Isotopes in Environmental and Health Studies

Transcription

1 Isotopes in Environmental and Health Studies Continuous field measurements of δ 13 C-CO 2 and trace gases by FTIR spectroscopy J. Mohn *, M. J. Zeeman, R. A. Werner, W. Eugster and L. Emmenegger Empa, Swiss Federal Laboratories for Materials Testing and Research Laboratory for Air Pollution & Environmental Technology Ueberlandstrasse 129 CH-8600 Duebendorf Switzerland ETH Zurich Institute of Plant Sciences Universitaetsstrasse 2 CH-8092 Zurich Switzerland *Joachim Mohn Tel.: Fax: joachim.mohn@empa.ch Number of Pages: 21; Figures: 5; Tables: 0

2

3 1 Abstract Continuous analysis of the 13 C/ 12 C ratio of atmospheric CO 2 (δ 13 C-CO 2 ) is a powerful tool to quantify CO 2 flux strengths of the two major ecosystem processes assimilation and respiration. Traditional laboratory techniques such as isotope ratio mass spectrometry (IRMS) in combination with flask sampling are subject to technical limitations that do not allow to fully characterize variations of atmospheric δ 13 C-CO 2 at all relevant timescales. In our study we demonstrate the strength of Fourier transform infrared spectroscopy (FTIR) in combination with a PLS-based calibration strategy for online analysis of δ 13 C- CO 2 in ambient air. The ability of the instrument to measure δ 13 C-CO 2 was tested on a grassland field-site and compared to standard laboratory-based IRMS measurements made on field-collected flask samples. Both methods were in excellent agreement with an average difference of 0.4 (n = 81). Simultaneously, other important trace gases such as CO, N 2 O and CH 4 were analysed by FTIR spectroscopy. Keywords: Stable isotope; Carbon dioxide; FTIR; Infrared spectroscopy; Isotope ratio mass spectrometry

4 2 1. Introduction Precise and fast measurement of the 13 C/ 12 C ratio of CO 2 (δ 13 C-CO 2 ) finds application in several research fields, including atmospheric chemistry, biochemistry and geochemistry. The standard analytical technique for compound-specific isotope analysis is isotope ratio mass spectrometry (IRMS), which achieves an impressive level of precision of up to In spite of this, inter-laboratory studies with different IRMS reveal discrepancies in the order of Additionally, contamination during flask sampling and sample conservation may change δ 13 C-CO 2 before IRMS analysis. Moreover, flask sampling followed by laboratory analysis limits the number and frequency of measurements and, therefore, may be inadequate to fully characterize variations of atmospheric CO 2 at all relevant timescales. It is mainly for the perspective of field applicability that various spectroscopic methods are attracting a growing interest. Infrared absorption spectroscopy has the potential to overcome most of the above limitations, especially those related to data coverage. The mid-infrared region is particularly well suited for isotopic measurements, because carbon dioxide exhibits highly characteristic rotational-vibrational bands of strong fundamental vibrations in this region. Isotopic substitution affects the vibrational and rotational energy states of a molecule, and thus each isotopologue or even isotopomer has its own ro-vibrational spectrum. Various optical techniques have been published to measure all or a selection of the most abundant CO 2 isotopologues, including non-dispersive infrared (NDIR) spectroscopy, nearinfrared (NIR) diode lasers at 1.6 µm and 2 µm and mid-infrared (MIR) diode

5 3 lasers around 4.3 µm. The laser based techniques typically use short spectral regions containing one or more ro-vibrational transitions of the isotopologues of interest. To our knowledge, the best results are currently obtained using quantum cascade laser absorption spectroscopy (QCLAS). Using QCLAS, Tuzson et al. found a precision of 0.16 (1 σ of a 500 s average) and an agreement of 0.3 with IRMS [19], while Nelson et al. [21] reached an significantly better precision (0.03 after 400 s) and a comparable long term accuracy of These instruments are however very costly and limited to the measurement of CO 2 isotope ratios. Alternatively, a broader part of a ro-vibrational band can be employed to determine isotopologue concentrations using a more common, commercially available spectroscopic instrumentation, such as Fourier transform infrared (FTIR) spectrometers. In the most straightforward approach, 12 CO 2 and 13 CO 2 concentrations are calculated from the overlapping ν 3 ro-vibrational absorption bands at 4.3 µm using a classical least squares (CLS) algorithm. However, for the parameterization of the CLS quantification algorithm, exceedingly precise single component calibration spectra are required which cannot be collected experimentally with an uncertainty of less than several percent. Therefore, most publications on δ 13 C-CO 2 using FTIR spectrometry so far suffered from low accuracy and precision. Using synthetic calibration spectra for the individual isotopologues calculated from the HITRAN database including environmental (pressure and temperature) and instrumental (resolution, line shape, wave number shift and path length) parameters, Esler et al. obtained a precision of 0.15 for δ 13 C-CO 2 at ambient CO 2 concentrations in a laboratory study.

6 4 However, the precision of the follow-up field campaigns was lower, between 0.2 and 0.5, and the accuracy of the complete analytical method was not determined. To overcome problems with inaccurate single component calibrations spectra we developed a calibration and quantification procedure that is distinctly different from the concepts published so far using a set of multicomponent standards in combination with an optimized partial least-squares (PLS) algorithm. In this paper, we demonstrate the capability of our approach to measure δ 13 C-CO 2 in real-time in ambient CO 2. Optimisation included temperature and pressure stabilisation, selection of a suitable calibration and measurement strategy, as well as automation for continuous field operation. Excellent accuracy and precision were shown during one week of field measurements in Central Switzerland using flask sampling followed by IRMS laboratory analysis as an independent analytical method. Simultaneously, other important trace gases such as CO, N 2 O and CH 4 were analysed by FTIR spectroscopy. 2. Experimental 2.1 Site description The field experiments were conducted at the Chamau research station of the ETH Zurich, during May The research station is located near Lake Zug in Central Switzerland at 400 m asl (47 12'34'' N / 8 24'36'' E). The site is an intensively managed agricultural plot on which a mixed ryegrass-clover vegetation was raised for hay. The dominant plant species are Italian ryegrass (Lolium multiflorum) and white clover (Trifolium repens) with a canopy height of

7 5 20 ± 8 cm at the time of our measurements. Air temperature during the experiment ranged from 10.1 to 26.6 C. Air samples for concentration measurement and isotopic analysis were collected at 1.83 m intake height. FTIR spectroscopic analysis was performed on-site, and compared to flask sampling followed by laboratory IRMS measurement. 2.2 FTIR spectrometry Setup In order to minimize disturbances by outside temperature changes, the FTIR was placed in an air conditioned trailer (22.5 ± 1 C) located at the edge of the grassland site in lateral direction of the mean wind. The sample gas was drawn from the sample inlet to the FTIR through a 60 m PVC tubing (ID 14 mm) with an oil free vacuum pump (Model X, Gast Manufacturing Corp., USA) at a flow rate of 150 l min -1. In front of the vacuum pump, a gas flow of approx. 2 l min -1 was extracted with a membrane pump (KNF Neuberger, CH), filtered with a sintered metal filter (SS-6F-MM-2, Swagelok, USA), dried with a Nafion dryer (PD SS, Perma Pure, USA) to a dew point of about -40 C and connected to a custom manifold system (figure 1). The lag time for a sample entering the intake and reaching the FTIR gas cell was 10 s. The manifold system was controlled by a digital I/O module (National Instruments, USA) and a Lab View programme. The sampling system switched periodically between high-purity nitrogen ( %), pressurised air used as working standard (451.8 ppm CO 2, δ 13 CO , ppm CH 4, ppb CO, ppb N 2 O) and sample gas.

8 6 As shown by laboratory experiments, the temperature stability of the sample gas is of utmost importance for the precise determination of δ 13 C-CO 2. Therefore, a thermoelectric chiller (ThermoCube 300, Solid State Cooling Systems, USA) was used to thermostat the sample gas as well as the multipath cell to 25.0 ± 0.1 C (figure 1). A constant flow of coolant was first circulated through a copper tubing (ID 4 mm, length 8.8 m) curled around the cell. Afterwards, the same coolant was used to condition the purge air and the sample gas in a heat exchanger with a total volume of about 3 litres. Sample gas and purge air passed the heat exchanger in two coiled stainless steel tubing (ID 4 mm, length 5 m). Gas temperature in the cell was continuously measured with a highprecision Pt100 thermometer (GMH 3750, Greisinger Electronic, D) with a resolution of 10 mk. The pressure in the multipath cell was automatically adjusted to 1000 ± 0.2 hpa with a digital back pressure regulator (P-702C, Bronkhorst, CH) and monitored with a digital manometer (HM 35, Revue Thommen AG, CH) mounted on the multipath cell. Insert Figure 1 about here FTIR spectrometric analysis The spectroscopic measurements were performed using a commercially available FTIR spectrometer (Nicolet Avatar 370, Thermo Electron, USA) equipped with a 20 m multipath White cell (The Foxboro Company, USA). A automatically refilled cryogenically cooled InSb detector (InfraRed Associates,

9 7 USA), with a specific peak detectivity (D*) of 2.03 x cm Hz 1/2 W -1 was used, giving a root-mean-squared (rms) noise level of ~ 3 x 10-5 absorbance units between cm -1. A spectral resolution of 0.5 cm -1 was selected, and the Happ-Genzel apodization function was employed. The spectrometer was purged with thermostated dry and CO 2 free air to eliminate residual CO 2 and water vapour absorptions in the optical path lying outside the White cell. The concentrations of the two isotopologues 12 CO 2 and 13 CO 2 were calculated from the absorption spectra using an optimized partial least squares (PLS) algorithm (TQ Analyst 7.0, Thermo Electron, USA) in combination with a set of multi-component calibration gases, which span the range of naturally occurring δ 13 C-CO 2 values between -8.7 and -19.0, and CO 2 concentrations between 364 and 530 ppm. Quantification of both isotopologues was done in the wavelength region between and cm -1, where both species have strong absorption bands and the influence of other trace gases can be neglected. The concentrations of CH 4, CO and N 2 O were determined using a conventional classical least squares (CLS) algorithm (TQ Analyst 7.0) and single-component calibration spectra for every species, which cover ambient concentrations. The following wavelength regions were used for quantification; cm -1 for CH 4, cm -1 for CO, and cm -1 for N 2 O. Interferences of H 2 O and CO 2 were corrected. The following measuring sequence was applied: (i) background spectra collection using high-purity nitrogen gas every 24 hours integrating over 256 spectral scans (12 minutes); (ii) before and after automatic liquid nitrogen refill

10 8 of the InSb detector, every six hours, collection of working standard spectra averaged over 512 scans (24 minutes) to check for drifts; and (iii) continuous recording of sample gas spectra of ambient air averaged over 16 scans (1 minute) for the determination of δ 13 C-CO 2, CO 2, CH 4, CO and N 2 O. 13 C/ 12 C isotopic ratios were corrected for temperature drifts and averaged over 24 minutes for comparison with IRMS data. CO 2, CH 4, CO and N 2 O concentrations were corrected for slight drifts based on the measurement of the working standard. As shown before, the FTIR system is very stable, which is illustrated by the fact that the standard deviation of the mean decreases with 1 / sqrt(n) for at least 16 minutes (see Allan plot in [28]) reaching a maximum precision for δ 13 C- CO 2 of Consequently, the uncertainty of a 1 minute average can be expected to be about 5 times (sqrt 24) larger than for 24 minutes averages. 2.3 Flask sampling and IRMS analysis Simultaneously to the FTIR measurements, the atmospheric CO 2 mixing ratio was monitored at the sampling height by a photoacoustic gas-monitor (Model 1412, Innova AirTech Instruments, DK). Furthermore, computer-controlled mobile air samplers (ASA) were used to collect flask samples in the field in periodic time intervals. Inside the ASA, the CO 2 concentration was determined (LI-840, LI-COR, USA) and the sample gas was dried by Mg(ClO 4 ) 2. Sample volumes and connecting tubes were purged with 0.9 l min -1 for one or two minutes before sampling, depending on whether sample containers of 10 ml (metal tubes) or 300 ml (glass flasks) were used.

11 9 The air sampler was brought to the laboratory for δ 13 C-CO 2 analysis. All gases, including laboratory standards with known δ 13 C-CO 2 values, were measured with a gasbench II equipped with a home-built liquid nitrogen cold trap coupled to a DeltaPlus XP isotope ratio mass spectrometer (both Finnigan MAT, Germany). The δ 13 C-values of the laboratory air standards were determined at the Max-Planck-Institute for Biogeochemistry (Jena, Germany) according to Werner et al Isotope convention Our calibration procedure and data interpretation are referenced to the Vienna Peedee Belemnite (VPDB) scale maintained by the International Atomic Energy Agency. The isotope relative abundance (δ 13 C-CO 2 ) of our standard gases and atmospheric samples were calculated as R C R sample δ 13 standard (1) where R 13 Sample is the molar ratio of the abundances of 13 C and 12 C of the sample and 13 R Standard the standard molar ratio. δ 13 C is reported in parts per thousand (per mil or ).

12 10 Results and Discussion 2.5 Temperature effect on δ 13 C-CO 2 measurement by FTIR Temperature and pressure of the sample gas influence the absorption spectra by changing line strength, as well as width and shape of the rotational lines. For our experimental setup the temperature dependence was determined in the laboratory to be 14 K -1. This value was confirmed during our field campaign by regular analysis of pressurized air with known isotopic composition (data not shown). The temperature and pressure stability of the experimental setup was carefully optimized to enable a precise analysis of component concentrations. An increase in the outside temperature of approx. 10 K, led to a temperature rise in the trailer of around 1 K and of less than 0.1 K in the gas cell (figure 2). For the remaining temperature changes in the gas cell, a correction was applied to the δ 13 C-CO 2 values. Insert Figure 2 about here 2.6 Field measurement of δ 13 C-CO 2 and CO 2 Intercomparison Figure 3 shows a comparison between δ 13 C-CO 2 values determined by online FTIR spectroscopy and flask sampling followed by IRMS laboratory analysis. Very good agreement was generally observed between both methods, with few distinct outliers (figure 3). The average difference between online FTIR

13 11 spectroscopy (24 minutes averaging) and flask sampling / IRMS for corresponding points in time was 0.4 (n = 81), which is identical to the average accuracy determined during external validation of the FTIR quantification algorithm in the laboratory. Differences between both methods were smaller at night-time (10 pm to 6 am), around 0.3 (n = 36), although variations in the δ 13 C-CO 2 ratio were higher. One reason might be the lower fluctuations in the outside temperature. Figure 3 also shows a time series of the CO 2 concentration analysed by FTIR, a LI-COR NDIR analyser and an Innova photoacoustic gas monitor. CO 2 mixing ratios as well as trends were in good agreement for all three methods. Unfortunately, a software failure of the manifold system of the FTIR spectrometer and the mobile air sampler caused a one day data loss, from 25 th to 26 th May for FTIR, and from 27 th to 28 th May for flask sampling / IRMS. Insert Figure 3 about here Carbon dioxide concentrations as well as δ 13 C-CO 2 in the grassland air showed typical diurnal variations. The diurnal range was from approx. -7 to -18 in δ 13 C-CO 2 and from 370 to 670 ppm in the CO 2 concentration. The data in figure 3 clearly show the well-known diurnal cycle with substantial additions of CO 2 from the grassland after sunset, as an effect of respiration of plant, root and soil micro-organisms, which have a more negative δ 13 C-CO 2 than ambient air. During calm nights, that is 24 th /25 th and 27 th /28 th May, a shallow stable boundary layer (SBL) builds up over the ground surface. The reduced vertical

14 12 mixing in this SBL leads to a strong vertical CO 2 gradient. Since this extra CO 2 accumulating in the SBL during the night stems from respiration with a distinctly more negative δ 13 C-CO 2 isotopic signature than what is found in ambient air, the nocturnal δ 13 C-CO 2 values in the SBL also become very negative. In other nights (e.g. 26 th /27 th May) strong turbulent mixing due to cloudy and relatively windy weather conditions caused a dilution of the photosynthetic signature, and thus the minimum nocturnal δ 13 C-CO 2 values are less negative. At daytime, the isotopic values of CO 2 are strongly influenced by photosynthesis driven assimilation and less so by respiration. The isotope effects of the photosynthetic process are an interaction between CO 2 diffusion into the leaf stomata and the isotopic fractionation of the Rubisco catalyzed carbon fixation (e.g. ). However, due to strong mixing within the atmospheric boundary layer during the day, δ 13 C-CO 2 predominantly represented the atmospheric background values (δ 13 C-CO 2 of approx. -8 ). The good agreement suggested by Figure 3 is only partly reflected when plotting IRMS vs. FTIR data (not shown). A linear regression yields a slope of 0.83 or 0.86 and a R 2 of 0.87 or 0.94 for δ 13 C-CO 2 and CO 2, respectively. We believe that this is mainly due to the fact that the FTIR and the IRMS sampling system did not have the same inlet and temporal resolution. Therefore, a direct comparison of the values is a very conservative approach. Integration periods of both methods were different, because sampling flasks are rapidly purged within less than one minute, while an exchange of the FTIR gas cell volume takes almost 10 minutes (t 95 ). A much better comparison is obtained when comparing the Keeling Plots obtained by IRMS and FTIR (Figure 4) [34]. This implies a

15 13 normalization of the δ 13 C-CO 2 values for 1/CO 2 and thus also accounts for differences in the sampling interval. Only slight, insignificant differences were observed for δ 13 C r determined by FTIR and flask sampling / IRMS, combining daytime and night-time data of the entire measurement campaign. The correlation of δ 13 C-CO 2 and 1/CO 2 data (R 2 ) was even stronger for the FTIR measurements, indicating its excellent overall performance. Insert Figure 4 about here 2.7 Trace gas concentration measurement by FTIR spectroscopy Figure 4 shows the trace gas concentrations determined from 22 nd to 29 th May by FTIR spectroscopy with an averaging time of 12 minutes (256 coadded scans). Analytical precision determined during laboratory experiments using the Allan variance method was 1.5 ppb for CH 4, 0.3 ppb for CO, and 0.8 ppb for N 2 O using 10 to 15 minutes of spectral averaging. Thus, a single FTIR spectrometer can simultaneously analyse these components with the same level of precision that is usually reached using three individual single component analysers. During the nights between 24 th /25 th and 27 th /28 th of May the micrometeorological conditions were favourable to accumulate CH 4, CO and N 2 O in the SBL, resulting in elevated concentrations (Figure 5). Although turbulent mixing in the SBL is the same for all trace gases and isotope ratios, the time courses of the concentrations show significant differences that indicate

16 14 the source of the origin of the respective trace gas. For example, CO typically results from combustion processes, whereas CH 4 is rather related to cattle live stock activity and soil microbial production or consumption. CO peak concentrations are seen well before midnight in the calm nights, whereas CH 4 concentrations peak a couple of hours later. The daytime methane concentrations declined from day-to-day as the soil moisture content decreased since the last rain event on the evening of 22 nd of May (data not shown), which is consistent with the fact that methane fluxes and soil moisture content are closely linked. A detailed ecological interpretation is, however, beyond the scope of this study and will the topic of a subsequent paper. Insert Figure 5 about here

17 15 3. Conclusions We have developed a precise and accurate method for continuous on-line analysis of δ 13 C-CO 2 in ambient air using a commercially available FTIR spectrometer in combination with a robust partial least squares algorithm. Optimisation included temperature and pressure stabilisation, as well as the selection of a suitable calibration and measurement strategy for long-term monitoring. Independent operation during field studies was possible by synchronizing the gas-flow system and the spectrometer operation, and by using an automatic liquid nitrogen refill for the detector. Using our analytical method the 13 C/ 12 C isotopic ratio of atmospheric CO 2 can be determined with an accuracy better than 0.4 (n = 81) compared to standard flask sampling followed by IRMS laboratory analysis as an independent analytical approach. With our experimental setup it is possible to determine, besides δ 13 C-CO 2 and CO 2, other important trace gases such as CO, N 2 O and CH 4 which is a clear advantage over standard single component techniques. The developed FTIR methodology has the ability to be employed for long-term flux measurements with good spatial and temporal averaging capabilities. For these applications, FTIR is an excellent alternative to IRMS, which requires flask sampling followed by laboratory analysis. Our calibration method is in principle applicable to other stable isotope ratio analyses; however, these may require the use of higher resolution FTIR spectroscopy.

18 16 Acknowledgments Kerstin Zeyer and Christoph Zellweger are acknowledged for CO 2, N 2 O, CO and CH 4 analysis of the multi-component standard gases. We thank Bela Tuzson for helpful discussion. The authors are grateful to Willi Brand and Michael Rothe from the Max Planck Institute for Biochemistry for δ 13 C-CO 2 and analysis of the IRMS laboratory standards. We acknowledge E. Pieper from the Section Mechanical Engineering / Workshop and W. Knecht and R. Gross from the Section Electronics / Metrology for technical assistance. M. J. Zeeman, R. A. Werner and W. Eugster were supported by the Swiss National Science Foundation, grant

19 References 17

20 18 4. Figure captions Figure 1. Schematic drawing of the experimental set-up used during the field campaign. Gas inlets were mounted on a micrometerological tower. Figure 2. Effects of changes in the outside temperature (a) on the temperature in the air-conditioned trailer (b) and in the gas cell (c). Figure 3. Time series of (a) δ 13 C-CO 2 determined by on-line FTIR spectroscopy as well as by flask sampling with subsequent IRMS analysis; two distinct outliers are marked with a circle; (b) CO 2 mixing ratio as analysed by FTIR, a photoacoustic monitor (Innova) and a NDIR analyser (LI-840). Figure 4. δ 13 C-CO 2 ratios plotted against the inverse of the CO 2 concentration. Intercept and slope of the linear regression are given together with their standard deviation for FTIR and flask sampling / IRMS measurements. Figure 5. Time courses of trace gas concentrations determined by FTIR spectroscopy. Short gaps correspond with the calibration periods, whereas the large gap between 25 th and 26 th is due to a software failure of the manifold system.