Environ. Sci. Technol. 2008, 42, 2041–2046
Evaluation of Laser Absorption Spectroscopic Techniques for Eddy Covariance Flux Measurements of Ammonia J A M E S D . W H I T E H E A D , * ,† MARSAILIDH TWIGG,‡ DANIELA FAMULARI,‡ EIKO NEMITZ,‡ MARK A. SUTTON,‡ MARTIN W. GALLAGHER,† AND DAVID FOWLER‡ School of Earth, Atmospheric and Environmental Sciences, The University Of Manchester, Oxford Road, Manchester, M13 9PL United Kingdom, and Centre for Ecology & Hydrology, Bush Estate, Penicuik, Midlothian, EH26 0QB United Kingdom
Received June 29, 2007. Revised manuscript received September 25, 2007. Accepted December 16, 2007.
An intercomparison was made between eddy covariance flux measurements of ammonia by a quantum cascade laser absorption spectrometer (QCLAS) and a lead-salt tunable diode laser absorption spectrometer (TDLAS). The measurements took place in September 2004 and again in April 2005 over a managed grassland site in Southern Scotland, U.K. These were also compared with a flux estimate derived from an “Ammonia Measurement by ANnular Denuder with online Analysis” (AMANDA), using the aerodynamic gradient method (AGM). The concentration and flux measurements from the QCLAS correlated well with those of the TDLAS and the AGM systems when emissions were high, following slurry application to the field. Both the QCLAS and TDLAS, however, underestimated the flux when compared with the AMANDA system, by 64%. A flux loss of 41% due to chemical reaction of ammonia in the QCLAS (and 37% in the TDLAS) sample tube walls was identified and characterized using laboratory tests but did not fully account for this difference. Recognizing these uncertainties, the agreement between the systems was nevertheless very close (R2 ) 0.95 between the QCLAS and the TDLAS; R2 ) 0.84 between the QCLAS and the AMANDA) demonstrating the suitability of the laser absorption methods for quantifying the temporal dynamics of ammonia fluxes.
1. Introduction Ammonia is the most important alkaline gas in the atmosphere. It plays a significant role in atmospheric chemistry and contributes to the formation of secondary particulate matter. The impacts of ammonia on the environment include acidification and eutrophication of sensitive ecosystems (1, 2), whereas ammonium aerosols affect local and global radiation budgets (3) and contribute to particulate matter that has been associated with human health effects (4). It has been * Corresponding author phone: 44 (0)161 306 3914; fax: 44 (0)161 306 3951; e-mail:
[email protected]. † The University Of Manchester. ‡ Centre for Ecology & Hydrology. 10.1021/es071596u CCC: $40.75
Published on Web 02/14/2008
2008 American Chemical Society
widely established that agriculture is the largest contributor to ammonia emissions, accounting for an estimated 80% of emissions in the UK (5). However, significant uncertainties in the magnitudes of these emissions remain an issue. Accurate measurements of ammonia fluxes provide a useful tool in estimating agricultural area sources, such as fertilized land, and dry deposition to sensitive seminatural vegetation. Micrometeorological measurements of NH3 fluxes have always proven difficult, in part because of the reactive nature of the gas; however, there have been many advances in this area in the last 10 – 15 years. A widely used technique is the Aerodynamic Gradient Method (AGM). Measurements by this method were initially limited to manual batch sampling with a time resolution of 1–2 h (6, 7). However, the development of continuous wet chemical sampling methods in the 1990s made long-term, high temporal resolution measurements of NH3 fluxes possible. The AMANDA (Ammonia Measurement by ANnular Denuder with online Analysis) (8) employs wet rotating denuders, sampling at three heights above the ground in order to quantify the vertical concentration profile, from which AGM fluxes of NH3 may be calculated. The disadvantage of relying on measurements at multiple heights to calculate the flux is that it is difficult to quantify flux divergence errors caused by advection or chemical interactions. Relaxed Eddy Accumulation (REA) provides a method by which fluxes of trace gases such as NH3 can be measured at a single height with a slow response analyzer. Nemitz et al. (9) describe an REA system for NH3 flux measurements and present an intercomparison with an AMANDA system. Good agreement was found between NH3 concentration measurements at the same height, but the REA flux exceeded the gradient flux calculated from the AMANDA data. REA, however, requires greater analytical precision in NH3 measurements than the AGM technique, and current instruments are not yet sensitive enough to study the flux divergence resulting from chemical interactions (see also (10)). The eddy-covariance (EC) technique is the most direct, least empirical method for measuring vertical fluxes, but its application to ammonia exchange has been hindered both by the lack of instrumentation capable of the required fastresponse detection, and by the effects of inlet wall interactions due to the highly reactive nature of ammonia (11). Laser absorption spectroscopy (LAS) shows promise as a technique suitable for EC flux measurements of ammonia (12), and this has been demonstrated with some success in the field (13). In September 2004 and April 2005, a Quantum Cascade Laser Absorption Spectrometer (QCLAS; Aerodyne Research Inc., ARI, Billarica, Massachusetts, USA) was deployed at Easter Bush, a grassland site south of Edinburgh, to measure EC fluxes of ammonia. During the measurement period, the field upwind of the instruments was fertilized with cattle slurry, and the resulting emissions were measured. The performance of the system is assessed with respect to inlet losses, and intercomparisons are made with EC flux measurements using a second LAS EC system, based on a leadsalt TDL (also ARI), and AGM flux measurements using an AMANDA system (ECN Petten, NL).
2. Materials and Methods 2.1. The Eddy Covariance Method. The EC method is based on the Reynolds decomposition of a turbulent quantity such as concentration (χ) into its time-averaged component (χj), and its instantaneous perturbation (χ′): VOL. 42, NO. 6, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2041
χ ) χ + χ′
(1)
The vertical flux of χ is then defined as the covariance between χ and the vertical component of wind speed, w: Fχ ) w′χ′ ) wχ - wχ
(2)
To capture the rapid fluctuations in the quantity being measured, fast instrumentation is required (typically several Hz). These have been available for several decades to measure properties such as wind speed (ultrasonic anemometers) and water vapor/CO2 (infrared gas analyzer), and LAS is capable of making high-resolution, fast-response measurements of a range of trace gases (12). However, wall interactions of some “sticky” gases such as NH3 attenuate the high frequency signal, resulting in a significant loss of flux information. Other requirements and corrections relating to the EC method are discussed by Foken & Wichura (14). 2.2. Laser Absorption Spectroscopy. Laser Absorption Spectroscopy (LAS) makes use of the rich spectral region in the mid-infrared, in which many atmospheric species of interest have resolvable vibrational–rotational absorption features. Lasers operating in this region are tuned to scan across a particular spectral feature. The concentration is then determined from the intensity of the absorption line using absolute spectroscopic data taken from the HITRAN database (15, 16). LAS has the advantages over other techniques of having a fast response time (up to 10 Hz), a high spectral resolution resulting in good selectivity between different species, and high sensitivity. The main difference between the two LAS systems used in this study is the type of laser used. The pulsed quantum cascade lasers (Alpes Lasers, Neuchatel, Switzerland) used in the QCLAS provide greater stability and output power than the lead-salt lasers without the need for cryogenic cooling. Instead, Peltier coolers are sufficient to control the QC laser temperatures. The instruments can both operate two lasers simultaneously. As well as measuring NH3 at 967 cm-1, the TDLAS monitored HNO3 at 1724 cm-1, and the QCLAS monitored NO2 at 1606 cm-1. The detection limits for the QCLAS measurements of NH3 and NO2 at 10 Hz are 0.34 µg m-3 and 1.1 µg m-3, respectively. Both instruments use cryogenically cooled HgCdTe detectors (Kolmar Technologies, Inc.). The instruments use 5 L astigmatic Herriott multipass absorption cells (12), with path lengths of 153 m, in order to maximize the path length for a given volume. The QCLAS is also equipped with a small (1 L, 56 m path length) absorption cell for faster response (0.05 s flush time compared to 0.5 s), but at the expense of sensitivity. The glass walls of the absorption cells are coated with a fluorinated silane compound in order to reduce the effect of wall interactions by reactive gases such as NH3 and HNO3 (17). The absorption cells are maintained at low pressure (20–30 Torr) in order to reduce absorption line broadening. The laser control, as well as the data acquisition, analysis and archiving, is all managed within a single software package called TDLWintel, developed by ARI (15). To calculate mixing ratios, the software performs a “fingerprint” fit in real time, using an iterative nonlinear least-squares minimization to fit the absorption spectra to a template of Voigt line profiles. This is determined by the HITRAN spectral database (16) as well as pressure and temperature in the sample cell, the lasertuning rate, and the observed laser line-width. The software calculates absolute concentrations, so calibration is in principle unnecessary. However, zeroing is routinely carried out in order to remove any unwanted spectral features, such as optical fringes, that may interfere with the absorption line of interest. This background subtraction is achieved in the QCLAS by flushing the system 2042
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 6, 2008
with a zero gas periodically (every 15 min in this case, usually dry N2), measuring the average spectrum and subtracting it from all subsequent spectra. 2.3. The Aerodynamic Gradient Method. The aerodynamic gradient method (AGM) is an alternative method used to calculate the flux where fast response sensors are unavailable for measuring NH3. The AGM is based on K-theory, which is analogous to Fick’s law of diffusion (18). The K-theory assumes the flux (Fx) is related to the vertical concentration gradient (∂z/∂z) and that the rate of transfer can be explained by an eddy diffusion coefficient (Kx) along the averaged concentration gradient
( z -L d ) ∂x∂z
Fx ) - Kx
(3)
where ((z – d)/L) is a nondimensional stability parameter. A detailed description of the AGM can be found in Thom (19) and for specific application to NH3 in Sutton et al. (6). The AGM is frequently coupled with wet chemistry techniques to measure the NH3 concentration gradient. The AMANDA system is an established wet chemistry method used to determine NH3 fluxes. The AMANDA samples the air at a rate of 28 l min-1 through a rotating annular denuder that is coated in a thin film of an acidic stripping solution (3.6 mM NaHSO4). The stripping solution is continually removed from the denuder and mixed with a solution of NaOH to convert the NH4+ back to NH3. The NH3 solution then passes a hydrophobic PTFE membrane, through which the NH3 selectively diffuses into a counterflow of deionized water. Here it is converted to NH4+ whose concentration is then measured by conductivity. For flux measurements, three annular denuders are placed at different heights, whose stripping solution is then drawn to the same detector, resulting in a full concentration gradient being recorded every 7.5 min. A detailed description of the AMANDA system has been provided by Wyers et al. (8). 2.4. Field Measurements. Field measurements took place at Easter Bush, an intensively managed grassland site south of Edinburgh, Scotland (55°52′N, 3°2′W, 190 m above sea level). Most of the area is gently undulating, whereas the Pentland Hills rise about 1 km to the west of the measurement site with peaks of up to 500 m. The site is made up of two fields, each approximately 5 ha, more than 90% of which is covered by Lolium perenne (perennial ryegrass), and is used for silage production and cattle and sheep grazing. Most of the surrounding fields are managed in a similar way. Easter Bush has been used extensively for micrometeorological flux measurements of NH3 in the past (9–11, 20). The experiments took place in September 2004 and again in April 2005. Measurements were conducted along the boundary between the two fields, with more than 200 m of fetch in the prevailing wind directions (SW, NE). The QCLAS and the TDLAS systems ran simultaneously, approximately 20 m apart along this boundary. In the first experiment, the QCLAS sampled down a 2 m length of ¼ in. OD Silcosteel tube (Thames Restek UK Ltd.), intended to minimize wall interactions, at about 20 l min-1. During the second experiment, a 2 m length of 1 in. OD fluorinated silane coated Pyrex tube was used instead, with a high flow rate (∼ 60 l min-1), from which the QCLAS subsampled with a short glass inlet at about 10 l min-1. The Pyrex tube was heated using a self-regulating heating cable (model BSX 3–2-OJ; Thermon, U.K., Ltd., Gateshead, Tyne and Wear) wrapped along its length. The high flow rate and fluorinated silane coating are again designed to minimize wall interactions. In both experiments, the mouth of the inlet was located at a height of 1.7 m near an ultrasonic anemometer (model HS; Gill Instruments Ltd., Lymington, U.K.). In both experiments, the TDLAS sampled air down a 2 m length of fluorinated silane coated quartz tube (5/8 in. ID; also heated) at a flow rate of 42 l min-1. The mouth of
the inlet system was colocated with an ultrasonic anemometer (model USA-1, METEK GmbH, Elmshorn, Germany) at a height of 2.5 m. No calibration was performed on either instrument other than the background subtraction for the QCLAS. Data from both the QCLAS and the ultrasonic anemometer were acquired at a frequency of 10 Hz using a LabVIEW Program (National Instruments) on a separate PC. Within this program, trace gas fluxes over an averaging period of 15 min were calculated online, as well as heat fluxes and other micrometeorological parameters. Trace gas fluxes were calculated from the TDLAS in a similar way. NH3 fluxes were also measured by the AGM using an AMANDA during the first experiment. This was located between the QCLAS and the TDLAS instruments, about 15 m from the QCLAS, with measurement heights at 0.38 m, 0.85 m, and 1.94 m in order to measure the concentration profile. Fluxes were calculated by AGM, and stability corrections (6) were applied, based on friction velocity and sensible heat flux measurements from the sonic anemometer. Calibration was done on a weekly basis using aqueous ammonium standards ranging from 0 to 500 ppb to produce a three point calibration curve, coupled with airflow measurements. During the period of high NH3 concentrations, a further calibration with ammonium standards in the range 1000–5000 ppb was carried out. For comparison, the concentrations from each instrument were interpolated to a common reference height of 1 m. During both experiments, cattle slurry was applied on the SW (mostly upwind) field, and measurements were taken before, during and after application. During the first experiment, approximately 70 kg N ha-1 was applied over two days from the 28th of September 2004. For the second experiment, this was 160 kg N ha-1 over three days from the 27th of April 2005. 2.5. Inlet Tests. The main limiting factor in making EC flux measurements of gases such as NH3 is the use of sample tubes. Underestimates in the flux result from the attenuation of high frequency fluctuations in trace gas concentration along the tubes. This may arise due to nonuniform velocity profiles across the sample tube or due to interactions of “sticky” gases with the inner surface of the tube. The former applies to any trace gas and can be minimized by maintaining turbulent flow through the sample tubes (21), while the latter is more difficult to address. Wall interactions in the sample tube may be reduced by minimizing its length and maintaining high flow rates. However, its importance will also depend on the trace gas of interest and the material of the tube. A number of experiments were carried out in the laboratory following the field experiments to test the response to step changes in NH3 concentration of sample tubes made from different materials. Gas from a cylinder of 10 ppm NH3 in N2 was diluted further with N2 to 100 ppb. A solenoid valve was used to allow the QCLAS to sample alternately between the diluted NH3 and a separate supply of N2 gas that acted as a zero. The sample tubes were 5 m lengths of ¼ in. OD Silcosteel, 316L stainless steel, PTFE and polyethylene tubes. In addition, Silcosteel and PTFE were tested while being heated to approximately 35 °C with heating tape along the length of the sample tube, in order to examine the effect of heating on the inlet response. The NH3 concentration was monitored with the QCLAS acquiring data at 10 Hz. A step-change in NH3 concentration was generated by toggling the solenoid valve. The inlet response was tested for both an increase and decrease in concentration (as these were found to differ), and each stepchange was repeated five times to determine an average. Each time, the valve was toggled, and then left for at least five minutes. This was found to be sufficient time to allow
FIGURE 1. Time series of NH3 concentration at 1 m and flux from the two LAS instruments and the AMANDA during the first Easter Bush experiment.
FIGURE 2. Time series of NH3 concentration at 1 m and flux from the two LAS instruments after slurry application during the second Easter Bush experiment. the NH3 concentration to reach equilibrium at either 0 or 100 ppb. Equilibrium was checked each time by measuring the slope of the concentration during the last ten seconds of each run. In all cases, the slope was not found to exceed (1 × 10-3 ppb/s, indicating equilibrium had been reached. Both the decay and the increase in NH3 concentration through the inlet could be described by a double exponential function (12) χ(t) ) A1e(-t / τ1) + A2e(-t / τ2)
(4)
where A1 and A2 are constants such that A1 + A2 ) χ(0) and τ1 and τ2 are the time constants. Of particular interest is the first, shorter time constant (τ1) because in most situations, this was found to govern a large part of the observed change in NH3 concentration (>75%). In a constantly fluctuating NH3 concentration time series, the slower time constant (τ2) is not expected to play a strong part. The decay/increase time series were first normalized, and then fit with the double exponential function (eq 4) to determine the time constants. The time constants were then averaged over the five runs.
3. Results and Discussion 3.1. Intercomparison of the Laser Absorption Spectrometers. The two LAS instruments operated simultaneously for much of both experiments. Figures 1 and 2 show the time series of NH3 concentration at 1 m and flux measured by both instruments for the first and second experiments, respectively. Included in figure 1 are the corresponding time series for the AMANDA, which are discussed below. Correlations between the QC- & TDLAS instruments are plotted in panels a and b in Figure 3 for the first experiment. The correlation parameters in all cases are listed in Table 1. Very good correlations are seen in the NH3 concentrations (R2 ) 0.97) and fluxes (R2 ) 0.95) between the two instruments VOL. 42, NO. 6, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2043
FIGURE 3. Scatter plots showing correlations of: (a) NH3 concentration (z ) 1 m) and (b) flux between the two LAS instruments; and (c) NH3 concentration (z ) 1 m) and (d) flux between the QCLAS and the AMANDA during the first Easter Bush experiment.
TABLE 1. Results of the Intercomparisons between the Instruments; the QCLAS is Compared to the AMANDA during High and Low Emission Periods, Separately concentration (1 m) (µg m-3)
1st experiment 2nd experiment
QCLAS vs TDLAS QCLAS vs AMANDA (high) QCLAS vs AMANDA (low) TDLAS vs AMANDA QCLAS vs TDLAS (2nd exp.)
slope
intercept
R2
1.3 0.60 0.74 0.44 1.1
-0.20 0.0 -0.15 0.33 5.2
0.97 0.98 0.71 0.96 0.83
in the first experiment. Slightly poorer, but still significant (at p < 0.001), correlations are seen in the relatively low NH3 emissions during the intercomparison period (which began after the spreading) of the second experiment. In both cases, the TDLAS underestimates the NH3 concentration compared to the QCLAS. This may be partly due to multimode operation of the TDLAS, or to an offset between the two instruments as a result of the lack of automatic background subtraction with the TDLAS measurements (this possibly explains the negative concentrations in Figure 2; however, this is not expected to significantly affect the flux measurements). Excellent agreement is seen between the EC flux measurements of NH3 by both instruments in the first experiment (slope ) 1.0, see Figure 3, Table 1). This, however, is primarily during the high NH3 emissions seen after the spreading began. The agreement in measured fluxes between the instruments declines with smaller emissions, as can be seen by comparing the correlations between the first and the second experiment. No correlation could be seen in either concentrations or fluxes between the two data sets prior to spreading, when flux magnitudes were low (
