Processes of ammonia air–surface exchange in a fertilized Zea mays canopy

. Recent incorporation of coupled soil biogeochemical and bi-directional NH 3 air–surface exchange algorithms into regional air quality models holds promise for further reducing uncertainty in estimates of NH 3 emissions from fertilized soils. While this represents a signiﬁcant advancement over previous approaches, the evaluation and improvement of such modeling systems for fertilized crops requires process-level ﬁeld measurements over extended periods of time that capture the range of soil, vegetation, and atmospheric conditions that drive short-term (i.e., post-fertilization) and total growing season NH 3 ﬂuxes. This study examines the processes of NH 3 air–surface exchange in a fertilized corn ( Zea mays ) canopy over the majority of a growing season to characterize soil emissions after fertilization and investigate soil–canopy interactions. Micrometeorological ﬂux measurements above the canopy, measurements of soil, leaf apoplast and dew/guttation chemistry, and a combination of in-canopy


Introduction
Livestock and crop production are responsible for 80 % of ammonia (NH 3 ) emissions in the US (US EPA, 2005). Fertilized soils account for 35 % (US EPA, 2005) of this fraction of total emissions. Fertilizer emissions vary temporally throughout the year in the US, reaching a maximum in the spring and a minimum in winter (Geobes et al., 2003). Approximately 60 % of emissions occur between March and June (Geobes et al., 2003). Temporal variability of NH 3 emissions influences the variability of ammonium (NH + 4 ), sulphate (SO 2− 4 ), and nitrate (NO − 3 ) aerosol formation in the atmosphere and wet and dry deposition of NH x (NH x =NH 3(g) + NH + 4(s, aq) ). Inorganic (i.e., NH + 4based) aerosol contributes significantly to total particulate matter with an aerodynamic diameter of < 2.5 µm (PM 2.5 ) (Edgerton et al., 2005;Malm et al., 1994;Chow et al., 2006). As a component of atmospheric nitrogen deposition, NH x contributes to soil acidification and aquatic and terrestrial eutrophication in sensitive ecosystems (Nihlgard, 1985;Paerl and Whitall, 1999).
Over the past decade, efforts have been made to incorporate the temporal variability of fertilizer emissions into inventories used for air quality modeling and assessments (Geobes et al., 2003;Gilliland et al., 2003). With respect to regional atmospheric models, emissions from fertilized soil have previously been processed and input separately from the physical and chemical components of the model that simulate transport, atmospheric transformations, and loss processes (Mathur and Dennis, 2003;Dennis et al., 2010). More recently, implementation of a bi-directional framework for NH 3 air-surface exchange in the Community Multi-scale Air Quality Model (CMAQ) (Bash et al., 2012) provides a basis for simulating fertilizer emissions in a process-oriented mode by incorporating the Environmental Policy Integrated Climate (EPIC) crop model, which includes soil biogeochemistry .
Calculation of soil emissions within the bi-directional flux framework implemented in CMAQ increases the temporal resolution of emissions and simulates the net soilcanopy-atmosphere exchange in a mechanistically representative manner. Representation of the air-surface exchange process in a resistance-based two-layer compensation point framework (Nemitz et al., 2001;Cooter et al., 2010) accounts for the competing processes of emission from the soil and uptake or emission by the leaf cuticle and stomata, which collectively produce the net exchange of NH 3 with the atmosphere. Accurate simulation of the complexities introduced by the bi-directional nature of NH 3 fluxes is particularly important for agricultural systems in which the overlying canopy may recapture a significant fraction of soil emissions, thereby regulating the net canopy-scale flux (Harper et al., 2000;Nemitz et al., 2000;Denmead et al., 2008;Bash et al., 2010). While this modeling approach represents a clear improvement over the temporally allocated fertilizer inven-tory approach previously used, testing and improvement of both the bi-directional flux and soil biogeochemical models requires process-level field measurements over extended periods of time that capture the range of soil, vegetation, and atmospheric conditions that drive short-term (i.e., postfertilization) and total growing season NH 3 fluxes within fertilized cropping systems. Further improvement of the biogeochemical model will require measurements for fertilizers currently in use and for those that will experience increased application in the future. For example, the use of nitrogen solutions mixed with urease inhibiting agents, which act to reduce NH 3 emissions, is likely to expand.
The purpose of this study was to examine the processes of NH 3 air-surface exchange in a fertilized corn (Zea mays) canopy over the majority of the growing season to characterize the dynamics of the emission process post-fertilization and investigate soil-canopy interactions once the canopy had reached peak leaf area. Here we describe the temporal variability of the net canopy-scale NH 3 fluxes, the quality of the flux and air concentration measurements, and the soil and vegetation chemistry underpinning the fluxes. We then focus on an intensive period of measurements within and above the canopy at peak leaf area to characterize soil/canopy interactions.

Site description
The measurement site was a flat, 200 ha agricultural field near Lillington, North Carolina, USA (35 • 22 35.7 lat., −78 • 46 45.1 long., 45 m elev.). Soils were primarily fine sandy loam (Exum series) with a texture of 21 %, 68 %, and 11 % sand, silt, and clay, respectively, and a bulk density of 1.42 g cm −3 over 0-10 cm depth. The entire field was planted in corn (Zea mays, Pioneer varieties 31G66 and 31P41, density of ≈ 70 000 plants ha −1 ) and fertilized with 20 kg N ha −1 ammonium polyphosphate (injected) over a period of 6 days from 18 April 2007 (DOY 108) to 23 April 2007 (DOY 113), starting at the northern perimeter of the field and ending at the southern perimeter. The field was fertilized again over the course of 5 days from 25 May 2007 (DOY 145) to 29 May 2007 (DOY 149), again from the north to south, with 134 kg N ha −1 surface applied urea ammonium nitrate solution (UAN) containing a urease inhibitor (Agrotain ® ). Fertilizer was drip applied ("side-dressed") to the soil surface adjacent to the plants. The canopy reached a peak leaf area index (LAI, single-sided) of 2.9 ± 0.6 m 2 m −2 and a maximum canopy height (h c ) of 2.2 m near 15 July 2007 (DOY 196)

Ammonia concentrations and fluxes
Above-canopy NH 3 fluxes were calculated using the modified Bowen ratio (MBR) method (Meyers et al., 1996). The MBR method assumes that the turbulent diffusivity of NH 3 is similar to the turbulent diffusivity of heat such that where F NH 3 is the air-canopy flux of NH 3 , w T is the kinematic heat flux, and C and T are co-located mean NH 3 concentration and air temperature differences between heights z 1 and z 2 above the canopy. The ratio of the heat flux to the temperature gradient is also known as the eddy diffusivity for heat. The vertical NH 3 gradient was measured with a continuous flow "AMANDA" (Ammonia Measurement by ANnular Denuder sampling with online Analysis; Wyers et al., 1993) wet denuder system. Gaseous NH 3 was collected from the sample airstream (30 L min −1 ) in a wetted continuous-flow annular denuder using a stripping solution of 3.6 mM NaHSO 4 . The aqueous NH 3 concentration was determined by an online detector based on a selective ion membrane and conductivity analysis (detection limit ≈ 0.02 µg NH 3 m −3 ) by sequentially sampling each denuder. In theory, this approach does not suffer from aerosol interference and because the denuders are located at the atmospheric sampling point and employ a short (30 cm (L) by 2.54 cm (O.D.)) Teflon inlet, sampling artifacts related to inlet surface losses are minimized. Fluxes were calculated using 30-min averaged concentrations measured between the lower and upper inlets, which increased in height from 0.3 to 2.5 m and 2.2 to 4.4 m, respectively, during the course of the experiment. The instrument was calibrated approximately bi-weekly using liquid NH + 4 standards of 0, 50 and 500 ppb (RICCA, Arlington, TX, USA), and systematic offsets between denuders were corrected by periodic collocation of the denuder sampling boxes. An R. M. Young (Model 81000V, Traverse City, MI, USA) sonic anemometer was positioned at z = 3.5 m to measure wind speed and direction, kinematic heat flux, and momentum flux. Processing of 10 Hz sonic data included spike removal, 30-min detrending and 2-D coordinate rotation prior to calculation of 30min averages. Data from the additional sonic anemometers referenced below were processed in the same manner. Air temperature was measured adjacent to each AMANDA sample inlet using fine wire copper-constantan thermocouples (OMEGA Engineering, Inc., Stamford, CT, USA). Fluxes and temperature profiles were reported as 30-min averages. The AMANDA Bowen ratio system was operational from DOY 142 to DOY 214 and is treated as the primary flux dataset in this analysis.
A second Bowen ratio system, operated for a shorter period of time , consisted of independent temperature and NH 3 gradients measured from a tower approximately 10 m west of the AMANDA tower. Ammonia concentration gradients were measured with duplicate dry annular denuders (URG Corp., Chapel Hill, NC, USA) at z = 1.7 m and 10 m. Two-hour samples were collected four times each day beginning at approximately 08:00, 10:00, 12:00, and 14:00 EDT. Glass impactors with cut-points of 2.5 µm aerodynamic diameter were used to remove particles from the air sample stream. Mass-flow controllers (Aalborg, Orangeburg, NY, USA) maintained sample flow rates at 20 L min −1 . Methods for annular denuder preparation and extraction were based on US EPA Compendium Method IO 4.2 (US EPA, 1997). Annular denuders were coated with a 1 % phosphorous acid solution and extracted with ∼ 10 mL of 18 M deionized water. Denuder extracts were analyzed by ion chromatography (Metrohm, Riverview, FL, USA) with a lower detection limit of 0.1 mg L −1 . Platinum resistance thermometers (Thermometrics Corp., Northridge, CA, USA) in aspirated radiation shields (Met One, Grants Pass, OR, USA) measured air temperature adjacent to denuder inlets at 0.1 Hz. A sonic anemometer (Model 81000V, R. M. Young) was placed at the upper height (10 m) to measure kinematic heat fluxes from 10 Hz wind speed and temperature data. Fluxes and temperature profiles were reported as 30-min averages.
In addition to above-canopy fluxes, an experiment was conducted in July, targeting peak leaf area, to examine in-canopy source/sink characteristics . In-canopy NH 3 concentrations were measured using duplicate phosphorous acid coated annular denuders (URG Corp.) mounted at 0.1, 0.3, 0.95, 1.5, and 2.25 m a.g.l., which were sampled for approximately 2 h each at an air flow rate of 20 L min −1 . Canopy height at the location of the profile measurements increased from 2.0 to 2.2 m during the incanopy experiment. Denuder preparation and extraction also followed US EPA Compendium Method IO 4.2 (US EPA, 1997). After sampling, denuders were extracted with 2.5 mL deionized water and analyzed for NH + 4 by ion chromatography (model DX120, Dionex Corporation, Sunnyvale, CA, USA). An ATI 3-D sonic anemometer (model SATI-3V, Applied Technologies, Inc., Longmont, CO, USA) was mounted on a height-adjustable bracket to measure wind speed and momentum fluxes within the canopy, and an R. M. Young sonic anemometer (Model 81000V) was located just above the top of the canopy. Soil and foliage fluxes were estimated from NH 3 concentration and turbulence profiles using an analytical first-order closure model .
Finally, 12-h integrated NH 3 concentrations were measured by annular denuder (URG Corp.) from DOY 142 to DOY 214 from on top of the mobile laboratory (4 m above the ground), which was approximately 30 m east of the AMANDA Bowen ratio system. Denuder protocols followed those of the in-canopy measurements, with the exception that air was sampled at a flow rate of 10 L min −1 through a Tefloncoated cyclone (URG Corp.) with an aerodynamic cut-point of 2.5 µm.

Foliage, leaf water and soil measurements
Single-sided leaf area index (LAI) was measured approximately bi-weekly by destructive (prior to canopy closure) and optical methods (LICOR Model LAI-2000, LICOR Biosciences, Lincoln, NE, USA) along with plant height. Stomatal resistance was measured by leaf cuvette on three occasions corresponding to LAI values of 0.25, 1.0, and 2.9 (peak) m 2 m −2 (LICOR Model LI6400 photosynthesis system). Measurements were taken on shaded and sunlit leaves over a range of photosynthetically active radiation (300-1800 µmol m −2 s −1 ). Canopy wetness was quantified by leaf wetness sensors (Campbell Scientific Model 237 wetness sensing grid) positioned below and at the midpoint of the canopy, which provided 30-min data. On three occasions after canopy closure, leaf wetness (g H 2 O m −2 ) was quantified by destructively sampling leaves from the lower, middle, and upper canopy from sunrise until the canopy was dry as confirmed by visual observation of the absence of droplets. Wetness was calculated as the difference in water mass between wet and dried leaves normalized for leaf area.
Leaf apoplast NH + 4 and H + concentrations were measured approximately biweekly using the vacuum infiltration technique of Husted and Schjoerring (1995). Leaf samples were taken from the upper canopy during late morning, after the canopy had dried, and extracts were composited until sufficient volume was collected for pH and NH + 4 analyses. Duplicate composite samples were analyzed, typically consisting of 5 mL of sample in each (≈ 50 g fresh vegetation). Dilution of the apoplast during extraction was quantified by extracting leaves with indigo carmine and measuring the difference in absorbance at 610 nm (Ocean Optics Model USB2000 spectrometer, Dunedin, FL, USA) between the infiltrate and extracted sample. A 1 mL subsample was analyzed immediately for pH using a PHR-146 microelectrode (Lazar Research Laboratories, Inc., Los Angeles, CA, USA). The remaining sample was then analyzed in duplicate for NH + 4 by ion chromatography (Dionex Model DX-120).
Leaf surface water consisting of dew and guttation droplets was collected on several occasions by direct syringe sampling and analyzed for pH and NH + 4 using the methods described above. An artificial leaf droplet experiment was conducted on DOY 217 (LAI ≈ 1.8 m 2 m −2 ) in which deionized water adjusted to a pH of 4.2 by addition of phosphorous acid was added to leaf surfaces as small droplets and recollected over a period of 5-35 min as a time series. The initial pH of 4.2 is consistent with the chemistry of local rainfall (National Atmospheric Deposition Program/National Trends Network, 2012). At the beginning of the experiment, approximately 50 mL of solution was quickly distributed via syringe among 20 green sunlit leaves selected randomly within a 10 m × 10 m section of the canopy. At each time interval, approximately 2 mL of the solution was retrieved from random locations. Aliquots of the composite sample at each time step were analyzed immediately for pH and frozen for later analysis of dissolved ions, including potassium (K + ), magnesium (Mg 2+ ), and calcium (Ca 2+ ), by chromatography. The experiment was conducted between 10:00 and 13:00.
Soil samples were collected approximately weekly at 12 locations within 100 m of the NH 3 flux measurement tower: 6 locations to the north and south of the tower, respectively. At each location, three soil samples (0 to 5 cm depth) were taken at the mid-point and sides of the planting row (i.e., adjacent to plants). Samples were composited and subsamples were analyzed for NH + 4 , NO − 3 , H + and moisture. A 5-g subsample of field moist soil was extracted within 1 h of collection in 25 mL of 1M KCl. Extracts were frozen and later analyzed for NH + 4 and NO − 3 by colorimetry (Lachat QuickChem Model 8000 Flow-Injection Autoanalyzer, Lachat Instruments, Loveland, CO, USA). Soil pH was measured within 1 h of sample collection in a 1 : 5 soil : deionized water mixture. Gravimetric soil moisture was determined by weight loss after heating 10 g of soil for 48 h at 60 • C. Particle size fraction and bulk density were also determined at 0-10, 10-20, and 20-30 cm layers using standard methods. Soil volumetric water was measured continuously at 10 cm and 20 cm depth using water content reflectometers (Model CS615, Campbell Scientific, Logan, UT, USA). Soil temperature averaged over 4-8 cm depth was measured continuously using chromel-constantan thermocouples (Campbell Scientific).
H + and NH + 4 concentrations in the soil solution, leaf apoplast, and leaf surface water were used to calculate respective NH 3 compensation points (χ) according to where T is soil or air temperature in kelvin and H + , NH + 4 , and χ concentrations are in mol L −1 (Nemitz et al., 2000). χ is subsequently converted to units of µg NH 3 m −3 . Plants may act as a source or sink for NH 3 , depending on the difference between the atmospheric concentration and the gaseous concentration in the leaf substomatal cavity. The gaseous concentration in the leaf substomatal cavity is known as the stomatal compensation point (Farquhar et al., 1980). Emission occurs if the mole fraction of NH 3 in the atmosphere is lower than the stomatal compensation point, while uptake (deposition) occurs when the mole fraction of NH 3 in the atmosphere is higher than the stomatal compensation point. Compensation points for soil and leaf surface water can be similarly calculated from measured NH + 4 and H + concentrations. The ratio of NH + 4 to H + concentration in Eq.
(2) is termed the emission potential ( ), which is unitless and temperature independent.

Data processing
Relative to the time-integrated manual denuder measurements, the AMANDA dataset is more complete with respect to temporal coverage and resolution and therefore is used as the primary dataset in the proceeding analysis of data quality and flux processes.
It is well known that the modified Bowen ratio technique is most uncertain under conditions of weak gradients and fluxes from which the eddy diffusivity is derived, particularly during periods of rapid stability transition such as evening and early morning. In the present study, eddy diffusivities were derived from the kinematic heat flux and temperature gradients. When both are near zero, the direction of the heat flux and temperature gradient may be opposite in sign, yielding a negative eddy diffusivity. Periods of negative eddy diffusivity account for ≈ 18 % of the total number of NH 3 gradient measurements (N = 2294), leaving a total of 1870 flux measurement periods for analysis. These remaining data were flagged according to a set of micrometeorological criteria to identify periods when the fluxes were subject to greater uncertainty due to limitations of the micrometeorological assumptions underlying the gradient method and fetch considerations. To indicate periods of limited turbulent mixing, data were flagged for wind speed (u) < 0.5 m s −1 , which corresponds to 22 % of observations. As mentioned above, stability transition in the morning and afternoon is characterized by low kinematic heat fluxes and small temperature gradients. These periods were identified as having an Obukhov length | L |< 0.2 m, which corresponds to 9 % of observations. Using the footprint model of Hsieh et al. (2000), fluxes were flagged if the corn field contributed < 80 % of the flux at the upper height of the NH 3 concentration gradient. Owing to the large size of the field, only another 8 % of observations failed this requirement after filtering for L. Periods in which the sampling tower fell within the aerodynamic footprint of the on-site mobile laboratory (wind direction = 85-95 • ) and when winds approached the sampling tower from the opposite side of the sample inlets (wind direction = 355-5 • ) were also flagged. These periods of potential flow obstruction collectively accounted for 3 % of observations. Accounting for periods of instrument maintenance and malfunction, as well as negative eddy diffusivities, flux data coverage between DOY 142 and 214 is 54 %. Of these observations, 70 % meet the meteorological criteria described above. Flagged data denote periods of greater uncertainty in the fluxes but were included in the following analyses.

Gradient detection limit and flux uncertainty
The quality of the fluxes was further assessed by estimating the percentage of observations above the operational gradient detection limit and total uncertainty in the flux. The ) measurements from which the uncertainty of the modified Bowen ratio NH 3 flux was calculated. In the top graph, upper and lower x-axes represent NH 3 air concentration and log-transformed concentration, respectively. In the lower graph, the x-axis represents the kinematic heat flux. Symbols represent bin-wise average values, and error bars represent ±1 standard deviation of x. gradient detection limit was quantified as outlined by Wolff et al. (2010) based on the precision of the air concentration measured during periods in which the denuder sample boxes were collocated (i.e., cosampling periods). The residuals of an orthogonal linear regression of the concentrations measured by the two sample boxes represent random error in the air concentration. The standard deviation of the residuals provides a measure of the total precision of the concentration gradient (σ C ), which is a measure of the gradient detection limit (Wolff et al., 2010). To examine the relationship between σ C and air concentration, the observations were ordered by air concentration (C = µg NH 3 m −3 ) and segregated into 5 bins of N = 45 observations, from which σ C was calculated. As illustrated in Fig. 1, σ C increases linearly with the log-transformed air concentration. The resulting function of σ C versus C was used to predict the gradient detection limit for each flux observation. Using this approach, 70 % J. T. Walker et al.: Ammonia fluxes over corn (N = 1292) of the observed 30-min average gradients from which the flux was determined were above the gradient detection limit.
Uncertainty in the flux (σ F ) was quantified as the combination of uncertainty in the concentration gradient (σ C ) and eddy diffusivity for sensible heat (σ K h ) following Gaussian error propagation (Walker et al., 2006): Uncertainty in the concentration gradient was determined as described above, and uncertainty in the eddy diffusivity (K h ) was determined from the uncertainty in the temperature gradient (σ T ) and the kinematic heat flux (σ w T ): For consistency, σ T and σ w T were determined using the orthogonal least squares regression procedure described above. σ w T was derived from collocated measurements of w T in which duplicate R. M. Young sonic anemometers were vertically matched in height and separated horizontally by approximately 0.5 m. To assess the potential relationship between the precision and magnitude of the heat flux, absolute values of w T were ordered and segregated into 5 bins of N = 30 observations. The linear relationship between the mean and standard deviation of the bin-wise heat fluxes ( Fig. 1) was then used to predict σ w T for each w T observation in the NH 3 flux dataset. Similarly, σ T was derived from collocated duplicate thermocouple measurements of air temperature (T ). σ T did not exhibit a dependence on T ; therefore, the overall standard deviation of the orthogonal least squares residuals (σ T = 0.024 K, N = 104) was used in the flux uncertainty calculation (Eq. 4). Following this approach, the median error of the flux (i.e., (σ F /F ) · 100) is 43 % considering only chemical gradients that are above the detection limit. Uncertainty is greatest at night when the heat flux and temperature gradient are small (Fig. 2). During these periods, the median hourly uncertainty ranges from 65-85 %. Hourly median uncertainty is lowest between 09:00 and 14:00 (25-30 %) when fluxes of heat and NH 3 are large. Median uncertainty for all fluxes, including chemical gradient below detection limit, is 59 %. However, this number represents periods of large uncertainty at night corresponding to fluxes near zero. Median uncertainty of all daytime fluxes only, which captures the majority (> 90 %) of the total daily flux, is 46 %.

Comparison of measurement techniques
AMANDA air concentrations were compared to manual denuder measurements comprising three sets of data: 12-h (D 12 ) integrated samples collected at a single height, 2-3 h samples from the in-canopy profile system (D IC ) and 2-h samples from the second Bowen ratio system (D BR2 ). Precision of the D IC and D BR2 systems, determined as the median relative percent difference of duplicate measurements, was 4.6 % (N = 45) and 11.6 % (N = 30), respectively. Precision of the D 12 integrated samples is ∼5 % (Robarge et al., 2002).
Prior to comparison, concentrations were normalized using the linearized vertical concentration gradient (dC/d ln z) to adjust for differences in measurement heights. For the D 12 comparison, the AMANDA gradient was interpolated for the D 12 height. For the D IC comparison, the concentration of the lower AMANDA denuder was compared directly to the uppermost denuder in the canopy profile system. AMANDA and D BR2 were compared by normalizing both gradients to 1 m. Data were then filtered for 80 % completeness of AMANDA coverage for a given period of denuder sampling, yielding 112 observations for comparison. Linear regression analysis of AMANDA versus manual denuder concentrations (Fig. 3) yielded a slope of 0.96 and a statistically non-significant (P = 0.1) intercept of −0.19 µg m −3 , indicating good overall agreement between methods. Scatter in the data is attributed to spatial variability in emissions, which would make agreement between methods sensitive to horizontal and vertical separation, particularly following fertilization or periods during which the flux changed rapidly.
Comparison of AMANDA versus manual denuder fluxes is limited to only 13 periods during which AMANDA coverage was complete. Given the difficulties inherent in measuring NH 3 fluxes and large natural variability in the data, such a limited dataset precludes rigorous comparison. However, mean fluxes were 64.0 and 67.0 ng NH 3 m −2 s −1 for the AMANDA and manual denuder systems, respectively. Using the nonparametric Wilcoxon rank sums test, the means are not significantly different at the 10 % level, suggesting that on average large systematic differences between the two techniques are not expected. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Fig. 3. Comparison of AMANDA NH 3 concentration measurements to 12-h (D 12 ) integrated samples, 2-3 h samples from the in-canopy profile system (D IC ) and 2 h samples from the second Bowen ratio system (D BR2 ). Concentrations were normalized to a common level using the linearized concentration gradient (dC/dlnz) to account for differences in measurement heights. Error bars represent ±1 standard deviation of the mean half-hourly AMANDA measurements. 39 Fig. 3. Comparison of AMANDA NH 3 concentration measurements to 12-h (D 12 ) integrated samples, 2-3 h samples from the in-canopy profile system (D IC ) and 2-h samples from the second Bowen ratio system (D BR2 ). Concentrations were normalized to a common level using the linearized concentration gradient (dC/d ln z) to account for differences in measurement heights. Error bars represent ±1 standard deviation of the mean half-hourly AMANDA measurements.

Fluxes
Cumulative fluxes (Fig. 4a) were calculated over a period of approximately 10 weeks from the point of UAN fertilization (DOY 149) to DOY 214. To avoid biasing the results toward hours of the day during which there are a higher number of valid observations, which may occur if daily mean or median fluxes are used, cumulative fluxes were calculated from mean and median diurnal flux profiles. To capture the general temporal features of the flux following fertilization, diurnal profiles were summarized by four periods (Fig. 4a): DOY 149 to 161, DOY 162 to 172, DOY 173 to 180, and DOY 181 to 214. Using this approach, daily mean and median NH 3 fluxes yield cumulative total N losses of 8.3 % and 6.1 %, respectively, of the 134 kg N ha −1 surface applied as UAN. For nitrogen containing solutions, including surface applied UAN, an NH 3 -N emission factor of 8 % has been used in recent US emission inventories (Geobes et al., 2003;ECETOC, 1994. This value is consistent with the controlled experiments of Watson et al. (2008), which examined fractional NH 3 -N loss (≈ 8 % averaged over all treatments) from UAN solutions in a variety of soil types at different temperatures.
The urease inhibiting action of nBTPT, the active ingredient in Agrotain ® , has been shown to reduce and delay NH 3 volatilization from urea in a number of field trials (Watson et al., 1994;Rawluk et al., 2001). NH 3 emissions from UAN + nBTPT have been studied less extensively. However, field (Chadwick et al., 2005) and laboratory (Watson et al., 2008) experiments indicate reduced and delayed emissions relative to UAN solutions without nBTPT, though % reduction of NH 3 volatilization is lower than observed for  urea + nBTPT owing to a lower % nitrogen associated with urea in UAN. In field trials on tillage soils, NH 3 volatilization from surface applied UAN + nBTPT was 44 % lower, on average, than volatilization from UAN only (Chadwick et al., 2005) over a period of 21 days after fertilization. Reductions were slightly larger in laboratory experiments, averaging 65 % across soil types, nBTPT concentration, and soil temperature (Watson et al., 2008), also over a 21 day period. Reducing the 8 % emission factor assumed for UAN solutions by 44 % and 65 % yields an expected range of percent NH 3 -N loss from UAN + nBTPT solutions of 5.3 % and 2.8 %, respectively. Our estimates of 8.3 % (6.1 %) total loss and 4.5 % (3.4 %) at 21 days from fertilization derived from daily mean(median) fluxes are consistent with this range of values. Note that the experiments of Watson et al. (2008) did not account for the recapture of soil emissions by overlying vegetation, which is in contrast to the net fluxes presented here. Though LAI (Fig. 4c) was still relatively low (≈ 1.75) at 21 days after fertilization, soil emissions were likely higher than the net canopy-scale emissions from which % N loss was calculated.
The majority of the cumulative flux occurred in two distinct periods (Fig. 4a). Fluxes increased immediately upon the completion of fertilization (DOY 149) and remained elevated for a period of several days. A decrease in emissions during the second week after fertilization was followed by a second period of increased emissions beginning in the third week after fertilization and lasting for approximately one week. We primarily attribute this temporal pattern to fertilizer characteristics. The emission pulse immediately following fertilization is likely primarily associated with the NH + 4 fraction of the UAN fertilizer. The highest daily median fluxes were observed following the first post-fertilization rainfall (DOY 153/154). Emissions decreased dramatically as soil temperatures decreased in response to the passage of a cold front on DOY 160 after which overcast skies persisted for a period of several days. The second period of elevated emission (DOY 172-180) developed as the soil warmed for several days following a light rainfall on DOY 162 (Fig. 4b).
This delayed second emission pulse likely reflects the urease inhibiting properties of the nBTPT (Agrotain), which has been shown to delay hydrolysis of urea for up to 3 weeks in some soils (Giaocchini et al., 2002).
The observed temporal pattern of emissions is to some extent supported by the corresponding pattern of soil extractable NH + 4 (0-5 cm depth) concentrations (Fig. 4c). As expected, soil NH + 4 concentrations were highest immediately following fertilization (DOY 149). The rainfall on DOY 153/154 likely moved some fraction of the fertilizer downward into the soil profile, potentially suppressing emission during the first week following fertilization relative to what would have been observed in the absence of rain. NH + 4 concentrations (0-5 cm) were much lower a week after fertilization and increased over the next two weeks before gradually decreasing over the remainder of the growing season. The apparent increase in NH + 4 between DOY 154 and 168 may reflect the conversion of urea to NH + 4 and the decreasing efficiency of the urease inhibitor with time, which in combination with warming of the soil stimulated the second period of emissions.
To examine the temporal variability of fluxes in more detail, data were divided into two periods. The first period includes DOY 149-180 (period A), during which the effects of fertilization are distinct. The second period covers DOY 181-213 (period B), during which the canopy reached maximum LAI (Fig. 4c) and the effects of fertilization on the canopy-scale flux had diminished. Fluxes from both periods are summarized in Table 1 along with soil, vegetation, and meteorological measurements. During period A, 53 % of the variability in daily median fluxes is explained by a linear combination of soil temperature (38 %) and volumetric water (15 %) at the 10 cm soil depth. Fluxes are positively correlated with both variables. Consistent with other studies, fluxes tended to increase exponentially with temperature (Roelle and Aneja, 2002), which results from the temperature dependence of the gas/aqueous partitioning of NH 3 in the soil pore water (Dawson, 1977) through Henry's law. The rate of urea hydrolysis and subsequent production of NH + 4 increases with increasing soil moisture (Vlek and Carter, 1983), while the resistance to transfer of NH 3 gas through the soil profile decreases (Moldrup et al., 1999). In combination, these processes yield a positive correlation between soil moisture and NH 3 flux, though their relative importance changes with time after fertilization as the effectiveness of the urease inhibitor decreases. Interpretation of the correlation between soil moisture and NH 3 emission during this period is therefore not straightforward.
At the diurnal timescale, maximum hourly average fluxes of ≈ 700 ng N m −2 s −1 occurred near mid-day, coincident with the daily maximum in friction velocity (u * , Fig. 5). Linear correlation with friction velocity explains 31 % of the variability in log-transformed half-hourly fluxes, which are also positively correlated with air temperature and soil temperature after accounting for their respective collinearity with friction velocity. Correlation with air temperature explains more of the flux variability (11 %) than soil temperature (6 %), suggesting that the NH 3 source is very near the soil surface during this period. Friction velocity and temperature are linked through the compensation point. Large u * flushes NH 3 out of the canopy, promoting more emission resulting from the compensation point, and higher temperatures correspond to higher compensation points. While including interaction terms for u * and temperature in the regression procedure should control for most of this collinearity, in reality the total flux variability may be more evenly partitioned between these two variables. Soil volumetric water does not explain a statistically significant fraction of the variability in half-hourly fluxes.
Ammonia air concentrations are generally higher at night when the boundary layer is shallow, though concentrations increase rapidly in the morning with the post-sunrise increase in the momentum flux. This spike in concentration, which is accompanied by an emission pulse, likely represents the upward mixing of NH 3 that accumulates near the ground (i.e., below the lowest NH 3 measurement height) under calm nighttime conditions  and emissions from drying moisture on the soil surface. This period of elevated concentration in the morning coincides with the drying of the surface and rapid decline in relative humidity (RH) (Fig. 5). The latter point is discussed in more detail below.
During period B, 5 to 10 weeks after fertilization, net emission was still observed though mid-day peak fluxes had declined to ≈ 125 ng N m −2 s −1 and day-to-day variability of the net canopy-scale flux was not as well correlated with soil temperature or soil moisture. At the diurnal timescale, fluxes were again larger during the day when friction velocity was high. A prominent feature of the diurnal pattern is a peak in the NH 3 air concentration and emission flux at approximately 09:00. This period coincides with surface layer mixing after sunrise as wind speed and friction velocity increased as well as drying of the surface (Fig. 5). This pattern likely results from a combination of venting of the canopy and release of NH 3 from moisture that accumulates overnight at the soil surface (e.g., Sutton et al., 1998). This process and Table 1. Summary of NH 3 fluxes (ng m −2 s −1 ), NH 3 concentration at (z − d) = 1 m ([NH 3 ], µg m −3 ), canopy temperature (T c , • C), soil temperature averaged over 4-8 cm depth (T soil , • C), wind speed at z = 3.5 m (U , m s −1 ), soil volumetric water at 10 cm depth (H 2 O vol , cm 3 cm −3 ), NH + 4 concentration in the leaf apoplast solution (NH + 4 ) s , leaf surface water (NH + 4 ) d , and soil pore water (NH + 4 ) soil (µM), along with pH and emission potential ( = NH + 4 /H + ) of the corresponding media. the variability of the fluxes during period B are further explored below through an examination of the in-canopy NH 3 source/sink characteristics and soil/vegetation chemistry.

Leaf apoplast
Over the course of the growing season, (Fig. 6), apoplast NH + 4 concentrations decreased gradually after fertilization until the onset of senescence (≈ DOY 200) at which time NH + 4 concentrations increased to levels observed just after fertilization. While apoplast extractions were performed only on green leaves during the senescing period, it is possible that the chemistry of the apoplast solution was influenced to some extent by cellular breakdown. Results during this period should therefore be interpreted cautiously. Apoplast pH exhibited the opposite pattern, increasing after fertilization until the onset of senescence, then decreasing slightly. The net effect was a relatively constant apoplast emission potential ( s = NH + 4 /H + ) though mean and median s were slightly higher during period A, after fertilization, compared to period B (Table 1). In general, the emission potential of the leaf apoplast was much lower than the soil pore or leaf surface water (Table 1). Over the typical range of daytime temperatures (25-35 • C), the overall median s (146) corresponds to stomatal compensation points (χ s ) in the range of 1-3 µg NH 3 m −3 (Fig. 7).
The range of measured apoplast s values (33-452) overlaps with the cuvette measurements of Farquhar et al. (1980) for corn, but observations from this study are lower than the average s (1186 ± 395) inferred from micrometeorological measurements by Harper and Sharpe (1995). Massad et al. (2010) point out that extraction techniques, such as employed in this study, generally yield lower estimates of Fig. 5. Mean hourly NH 3 flux, NH 3 air concentration at (z − d) = 1m, friction velocity (u * ), soil temperature averaged over 4-8 cm depth, relative humidity (RH), and leaf wetness for Periods A and B. Leaf wetness represents the fraction of the hourly period during which the leaf wetness sensor indicated that moisture was present. Error bars represent ±1 standard deviation of the mean. 41 Fig. 5. Mean hourly NH 3 flux, NH 3 air concentration at (z − d) = 1 m, friction velocity (u * ), soil temperature averaged over 4-8 cm depth, relative humidity (RH), and leaf wetness for Periods A and B. Leaf wetness represents the fraction of the hourly period during which the leaf wetness sensor indicated that moisture was present. Error bars represent ±1 standard deviation of the mean. s compared to gas exchange approaches, possibly as a result of regulation of H + and NH + 4 post-collection of the leaf material and during extraction or because gas exchange is dominated by localized hotspots. Differences between our results and those of Farquhar et al. (1980) and Harper and Sharpe (1995) may also be related to fertilizer amount and plant growth stage.

Leaf surface water
The leaf surface water, composed of dew and guttation, exhibited mean and median d of 4327 and 4502, respectively. NH + 4 concentrations were approximately a factor of 2 lower on average than concentrations measured in the leaf apoplast, ranging from 0.6 to 166 µM. The much higher emission potential relative to the apoplast was instead a result of higher pH, which ranged from 6.6 to 8.7. The observation of high pH in natural dew and guttation was further explored in a controlled experiment in which deionized water, adjusted to a pH of 4.2 by addition of phosphorous acid, was added to green sunlit leaves as small droplets and re-collected over a period of 5-35 min as a time series.
The pH of the artificial leaf droplets increased rapidly and appeared to reach a maximum increase of 2 to 4 pH units within 30 min; droplets applied to a clean surrogate surface maintained a constant pH (Fig. 8). Final pH values after 35 min were slightly lower than natural dew droplets and guttation but confirmed a large neutralizing capacity of the leaf surface. While consistent temporal trends in dissolved ions were not clear from this limited dataset, general increases in K + , Mg 2+ and Ca 2+ were observed, with very high concentrations of K + noted in a few samples ( Table 2). The high concentrations of these base cations suggest two possible mechanisms responsible for the high observed pH, the first being H + exchange. As explained by Mecklenburg et al. (1966), water in dew and guttation dissolves CO 2 from the atmosphere to form carbonic acid, which then dissociates, releasing H + . These protons are exchanged with base cations through the cuticle on sites within the leaf cell wall to form alkaline carbonates, which may remain in solution or precipitate back to the cuticle surface. Second, solubilization of pre-existing alkaline particles, from sources such as soil, on the cuticle would also increase the pH (Hutchinson et al., 1986). This process is very likely, given that the system is composed of a short canopy over tilled soil that was relatively dry for extended periods of time during the growing season. Furthermore, the presence of soil particles on the leaves was visually observed, particularly in the lower canopy. Our measurements were not sufficiently detailed to resolve possible variations in pH with vertical position in the canopy, though a higher surface neutralizing capacity of leaves in the lower canopy is expected given that rain is more likely to clean leaves in the upper canopy more completely. Interaction of other trace compounds with the water layers will further complicate the leaf water chemistry (Flechard et al., 1999;Burkhardt et al., 2009). Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |  High neutralizing capacity of leaf surface droplets has also been observed for leaves of other crops (Hutchinson et al., 1986), particularly species with wettable leaf surfaces. However, the pH of leaf surface water on trees and grasses is typically lower, in the range of 3.5 to 6.5 (Hutchinson et al., 1986;Burkhardt and Eiden, 1990;Burkhardt et al., 2009). In general, measurements of leaf droplet pH reported in the literature are highly variable, which results not only from  vegetation characteristics but also the chemical characteristics of dry deposition to leaf and needle surfaces. High pH corresponds to a high emission potential, with d producing compensation points for leaf surface water (χ d ) of 10 to 30 µg NH 3 m −3 over the range of typical observed nighttime temperatures (18-28 • C, Fig. 7). As described in the following section, NH 3 concentrations within the vegetation canopy were highest immediately above the soil surface and decreased exponentially with height. During the four periods in which in-canopy concentrations were measured at night and under wet or partially wet canopy conditions in the early morning, concentrations at 10 cm above the soil were <10 µg NH 3 m −3 (2.7 to 7.8), indicating that, late in the growing season, leaf wetness would not have played a large role in the recapture of soil emissions at night.
To explore further the role of canopy wetness in the net canopy-scale flux, nighttime fluxes were compared between periods when the canopy was wet versus dry. At night, stomata are closed and the leaf level exchange should therefore be dominated by cuticular processes. To normalize the comparison for turbulence conditions, data were filtered to include only 0.01 < u * < 0.2 m s −1 , which produced approximately equivalent mean u * between the dry and wet periods. Comparison of leaf-level processes also assumes that differences between soil emissions are negligible. Because soil emissions should be larger immediately following rain events (i.e., high soil volumetric water content), data were also filtered to exclude days on which rain occurred. As shown in Table 3, net canopy-scale emissions at night were larger when the canopy was wet, indicating that the wetting of the canopy, and the accompanying high pH of the surface water, increased the cuticular resistance to NH 3 uptake and reduced the capacity of the canopy to recapture soil emissions. Because air concentrations were similar between wetness categories, differences in fluxes translate into differences in exchange velocities (V e = flux/concentration), with higher exchange velocities observed when the canopy was wet. Comparing periods, the percent difference between the mean fluxes during dry versus wet periods is much larger during period B, later in the growing season, compared to period A. This likely reflects the lower overall leaf surface area during period A and higher in-canopy concentrations of NH 3 resulting from larger soil emissions, which may exceed χ d more frequently compared to period B. It is also possible, however, that soil emissions were larger during period B wet conditions due to higher soil volumetric water. During period A, soil volumetric water content was similar during wet and dry canopy conditions (Table 3). Differences in canopyscale fluxes are therefore not likely due to differences in soil emissions. During period B, a difference in soil volumetric water between canopy wetness categories persists even after eliminating days on which rain occurred, with higher soil volumetric water content corresponding to wet canopy periods. We can not rule out the possibility that the higher fluxes during wet canopy conditions were partly due to higher soil emissions.
The relationship between leaf wetness and flux described above differs from previous studies, in which leaf wetness typically reduces the cuticular resistance to NH 3 deposition (see review of Massad et al., 2010). The most notable contrast to previous work is the very high pH of leaf surface droplets observed in the present study, with corresponding high d . Additional measurements are needed to assess the extent to which our observations of high surface water pH are site specific (i.e., driven by soil particles persisting on leaf surfaces) rather than a general feature of corn plants (i.e., driven by characteristics of transcuticular ion exchange). Furthermore, the extent to which bulk sampling of leaf surface Table 3. Summary of fluxes (ng NH 3 m −2 s −1 ), NH 3 concentration at z − d = 1 m (µg m −3 ), exchange velocity (V e , cm s −1 ), u * (m s −1 ), and soil volumetric water (S.V.W., cm 3 cm −3 ) during wet and dry night-time conditions for periods A and B. Data were filtered to include only 0.01 < u * < 0.2, which established approximately equal u * between wetness categories, thereby essentially normalizing the comparison for turbulence conditions. Data were also filtered to exclude days on which rain occurred. S.D. and N represent standard deviation and number of observations, respectively. water represents the cuticular NH 3 exchange characteristics of the entire canopy is unknown. It is possible that bulk sampling does not resolve the spatial variability in H + and NH + 4 concentrations that also affect the net canopy-scale flux. Microscopic (invisible) leaf water layers control the cuticular exchange at high RH when no visible leaf water is present, but their chemistry cannot be measured directly for comparison. While our comparison of fluxes during wet and dry periods suggests that the bulk chemistry of the leaf surface water is linked to the net canopy-scale flux, the representativeness of the bulk measurements with respect to overall cuticular exchange characteristics is uncertain. In addition to the uncertainties described above, there are two notable limitations of this assessment. First, U and u * are typically low at night in Eastern North Carolina during the summer. The mean u * values in Table 3 are < 0.1 m s −1 , corresponding to conditions of small heat fluxes and temperature gradients under which the modified Bowen ratio technique becomes highly uncertain (Fig. 2). Second, the majority of the leaf droplet measurements were collected late in the growing season (Fig. 6), which precludes a robust characterization of potential temporal variability of leaf surface J. T. Walker et al.: Ammonia fluxes over corn 993 water chemistry as the canopy grew. d was similar during periods A and B (Table 1).

Soil
The soil pore solution exhibited mean and median soil values of 120 000 and 15 400, respectively, which is consistent with the observations from fertilized soils compiled by Massad et al. (2010). Over the range of observed soil temperatures (20-40 • C), the median soil corresponds to soil compensation points (χ soil ) in the range of 60-500 µg NH 3 m −3 (Fig. 7). soil was considerably higher during period A compared to period B (Table 1), reflecting the high concentrations of soil NH + 4 measured immediately after fertilization. Over the course of the growing season, variability in soil was primarily driven by variability in soil NH + 4 , while soil pH remained relatively constant throughout the growing season. Soil NH + 4 concentrations were highest immediately following fertilization and decreased dramatically between the first and second sampling periods post-fertilization (Fig. 6), most likely in response to volatilization losses of NH 3 from the NH + 4 fraction of the UAN solution and heavy rainfall on DOY 154, which would have diluted the surface fertilizer and moved it down into the soil profile. NH + 4 concentrations then increased for a period of 2 to 3 weeks, likely as a result of conversion of urea to NH + 4 as the urease inhibiting efficiency of the Agrotain nitrogen stabilizer declined. NH + 4 decreased late in the growing season, after the onset of canopy senescence, to pre-fertilization levels.

Net canopy-scale fluxes
An intensive in-canopy experiment was conducted between DOY 187 and 213, capturing peak LAI (Fig. 4), to quantify the relative importance of the soil and foliage NH 3 exchange processes with respect to the net canopy-scale flux. The source/sink profile within the canopy was estimated from measured vertical NH 3 concentration profiles and turbulence parameters using a first-order analytical closure model, which is described in detail by Bash et al. (2010). A brief description of the model is included as Supplemental Material.
In-canopy concentration profiles consisting of 15 measurement periods are summarized by time of day in Fig. 9. All profiles followed a pattern of highest concentrations near the soil surface, decreasing rapidly with height in the lower quarter of the canopy, then more slowly to the top of the canopy. Based on the consistency of the profile shapes, the soil appeared to still be a constant source of NH 3 at this stage of the growing season. The shape of the profiles was driven by a combination of turbulent mixing within the canopy and the concentration of NH 3 just above the soil surface. Bash et al. (2010) showed that the source/sink profiles were indica- Fig. 9. In-canopy NH 3 concentration versus measurement height above the ground (z). Symbols re resent the mean and error bars represent ±1 standard deviation of the mean. N represents number observations at each height. Lines connecting observations are included only as a visual aid. "Ear morning" includes profile measurement periods ending before 0900, "late morning" includes profi measurement periods ending between 0900 and 1300, and "afternoon" includes profile measureme periods ending between 1300 and 1800. 45 Fig. 9. In-canopy NH 3 concentration versus measurement height above the ground (z). Symbols represent the mean, and error bars represent ±1 standard deviation of the mean. N represents number of observations at each height. Lines connecting observations are included only as a visual aid. "Early morning" includes profile measurement periods ending before 09:00, "late morning" profile measurement periods ending between 09:00 and 13:00, and "afternoon" profile measurement periods ending between 13:00 and 18:00. tive of emission from the soil and recapture of a varying fraction of emissions by the overlying canopy. The modeled average flux from the soil was 155 ng m −2 s −1 , and the net flux to the atmosphere was 37 ng m −2 s −1 , which agreed reasonably well with the micrometeorological flux measurements during the corresponding in-canopy measurement periods. The average total flux to the foliage (cuticular plus stomatal fluxes) was −118 ng m −2 s −1 , equivalent to 76 % of the average soil flux. The slight difference between our value of 76 % canopy recapture and the value of 73 % reported by Bash et al. (2010) reflects the use of average fluxes in the present analysis. Our results are consistent with other studies that have shown the recapture of NH 3 emissions from fertilized soil by the overlying crop canopy (Harper et al., 2000;Nemitz et al., 2000).

Soil flux
The flux from the ground (F g ) is driven by the difference between the soil (χ soil ) and canopy (χ c ) compensation points, the resistance to diffusion through the soil profile (R soil ), the resistance to diffusive transport across the air-side laminar boundary layer at the soil surface (R bg ) and the resistance to turbulent transfer within the canopy (R inc ), all of which can be viewed within a resistance modeling framework: Following the parameterizations for R inc , R bg , and R soil described by Pleim (2006) and Cooter et al. (2010), it is apparent that R soil is the limiting resistance. While the dynamics of the in-canopy atmospheric resistances will drive the variability of the soil flux on a short timescale (Bash et al., Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Fig. 10. Time series of soil volumetric water during in-canopy measurement period along with NH 3 air concentration at z = 10cm above the soil surface (top graph); correlation between NH 3 concentration at z = 10cm and soil volumetric water at 10 and 20 cm depth (bottom graph). Soil volumetric water data represent 30 minute averages and air concentrations represent 2 hour integrated samples. 46 Fig. 10. Time series of soil volumetric water during in-canopy measurement period along with NH 3 air concentration at z = 10 cm above the soil surface (top graph); correlation between NH 3 concentration at z = 10 cm and soil volumetric water at 10 and 20 cm depth (bottom graph). Soil volumetric water data represent 30-min averages, and air concentrations represent 2-h integrated samples. 2010), we find that the concentration of NH 3 measured just above the soil surface (z = 0.1 m) was strongly linked to soil moisture on a timescale of days to weeks. Figure 10 shows a time series of soil volumetric water content at 10 and 20 cm depth along with the concentration of NH 3 at 10 cm above the soil surface over the 26-day period during which in-canopy concentrations were measured. The surface air concentration follows the pattern of soil moisture quite closely, exhibiting high correlation with moisture at both 10 and 20 cm depths (Fig. 10). Ammonia concentrations just above the soil are higher immediately after the rainfall on DOY 191 relative to the samples preceding the rain event and decrease over time as the soil dries.
The correlation between the concentration of NH 3 just above the soil surface and soil moisture likely reflects a combination of chemical and physical processes. The NH 3 flux from the soil surface is linked to soil moisture via resistance to NH 3 diffusion through the soil profile, which decreases non-linearly with increasing soil moisture. As described by Sakaguchi and Zeng (2009), the soil resistance is a function of the ratio of the length of the dry soil layer, through which a gas must diffuse to the atmosphere, to the gas diffusion coefficient. As the soil dries, the length of the diffusion path increases, thereby increasing the resistance to the diffusive flux. This effect may be reflected in Fig. 10 as the soil dries following the heavy rainfall on DOY 191. On a timescale of days, the supply of NH 3 in the lower canopy, and therefore available for transport to the atmosphere, was apparently limited by the soil resistance as regulated by soil water content. It is possible, however, that the relationship shown in Fig. 10 represents the dynamics of both the soil resistance and NH + 4 pool. As indicated in Fig. 6, soil decreases during the period shown in Fig. 10, though only two soil samples are available. The temporal pattern of the NH 3 air concentrations shown in Fig. 10 may therefore partly reflect a coincident decrease in the soil emission potential. Urea hydrolysis would also exhibit a positive relationship with soil moisture. If significant concentrations of urea persisted this late into the growing season, the decrease in soil may partly reflect the relationship between urea hydrolysis and soil moisture content.

Foliage
The analysis presented by Bash et al. (2010) established that the foliage was a large sink for NH 3 emitted from the soil during the day. Here we extend that analysis by partitioning the bulk foliage flux into cuticular and stomatal components, using measured stomatal resistances for H 2 O and NH 3 compensation points derived from measured apoplast chemistry. The cuticular and stomatal fluxes are constrained by the total (net) canopy-scale and ground fluxes estimated from the analytical source/sink model of Bash et al. (2010). Following the resistance framework employed within CMAQ (Bash et al., 2012), the total (net) flux (F t ) above the canopy is equivalent to where χ c is the canopy compensation point (µg NH 3 m −2 ), χ a the air concentration measured at the top of the canopy (µg NH 3 m −2 ), F g the ground flux (ng m −2 s −1 ) estimated via the analytical source/sink model, F s the stomatal flux (ng m −2 s −1 ) and F w the cuticular flux (ng m −2 s −1 ). In this analysis, F t is also estimated via the analytical source/sink model and F s is calculated as where χ s is the stomatal compensation point estimated from measured apoplast H + and NH + 4 concentrations during period B (Table 1) using Eq. (2).
Equations for the aerodynamic (R a ), boundary layer (R b ) and in-canopy (R inc ) resistances were summarized by Pleim (2006). The NH 3 concentration profile in vegetation canopies is not constant, and NH 3 can be exchanged with foliage along this gradient . In order to reconcile this with a resistance model framework, R inc is split in Eq. (6) such that half of the resistance is applied between the soil and the canopy compensation point (χ c ) and the other half from the canopy to the atmosphere above. This was done to account for NH 3 emitted from the soil that is not subjected to the entire in-canopy aerodynamic resistance before exposure to foliage surfaces. After calculating χ c from Eq. (6) and F s from Eq. (7), F w is calculated as Ignoring one measurement period conducted entirely at night, the remaining (N = 14) in-canopy periods yielded mean (median) stomatal and cuticular fluxes of −32(−10) and −94(−84) ng NH 3 m −2 s −1 , respectively. The ratio of median and mean fluxes, respectively, suggests that stomatal uptake accounts for 12-34 % of the total foliage flux during the day, the balance being attributed to the cuticular flux. Dominance of the cuticular pathway is partly due to the relatively large minimum stomatal resistance of NH 3 , which was 154 ± 17 s m −1 at a PAR level of 1800 µmol m −2 s −1 and 30 • C as derived from measured leaf-level H 2 O resistances adjusted for differences in molecular diffusivity. The flux budget derived by Bash et al. (2010) in combination with the analysis presented here is summarized in Fig. 11.
Given the relatively high emission potential measured in the leaf surface water, apparent dominance of the cuticular flux during the day suggests that the chemistry of the wet cuticle is not necessarily indicative of the resistance characteristics of the dry cuticle. This is somewhat supported by the results in Table 3, which indicate lower net-canopy emissions (i.e., greater recapture by the leaf cuticle surface) when the canopy is dry. The in-canopy measurements summarized by Bash et al. (2010) are limited to a single measurement (DOY 199,(5)(6)(7)(8) during which the canopy was wet for the entire observation period. The source/sink modeling results of Bash et al. (2010) indicate that the canopy recaptured 43 % of the NH 3 emitted from the soil during this observation period, compared to 78 % recapture derived from median fluxes during the six observation periods in which the canopy was completely dry. During the DOY 199 morning measurement period, a canopy concentration (χ c ) of 7.9 µg m −3 is estimated from the source/sink resistance modeling described above, compared to a leaf droplet compensation point (χ d ) of 6.3 µg m −3 derived from leaf droplet chemistry measured on the morning of DOY 200, indicating that the surface droplets were capable of NH 3 uptake. Though the canopy was wet, partitioning of the foliage flux as described above indicated that F w (−8 ng m −2 s −1 ) was half of F s (−16 ng m −2 s −1 ). This single observation is generally consistent with the pattern of larger cuticular resistances under wet conditions reflected in Table 3.
Regarding the leaf surface water, the most prominent feature of the diurnal profile of the net canopy-scale flux (Fig. 5) during period B is a large emission peak in the morning, coincident with drying of the canopy and rapid mixing of the surface layer. Based on destructive sampling and weighing of wet leaves on two occasions between sunrise and 10:00, we estimated the water content of the middle and upper canopy to be 25-35 g H 2 O m −2 , which typically dried completely by 09:30. Based on the maximum measured concentration of NH + 4 (2990 µg L −1 ) in dew and guttation and a LAI of 3.0 m 2 m −2 , the maximum possible flux associated with complete drying of the canopy between sunrise and 09:30 is 17.6 ng NH 3 m −2 s −1 , compared to mean and median fluxes between 07:00 and 10:00, which encompasses the period during which the flux typically increases to its morning peak, of 137 and 43 ng NH 3 m −2 s −1 . While a much larger contribution to the net flux may be observed if the majority of the NH 3 volatilization were confined to the late stages of the canopy drying, our results suggest that the morning emission peak cannot be entirely explained by desorption associated with the drying of the canopy. It is more likely that the morning peak in NH 3 air concentrations and fluxes above the canopy results from a combination of canopy drying, drying of the moisture that accumulates on the soil surface at night, and venting of NH 3 that accumulates within the canopy and below the lowest NH 3 sampling inlet under stable nighttime conditions. This hypothesis is supported by the temporal pattern of changes in in-canopy concentration profiles during 6 sequential periods from 01:00 to 14:30, which showed a reduction in the in-canopy concentrations post-sunrise consistent with the above-canopy late morning peak in concentration .
Overall, the comparison of the emission potentials of soil, leaf surface water (dew/guttation) and foliage suggests that any emission from the canopy does not originate from the foliage. This is consistent with an intensive examination of