Errors in flux calculations can similarly be determined through the Gaussian error propagation method applied to Eq. (1). Wolff et al. (2010b), using an analogous form of this equation, showed that total error in the flux is composed of (σ c ) and the error in the flux-gradient relationship (expressed as a transport velocity by Wolff et al., 2010b), which is dominated by the error in u * (σ u * ).
This simplification neglects the detailed secondary errors associated with the stability correction which to quantify fully is beyond the scope of this paper. σ u * is dependent upon the sonic anemometer used and whether conditions are neutral or non-neutral (Foken, 2008;Nemitz et al., 2009b). For neutral conditions, and based on the sonic anemometer used, σ u * was estimated at ≤ 10 %. For non-neutral conditions, σ u * was estimated at 12 % median, which, in combination with σ c , was used to calculate σ F . Throughout this paper, stated errors for concentration measurements are derived from the measurement uncertainty as calculated by Eq. (10), while stated errors for flux calculations are derived from the flux uncertainty as calculated by Eq. (11). Calculated errors for the uncertainty in concentration measurements, the error in the concentration difference and the error in the calculated fluxes for all species measured are similar to values determined by previous studies which have used the GRAEGOR successfully to measure flux gradients (Thomas et al., 2009;Wolff et al., 2010b;Twigg et al., 2011).

Data postprocessing
Concentrations that were less than 5 times the limit of detection as calculated before the campaign began (20 March) were discarded. Calculated fluxes were filtered according to a standard protocol. Fluxes were not calculated for periods of low wind speed (u < 1 m s −1 ), low friction velocity (u < 0.15 m s −1 ) and very stable conditions as indicated by the Obukhov length absolute value (|L| < 5 m). Fluxes were also discarded for periods when the wind was obstructed by the measurement cabin and other towers (270 • > wd < 320 • , and 120 • > wd < 160 • ). Calculated fluxes which were below the minimum detectable flux value for their respective trace gas or aerosol species were discarded. Figure 2 shows time series of the rainfall, radiation, relative humidity, air temperature, and wind speed and direction measured during the campaign. The meteorology splits into two episodes. From 24 May to 5 June 2016, the dominant prevailing wind direction was north-easterly, accompanied by dry and sunny conditions with air temperature displaying a characteristic diel cycle that increased each day. Following a period of cloudier conditions from 6 to 10 June, the prevail- ing wind direction shifted to south-westerly for the remainder of the measurement period. Conditions became wetter and the diel air temperature amplitude was reduced. Relative humidity remained high throughout the campaign, with only occasional periods < 70 %, such as 3-4 June and 21-23 June. Wind speed was variable throughout, ranging between 0.05 and 5.87 m s −1 , with a median value of 2.16 m s −1 . During the fertilisation period, the prevailing wind direction was from the SW, and therefore over the south field, with no precipitation but high (> 90 %) relative humidity.

Concentrations of trace gases and water-soluble aerosols
Summary statistics for the concentrations of the trace gas and water-soluble aerosol species measured at 2.4 m during the campaign are presented in Table 1. Median values for the concentrations of water-soluble aerosol species were similar to those measured in PM 10 at the nearby rural background monitoring site of Auchencorth Moss (Twigg et al., 2015). The time series of the measured aerosol and trace gas concentrations are displayed in Figs. 3 and 4, respectively. Data gaps in the time series are due to in-field calibrations, poor chromatograms, or instability in liquid or airflow. Mean concentrations of NO − 3 were 1.53 µg m −3 (2.4 m), whereas its gaseous counterpart, HNO 3 , had mean concentrations of 0.19 µg m −3 (2.4 m). The mean particulate NO − 3 concentrations were therefore almost 6 times greater than the gaseous HNO 3 counterpart. The same dominance of particulate SO 2− 4 concentrations over gaseous SO 2 concentrations was also observed.
Median concentrations of particulate Cl − were 0.37 and 0.36 µg m −3 at 0.6 and 2.4 m, respectively. The mean concentrations of Cl − were also similar at both heights at 0.89 and 0.91 µg m −3 , respectively. Variation in HCl concentrations at each height was more pronounced, with a mean value of 0.16 µg m −3 at 0.6 m and 0.20 µg m −3 at 2.4 m, and a median value of 0.12 µg m −3 at 0.6 m and 0.15 µg m −3 at 2.4 m. As for particulate NO − 3 and gaseous HNO 3 , measured particulate Cl − concentrations were greater than those of gaseous HCl, by about a factor of 2 at each height.
In contrast, NH 3 concentrations were larger than those of particulate NH + 4 ; median concentrations of NH 3 were 1.15 µg m −3 (2.4 m), while median concentrations of NH + 4 were 0.64 µg m −3 (2.4 m). The average concentrations of NH 3 were similar to those reported previously at the same site for the same time of year (Milford, 2004). Similarly, the average concentrations of HONO are higher than those of its particulate counterpart, NO − 2 , with median concentrations for HONO of 0.04 µg m −3 (2.4 m) and corresponding concentrations for NO − 2 of 0.02 µg m −3 (2.4 m), respectively. Maximum concentrations for NH 3 and HONO at 0.6 m were 21.4 and 0.15 µg m −3 . At 2.4 m, the maximum concentrations for NH 3 and HONO were 13.8 and 0.12 µg m −3 .   The maximum values at each height occurred at 11:00 on the 13 June for NH 3 , 1 h after fertilisation of the south field, and at 13:00 on 13 June for HONO, 4 h after fertilisation of the south field.
The time series of measurements presented in Figs. 3 and 4 show that both aerosol and trace gas concentrations are affected by prevailing meteorological conditions, with larger concentrations for each species during the drier, warmer period of 28 May to 6 June, followed by decreased concentrations from 6 to 10 June when precipitation increased and temperature decreased. Concentrations were lower -except for the peaks in NH 3 and HONO after fertilisation on the 13 June -during the period from 10 June to the end of the campaign, concurrent with the change in prevailing wind direction from the NE to the SW.
The concentrations of HNO 3 and SO 2 showed a strong diel cycle (Fig. 4) from 26 May to 9 June, with maxima at both measurement heights occurring between 11:00 and 14:00 and minima occurring at night between 03:00 and 06:00. A similar, but weaker, inverted pattern was exhibited by their particulate counterparts, with NO − 3 concentrations at both heights (Fig. 3) having maxima between 02:00 and 04:00 and minima between 12:00 and 15:00. Figure 5 shows the median diel concentrations of NH 3 , HCl, HONO, HNO 3 and SO 2 at 2.4 m prior to fertilisation. The median concentrations of HONO remained above the detection limits of the instrument even during daytime, contrary to its expected photochemistry. While concentrations of HONO peaked during nighttime and decreased during the day as incoming solar radiation increased, there remained a detectable concentration of HONO at both heights even for the measurement minima at 15:00. The median diel concentrations for HCl, HNO 3 and SO 2 show a shared pattern, with concentrations peaking during the day to reach a maximum between 11:00 and 14:00, followed by a decrease during the night, reaching minima between 02:00 and 04:00. The concentrations of NH 3 showed little variation across the day. Figure 6 shows the median diel concentrations of NH + 4 , Cl − , NO − 3 and SO 2− 4 at 2.4 m prior to fertilisation. The median diel concentrations of NH + 4 reach a minimum at 16:00, with a maximum at 02:00. The concentrations of NO − 3 show a similar pattern of early morning median maxima (04:00) and afternoon minima (13:00). The median diel SO 2− 4 concentrations had maxima at midnight and a minimum at 16:00. The Cl − concentrations reached a maximum at 03:00 and a minimum at 13:00; however, the upper quartile range was high across all hours, with the maximum concentration of 7.88 µg m −3 recorded at 03:00 (median at this time is 0.5 µg m −3 ). Figure 7 shows the time series of the fluxes for the traces gases measured during the campaign. Data gaps are due to either absent data points (unpaired concentrations) or periods where data were filtered (refer to Sect. 2.3.4).

Fluxes of trace gases
Bidirectional fluxes were present for both NH 3 and HONO, with emission events for each gas occurring dur- ing the period of fertilisation of the south field. For the other trace gases -HCl, HNO 3 and SO 2 -the flux was unidirectional, with deposition occurring throughout the campaign. The deposition for HCl, HNO 3 and SO 2 varied, with larger deposition fluxes occurring during the warmer, drier periods, particularly during the period 1-8 June, and smaller deposition fluxes close to zero during the colder, wetter period at the end of the campaign (15-24 June).
Summary statistics for the trace gas fluxes, deposition velocities, theoretical maximum deposition velocities and canopy resistance values are presented in Table 2. The maximum NH 3 flux was +1460 ng m −2 s −1 , recorded at 12:00 on 13 June, 3 h after fertilisation. The mean flux for NH 3 was +15.24 ng m −2 s −1 , suggesting that emission was the predominant flux for NH 3 during this campaign. For all other gases, the mean flux values were negative, suggesting that deposition was the net flux process overall. However, a max-imum flux for HONO of +4.92 ng m −2 s −1 , recorded 5 h after fertilisation, highlights the bidirectional flux pattern for HONO during the campaign. The maximum HONO flux measured here was particularly large. Nitrous acid emissions have previously been reported postfertilisation of grassland using cattle slurry at the same field site ranging from +1.0 to +1.5 ng m −2 s −1 (Twigg et al., 2011). Table 2 also shows the median relative flux error, the typical flux detection limits (F LOD ) and the fraction of 60 min flux values that exceed the F LOD of that period, based on actual concentration and turbulence. It should be noted that the uncertainty of the campaign averages is much smaller as random uncertainty reduces with the square root of the number of observation that enter the calculation of the ensemble average (e.g. Langford et al., 2015).
Median diel cycles for the deposition velocity and calculated theoretical maximum deposition velocity for the trace gases HCl, HONO, HNO 3 and SO 2 are shown in Fig. 8. The calculation of median diel values for trace gas deposition velocities and canopy resistances excludes the period of flux divergence which occurred during fertilisation. The diurnal deposition velocities for HCl and HNO 3 were very close to the calculated maximum deposition velocities, which is expected as a result of their reactivity and high water solubility. The deposition velocity for SO 2 is near the theoretical maximum during nighttime but is lower during daytime. The deposition velocity for HONO was consistently lower than its theoretical maximum throughout the entire day. While median values for the V d for HNO 3 are close to the values for V max , deposition velocities that exceeded their corresponding theoretical maximum were recorded. While most exceedances fall within the uncertainty range of the measurement, a maximum deposition velocity of 56.8 mm s −1 was recorded at 14:00 on 13 June, 4 h after fertilisation.

Fluxes of water-soluble aerosol components
The measured surface fluxes of the aerosol species Cl − , NO − 3 , SO 2− 4 and NH + 4 are shown in Fig. 9, as well as the summary statistics for the fluxes and deposition velocities in Table 3. A large data gap in NH + 4 fluxes from 31 May to 10 June 2016 was due to NH + 4 only being measured at one height on account of unreliable data for NH + 4 at the lower height of 0.6 m.
Pre-fertilisation, all aerosol species exhibited deposition fluxes. The deposition fluxes were larger during the drier, warmer period from 31 May to 6 June and close to zero during the wetter conditions at the end of the campaign. An important exception was the emission of NH + 4 and NO − 3 from 13:00 on 13 June to 02:00 on 14 June, starting 4 h after fertilisation of the south field.
Summary statistics for the fluxes and deposition velocities for the aerosol species measured are shown in Table 3. As for the trace gases, the median deposition velocities for the aerosol species exclude the period of flux divergence which occurred during fertilisation. The maximum flux for NH + 4 of +18.16 ng m −2 s −1 was recorded at 16:00 on 13 June, 7 h after fertilisation of the south field. Similarly, the maximum flux for NO − 3 (+31.84 ng m −2 s −1 ) was also recorded soon after fertilisation, at 18:00 on the 13 June. Overall, however, the mean fluxes for all aerosol species were negative, confirming a predominant net deposition to the surface.

HONO comparison study between GRAEGOR and LOPAP
A comparison of HONO measurements from the GRAEGOR and two LOPAP instruments was conducted from 26 May to 6 June to investigate the potential artefacts in the WRD method used by the GRAEGOR. The LOPAPs were part of a study to investigate the mechanisms controlling HONO fluxes over managed grassland, including investigating the potential ground sources of HONO, details of which are presented in Di Marco et al. (2018). A series of simple linear regression analyses was conducted to determine the level of agreement between the concentrations of HONO measured by each sample box of the GRAEGOR and each of the LOPAPs. The two LOPAP instruments were operated at the two heights of 0.6 and 2.0 m (hereafter referred to as LOPAP (0.6 m) and LOPAP (2.0 m), respectively). In all comparisons, the GRAEGOR recorded a higher concentration of HONO than either of the LOPAPs. The linear regressions suggest that there is a consistent offset in all GRAE-GOR concentrations, varying between 0.01 and 0.02 µg m −3 . In comparisons between the GRAEGOR sample box 1 at 0.6 m and both LOPAPs, the linear concentration relation for HONO varies from 0.92 to 0.97. The comparisons between the GRAEGOR sample box 2 (2.4 m) and the LOPAPs suggest that the linear concentration relation for HONO is 1.06 and 1.01 for LOPAP (2.0 m) and LOPAP (0.6 m), respectively. In all comparisons, however, there exists a constant concentration offset, which results in a constant higher concentration recorded by both GRAEGOR sample boxes. The closest agreement is between GRAEGOR sample box 2 (set at height 2.4 m) and LOPAP (2.0 m), where the HONO concentration recorded by the GRAEGOR sample box 2 (2.4 m) is 1.06 that of LOPAP (2.0 m). This comparison also has the best statistical agreement, with an R 2 value of 0.67, suggesting a reasonable agreement between the GRAEGOR sample box 2 and LOPAP (2.0 m) measurements.

NH 3 comparison study between GRAEGOR and QCL
On 7 June, a QCL with inlet at height 1.6 m was installed at the Easter Bush site and took measurements of NH 3 from 19 June to 7 August. A total of 3 days of concurrent NH 3 measurements taken by the GRAEGOR and the QCL were recorded in the period 21-24 June. The time series of the NH 3 measurements by each instrument are shown in Fig. 10a. An averaged NH 3 concentration at 1.0 m (χ (1 m)) taken by the GRAEGOR was compared with the NH 3 concentrations taken by the QCL in a simple linear regression analysis, displayed in Fig. 10b. The linear regression shows   that the GRAEGOR recorded a factor of 1.22 higher concentrations of NH 3 than the QCL, with an associated R 2 value of 0.76. However, the number of concurrent measurements is small, with only 41 shared hourly measurement values across 3 days and a period of 19 continuous hours missing between 02:00 and 23:00 of 23 June.

Ion balance
The ion balance for the hourly-measured cation (NH + 4 ) and anion (NO − 3 and SO 2− 4 ) aerosol species pre-fertilisation is shown in Fig. 11. Values are shown as molar equivalent concentration, derived from aerosol mass concentrations converted to molar concentrations and subsequently multiplied by their charge. Cl − charge was not included, under the assumption that it would be entirely associated, in the form of sea salt, with Na + which was not measured by the GRAEGOR. While the correlation between cation and anion species is very good (R 2 = 0.71), the linear regression suggests a deficit of NH + 4 , suggesting that some of the NO − 3 and/or SO 2− 4 was balanced by ions other than NH + 4 . A likely candidate is Na + : some of the SO 2− 4 is likely to have represented sea-salt SO 2− 4 and some NaNO 3 is formed by reaction of NaCl with HNO 3 . Figure 11 is coloured by Cl − concentration, and periods of anion excess tend to be associated with elevated Cl − concentrations.
The formation of NaNO 3 through the reaction of HNO 3 or NO x with sea salt has been previously observed in coastal sites (Andreae et al., 1999(Andreae et al., , 2000Bardouki et al., 2003;Dasgupta et al., 2007;Kutsuna and Ibusuki, 1994) and within the UK and Ireland, where the interaction with marine air with polluted air masses at coastal sites was shown to significantly shift the aerosol NO − 3 to the coarse mode (Yeatman et al., 2001;Twigg et al., 2015). Scavenging of atmospheric H 2 SO 4 , formed from SO 2 (O' Dowd and de Leeuw, 2007), by sea salt may also be occurring, which would also shift some of the SO 2− 4 from the fine to the coarse mode.

Fluxes of water-soluble aerosols and trace gases
Fluxes of SO 2− 4 and Cl − throughout the campaign were deposited unidirectionally towards the canopy surface. However, during the fertilisation period of the south field, bidirectional fluxes of NH + 4 and NO − 3 were observed. Prior to fertilisation these species were deposited to the site. An apparent emission flux of NO − 3 is consistent with the possibility of NH 4 NO 3 formation above grassland suggested by the divergence of HNO 3 V d from V max (Nemitz et al., 2009a) in the presence of high concentrations of NH 3 near the surface. Concentrations of NH 3 peak at 21.4 µg m −3 on 13 June, 11:00, which occurs 3 h before peak HNO 3 V d and 7 h prior to the apparent peak in emissions of NO − 3 at 18:00. Fluxes for the trace gases were bidirectional for NH 3 and HONO, with deposition for all other species measured. Emissions of NH 3 and HONO occurred throughout the campaign, with HONO emissions particularly present during the early morning. Both species reached peak emissions soon after fertilisation. Increases in atmospheric NH 3 concentration and emissions of NH 3 resulting from the application of solid urea fertiliser have been previously established (Akiyama et al., 2004;Sommer and Hutchings, 2001), with losses from volatilisation increased if the urea pellets are poorly mixed into the soil and if conditions are dry and warm. While conditions prior to the fertilisation event were cool, temperatures increased quickly throughout the day, peaking at 19.2 • C at 13:00, 4 h after fertilisation. Volatilisation was likely exacerbated by the dry conditions throughout 13 June. The increase in concentration and upward flux of NH 3 provides the source for the formation of NH 4 NO 3 in the presence of HNO 3 . The mechanisms of the HONO emission fluxes are not discussed here but can be found in Di Marco et al. (2018).

Aerosol deposition velocities
Deposition velocities for NO − 3 reached a maximum value of 9.8 mm s −1 during daytime and a minimum of 0.2 mm s −1 outside the period of apparent emission fluxes at night. A similar pattern was observed for sulfate, which reached Atmos. Chem. Phys., 18, 16953-16978, 2018 www.atmos-chem-phys.net/18/16953/2018/ a maximum value of 9.5 mm s −1 during daytime and a minimum value, outside of apparent emission events, of 0.15 mm s −1 during night. Median V d values for NO − 3 and SO 2− 4 were 1.52 and 1.45 mm s −1 , respectively. For Cl − , the median V d was 3.14 mm s −1 . The deposition velocities for SO 2− 4 where larger than those previously observed and derived for accumulation mode particles from theoretical considerations (Petroff et al., 2008). For sulfate in the fine (< 0.1 µm diameter) range, Allen et al. (1991) recorded a mean value of 1 mm s −1 for deposition velocity over short grass, similar to observations made by Gallagher et al. (2002), who reported a mean value of 0.9 mm s −1 .
The dry deposition of particles can be modelled using a process-orientated approach, which describes the deposition velocity as a function of particle size based on removal mechanisms acting within the vegetation canopy, such as Brownian diffusion, impaction, interception and sedimentation (Slinn and Slinn, 1980;Davidson et al., 1982;Slinn, 1982). The models predict that for particles > 0.1 µm in diameter deposition velocity increases with increasing particle size. Vong et al. (2004) recorded deposition velocities greater than 2 mm s −1 for PM 10 particles over grassland. If the sulfate and chloride were in particularly coarse particles, deposition velocities would potentially be skewed towards a higher deposition velocity.
Secondary ammonium compounds are typically found in the accumulation mode (0.1 to 1 µm), while sea salt is found in super-micron particles (Myhre et al., 2006). Although measurements of particle size were not made during this campaign, measurements of aerosol species (including Cl − and SO 2− 4 ) in the PM 2.5 and PM 10 size fractions were taken by a two-channel Monitor for Aerosols and Gases in Ambient Air (MARGA, Applikon B.V., the Netherlands) instrument located at Auchencorth Moss, 12 km south-west of Easter Bush. Aerosol concentration data were taken from an online database of MARGA measurements (DEFRA, 2018). Agreement between MARGA and GRAEGOR aerosol concentrations was excellent (with correlations for SO 2− 4 with R 2 = 0.95, and for Cl − with R 2 = 0.91 between MARGA PM 10 and GRAEGOR TSP). As proxy for a particle size measurement, the proportion of PM 2.5 to PM 10 was used, with a lower proportion of PM 2.5 indicating a greater proportion of coarse aerosol and a corresponding larger deposition velocity based on process-orientated modelling. To a first-order approximation, particle deposition velocities scale with u * (Pryor et al., 2008). Figure 12 shows plots of the normalised deposition velocities (V d /u * ) against the fraction of the PM 10 mass contained in PM 2.5 at Auchencorth Moss (f PM 2.5 = PM 2.5 /PM 10 ) for nitrate (a), sulfate (b) and chloride (c).
While the dynamic range of f PM 2.5 varied between compounds, third-order polynomial curves consistently describe the relation between the proportion of PM 2.5 to overall PM and the normalised V d for nitrate, sulfate and chloride, suggesting -in line with Slinn (1982) -that deposition velocity increases strongly with increasing particle size above 0.1 µm particle diameter. However, the relationship -although statistically significant -shows significant variability, which may be due to measurement uncertainty, but might also reflect the additional effect of atmospheric stability on particle fluxes (e.g. Wesely et al., 1985;Petroff et al., 2008) or differences in the size distribution between the Auchencorth and Easter Bush measurement sites. It must be stressed that the proportion of PM 2.5 to PM 10 is a proxy measurement for particle size and can only differentiate the proportions of aerosol of diameter less than or greater than 2.5 µm.
By contrast, the median deposition velocity of 0.37 mm s −1 for NH + 4 was much smaller and within the range of previous measurements of dry deposition velocities of accumulation mode particles to grassland. The average f PM 2.5 for NH + 4 recorded was 96 %, compared to 78 % for NO − 3 and 86 % for SO 2− 4 , suggesting that virtually all of the NH + 4 measured was contained in fine particles, within the measurement error. The average normalised deposition velocity (V d /u * ) of NH + 4 of 0.04 was in the range of the values for the other compounds evaluated at f PM 2.5 = 100 %.
Thus, the relatively high deposition velocities for Cl − , NO − 3 and SO 2− 4 (compared with NH + 4 ) are a result of some of these compounds being contained in coarse aerosol. This is consistent with the ion balance (Fig. 11), which suggests that some of these compounds are balanced by sea-salt Na + , which is found mostly in the coarse fraction.
It should be noted that the increase in V d with increasing contribution of coarse aerosol only accounts for the size dependence of the processes of impaction and interception. As a non-turbulent process, gravitational sedimentation is not reflected in micrometeorological flux measurements and the sedimentation velocity would need to be added to the deposition velocity derived here.

Trace gas deposition velocities
Median diel deposition velocities for HNO 3 and HCl closely matched the theoretical maximum deposition velocities within the uncertainty of the measurement (Fig. 8), which closely conforms to their expected physico-chemical behaviour. Both HNO 3 and HCl are reactive and highly water soluble, and consequently it is expected that their deposition velocities should equal the theoretical maximum and that the canopy resistance for these species should be equal to zero (Dollard et al., 1987;Muller et al., 1993). The close agreement between V d and V max during most of the campaign suggests that chemistry is not affecting the fluxes and that the standard aerodynamic gradient method is therefore applicable. However, significant deviations of the calculated deposition velocity from the theoretical maximum for HNO 3 exist: R c values for HNO 3 were particularly large 40 h after fertilisation, from 15 to 16 June, when the mean R c value was 14.8 s m −1 . Conversely, there were periods when the V d for HNO 3 exceeded the V max , such as on 13 June at 13:00 h, when V d for HNO 3 was recorded as 56.8 mm s −1 compared with a calculated maximum of 17.5 mm s −1 .
Reductions in V d for HNO 3 (or in other words a non-zero R c ) have been linked to ground-level sources or non-zero vapour pressures of HNO 3 over nitrate-containing aerosol (particularly, NH 4 NO 3 ), which may evaporate from aerosol within the air space below the measurements or previously deposited to leaf surfaces (Brost et al., 1988;Kramm and Dlugi, 1994;Nemitz et al., 2000b;Nemitz and Sutton, 2004). By contrast, values of V d for HNO 3 that exceed the theoretical maximum could suggest the presence of an additional sink for HNO 3 , which would potentially arise as the result of NH 3 reactions with HNO 3 to form NH 4 NO 3 (Nemitz et al., 2000a;van Oss et al., 1998). The higher V d values for HNO 3 during the fertilisation period, followed by a higher R c value 40 h afterwards, could suggest the formation of NH 4 NO 3 immediately following fertilisation followed by its volatilisation soon after. Indeed, the exceedance of V max coincided with upward fluxes of NH + 4 and NO − 3 (Fig. 9) and this suggests that, during the period after fertilisation, the increase in NH 3 concentration led to an exceedance of the equilibrium vapour pressures of NH 4 NO 3 near the ground, resulting in partitioning of NH 3 and HNO 3 into the aerosol phase. This would have constituted an additional airborne sink for HNO 3 (V d > V max ) as well as a source (apparent emission) for NH + 4 and NO − 3 as previously reported by Nemitz et al. (2009a). It should be noted that during this period the aerodynamic gradient method does not derive accurate fluxes because the condition of flux conservation is not met (Wolff et al., 2010b), and this period has therefore not been included in the diel cycles and summary statistics presented above.
By contrast, fluxes of total ammonium (tot-NH + 4 = NH + 4 + NH 3 ) and total nitrate (tot-NO − 3 = NO − 3 + HNO 3 ) would be conserved, as the effect of gas-particle interactions are not considered, and their assessment provides additional information on the processes occurring during periods when fluxes are not conserved with height.
The time series for tot-NO − 3 and tot-NH + 4 fluxes are shown in Fig. 13. Prior to the fertilisation event on 13 June, the fluxes for tot-NO − 3 were universally depositional to the surface, while fluxes of tot-NH + 4 were bidirectional with significant variation. However, 6 h after fertilisation, a significant emission event of tot-NO − 3 was observed lasting for 6 h. Interestingly, this indicates that the apparent NO − 3 emission during this period (Fig. 9) exceeded the measured deposition of HNO 3 and that there must have been a net source of NO − 3 at the surface during this period. Upward fluxes have previously been reported in the literature where they were attributed to the volatilisation of NH 4 NO 3 from leaf surfaces (Neftel et al., 1996) or alkyl nitrate chemistry (Farmer and Cohen, 2008). Primary emissions of HNO 3 could arise from the heterogeneous reaction of NO 2 with water (Harrison et al., 1996): Kleffman (2007) suggests that HNO 3 could be formed by the reduction of NO 2 on organic sources of humic acid, a process that would also lead to the production of HONO. The formation of HNO 3 inferred from observations coincided with emissions of HONO postfertilisation. However, as discussed previously, this reaction is slow and, while possibly contributing to some of the observed HONO emission, may not be able to account for the majority of observed emissions.
A second potential pathway is the emission of HONO from the soil. As described by Scharko et al. (2015), the oxidation of ammonium by microbes in soils with high nitrification rates can lead to biogenic emissions of HONO. The addition of urea to the agricultural soil at Easter Bush would lead to an increase in soil NH + 4 concentrations and subsequently, through oxidation by soil microbes, the observed emission of HONO. Further discussion of the sources of HONO emissions at Easter Bush will be provided in a future paper by Di Marco et al. (2018).

Daytime source of HONO
As shown in Fig. 5, the median diel concentrations for HONO recorded by the GRAEGOR at 2.4 m do not drop below the detection limits of the instrument, determined to be 30 ng m −3 from calibrations carried out during the campaign. This is contrary to what would be expected based solely on the photolysis rate of HONO, which would suggest that, after accumulation of HONO during nighttime, rapid photolysis should reduce concentrations to below the detectable levels for measurement during early morning (Pagsberg et al., 1997). As measurement approaches have improved over the past 10 years, a growing number of measurements have revealed non-negligible HONO daytime concentrations at rural (Acker et al., 2006;Su et al., 2008;Sörgel et al., 2011), agricultural (Laufs et al., 2017) and urban (Lee et al., 2016) sites, including previous studies at the Easter Bush site (Twigg et al., 2011). Details on the discussion of a potential daytime source of HONO are further discussed in Di Marco et al. (2018).

Comparison of nitrous acid measurement between GRAEGOR and LOPAP
The comparison between the LOPAPs and the GRAEGOR revealed that both sample boxes of the GRAEGOR measured higher HONO concentrations than the LOPAP, principally due to the presence of a constant concentration offset of 0.01 to 0.02 µg m −3 of HONO. Previous comparisons of measurements of HONO have been between the wet annular rotating denuder, as used in the GRAEGOR, and optical absorption techniques, primarily differential optical absorption spectroscopy (DOAS) instruments. In those comparisons, it has been found that HONO measurements by WRD, particularly during daytime and at low concentrations, tend to be significantly higher than DOAS measurements (Appel et al., 1990). By comparison, the LOPAP shows good agreement in HONO measurements with the DOAS (Kleffman et al., 2006), as the DOAS method is a molecule-specific method and the LOPAP method measures any potential NO x artefact. The higher concentrations recorded by the GRAEGOR can be explained by the presence of chemical interferences that occur on the inlet, at the air-liquid interface and within the sampling solution. As the WRD uses a liquid film to sample HONO, and as HONO can form heterogeneously on such surfaces, overestimation of HONO can occur. Furthermore, interferences by chemical reactions of NO 2 with hydrocarbons within the sampler can lead to a further interference (Gutzwiller et al., 2002), particularly in proximity to diesel emissions. It has also been shown that, in high-alkalinity sampling solutions, mixtures of SO 2 and NO 2 can add a further interference to measurements (Spindler et al., 2003). Finally, photolytically induced artefacts can be introduced in the sampling lines that connect the GRAEGOR sampling box to the detector unit (Kleffman and Wiesen, 2008). The LOPAP, which is also a wet-chemistry-based instrument, is designed to minimise the chemical interferences and artefacts that can be introduced in other wet-chemistry instruments.
A comparison between daytime (06:00 to 18:00) GRAE-GOR HONO concentrations and LOPAP HONO concentrations found only a slightly greater difference than the comparison between nighttime (19:00 to 05:00) concentrations recorded by the GRAEGOR and LOPAP. While previous comparisons between the DOAS and the WRD found that daytime concentrations measured by the WRD were higher than the DOAS compared to nighttime measurements, these studies were generally conducted in urban areas where both HONO and NO x concentrations were high (Febo et al., 1996), in contrast to the low concentrations at Easter Bush. The implementation of thermal insulation material around the GRAEGOR sampling lines may have also reduced the influence of photolytic artefacts in exposed sampling lines during the day, which would have elevated daytime HONO measurements recorded by the GRAEGOR. Spindler et al. (2003) developed the following quantification of the chemical artefact produced by the mixing of NO 2 and SO 2 in highly alkaline sampling solutions for HONO measurements in their investigation of SO 2 and NO 2 chemical interference, with all concentrations measured in ppb.
[HONO] artefact = 0.0056 [NO 2 The first term describes the heterogeneous formation of NO 2 with water alone, and the second describes the aqueousphase reaction of NO 2 and SO 2 . Using measurements of SO 2 and NO 2 concentrations, the HONO artefact for the period of the GRAEGOR-LOPAP comparison was calculated and subtracted from the HONO concentrations recorded by the GRAEGOR. A linear regression between the concentrations recorded by GRAEGOR sample box 2 and LOPAP (2.0 m), which had the best agreement without artefact reduction, indicated better agreement after the correction for the artefact (GRAEGOR artefact (2.4 m) = 1.02 · HONO(HONO(LOPAP(2.0 m))), intercept = 5 × 10 −3 µg m −3 , R 2 = 0.72). Coefficient values were altered to produce the best possible agreement between GRAEGOR and LOPAP HONO values, arriving at a final artefact quantification of Use of these altered coefficients further reduced the offset in GRAEGOR HONO measurements but also reduced the statistical agreement between GRAEGOR and LOPAP HONO measurements (GRAEGOR artefact (2.4 m) = 0.98 · HONO(HONO(LOPAP(2.0 m))), intercept = 2 × 10 −3 µg m −3 , R 2 = 0.57). Figure 14 shows the results of these analyses, with the linear regression between GRAEGOR sample box 2 (2.4 m) and LOPAP (2.0 m) without the artefact reduction applied to GRAEGOR sample box 2 (2.4 m) HONO concentrations, with Spindler's artefact reduction and with the modified Spindler's artefact reduction. While the agreement between the LOPAP and GRAEGOR is improved by the introduction of an artefact reduction value, it does not fully close the gap even with altered coefficient values, with a constant concentration offset still present in measurements. The possibility that a further artefact is introduced from NO 2 mixing with hydrocarbons would require further investigation, with concurrent measurements of hydrocarbons. To determine whether the HONO concentration offset in the GRAEGOR measurements impacted on the measurements of HONO flux, a comparison between the HONO flux values derived from GRAEGOR and LOPAP measurements was conducted. Concurrent fluxes of HONO derived from LOPAP and GRAEGOR measurements exist for 72 hourly measurements, from 26 May to 6 June. Figure 15 shows (a) the full time series of concurrent HONO flux values derived from GRAEGOR and LOPAP measurements and (b) a scatter plot of GRAEGOR against LOPAP HONO flux values. Overall, GRAEGOR HONO fluxes are biased towards deposition, with greater deposition values and lesser emission values compared to concurrent LOPAP values. This pattern would be consistent with the concept of an artefact formation dependent upon SO 2 and NO 2 . SO 2 fluxes were unidirectionally depositional at Easter Bush during the campaign as measured by the GRAEGOR. Deposition of SO 2 would lead to greater formation of artefacts within the sample box set at a higher height, which is consistent with comparisons of HONO concentrations between each sample box of the GRAEGOR and each LOPAP instrument. In turn, this would lead to a bias in HONO flux values, resulting in a skew towards deposition. It should be noted that the sample size of concurrent measurements of HONO flux from GRAEGOR and LOPAP measurements is limited (n = 72).

Comparison of ammonia measurements with GRAEGOR and QCL
The comparison between the GRAEGOR and QCL found that, while there was reasonable agreement between the instruments, the GRAEGOR measured somewhat higher NH 3 concentrations than the QCL, by a factor of 1.2. Due to a lack of ancillary micrometeorological data during this campaign, the short overlap in measurements and necessary filtering of unreliable data, there are too few concurrent measurements (15 h) of flux between the QCL and the GRAEGOR for a reliable comparison. There are also only 41 h of concurrent concentration measurements between the two instruments, which overlapped with a period of low NH 3 concentrations. A similar comparison between a WRD system (the Ammonia Measurement by Annular Denuder with Online Analysis, AMANDA) and the QCL system was conducted at the same site in 2004 and 2005 by Whitehead et al. (2008). This comparison also found that the WRD system measured higher concentrations of NH 3 compared to the QCL, but at a far greater factor of 1.67. This difference was particularly pronounced during periods of low NH 3 concentrations, with better agreement recorded during a fertilisation and cutting event that occurred during that study. The older (pumped) QCL used during the earlier campaign did not derive its concentrations from first principles, in contrast to the QCL used during the comparison with the GRAEGOR reported here, which should be within 3 % of the absolute value without further calibration, according to the manufacturer. An inter-comparison between 11 different measurement techniques for NH 3 -including the AMANDA and two QCL instruments (the DUAL-QCLAS and the compact-QCLAS) -was conducted at the Easter Bush site in 2008 (von Bobrutzki et al., 2010). While good statistical agreement was found in linear regression between the AMANDA and both QCL instruments for NH 3 concentrations throughout the entirety of the campaign (R 2 = 0.92 and R 2 = 0.97 for the compact-QCL and DUAL-QCLAS, respectively), there was less agreement between the instruments during periods of low (< 10 ppb) NH 3 concentrations (R 2 = 0.81 and R 2 = 0.52 for the compact-QCL and DUAL-QCLAS, respectively). During periods of low concentration, the QCL systems also underestimated NH 3 concentrations compared to the AMANDA.
Any errors in the GRAEGOR's internal NH 3 calibration system are unlikely to have an effect at low NH 3 concentrations. As a test, the calibration values obtained from all the internal calibration checks which were carried out through the campaign (total calibrations = 5) were used to calculate the NH 3 concentrations during the period of QCL measurements. No significant concentration difference was found between the concentrations obtained by different calibration values, due to no systematic difference in agreement between the different calibration periods.
While there remain significant differences in measured NH 3 concentrations between the GRAEGOR and QCL, the improved agreement between those concentrations, particularly at low values, compared with the results from 2004 and 2005 suggests an improved methodology in use by the QCL system in place at Easter Bush. Further measurements, particularly of fluxes and during periods of high NH 3 concentrations, would be required for a more detailed analysis.

Conclusion
In this paper, we have presented for the first time simultaneous measurements of the trace gases HCl, HONO, HNO 3 , SO 2 and NH 3 as well as their associated water-soluble aerosols counterparts Cl − , NO − 2 , NO − 3 , SO 2− 4 and NH + 4 , before and after urea fertilisation of an agricultural grassland. The main findings for this study are described as follows.
1. Simultaneous measurements of the components of the NH 3 -NO 3 -NH 4 NO 3 triad suggested formation of ammonium nitrate postfertilisation. The use of the conservative exchange fluxes tot-NH + 4 and tot-NO − 3 indicates the presence of a ground source of HNO 3 postfertilisation, which would be rapidly scavenged by high post-fertilisation concentrations of NH 3 to form NH 4 NO 3 . Through this mechanism, use of urea fertiliser becomes a source of regional, rather than local, pollution.
2. The deposition velocities measured for the aerosol compounds Cl − , NO − 3 and SO 2− 4 were significantly larger than those measured for NH + 4 . After normalisation by turbulence, the measurements suggested a clear relationship between deposition velocity and particle size for Cl − , NO − 3 and SO 2− 4 , as parameterised using the proxy of compound in PM 2.5 /PM 10 , although the relationship shows significant variability. Therefore, the high deposition velocities for aerosol compounds recorded at the site are a result of a fraction of the compounds being contained in super-micron aerosol, such as sea-salt sulfate and sodium nitrate.
3. Evidence for a HONO daytime source at the site throughout the campaign adds to the growing body of past measurements that has found evidence for HONO daytime formation in rural, urban and agricultural areas. There is also evidence for the emission of HONO postfertilisation at the site.
This also appears to be the first time a comparison between measurements of HONO concentrations determined by the LOPAP and the GRAEGOR instruments has been documented. While good linear agreement exists between HONO measurements taken by GRAEGOR and LOPAP at both measurement heights, a consistent offset in GRAEGOR HONO measurements suggests the presence of chemically induced artefacts within the GRAEGOR system. This is potentially linked to atmospheric SO x and NO x concentrations. Furthermore, this paper presents a comparison between measurements of NH 3 concentration determined by the GRAEGOR and a QCL system. While changes to the QCL operation system compared to previous studies conducted at the site have resulted in better agreement between the GRAE-GOR and QCL, particularly for low NH 3 concentrations, there still remain significant differences in NH 3 concentrations with larger values reported by the denuder system. Future measurements of aerosol deposition velocities should aim to investigate the effect of particle size upon deposition velocity, using a more robust measurement of particle size than used here. In addition, the ability of urea pellets to act as a potential surface on which heterogeneous formation of HONO and HNO 3 occurs should be investigated, particularly as the formation of these compounds can give rise to the formation of the regional pollutant NH 4 NO 3 .
Data availability. Data from the MARGA were obtained from DE-FRA (2018) and are subject to Crown copyright, DEFRA, licenced under the Open Government Licence (OGL).
Author contributions. EN, CFDM and WJB devised the study and secured some of the funding. GRAEGOR measurements were taken by RR and CFDM. GRAEGOR data were processed by RR with input from CFDM, MMT and EN. LOPAP HONO measurements were made by CFDM, MMT, LJK and LC. The QCL NH 3 measurements were made by NC and the MARGA measurements by SRL and MMT. MRJ and SRL provided field site maintenance and ancillary measurement data. RR interpreted the data with contributions from EN, CFDM, MRH, MMT, WJB, MS and MA. RR led the manuscript with contributions from all the authors.
Competing interests. The authors declare that they have no conflict of interest.