Recent advances in laser spectrometry offer new opportunities to
investigate ecosystem–atmosphere exchange of environmentally relevant trace
gases. In this study, we demonstrate the applicability of a quantum cascade
laser (QCL) absorption spectrometer to continuously measure ammonia
concentrations at high time resolution and thus to quantify the net exchange
between a seminatural peatland ecosystem and the atmosphere based on the
eddy-covariance approach. Changing diurnal patterns of both ammonia
concentration and fluxes were found during different periods of the
campaign. We observed a clear tipping point in early spring with decreasing
ammonia deposition velocities and increasingly bidirectional fluxes that
occurred after the switch from dormant vegetation to CO
Increased agricultural production and energy consumption over the last
century led to a dramatic increase in anthropogenic reactive nitrogen (N)
production (Erisman et al., 2008). Atmospheric N deposition can be a major
driver of change in most natural and seminatural ecosystems and may
considerably alter species composition, biodiversity, and ecosystem
functioning with regard to causing nutrient imbalances. As ammonia
(NH
Until now only little was known about the temporal and spatial variability of
NH
In this study, we demonstrate the applicability of a QCL absorption spectrometer to continuously and quickly measure turbulent
fluctuations at background-level NH
Tower-based fast response measurements of ammonia concentrations were
conducted from 18 February to 8 May 2014 at an ombrotrophic, moderately
drained peatland located in northwestern Germany near the city of Meppen
(52
From 1981 to 2010 the annual average air temperature (
Ammonia concentrations were measured at high temporal resolution with a QCL
absorption spectrometer (model mini QC-TILDAS-76) from Aerodyne Research,
Inc. (ARI, Billerica, MA, USA). They were recorded as mol NH
To prevent damage to the laser cell from particles, it is necessary to
filter the ambient air. Because of the stickiness of NH
Schematic overview of the measurement setup. An ultrasonic
anemometer (measurement height 2.5 m) is mounted closely to the heated “inertial
inlet”
box (39 cm
An ultrasonic anemometer (model R3, Gill Instruments, Lymington, UK) was installed at 2.5 m above ground. The inertial inlet box was mounted next to it on the northern side, a less common wind direction at the site (5.6 % of the campaign), placing the sample air inlet westward 40 cm below the center of the sonic anemometer array. To protect the inlet from rain, a tee inlet was attached, which allowed to feed in calibration gas (see above).
Additional measurements of NH
Sonic anemometer data were recorded by the EddyMeas software, which is an embedded application of the software package EddySoft (Kolle and Rebmann, 2009). The QCL was controlled by Aerodyne's TDL Wintel software with ammonia concentration data being recorded at a frequency of 10 Hz on the QCL computer. Anemometer and concentration time series needed alignment to a reference timestamp before the software EddyPro (LI-COR Inc.) could be used to compute half-hourly exchange fluxes. Block averaging and 2-D coordinate rotation were applied.
The effective time lag between the data series of the vertical wind speed
(
The ogive method from Ammann et al. (2006) was applied to empirically
investigate the high-frequency damping of NH
The random flux error was computed according to the method of Finkelstein
and Sims (2001) and the corresponding limit of detection was determined
from Langford et al. (2015) as 1.96 times the flux error (95 % confidence
limit), resulting in a median value of 7.75 ng N m
To guarantee a high level of data quality, fluxes were flagged according to
criteria presented in Mauder and Foken (2006). Data from 11 to 28 April were
excluded from analysis due to several technical difficulties such as power
outages, pump failure, or insufficient temperature control inside the
analyzer housing. 33.9, 44.3, and 10.9 % of data were flagged
with grade 0, 1, and 2, respectively. Data with quality flags 0 and 1 were
used for further analysis, while flag 2 data were discarded. Because the
stationarity tests included in the flagging protocol by Mauder and Foken (2006) might not be applicable for NH
Characterization for four subperiods of the measurement campaign (I to IV)
with different NH
Half-hourly measured ammonia concentrations and fluxes. Vertical lines indicate beginning of periods listed in Table 1. Each period is represented by a box plot with bold horizontal lines showing the median, fine horizontal lines indicating lower and upper quartile values, whiskers representing 1.5 times the interquartile range, and dots with arrows indicating the mean and standard deviation.
Mean diurnal variation of ammonia concentrations separated into different periods. Dates are given as month/day.
A state-of-the-art dry deposition inferential model driven by measured
NH
On a half-hourly basis, ammonia concentrations ranged from 2 to 85 ppb with short-term (10 Hz) maxima reaching up to 110 ppb. The highest values were found at the beginning of March (period II) and at the beginning of April (end of period III), which coincided well with the peak of fertilization activities on nearby agricultural fields, including the spreading of organic manures from livestock farming (Fig. 3; for details on farming practices see Hurkuck et al., 2014). The base concentration level outside the fertilization periods ranged mostly between 7 and 15 ppb and is well represented by mean values of periods I and IV.
The highest mean diurnal variability of ammonia concentrations was found in period II with peak concentrations being observed in the late afternoon (> 30 ppb from 5 to 15 March; Fig. 4). Concentrations were lowest at night during the whole observation period. While in period III the average mean diurnal course exhibited less variability than in period II, almost stable concentrations at a low level (7 to 12 ppb) were found in periods I and IV.
The frequency distributions of wind directions and ammonia concentrations
for the whole observation period are shown in Fig. 5. The typical main wind
direction from the 200 to 280
Similarly diurnal variability of ammonia concentrations with peak values during afternoon over a variety of ecosystems has been observed by other authors (e.g., Sutton et al., 2000; Wolff et al., 2010). There are most likely several reasons for the observed pattern in this study. First and foremost, concentration levels are highly influenced by agricultural activities in the surrounding area. Farmers usually fertilize their land during the day, thereby causing ammonia volatilization, which is then transported and detected at the study site. With a stable nighttime planetary boundary layer, ammonia in the lower atmosphere is likely being deposited, causing decreasing concentrations. With no further penetration from higher layers containing higher loads, concentrations remain low from midnight to sunrise (Fig. 4). When temperatures rise, turbulent mixing of the planetary boundary layer starts and vertical exchange with higher layers increases, which, consequently, leads to rising ammonia concentrations over the day. In contrast, continuous sources like mechanically ventilated stables could cause an opposite pattern with the planetary boundary layer acting as a lid and leading to a concentration build up at night. As no information was available of ventilation types of stables, we assume that land-applied manure during the day dominates the concentration signal in Fig. 4. The conspicuous afternoon peak cannot be explained by turbulent mixing, but it is strongly linked to the northeastern wind sector, where agricultural sources are closest (approx. 1.5 km distance, see also Fig. S1 in the Supplement). Therefore this is assumed to be an artifact caused by these agricultural point sources.
Beside the strong influence of agricultural management on seasonal
concentration variability, temperature is usually a substantial driver.
Higher temperature indirectly leads to higher NH
Frequency distribution of wind direction for each period (left
panel: PI (18 February–4 March) in black, PII (5–15 March) in blue, PIII (16 March–10 April) in red, PIV (29 April–7 May) in orange) and ammonia concentration (right
panel) for each 10
Another driver for the observed concentration pattern might be leaf surface wetness. Peatlands in general, particularly during colder parts of the year, are moist environments where ammonia can easily be taken up by wet surfaces. However, it is released back to the atmosphere when surface water, e.g., dew, evaporates during morning and midday hours (Walker et al., 2006; Wentworth et al., 2014; Flechard et al., 1999; Wu et al., 2009; Burkhardt et al., 2009), which then causes again rising ammonia concentrations as observed in this study (Fig. 4). Other authors observed the concentration peak earlier in the morning, e.g., Walker et al. (2006) and Wolff et al. (2010), at arable land and grassland, respectively. The observed peak in our study might have been shifted because of the much higher humidity at our peatland site, indicated by small water pools, causing a longer duration of the evaporation process.
Mean diurnal variation of ammonia fluxes with standard deviation (upper left panel), separated by periods (upper right panel), precipitation (lower left panel), and days after last rain (lower right panel). Dates are given as month/day.
Mean diurnal cycles of ammonia concentration, ammonia flux, net radiation, air temperature, and friction velocity.
The monthly integrated ammonia concentration of 16.8 ppb in March from QCL measurements was in good agreement with those values measured by DELTA denuder and passive samplers. The latter approaches resulted in 14.5 and 15.2 ppb for DELTA and passive samplers, respectively, indicating their robustness and validity as low-cost methodologies for long-term air quality monitoring. As the time of exposure of DELTA denuders and passive samplers was not consistent with our QCL measurements due to instrument failures and campaign duration in February, April, and May, as well as due to highly variable concentrations during that time, we could not directly compare QCL numbers with those from the monthly integrating methods.
Half-hourly measured ammonia fluxes ranged mainly within
We observed considerable diurnal variability in ammonia fluxes throughout
the campaign. The average diurnal flux showed moderate uptake around
Springtime ammonia uptake at sites that were highly influenced by
fertilization and other local sources of ammonia in the surrounding area has
been reported earlier, e.g., by Mosquera et al. (2001), who found
considerable average deposition fluxes at their seminatural grassland site.
Beside management, they showed that higher surface wetness in spring
amplified local ammonia deposition, while net emission was typically found
only in summer (see also Wichink Kruit et al., 2007). An undisturbed
peatland site is likely to be a higher ammonia sink than managed grasslands
due to a lower nitrogen status and therefore lower ammonia compensation
point. However, with the chronically high atmospheric nitrogen loads caused
by agriculture over several decades our peatland is presumably not such an
efficient sink anymore. With this assumption and the decreasing uptake over
the progression of the measurement campaign (box plots in Fig. 3),
saturation effects might have played a role in biosphere–atmosphere exchange
characteristics. Flechard and Fowler (1998) already showed that peatland
vegetation under its common wet conditions may not necessarily be an “almost
perfect sink” as was reported by Duyzer (1994) due to nitrogen saturation
effects in heathland plants caused by persistently high ammonia deposition
(approx. 13 kg N ha
Regarding the diurnal flux patterns observed in this study, significantly different exchange characteristics have been reported elsewhere. For example, Horvath et al. (2005) and Wichink Kruit et al. (2007) found highest deposition rates in early morning hours due to dew formation at their grassland sites with decreasing deposition and even emission afterwards through drying leaves and stomatal release. Their finding, i.e., the ecosystem emits only under dry conditions, is contrary to our observations (Figs. 6 and 7).
Upper panel: deposition velocity; negative values indicate emission and maximum deposition velocity. Middle panel: canopy resistance; negative values are not shown, as they are not defined in the resistance framework and mostly correspond to phases of emission. Lower panel: canopy compensation point derived from measured data. Note that the upper and middle panels are based on a one-layer deposition only and the lower panel to a one-layer canopy compensation point framework (cf. Sutton et al., 1993).
Regarding the entire campaign, a significant tipping point in ammonia
exchange was found on 15 March, when higher deposition rates changed to
much lower deposition and bidirectional exchange (cf. Figs. 8 and 11;
Sect. 3.3). This tipping point followed 10 days after the onset of spring,
occurring on 5 March, as indicated by CO
The tipping point from higher to much lower ecosystem ammonia uptake was
also observed in the course of deposition velocities (
A continuous time series of the canopy compensation point was derived (Fig. 8, lower panel), using the relation by Nemitz et al. (2000) for a
single-layer canopy compensation point resistance model
The statistical significance of the ammonia flux dependency on
meteorological variables when classified into different ranges of values was
checked by means of a Kruskal–Wallis test (Table 2, see also Table S1 in the Supplement for
deposition velocity, canopy resistance, canopy compensation point, and
emission potential). The null hypothesis of identical population was
rejected in all cases when the
Data classification and results of Kruskal–Wallis test on the NH
Radiation and temperature are also driving local turbulence, i.e.,
Dependency of ammonia concentrations and fluxes on
Dependency of NH
Comparison of measured and modeled daily mean NH
The interdependency of the ammonia flux and concentrations is shown in Fig. 10. We separated emission from deposition periods and bin-averaged the concentration data (for an extract without bin-averaging see Fig. S2). Our observations are consistent with Milford (2004) who also found both increasing emission and deposition under elevated concentration, also when separating by wind direction (data not shown). It remains a matter of speculation whether the flux controls the concentration or vice versa as the relationship is highly controlled by plant nitrogen status and at least to some extent by biometeorological variables as mentioned above. However, Milford's (2004) statement that the concentration may still determine the flux during deposition periods, whereas during emission periods it may be the ammonia flux itself that is controlling the concentration, is also likely be applicable for our site during deposition periods. In emission periods, a coincidence of flux drivers and high concentration levels, which were high due to advection from the local sources, is the more realistic reason for the relationship in Fig. 10 (right panel). The nearest agricultural ammonia point source was 1.5 km away from the tower. With a measurement height of 2.5 m none of the sources were located within the flux footprint, and thus we can largely exclude effects from flux heterogeneity such as a direct contribution of the sources to the measured vertical fluxes. However, there might be still large-scale transport processes as outlined in a study by Mohr et al. (2015) that influence ammonia concentrations at the site. What is needed at this point is not just an observation on one field and one ecosystem, but a landscape-scale or regional-scale model of emission, dispersion, chemistry, exchange, and deposition that makes it possible to work on this question interactively.
The comparison of measured and modeled daily mean and cumulative half-hourly
ammonia fluxes is given in Figs. 10 and 11 (for 1 week with half-hourly
values see Fig. S4). We found a considerable mismatch between
modeled and measured fluxes with the latter showing higher uptake in period
I and lower uptake in period III than model outputs. In contrast, during
periods II and IV measured and simulated fluxes run fairly parallel, with
the exception of a short period of overestimated deposition during last week
of March, indicating that, on average, during these times the model is able
to reproduce the measured fluxes well. The interdependency of modeled
ammonia fluxes and measured concentrations is very similar to the measured
ones except that the model does not exhibit any emission and the fluxes are
generally lower with increasing concentrations (Fig. 10). We do not see the
larger measured deposition fluxes during periods I and II as being
indicative of faulty measurements. Instead, we suspect that under the local
pollution climate during the measurement period, the model predicts a too
large non-stomatal resistance (
Regarding the canopy compensation point (Fig. 8, lower panel) it must be
considered that in a unidirectional framework, this high canopy compensation
point increases the effective canopy resistance
The cumulative exchange after approx. 9 weeks of observation resulted in
total deposition of 911 and 857 g NH
EC flux measurements of ammonia using a QCL in combination with
an “inertial inlet box” were conducted at a peatland site in an agricultural
landscape. This methodology has high potential for (1) establishment in
long-term observation networks with the aim to improve nitrogen budgets and
transfer calculations at local and regional scale as well as (2) providing
deeper insight into the mechanisms of ammonia transfer and the ecosystems'
responses to ammonia loads in the atmosphere by offering continuous flux
observations at unprecedentedly high temporal resolution. In the present
study, we interpret changing diurnal patterns of ammonia concentration and
fluxes as well as a tipping point followed by decreasing deposition
velocities and increasing canopy resistance, as a sign of non-stomatal leaf
surface NH
Data will be archived and are available from the corresponding author on request.
Funding for this study from the German Federal Ministry of Education and Research (BMBF) within the framework of the Junior Research Group NITROSPHERE under support code FKZ 01LN1308A is greatly acknowledged. We thank Jeremy Smith, Jean-Pierre Delorme, as well as Ute Tambor, Andrea Niemeyer, and Daniel Ziehe for excellent technical support and conducting laboratory analyses of denuder and filter samples, respectively. Edited by: L. Zhang Reviewed by: two anonymous referees