Dynamics of ammonia exchange with cut grassland: strategy and implementation of the GRAMINAE Integrated Experiment

Abstract. A major international experiment on ammonia (NH3) biosphere-atmosphere exchange was conducted over intensively managed grassland at Braunschweig, Germany. The experimental strategy was developed to allow an integrated analysis of different features of NH3 exchange including: a) quantification of nearby emissions and advection effects, b) estimation of net NH3 fluxes with the canopy by a range of micrometeorological measurements, c) analysis of the sources and sinks of NH3 within the plant canopy, including soils and bioassay measurements, d) comparison of the effects of grassland management options on NH3 fluxes and e) assessment of the interactions of NH3 fluxes with aerosol exchange processes. Additional technical objectives included the inter-comparison of different estimates of sensible and latent heat fluxes, as well as continuous-gradient and Relaxed Eddy Accumulation (REA) systems for NH3 fluxes. The prior analysis established the spatial and temporal design of the experiment, allowing significant synergy between these objectives. The measurements were made at 7 measurement locations, thereby quantifying horizontal and vertical profiles, and covered three phases: a) tall grass canopy prior to cutting (7 days), b) short grass after cutting (7 days) and c) re-growing sward following fertilization with ammonium nitrate (10 days). The sequential management treatments allowed comparison of sources-sinks, advection and aerosol interactions under a wide range of NH3 fluxes. This paper describes the experimental strategy and reports the grassland management history, soils, environmental conditions and air chemistry during the experiment, finally summarizing how the results are coordinated in the accompanying series of papers.


Introduction
Atmospheric ammonia (NH 3 ) is a trace pollutant that has major impacts on terrestrial ecosystems and on the formation of secondary particulate matter. In the atmosphere, ammonia acts as a base, reacting to form ammonium (NH + 4 ) aerosol. By contrast, deposition of both forms of reduced nitrogen (NH x ) can lead to soil acidification (e.g., van Breemen et al., 1982;Sutton et al., 1993c). In addition, NH x leads to ecological changes through eutrophication as well as through direct effects of gaseous NH 3 (e.g., Achermann and Bobbink, 2003;Erisman et al., 2007;UNECE, 2007).
As a result of reductions in European emissions of sulphur and nitrogen oxides over the last decade, the relative importance of NH 3 has grown substantially, so that it is estimated that NH 3 will be the largest contributor to acidifying deposition in Europe in the next decade (Amman et al., 2005;Simpson et al. 2006). In addition, the role of NH 3 as a precursor for secondary aerosol formation, with effects on human health, visibility and radiative forcing is increasingly being recognized (e.g., Adams et al., 2001;Vayenas et al., 2005;. The exchange of NH 3 between the land surface and the atmosphere has long been a key uncertainty in quantifying atmospheric budgets and nitrogen deposition (e.g., Eriksson, 1968;Söderlund and Svensson, 1976;Fowler et al., 1989;Langford et al., 1992;Sutton et al., 1993cSutton et al., , 1994Hertel et al., 2006;Kugler et al., 2008). Biosphere-atmosphere exchange of NH 3 is complicated by the existence of bi-directional fluxes for a wide range of land surfaces: the net flux at any time represents the combination of several opposing component fluxes, with leaf surfaces, leaf intercellular matrix, leaf litter and the soil surface (e.g., Denmead et al., 1978;Sutton et al., 1995aSutton et al., , b, 2000bSutton et al., , 2001aHarper et al., 1996;Wyers and Erisman, 1998;Nemitz et al., 2000aNemitz et al., , 2004aMassad et al., 2008). These component fluxes are affected by a range of environmental, plant, soil and management factors, so that the overall NH 3 flux for a given surface may switch between net emission and net deposition on sub-hourly, diurnal and seasonal scales.
Grasslands represent a major uncertainty in quantifying regional NH 3 fluxes and parametrizing atmospheric models. This is both due to the large area extent of grasslands and because of the wide range of grassland types and management practices. In the context of bi-directional NH 3 exchange, grasslands are of particular interest. Firstly, intensively managed grasslands are frequently net NH 3 sources (e.g., Sutton et al., 1993bSutton et al., , 2001aBussink et al. 1995;Harper et al. 1996, Plantaz et al., 1997Milford et al. 2001b;Daemmgen and Webb, 2005;Misselbrook et al., 2005;Kruit et al., 2007), resulting in grasslands impacting on atmospheric NH 3 composition. Secondly, semi-natural grasslands are mainly net NH 3 sinks (Duyzer et al., 1994;Sutton et al., 1993a;Flechard and Fowler, 1998;Spindler et al., 2001;Horváth et al., 2005), resulting in atmospheric NH 3 impacting on the ecological functioning of grasslands.
A limitation of such a dispersed network approach, however, is that it is not economically feasible to consider in an integrated way all of the interacting factors that affect NH 3 exchange at all the individual measurement sites. As part of the GRAMINAE effort, a major experiment was therefore designed to consider the integration of different processes affecting NH 3 exchange. The experiment took place at the German Federal Agricultural Research Centre Bundesforschungsanstalt für Landwirtschaft, (FAL) near Braunschweig in 2000, and linked micrometeorological, cuvette and bioassay analyses for NH 3 , gas-particle interactions, effects of management practice and the role of advection from agricultural sources on NH 3 fluxes in the rural landscape. Overall, the experiment included fifty scientists in the field.
The present paper describes the objectives, strategy and implementation of the GRAMINAE Braunschweig Experiment. The method developments and prior modelling tasks necessary to refine the measurement strategy are summarized, followed by a description of the spatial and temporal design of the experiment, including the soil and meteorological conditions prevailing during the experiment. Finally, the paper outlines the inter-dependence of the detailed analyses arising from the experiment, and reflects on the merits and challenges of such an integrated approach. 4. How much does the near-surface perturbation of the equilibria between ammonia, acid gases and aerosols lead to non-conservation of vertical ammonia fluxes, and is this a relevant mechanism for particle formation/growth?

5.
To what extend can divergence in vertical ammonia fluxes (either due to advection or chemical reactions) be directly measured by available techniques?
All of these questions are closely linked. Although first attempts to measure NH 3 fluxes by eddy covariance have been made (Shaw et al., 1998;Famulari et al., 2004;Whitehead et al., 2008), most studies have applied the aerodynamic gradient method (AGM), which is based on determination of the vertical concentration profile of NH 3 above the canopy (e.g., Sutton et al., 1993a;Wyers and Erisman, 1993;Kruit et al., 2007;Neirynck and Ceulemans, 2008). The gradient method is not well suited to the quantification of divergence in the vertical flux (compared with the flux at the surface), since it estimates an average vertical flux based on measurements at different heights. Flux measurements methods that require sampling at a single height are therefore preferred (e.g., eddy covariance, EC; relaxed eddy accumulation, REA). Given the challenges for accurate EC measurements of NH 3 , a priority was identified to the application of REA for NH 3 , with the ultimate goal to deploy several REA systems at different heights and thereby directly measure vertical flux divergence (change in the flux with height) where this is significant. Both advection of NH 3 from nearby sources and reactions with acid gases are possible sources of flux divergence and can lead to errors in the determination of vertical fluxes using micrometeorological methods. The tendency in micrometeorological studies has been to try to avoid such effects through careful selection of field sites and experimental conditions. However, it often proves difficult to estimate the importance of advection and chemical interactions a priori. Advection from nearby sources is often thought to be of minor importance, as long as the flux footprint falls within the study area, while Loubet et al. (2001Loubet et al. ( , 2006 have shown that downwind of ground-level NH 3 point-sources, advection errors can be significant for several hundred metres. Similarly, Nemitz et al. (2004a, b) and Nemitz and Sutton (2004) showed that evaporation of NH 4 NO 3 near the ground can alter NH 3 exchange fluxes, even where this effect might be expected to be insignificant.
In addition, the conscious minimization of advection errors in the choice of field sites means that the "ideal" field sites typically selected by researchers are far from representative. By contrast, such non-stationarities are "real world" issues that need to be quantified rather than avoided. This is particularly relevant for NH 3 where near-by agricultural point sources are often impossible to avoid in many European landscapes. Advection and chemistry effects are also of interest in their own right as processes that affect the net exchange of NH 3 , HNO 3 and NH + 4 containing aerosol (e.g. Brost et al., 1988;Milford et al., 2001a;Nemitz et al., 2004a, b;Nemitz and Sutton, 2004;Loubet et al., 2006). The GRAMINAE Braunschweig experiment was therefore designed to investigate the extent of vertical flux divergence due to advection and chemistry effects.
Finally, a core issue of NH 3 bi-directional exchange is the quantification of how different component fluxes interact to generate net NH 3 exchange. To understand these effects requires analysis of the interactions with plant physiology, litter dynamics and interactions with soil conditions. In particular, the vertical structure of a plant canopy can lead to an internal cycle, so that NH 3 released from litter or the soil surface may be recaptured by overlaying leaves (Denmead et al., 1976;Nemitz et al., 2000a, b;Harper et al., 2000). The understanding of these interactions can be improved by the integration of above-canopy measurements of net NH 3 exchange with cuvette measurements, bioassay determinations and analysis of within-canopy NH 3 fluxes.
Based on the above concerns, the Primary Scientific Objectives were established as: 5. to integrate the quantification of component processes and fluxes in the canopy, including measurements of soil N availability/uptake, apoplastic NH + 4 concentrations, and within-canopy source/sink distributions; 6. to quantify particle size-distributions, particle number fluxes and chemical fluxes of acid gases and inorganic aerosols; 7. to quantify the extent of chemical reactions of NH 3 in the air within and immediately above the plant canopy to form or evaporate particulate matter, and to consider the consequence of this process for net NH 3 and particle fluxes; 8. to quantify the interactions of NH 3 fluxes with the chemistry of leaf surfaces wetted by dew or rain; 9. to determine horizontal concentration gradients of NH 3 downwind of a farm source allowing quantification of the advection error in the vertical flux and analysis of the consequences for net NH 3 fluxes; 10. to compare NH 3 emissions from a farm source estimated by plume measurements and inverse dispersion modeling, as a basis to assess inventory-based emission estimates for cattle and pigs.
The Technical Objectives that provide the basis to address primary objectives included: 1. to inter-compare above-canopy estimates of turbulent exchange for momentum, sensible and latent heat, identifying the implications for sampling precision in situations without replication and providing robust estimates for this study; 2. to inter-compare estimates of NH 3 fluxes derived from different continuous NH 3 gradient methods, identifying the implications for un-replicated measurements and to provide a robust best-estimate of the net NH 3 flux; 3. to evaluate and inter-compare several continuous REA systems for NH 3 flux measurement and, if sufficiently precise, to deploy these to measure vertical flux divergence; 4. to determine the turbulent structure within the grass canopy and to quantify diffusive time-scales, comparing direct turbulence measurements with estimates derived from application of a Radon tracer method; 5. to develop and test a method for estimating total soluble nitrogen concentration of grass leaves as an estimate of "substrate nitrogen" for comparison with ecosystem model estimates; 6. to refine an approach to estimate farm-scale NH 3 emissions (applied to the dairy and pig unit of the FAL) from fixed and mobile NH 3 concentration measurements downwind of the farm coupled with inverse dispersion modelling.
These objectives relate both to the measurements and model development. Firstly, models were used as analysis tools to derive further measurement-based estimates, e.g., landscapescale inverse dispersion and advection modelling (Loubet et al., , 2006, within-canopy inverse Lagrangian modelling (Nemitz et al., 2000a), modelling of gas-particle interactions (Nemitz and Sutton, 2004) and the calculation of in-canopy eddy diffusivities from Radon profiles (Lehmann et al., 1999). Secondly, other models were applied with basic site, management and environmental input data to compare with the measured results, e.g. soil-vegetation-atmospheretransfer (SVAT) modelling for NH 3 net exchange (Sutton et al., 1995b(Sutton et al., , 1998Flechard et al., 1999;Nemitz et al., 2001b), ecosystem modelling of NH 3 exchange with grassland functioning (Riedo et al., 2002;Massad et al., 2008) and aerosol deposition modelling .

Prior assessment and development
Given the interaction of objectives, it was important to design the GRAMINAE experiment so that certain tasks would not compromise others, to refine sampling methods and to establish that the effects of interest would be detectable. This required both prior modelling of advection effects, plus several prior method developments in preparation for the experiment: continuous REA measurement for NH 3 , foliar N bioassays, and measurement of vertical acid gases and aerosol concentration profiles.

Prior modelling of advection effects
The field site at FAL Braunschweig was initially selected because of the good local infrastructure, control of field management, large field size and presence of a nearby farm NH 3 source. Before confirming the exact location of the measurements, it was necessary to show that advection effects would be significant, but not so large as to dominate the flux and uncertainties, which would prevent assessment of other interactions. Based on housed animal numbers (120 cows, 130 other cattle, 100 pigs), NH 3 emissions from the farm unit were estimated using standard emissions factors to be 10.1 kg NH 3 day −1 . The possible measurement locations in the main grass field ("Kleinkamp") (see Sect. 4.1) were ∼400-800 m east of the farm unit, allowing for sufficient fetch in the field under westerly air flow. The MODDAS model (Loubet et al., 2006) was applied to estimate the horizontal gradients in NH 3 concentrations and the advection errors in vertical NH 3 fluxes (F z(ad) ) with distance from the farm source (Figs. 1, 2). The modelled concentrations were found to be in reasonable agreement with preliminary measurements of NH 3 concentrations (October-November 1999). Monthly mean NH 3 values at different locations around the farm (see Sect. 4.1) ranged 12-33 µg m −3 , while at 240 m east of the farm the measured mean concentration was 3.5 µg m −3 compared with a background of 2.2 µg m −3 . The model analysis showed that advection errors would dominate NH 3 fluxes at distances closer than ∼100 m from the farm (F z(adv,2.7 rmm−hc) >400 ng m −2 s −1 ), with the increased values of F z(adv) for larger measurement heights. By contrast, at the furthest measurement distance (780 m) calculated values of F z(adv,2.7 m−hc) were ∼20 ng m −2 s −1 , while the value for 550 m was 28 ng m −2 s −1 (for periods when the measurement sites would be downwind of the farm source). The values at these larger distances are small enough not to be important in periods of high NH 3 emission (e.g., after fertilization), but would be significant for periods of smaller NH 3 fluxes (e.g., prior to cutting). This analysis supported the use of the FAL Braunschweig field, and it was concluded to use two distances for flux measurement (550 m, 780 m, Fig. 3), supported by measurement of NH 3 concentrations closer to the farm.

Development of relaxed eddy accumulation systems for ammonia
Continuous systems to measure vertical concentration gradients have largely used wet chemical methods and have become widely available Sutton et al., 1995bNeftel et al., 1998). By contrast, although traditional batch sampling techniques have been applied to measure NH 3 fluxes by REA (Zhu et al., 2000), continuous REA systems for NH 3 represent an ongoing challenge. A significant effort was therefore invested in GRAMINAE to refine and test several continuous implementations for REA (e.g., Neftel et al., 1999;Erisman et al., 2001b;Sutton et al., 2001a;Nemitz et al., 2001a, Pryor et al., 2002, which were then compared in the GRAMINAE experiment.

Development of bioassays of ammonia emission potential
It is well known that "compensation point" concentrations apply to NH 3 exchange with vegetation, these being the air concentrations of NH 3 at which no net flux occurs, with component emission fluxes balancing uptake at different scales (e.g. Farquhar et al., 1980;Sutton et al. 1995b). The stomatal compensation point (χ s ) represents the air concentration in equilibrium with the apoplastic (intercellular) fluids of plant leaves. The temperature dependent value of χ s can be calculated from the thermodynamic equilibria and the ratio the ratio [NH + 4 ]/[H + ] in the apoplast, denoted s Nemitz et al., 2000bNemitz et al., , 2001b  [H + ]. A critical experimental evaluation of such methods was therefore made , including analysis of practical issues such as the extent of pH buffering on extracted (diluted) apoplastic solutions (Hill et al., 2001). Prior development of an ecosystem model of coupled grassland C-N cycling and NH 3 exchange Riedo et al., 2002) also highlighted the importance of different plant N pools, including "substrate nitrogen" being the nitrogen available for plant growth, as compared to that built into plant structures. Although this term represents a model construct, efforts were made in preparation for the GRAMINAE Experiment to develop a measurement of total soluble nitrogen concentrations in the plant as an estimate of substrate nitrogen, to complement other indicators, such as total foliar ammonium and nitrate (Husted et al., 2000a, b;Loubet et al., 2002;Mattsson and Schjoerring, 2003).

Measurement of ammonia interactions with particles
Several fast response optical methods were already available for the measurement of particle fluxes, by EC, including condensation particle counters (CPC) and Active Scattering Aerosol Spectrometer Probes (ASASP-x) (e.g., Gallagher et al., 1997;Nemitz et al., 2002Nemitz et al., , 2004b. A key challenge, however is the continuous measurement of chemically speciated aerosol fluxes. For this purpose the steam jet aerosol col-lector (SJAC) approach (Khlystov et al., 1995) coupled with a denuder system for acid gases, running with on-line chromatography is well suited (Jongejan et al., 1997;Wolff et al., 2007). In preparation for the GRAMINAE experiment, development efforts focused on building a full dual channel system (cf. Nemitz et al., 2004b), permitting measurement at two heights simultaneously, which allowed fluxes to be calculated using the AGM.

Description of the field site
The site at FAL is located at latitude 52 • 18 N and longitude 10 • 26 E, being rather flat at 79 m above mean sea level. FAL is situated at the fringe of the Lower Saxonian city of Braunschweig (Brunswick) with 245 000 inhabitants, 12 km WNW of the city centre. It is therefore subject to mixed agricultural air masses for SW to N wind directions and more urban air masses for easterly winds. The main field of the experiment, "Kleinkamp" (Fig. 3), is surrounded by arable and other managed grassland to the N, woodland, by grassland and sub-urban development to the E and S, and by experimental arable plots and the FAL farm to the west.
Biogeosciences, 6, 309-331, 2009 www.biogeosciences.net/6/309/2009/ The FAL Farm is the main NH 3 point source in the immediate locality, which typically houses (according to the prior estimates) 120 dairy cows, 130 other cattle and 100 pigs (50 being in a house fitted with NH 3 scrubbing technology). Based on emission factors for housing and manure storage, 92% of the NH 3 emission was estimated to result from cattle excreta. A further source of NH 3 was liquid cattle manure (slurry) spreading on Field II (see Fig. 3) at ∼17:00 GMT 22 May, leading to NH 3 plumes over Kleinkamp during spreading and at ∼18:00 24 May. Emissions from this source ceased on 25 May, following irrigation with dirty water.
The Kleinkamp field (12 ha) has a sandy soil, with an upper rooting and ploughing A horizon with a high organic matter content, a yellow C1 horizon with occasional worm and root penetration, and below this a red coloured C2 horizon, including ∼20% gravel and flints. Following FAO nomenclature, the soil is a dystric-luvic-cambisol on silty-sandy, cover-sediments overlying gravelly-sandy moranic deposits and glacifluviatile sands. Further details are given in Table 1.
From 1996 to 1999, the field typically received 1 application of liquid animal manure (slurry) each spring (between 25 January and 20 March) at 30-50 tonne ha −1 with an estimated N content of 1 kg N tonne −1 . Annually, the field received 2-4 applications of mineral fertilizer between March and September (usually urea -ammonium nitrate solution). Overall, the field received annual mineral fertilizer inputs of 144, 220, 212 and 288 kg N ha −1 for the years 1996, 1997, 1998 and 1999, respectively, with an additional average slurry input of 37.5 kg N ha −1 year −1 . In order to simplify conditions for the GRAMINAE experiment, no animal manure was applied in 2000 prior to June. Instead only calcium ammonium nitrate was used, with 100 kg N ha −1 being applied on 27 March 2000 and 108 kg N ha −1 being applied on 5 June 2000 (as part of the experiment). The field was ungrazed since sowing, being cut for silage once in June 1996, twice in 1997 (May and June), and three times each in -1999. In 2000 the first cut was made on 29 May 2000 (as part of the experiment).

Spatial measurement strategy
Measurements were made at a series of 7 main sites in order to meet the different objectives of the experiment, as shown in Fig. 3. Mains electricity was provided to all sampling sites, with temporary installations established for the experiment at Sites 1-4. The field measurements at Site 1 were supported by two mobile laboratories (Fig. 4b) for deployment of particle-sampling equipment (UoM Lab., 30 m into the field) and for chemical preparation (ECN Mobile Lab., at the N edge of the field). Details of the measurements made are given in Sect. 5.

Site 4: Small plots
In order to address the effect of different management options on net ammonia exchange (Sect. 4.3), 4 small plots, each of ∼10 m×10 m, were marked out from the Kleinkamp field at the start of the experiment. These plots were located at the N edge of the field, NE of Site 1 (Figs. 3 and 4). The main field was designated as Plot 1 (periods A and B), with the small plots designated as Plots A2, A3, B2, B3. In addition, a further 15 m×20 m was left uncut and unfertilized (Plot C) for determination of vertical profiles of turbulence, temperature and NH 3 within the canopy after the cut (Nemitz et al., 2009c). Measurements in the small plots focused on cuvette measurements of NH 3 fluxes (David et al., 2009a, b) and foliar N bioassays in relation to altered soil, vegetation and microclimate (Mattsson et al., 2009a) Site 4 was supported by two further mobile laboratories (Fig. 4b) for apoplastic extractions and plant handling (UoC/FAL-D Lab) and for measurements related to the small plots (INRA Lab).

Site 5: FAL Farm buildings
A record was kept of the numbers of different types of livestock in each of the farm buildings during the experiment, allowing detailed calculation of NH 3 emissions according to the inventory approach ). The sources included both livestock houses and manure stores and the location of the main sources is shown on Fig. 3. Ammonia concentrations were measured at several points around the farm using passive samplers, according to the method of Ferm and Rodhe (1997

Temporal measurement strategy
The experiment was divided into three measurement periods of approximately one week to study the effects of cutting and fertilization on ammonia fluxes and exchange processes. The field site was measured out, electricity installed and equipment established on 14-22 May 2000, with measurements started at 12:00 Greenwich Mean Time (GMT) on 22 May. This allowed seven days of measurements before the grass was cut (from ∼0.75 to 0.07 m height) at 06:00-10:00 GMT on 29 May. On the morning of 30 May, the grass was turnedover using mechanical "spinners' to help dry it and was then lifted from the field 06:00-10:00 GMT on 31 May. Hand raking of the grass around the sampling equipment at Site 1 was complete by 12:00 GMT. Seven days after cutting, the field was fertilized with nitrogen at 06:00-07:00 on 5 June (Sect. 4.1), allowing the effects of fertilization of the main field to be followed for 10 days, before measurements were stopped at 12:00 GMT on 15 June.

Measurement strategy on the main field
Micrometeorological measurements of net NH 3 fluxes were made throughout the experiment using the AGM, while additional resources were deployed to address different questions before and after the cut. Before the cut, particular attention was given to measuring within-canopy turbulence and NH 3 concentration profiles (Nemitz et al., 2009c) investigating the nature of canopy sources and sinks of NH 3 . This also included attention to vertical profiles of apoplastic bioassays (Hermann et al., 2009) and differences between plant species (Mattsson et al., 2009b). After the cut, the NH 3 detection systems (ECN, FAL-CH) used for within-canopy profiles, were deployed in the intercomparison of REA methods , while bioassay measurements focused on measuring the consequences of different management options using the small plots (Sect. 4.4.2) and on determining the extent of diurnal variability of apoplastic [NH + 4 ] and [H + ] and total tissue NH + 4 and NO − 3 . Diurnal courses of 3-hourly bioassay sampling were made on 26-27 May, 4-5 June and 12-13 June ).

Measurement strategy of the small plots
Measurements on the small plots and comparison with the main field allowed the effects of hay/silage management, different levels of N supply, and the longer term effects after cutting to be tested. The following 10 m×10 m plots were defined: Plot A1: Main field not cut, ∼200 kg N ha −1 . While cuvette NH 3 flux measurements focused sequentially on Plots A1-3 and then Plots B1-3 (David et al., 2009a, b), the bioassay measurements allowed the responses of Plots A2 and A3 to be followed for up to 3 weeks after cutting (Mattsson et al., 2009a). In addition, supplementary plots (C1, C2, C3) were delineated during the experiment with an additional cut on 10 June (with cut grass removed), the following conditions: Plot C1: Regrowth of grass, no fertilization, Plot C2: As C1, but bare soil from 12 June, Plot C3: As C1, but bare soil from 12 June, plus green leaves added to surface.

Data acquisition and reporting
The Integrated Experiment naturally included a very wide range of measurement types, with different acquisition systems being applied by different groups according to data type. Continuous datasets were logged on instrument dataloggers or PC-based systems by each of the research groups. To ensure comparability of the measurements, data were collected and reported according to common standards: a) for continuous measurements, data were collected on a 15 min basis (or hourly basis, where this was not possible), b) stated sample times refer to the end-time of the sampling period (i.e., 1230 would denote 1215-1230); c) all measurements were recorded according to GMT (being the Universal Time Constant), which equated to local time minus 2 h. The consolidated datasets were provided as a CD and on a shared file server to participants .

Quality assurance and quality control
Many of the measurements made during the experiment are known to be very challenging, especially for continuous determination (e.g., NH 3 , HNO 3 , NH + 4 and anion aerosols, apoplastic bioassays etc.) and when considering fluxes and gradients rather than just concentrations. In addition to following good experimental practice, the availability of different scientists on-site allowed for regular trouble-shooting and repair of instruments. A key QA element was the use of replicated measurement of the same parameter by different groups using either the same or similar sampling methods. Thus the determination of turbulent exchange parameters (Nemitz et al., 2009b) was made using 9 different ultrasonic anemometers, as well as wind/temperature profiles etc. Similarly, the AGM for NH 3 fluxes was applied using 4 independent NH 3 detectors , while assessment of NH 3 fluxes by REA also used 4 independent NH 3 detectors . The continuous measurements of trace gas and aerosol concentrations were also supported by FAL permanent daily denuder/filterpack measurements (300 m N of Site 3). Such inter-comparisons enabled screening of the datasets for outliers, while standard filters were applied for flux calculations (e.g., low windspeed, atmospheric stability and flux footprint, Nemitz et al., 2009b).
Gas-phase calibrations were used for CO 2 and SO 2 measurements. However, such calibration is uncertain for NH 3 , and in this case liquid phase calibration of the chemical detectors used was applied, using common aqueous standards for all groups. At three times during the experiment, "mystery solutions" were provided blind to test the performance of the different analysis systems for aqueous NH + 4 .

Summary of measurement methods applied
As noted above, a large number of sampling systems were applied by the different groups. These are summarized in Table 2, which also notes where the methods are described in more detail. Overall, 14 continuous NH 3 measurement systems with online chemical analysis and 2 automatic batch systems were applied during the experiment, relating to 9 different methods/implementations. Four optical measurement systems were used for aerosol physical properties and number concentrations/fluxes, plus two chemical methods. Ten systems (inc. EC, AGM and energy balance methods) were compared for calculation of sensible heat flux, of which four also provided assessment of latent heat fluxes. Within-canopy micrometeorological measurements included the application of a micro-ultra sonic anemometer and its comparison with a 222 Radon tracer method to estimate within-canopy turbulence and exchange coefficients (Nemitz et al., 2009c). A wide range of other surface parameters were measured including temperature profiles, leaf temperatures (by attached sensors as well as infra-red detection), Biogeosciences, 6, 309-331, 2009 www.biogeosciences.net/6/309/2009/ soil temperatures and leaf wetness, using leaf surface conductivity (Burkhardt and Eiden, 1994). Soil moisture was sampled continuously using TDR (Time Domain Reflectometry), as well as a theta capacitance probe and by regular gravimetric measurements, focusing on the A horizon at 0-10 cm and 10-30 cm. Finally, plant material was sampled for N bioassays being prepared and extracted on-site (UoC lab), including pressure impregnation and extraction of apoplastic extracts using a cooled centrifuge (4 • C) and measurement of apoplastic pH (Mattsson et al., 2009a). Leaf extracts for analysis of NH + 4 and other N components were frozen in liquid nitrogen and subsequently stored on site at −18 • C.

Synoptic meteorology during the experiment
The meteorology was characterized by mainly W flow, plus shorter periods of S and SE flow, with several frontal systems passing during the experiment. The NOAA HYSPLIT model was used to calculate daily 24-h back-trajectories for air masses arriving at 12:00 and 00:00 GMT (Czik et al., 2000) and a daily summary has been tabulated by . Changing conditions on 22 May were followed by W flow on 23-27 May. Surface air masses (50 m height on arrival) on 23-27 May were estimated to originate over NW Germany while higher air masses (1000, 1500 m on arrival) originated from S England and NW France. Warm fronts passed on 23 and 26 of May, followed by a cold front on 27 May.
Following a brief period of S flow on 28 May, W flow resumed on 29-30 May with surface air masses (50 m) originating 24 hours previously over the Benelux countries, while higher air masses originating over N England. A further period of change on 31 May, was followed by 5 days (1-4 June) of SW flow, with surface air originating 24 h previously over Germany and higher air masses originating over N France. Warm fronts passed on 1 and 3 June.
The flow changed on 5 June associated with a cold front, to be followed by WNW flow on 6-8 June from over the Netherlands, with higher air masses originating 24 previously over the North Sea. For air masses arriving on 9-10 June, the flow was from the SE over the Czech Republic and S Germany, with a cold front and associated rain on 10 June. Finally, following change on 11 June, W flow was resumed for 12-15 June, with variable 24 hour backtrajectory origins of surface (50 m) air being over NW Germany, while higher air masses (1000, 1500 m) originated from the S North Sea and S England.

Environmental conditions before and during the experiment
An overview of the environmental conditions prior to and during the experiment is shown in Fig. 6. Before the experiment (2-17 May) the conditions were largely dry and sunny, with maximum air temperature (2 m, DWD data) between 20-30 • C. These conditions resulted in the farmer irrigating Kleinkamp on 15-17 May, before the experiment started.
Conditions were more mixed during the experiment, with in excess of 5 mm precipitation on 7 days, and maximum temperatures ranging from 14 • C on 29 May to 31 • C on 10 June. Mean daily windspeeds during the experiment ranged 2.0-5.4 m s −1 (overall mean 3.4 m s −1 ), with mean daily relative humidity in the range 49-83%. The hourly structure of global radiation, wind direction (10 m), friction velocity (u * ) and air/canopy temperatures are shown in Fig. 7 canopy temperatures occurred on 1-4 and 10 June, with values being larger after cutting due to lower transpiration fluxes (Nemitz et al., 2009b).

Air chemistry during the experiment
The different air-mass origins and local meteorological and environmental conditions provided a wide range of air chemistry conditions during the experiment. This was of great benefit for the experiment, allowing consideration of interactions of gases and aerosols under situations of both potential aerosol growth and evaporation, as well as of differing leafsurface and biological interactions in the presence of altering amounts of acidic (HNO 3 , SO 2 ) and basic (NH 3 ) gases. The time course of aerosol and gaseous pollutant concentrations during the experiment is shown in Fig. 8. The three physical measurements of particle number concentrations reflect different particle sizes, with categories presented in Fig. 8a including the fractions 0.01-3 µm (CPC), 0.1-3 µm (ASASP-x) and 0.5-2.5 µm (APS). Since these are number concentrations, the values are dominated by the smallest particle sizes in each range. Thus, the CPC counts are dominated by the nucleation and Aitken modes, the ASASP-x reading by the smaller accumulation mode particles and the APS concentration by the larger accumulation mode. These differences help interpret the aerosol processes occurring at different times. For example on 5 June, surface air was moving little  with regional pollution providing a high correlation between accumulation (APS) and Nucleation/Aitken mode (CPC) aerosol. By contrast, on the 9 June, flow from the SE over S. Germany and the Czech Republic (Table 2) brought SO 2 -and HNO 3 -rich air masses with a low APS reading over Braunschweig to the site, apparently favouring regional nucleation of new particles. The aerosol appears to age over the 10 June, with higher values in the larger size classes. This interpretation is supported by the results of the Differential Mobility Particle Sizer (DMPS), which recorded particle size spectra throughout the experiment. Following a period of small concentrations on 8 June, a nucleation event occurred at 9 am on 9 June, followed by particle growth (through coagulation and condensation) from 8 nm at 00:00 9 June to 100 nm by 12:00 10 June (Fig. 9). The "chimney" or "banana" shape of the event on the threedimensional presentation is consistent with nucleation events observed around the globe (e.g., Kulmala et al., 2004).
The ASASP-x's provided the size categories that are most consistent with the mass concentrations of aerosol recorded by the SJAC (Fig. 8b). Although there is a reasonable correlation with SO 2− 4 mass concentrations, this makes up only 46% of the measured inorganic anion aerosol mass on average, with NO − 3 being the largest component, with the SJAC recording only small concentrations (<1 µg m −3 ) of Cl − and NO − 2 aerosol. The highest NO − 3 concentrations were associated with the SE and S flow on 10 June, with high values also at the start of the experiment 23 May under SW flow, which passed over the industrial Ruhr area of Germany.
These peaks are clearly reflected in the SO 2 and HNO 3 concentrations (Fig. 8c), which were rather similar, at up to 6-8 µg m −3 on the 23 May, 9 and 10 June, while the latter event was also associated with high O 3 mixing ratios of up to 70 ppbv indicating significant photochemical activity. These peaks may be contrasted with cleaner periods (e.g., 5-6 June) with only 0.8 and 0.4 µg m −3 SO 2 and HNO 3 , respectively.
The "best estimate" of NH 3 from the continuous gradient measurements  was interpolated to 1 m above canopy, based on the mean of up to four instruments (Fig. 8c). There was some doubt about the performance of two of the AMANDA denuders on selected days (due to disagreement with the two other systems), and for this reason an "alternative estimate"  is also shown. The 1 m NH 3 surface air concentrations agree very closely with the background air at 43 m prior to 29 May, while after 5 June, the NH 3 concentrations at 1 m are much larger. This illustrates how, prior to 29 May, NH 3 concentrations at 1 m were largely controlled by external sources, while the later larger values, especially after the 5 May, are mainly due to NH 3 emissions from the field itself. The increased NH 3 concentrations following cutting and fertilization thus provide the opportunity to investigate consequent interactions Biogeosciences, 6, 309-331, 2009 www.biogeosciences.net/6/309/2009/ Site 1 (7.5 min.) FAL-D, FRI  NH 3 (5) Continuous mini-WEDD Site 1 (7.5 min.) FAL-CH (Neftel et al., 1998) NH 3 (3) Continuous AMANDA denuders with triple denuder inlets Site 2 (7.5 min.) CEH  Gaseous  (David et al., 2009a) Physical aerosol measurement (2 m) Size distribution of particle numbers (2.9 nm<D p <0.45 µm) Differential mobility particle-sizer (DMPS) Site 1 (10 min.) UoM (Williams, 1999;Coe et al., 2001)   T and E profiles (2 heights), net radiation (Rn), total solar radiation (S t ), ground heat flux (G) and precipitation (2 systems) Campbell "Bowen Ratio" systems (fine thermocouples, dewpoint meter and additional parameters) Site  (Lehmann et al., 1999) with the acid gases and aerosols. Not surprisingly, the time course of NH + 4 aerosol follows that of aerosol anions, and for much of the time 4 June,(8)(9)(10) completely neutralized the equivalent anions (average neutralization 94%).
While there were clear relationships of the aerosol concentrations to air mass origin, there was only a weak relationship to local (10 m) wind direction (Fig. 10) indicating a significant contribution from long-range transport. By con-trast, NH 3 , SO 2 and HNO 3 were largest for SE winds, with the latter linked to the episode on 8-10 June, indicating pollution of more local/regional origin.

Plant and soil conditions during the experiment
At the time of cutting, the grass was 0.75 m tall, with a single sided Leaf Area Index (LAI) of 3.06. Cutting reduced the sward to 0.07 m tall with a LAI of 0.14. Biogeosciences, 6, 309-331, 2009 www.biogeosciences.net/6/309/2009/ Ample soil moisture was available for plant growth, due to the prior irrigation and rain events at several times during the experiment. Figure 11 shows the time course of volumetric soil water content, with values in the range 10-18% vol. according to TDR measurements at 0.15 m, and 12-20% vol. by gravimetric determination for 0-0.1 m, with a similar range from theta probe measurements. These values represent 46-91% of field capacity (Table 1), with averages of 63% (TDR) and 73% (gravimetric) field capacity for the experiment. The moisture content at 0.1-0.3 m was only slightly higher at 66% (TDR, 0.3 m) and 76% field capacity (gravimetric, 0.1-0.3 m). The TDR results are considered the more reliable and outlier gravimetric results (>field capacity) for wet conditions on 23 May are excluded from these means. TDR values for the C1 horizon were in the range 12.8-16.5, representing 77-99% field capacity.
A survey of soil moisture across the field (theta probe, 0-0.1 m depth, 50 m intervals, 3-4 June) showed no major E-W gradients (11.5±2.5% vol., ±SD), although the field was slightly drier at the N edge of Kleinkamp, E of Site 1, and at the S edge, E of the DWD. This did not significantly affect the flux foot-prints of Sites 1 and 2, but resulted in a slightly drier soil (9.4% vol.) at Site 4.
The cutting and fertilization had major effects on the plant and soil N status conditions, providing a significant perturbation for NH 3 exchange processes. Figure 11 shows that cutting the sward had no significant effect on soil NH + 4 , although soil NO − 3 (0-0.1 m) increased from 5 to 18 mg N g 1 dry soil after 4-5 days, with a smaller increase for 0.1-0.3 m). Much larger responses were recorded following fertilization, with soil NH + 4 and NO − 3 (0-0.1 m) increasing to 139 and 146 mg N g −1 dry soil, respectively, 3 days after fertilization. The decrease by 10 June may be due to plant up-take and microbial immobilization, while rain later on 10-11 June may have led to remobilization, increasing NH + 4 and NO − 3 values on 12-14 June. Even 9 days after fertilization, a strong vertical profile was present, with ∼200 mg N g −1 each of NH + 4 and NO − 3 for the layer 0-0.02 m. This contrasts with 1 June where there was little gradient in the layer 0-0.1 m.
While there was little effect of cutting on available soil N, a clearer effect was recorded for total N content of grass foliage. Before cutting, the mature leaves contained 2.0% N, and, where the grass was left uncut and unfertilized, this decreased slightly to 1.8%. By contrast, the short re-growing sward (0.13-0.18 m) measured on 2-4 June had a N content of 3.0-3.4% N, presumably to N re-mobilization within the cut plants. Foliar N increased further following fertilization, with the values representing a balance between the effects of soil N uptake and growth dilution. A further increase occurred following the rain and increased soil N mobilization after 10 June.

Discussion and conclusions
This paper has introduced the rationale, questions and objectives of the GRAMINAE Braunschweig Experiment. It has described the strategy taken to address these multiple objectives in an integrated way, the nature of the measurements made, and the environmental, plant and soil conditions achieved during the experiment.
Integration is one of the key challenges in understanding environmental processes. Because there are so many interactions, addressing them necessarily increases the scale of the required to meet some objectives may compromise others. Careful design of the GRAMINAE experiment was therefore essential. In particular, some would say that it is unwise to address advection issues in an experiment that also examines sources and sinks, net NH 3 biosphere/atmosphere exchange and its interactions with aerosol chemistry. However, this misses the challenge faced: that these are real world issues to be addressed rather than avoided.
To allow consideration of advection effects, distances were therefore selected based on the prior modeling (Sect. 3.1) so that these would not dominate fluxes. In addition, it was recognized that advection effects would only occur for selected wind directions during the experiment. By contrast to the potential for conflict, addressing multiple questions also has the benefit that the researcher cannot anticipate which measurements will be least and most successful.  Fig. 9. Changing particle diameter (d p , nm) distribution from the Differential Mobility Particle Sizer (DMPS) during a regional nucleation event following a period of clean air from the SE on 8-10 June 2000. Increasing particle numbers (N ), represented as blue to green to red, are expressed as dN/dlog(d p ) (cm −3 ). The basic site conditions presented here of synoptic meteorology, micrometeorological conditions, air chemistry, soil and plant conditions, show that the experimental set up was successful in obtaining a wide range of conditions to assess NH 3 exchange processes. Firstly, the prevailing meteorology provided the range of turbulence, temperatures, wetness needed to address the interaction of plant physiological, leaf surface and canopy exchange processes. Secondly, different air mass origins provided a range of air chemistry, including conditions for both formation and evaporation of NH + 4 aerosol, while varying local wind direction allowed comparison of periods with and without NH 3 significant advection errors in the vertical fluxes. Thirdly, the management of the main field and the management treatments provided a major perturbation to the grassland N cycling, allowing the processes of NH 3 fluxes to be investigated in great detail.
The present paper provides the necessary background and introduction to the resulting series of papers arising from the experiment, presented in this special issue. A logical overview of the different paper contributions is provided in Fig. 12. These can be considered in four groups: 1) Mi- crometeorology and net NH 3 fluxes, 2) Biospheric controls, including within canopy turbulent exchange, 3) Interactions with acidic and other gases at the canopy surface, and 4) aerosol fluxes and NH 3 gas-particle interactions.

Micrometeorology and net NH 3 fluxes
The quantification of turbulent exchange fluxes and the energy balance is fundamental to any micrometeorological assessment. This is analyzed by Nemitz et al. (2009b), who use an inter-comparison of the different measurement systems applied to provide 'best estimates' of the exchange terms for the experiment. At the same time, they provide an assessment of the uncertainty in such measurements when (as in most studies) they are not replicated. Using these turbulent exchange estimates, Milford et al. (2008) compare the different AGM implementations for measurement of net NH 3 fluxes, deriving both best estimates and uncertainties in the measured fluxes. Milford et al. thus provide a robust quantification of the response of NH 3 exchange to environmental conditions and the grassland management events. These robust ammonia flux estimates by the AGM provide a reference for the assessment of four continuous REA systems for NH 3 by Hensen et al. (2008). Two papers address NH 3 emission from the adjacent FAL farm. Firstly, Hensen et al. (2009) apply the plume NH 3 measurements to estimate the farm NH 3 emissions, in comparison with the inventory approach. Secondly,  apply the NH 3 concentrations measured at Sites 1, 2 and 3 in a Lagrangian stochastic dispersion model (MOD-DAS) to quantify the effects of the farm, and the emitting field itself, on advection errors in measured vertical fluxes.

Biospheric controls on NH 3 exchange
The bioassay measurements of potential for NH 3 emission are reported by Mattsson et al. (2009a, b) and Hermann et al. (2009). The interaction of plant N bioassays with changing soil conditions through the experiment, including determination of s from apoplastic extractions, is presented by Mattsson et al. (2009a). Hermann et al. (2009) then report on vertical changes within the plant canopy, while Mattsson et al. (2009b) describe the inter-species differences in plant N bioassays and NH 3 emission potential.
The effects of the grassland management options on bioassays are reported by Mattsson et al. (2009a), while David et al. (2009b) address the effects on net NH 3 fluxes using the cuvette approach. David et al. (2009a) also use this approach to investigate the nature of the NH 3 sources and sinks within the grass canopy. The cuvette approach is complemented by the investigation of within-canopy micrometeorology by Nemitz et al. (2009c). Their investigation applies the measured vertical profiles of radon activity to estimate within-canopy exchange and estimates turbulent and heat exchange using measurements from a micro-sonic anemometer and hotwire anemometer located in the tall grass canopy. These approaches provide the basis to estimate within-canopy NH 3 fluxes for selected periods.
The integration of biospheric and micrometeorological conditions in defining net NH 3 fluxes is considered by Personne et al. (2009) and by Sutton et al. (2009). Firstly, Personne et al. apply the "canopy compensation point" resistance modelling approach (Sutton et al., 1995b; to described the measured net NH 3 fluxes, deriving stomatal resistances and values of s , soil and the leaf cuticular uptake resistance for NH 3 (R w ). Secondly, Sutton et al. (2009) apply the coupled grassland C-N and NH 3 exchange model, PASIM (Riedo et al., 2002) to the conditions of the experiment, comparing the results with the measured net NH 3 fluxes. Together with the foregoing analyses, these examine the canopy scale trade-offs that define net NH 3 fluxes.

Interactions with acidic and other trace gases
The performance of the profile SJAC system for measuring concentrations and fluxes of acidic and basic species is summarized by Nemitz et al. (2009a). The species measured included gaseous SO 2 , HNO 3 , HCl, HNO 2 , NH 3 and aerosol SO 2− 4 , NO − 3 , Cl − and NH + 4 . The availability of Biogeosciences, 6, 309-331, 2009 www.biogeosciences.net/6/309/2009/ data on NH 3 , the acidic gases and aerosols, together with the results of dew chemistry sampling during the experiment allowed the leaf surface and flux interactions to be addressed by Burkhardt et al. (2009) using a canopy compensation point model that incorporates mixed chemistry dynamics on leaf surfaces (Flechard et al., 1999). In parallel, the measurements of CO 2 and O 3 exchange, and their relationship to grassland management are presented by Mészáros et al. (2009).
6.4 Aerosol fluxes and NH 3 gas -particle interactions The large NH 3 emissions observed after fertilization of the grass, as well as the warm canopy temperatures, provide conditions with the potential for both formation/growth and evaporation of NH + 4 -containing particles. Nemitz et al. (2009a) investigate the aerosol dynamics including the size resolved EC measurements of physical particle fluxes. They then combine these data with the results of the NH 3 and NH + 4 aerosol fluxes to assess the interactions between NH 3 fluxes and gas-particle inter-conversion.

Conclusions
This introduction and the series of papers noted above demonstrates the multiple interactions that occur in defining net NH 3 fluxes between the biosphere and atmosphere. The GRAMINAE Braunschweig experiment was designed to assess many of these in an integrated way, providing substantial synergy between different measurement groups and scientific issues being addressed. The experiment was successfully implemented with measurements over a 25-day period, covering a range of meteorological conditions and atmospheric chemistry, as well as plant and soil conditions associated with management perturbation of the grassland nitrogen cycle. It is shown how such an integrated strategy can be developed, which allows multiple objectives to be addressed through careful placement of sampling locations and timing of the different activities.
The detailed papers resulting provide the most comprehensive analysis of NH 3 exchange processes that has been attempted to date. These cover issues of a) quantifying net NH 3 fluxes in a real landscape and using new methods, b) the biospheric controls on NH 3 fluxes, using a combination of flux measurement approaches, bioassays and models, c) the interactions that occur with acidic gases at the canopy surface, and d) the potential for NH 3 fluxes to interact with particle fluxes within and immediately above a grassland canopy. The results of these detailed analyses are brought together at the end of this Special Issue in the synthesis presented by Sutton et al. (2009).