Field Testing of Cavity Ring-Down Spectroscopy Analyzers Measuring Carbon Dioxide and Water Vapor

Scott J. Richardson Department of Meteorology, The Pennsylvania State University, University Park, Pennsylvania

Search for other papers by Scott J. Richardson in
Current site
Google Scholar
PubMed
Close
,
Natasha L. Miles Department of Meteorology, The Pennsylvania State University, University Park, Pennsylvania

Search for other papers by Natasha L. Miles in
Current site
Google Scholar
PubMed
Close
,
Kenneth J. Davis Department of Meteorology, The Pennsylvania State University, University Park, Pennsylvania

Search for other papers by Kenneth J. Davis in
Current site
Google Scholar
PubMed
Close
,
Eric R. Crosson Picarro, Inc., Sunnyvale, California

Search for other papers by Eric R. Crosson in
Current site
Google Scholar
PubMed
Close
,
Chris W. Rella Picarro, Inc., Sunnyvale, California

Search for other papers by Chris W. Rella in
Current site
Google Scholar
PubMed
Close
, and
Arlyn E. Andrews National Oceanic and Atmospheric Administration/Earth System Research Laboratory/Global Monitoring Division, Boulder, Colorado

Search for other papers by Arlyn E. Andrews in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

Prevalent methods for making high-accuracy tower-based measurements of the CO2 mixing ratio, notably nondispersive infrared spectroscopy (NDIR), require frequent system calibration and sample drying. Wavelength-scanned cavity ring-down spectroscopy (WS-CRDS) is an emerging laser-based technique with the advantages of improved stability and concurrent water vapor measurements. Results are presented from 30 months of field measurements from WS-CRDS systems at five sites in the upper Midwest of the United States. These systems were deployed in support of the North American Carbon Program’s Mid-Continent Intensive (MCI) from May 2007 to November 2009. Excluding one site, 2σ of quasi-daily magnitudes of the drifts, before applying field calibrations, are less than 0.38 ppm over the entire 30-month field deployment. After applying field calibrations using known tanks sampled every 20 h, residuals from known values are, depending on site, from 0.02 ±0.14 to 0.17 ±0.07 ppm. Eight months of WS-CRDS measurements collocated with a National Oceanographic and Atmospheric Administrations (NOAA)/Earth System Research Laboratory (ESRL) NDIR system at West Branch, Iowa, show median daytime-only differences of −0.13 ±0.63 ppm on a daily time scale.

Corresponding author address: Dr. Scott J. Richardson, 414 Walker Building, Dept. of Meteorology, College of Earth and Mineral Sciences, The Pennsylvania State University, University Park, PA 16802. E-mail: srichardson@psu.edu

Abstract

Prevalent methods for making high-accuracy tower-based measurements of the CO2 mixing ratio, notably nondispersive infrared spectroscopy (NDIR), require frequent system calibration and sample drying. Wavelength-scanned cavity ring-down spectroscopy (WS-CRDS) is an emerging laser-based technique with the advantages of improved stability and concurrent water vapor measurements. Results are presented from 30 months of field measurements from WS-CRDS systems at five sites in the upper Midwest of the United States. These systems were deployed in support of the North American Carbon Program’s Mid-Continent Intensive (MCI) from May 2007 to November 2009. Excluding one site, 2σ of quasi-daily magnitudes of the drifts, before applying field calibrations, are less than 0.38 ppm over the entire 30-month field deployment. After applying field calibrations using known tanks sampled every 20 h, residuals from known values are, depending on site, from 0.02 ±0.14 to 0.17 ±0.07 ppm. Eight months of WS-CRDS measurements collocated with a National Oceanographic and Atmospheric Administrations (NOAA)/Earth System Research Laboratory (ESRL) NDIR system at West Branch, Iowa, show median daytime-only differences of −0.13 ±0.63 ppm on a daily time scale.

Corresponding author address: Dr. Scott J. Richardson, 414 Walker Building, Dept. of Meteorology, College of Earth and Mineral Sciences, The Pennsylvania State University, University Park, PA 16802. E-mail: srichardson@psu.edu

1. Introduction

The interest in deploying regional measurement networks to quantify carbon dioxide (CO2) fluxes, both biogenic and anthropogenic, is growing as the need to regulate CO2 emissions grows. Inverse studies of CO2 mixing ratio have traditionally been conducted at coarse spatial and temporal resolution because of both computing restraints and a scarcity of measurements. Until fairly recently the majority of high-accuracy field CO2 measurements have been made using sensors based on nondispersive infrared (NDIR) spectroscopic gas detectors (e.g., Zhao et al. 1997; Bakwin et al. 1998). Wavelength-scanned cavity ring-down spectroscopy (WS-CRDS), however, has several key advantages over the NDIR technique, including improved stability (resulting in the reduced need for calibration gases) and concurrent water vapor measurements, potentially eliminating the need for drying (Busch and Busch 1997; Crosson 2008). Consequently, the simpler field deployment requirements are facilitating deployment of regional greenhouse gas measurement networks (e.g., Miles et al. 2012). These data will be used to address the need, as described by Wofsy and Harriss (2002) and the National Research Council (2010), to better quantify regional fluxes, both biogenic and anthropogenic, using atmospheric inversion methods.

Crosson (2008) documents a 45-day trial, comparing WS-CRDS with National Oceanographic and Atmospheric Administrations (NOAA)/Earth System Research Laboratory (ESRL) NDIR measurements. Even without sample drying, the root-mean-square difference was less than 0.18 ppm, and the drift is less than 0.4 ppm over the duration of the trial. Using laboratory tests, Chen et al. (2010) developed a general equation to account for water vapor effects on the CO2 and CH4. Separate calibration of each instrument for water vapor may improve the CO2 accuracy by a small but significant amount, however. Chen et al. (2010) demonstrate the mean difference to be 0.22 ±0.09 ppm CO2 between WS-CRDS and NDIR for flights over the Amazon rain forest during the Balanço Atmosférico Regional de Carbono na Amazônia (BARCA) phase B campaign before applying any field calibrations for the WS-CRDS. The average error is reduced after applying calibrations at the end of the campaign using synthetic calibration gases (accounting for isotopic effect and pressure broadening effect resulting from variations of compositions in air).

The North American Carbon Program (NACP) Mid-Continent Intensive (MCI) was designed as a methodological test to compare atmospheric inversions with agricultural and forest inventory estimates of regional CO2 emissions (Ogle et al. 2006). Evaluating both strategies to estimate regional fluxes is a key element of the NACP Science Implementation Strategy (Denning et al. 2005). This paper describes instrument performance of five Picarro, Inc. (Sunnyvale, California), WS-CRDS systems, deployed near Centerville, Iowa; Galesville, Wisconsin; Kewanee, Illinois; Mead, Nebraska; and Round Lake, Minnesota, in support of the NACP MCI from May 2007 to November 2009. Predeployment calibrations and deployment details are described, followed by results from laboratory precision tests, round-robin field tests, analyzer drift, and, finally, an 8-month comparison of WS-CRDS to NOAA/ESRL NDIR measurements.

2. Instrument and deployment description

The WS-CRDS systems described in this paper (Picarro, Inc., CADS models CADS01, CADS03, CADS05, CADS06, CADS07, CADS08, CADS09, and CADS10) were developed as part of a small business innovative research grant, and they are the basis for the company’s current G1301, G2301, G2302, and G2401 systems.

The WS-CRDS instruments scan a 12C16O2 line of carbon dioxide, which is used to infer the total carbon dioxide content of the sample gas. Variations in the isotopic content of the sample gas will cause a systematic, but very small, error in the total carbon dioxide, because 13C16O2 and other isotopologues are not measured by the instrument. The 13C16O2 exists at approximately 1.1% of the total carbon dioxide in the atmosphere; typical daily and seasonal variations of that ratio are 1‰–2‰. The resulting error is thus up to 2.2 × 10−5 of the reported carbon dioxide mixing ratio, which at 400 ppm corresponds to an error of less than 0.01 ppm. The next most abundant isotopologue 12C16O18O is 2.5 times less abundant, and has a 3.6 ppb systematic shift for a 2‰ change in the isotopic ratio.

The instruments described in this paper are nearly identical in concept to the current models. One significant difference, however, is that the systems deployed in this work measure water vapor content using an HDO line instead of an H2O line, necessitating an empirical correction to the data to be discussed in section 3e.

a. WS-CRDS laboratory calibration

Prior to deployment for the MCI, the WS-CRDS systems were tested in The Pennsylvania State University (PSU) CO2 calibration facility, and a calibration was developed using four NOAA/ESRL CO2 calibration standards with values of approximately 340, 370, 400, and 430 ppm CO2. All five systems simultaneously sampled the same calibration gas using ⅛″ OD stainless steel tubing and three-way compression fittings to split the flow. Parker, Inc. (Cleveland, Ohio, Part 003-0216-900), valves and Scott Specialty Gas [now Air Liquide (Plumsteadville, Pennsylvania) Model 51-14B-590] regulators were used for gas handling. Each calibration gas was sampled for 8 min, and the first 3 min of data were ignored to allow for the equilibration of the sample. Six complete cycles through the four calibration gases were used to develop linear calibration equations that were used for the 30-month MCI deployment.

b. Water vapor correction to CO2

Because the air sample is not dried during the MCI deployment of the PSU WS-CRDS systems, the water vapor dilution and line broadening effects are accounted for by measuring the water vapor content and correcting the raw CO2 measurement. Therefore, any error in the water vapor measurement results in an error in the corrected CO2. An HDO line was originally chosen to measure water vapor because of its proximity to the line used for CO2. The measured HDO was calibrated to H2O in a manner that inherently assumed a constant isotopic ratio. However, during the deployment, errors associated with the use of an HDO line became apparent. In section 3e, we develop an empirical correction based on measurements of both H2O and HDO using one of the WS-CRDS instruments relocated to the NOAA/ESRL West Branch, Iowa, tall tower site after the conclusion of the MCI. This empirical relation accounts for the fact that, in general, the deuterium isotopic ratio is correlated to concentration, that is, more enriched deuterium tends to occur at higher levels of humidity.

To correct for water vapor effects we use the following relationship (Chen et al. 2010):
eq1
with H in percent. We note that, to first order, the correction of CO2 resulting from H2O dilution and line broadening effects is 0.012CO2wetH.
The expression above found in Chen et al. (2010) uses the same carbon dioxide spectral line as the five instruments used in this study but a different 1H216O line. An additional laboratory test was performed on a specially modified G1301 to obtain the cross calibration of these two spectral lines. This work resulted in the following linear expression:
eq2
where HH2O is the measurement of 1H216O used in these five instruments, and H is the H2O calibrated to the G1301 instruments.

With these two measurements, it is possible to correct the carbon dioxide data reported by the instrument to obtain the dry mole fraction when the instrument is measuring the 1H216O line. To correct the data that are collected when the instrument was measuring the HDO line (nearly all of the results reported here), we generated an empirical correction factor using in situ data collected at West Branch with a modified instrument running in a mode where both water lines were scanned. This correction is described in greater detail in section 3e.

c. Deployment details

The initial laboratory CO2 calibration was used for the duration of the project, with further field calibration performed at each site using two NOAA/ESRL CO2 calibration standards (see Table 1 for calibration tank values) sampled every 20 h for 10 min each. The field standards were prepared by NOAA–ESRL (and thus contain near-atmospheric values of CO2 isotopic ratios) and were calibrated at PSU. Predeployment tests sampling a tank for 5 days indicated typical drifts of 0.004 ppm day−1; thus, quasi-daily calibrations are sufficient for the purpose of correcting for drift. The 20-h field calibration cycle was chosen such that the calibration time propagated throughout the day in order to detect time-of-day-dependent calibration errors; none were detected. As for the length of time to sample each of the calibration tanks, Allan variance (Allan 1966) plots of 5-day predeployment tank tests indicate an optimal calibration averaging time of 30–60 min. However, 2σ of even the native (1.5 s) data are within 0.06 ppm of the mean and, as the averaging time is increased, 2σ of the averages are within 0.03 ppm for 3-min averages. We chose a 10-min calibration sampling time in order to allow more-than-sufficient time for transitions. For each day, the mean of the residuals of the two tanks was used to correct for the zero drift of the instrument. Although it is likely that the slope and zero of the linear calibration drifted over the deployment period, the error associated with correcting with an offset rather than a slope and offset is small (e.g., 0.03 ppm at Kewanee and 0.02 at Mead over a 370–400 ppm range). In addition, a zero drift correction was chosen over a linear correction because only two calibration tanks were deployed with each instrument. Except during hours in which the field standards were sampled, the upper level (110–140 m AGL, see Table 1) at each site was sampled for 45 min and the lower level (30 m AGL) was sampled for 15 min. Parker, Inc., (Cleveland, Ohio, Part 091-0094-900) valves and ¼″ OD Synflex tubing (Part 1300-04403) from the sample level on the towers to the instrument were used.

Table 1.

Location, elevation, installation date, sampling heights, and calibration tank values for the PSU WS-CRDS sites during the MCI.

Table 1.

Table 1 shows the location, elevation above sea level, installation date, sampling heights, and calibration tank values for all five PSU MCI sites. The deployment strategy was to locate the five WS-CRDS systems at existing communication towers that were at least 100 m tall, with climate-controlled facilities and line power. To enable real-time trouble shooting and daily data downloads, Verizon Wireless cell phone coverage and Airlink Raven Evolution-Data Optimized (EVDO) modems were used.

3. Results

a. Predeployment laboratory testing

In April 2007, prior to field deployment, the performance of the WS-CRDS instruments was assessed via laboratory tests. Three WS-CRDS systems (CADS03, CADS05, and CADS06) sampled indoor air from a 4-L buffer volume for 3 days. The CO2 mixing ratio varied from 390 to 485 ppm and water vapor ranged from approximately 0.4% to 1%. The difference from the mean of the 1-min averages is less than 0.1 ppm for 98% of the values (Fig. 1).

Fig. 1.
Fig. 1.

The CO2 mixing ratio cumulative difference from the mean of the three WS-CRDS systems available for testing. All units sampled from a 4-l buffer volume during a 3-day laboratory test. No further calibrations were performed beyond the initial characterization. Results shown are 1-min averages.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00063.1

b. Round-robin field testing

Round-robin tests were performed in February and November 2008, in which four standard tanks that were filled and calibrated by NOAA/ESRL were tested at each of the five PSU MCI sites. The tanks are considered unknowns and corrected using the field standards using the same method applied to atmospheric samples. Results from the February tests are shown in Fig. 2; similar results were obtained from the November tests. Each tank was sampled 2–4 times at each site for 5 min. Tests that, upon postanalysis, did not equilibrate are not included in the mean. The mean difference between the field-calibrated round-robin tanks and the known values is –0.11 ±0.09 ppm.

Fig. 2.
Fig. 2.

Differences between calibrated results and known values (measured − known) for each PSU site during a round-robin test completed in February 2008. Tanks 2 and 4 were not tested at Round Lake and unequilibrated tests are not shown. The CO2 levels are 338.81, 369.39, 401.68, and 431.78 ppm for tanks 1, 2, 3, and 4, respectively.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00063.1

c. Long-term stability

Errors associated with the use of an HDO line became apparent during the deployment, although the magnitude was not yet determined. In April 2009, Nafion dryers (Permapure, Toms River, New Jersey, Part MD-110-24S-2) with molecular sieves [Air Liquide America Specialty Gases (formerly Scott Specialty Gases, Plumsteadville, Pennsylvania)], to provide dry counterflow, were used to dry the sample air and (slightly) humidify the dry field calibration standards. By June 2009, the molecular sieve was depleted and the field standards were subsequently humidified to the ambient moisture levels, leading to differences between the measurements and known tank values of up to 0.6 ppm (Fig. 3). Moistening the field standards reduces the reported CO2 between June and November 2009. The error is strongly correlated with the water vapor mixing ratio, and an empirical correction (the quadratic fit of the error as a function of water vapor) was applied to the residuals for June–November 2009 for each site. Note that the equilibration time for Nafion dryers is on the order of days, and therefore the standards and the sample gas had essentially the same water vapor content, thus calibrating out effects of the moist Nafion dryers.

Fig. 3.
Fig. 3.

Residuals (measured − known) for the higher of two NOAA/ESRL field standards sampled every 20 h at (a) Centerville, (b) Galesville, (c) Kewanee, (d) Mead, and (e) Round Lake. These data are used to apply quasi-daily field calibrations. Beginning in June 2009, Nafion dryers humidified the field standards, and the residual values prior to empirical correction are indicated (gray symbols). Large deviations in the residuals on 17–30 Nov 2007 and 2 Nov–3 Dec 2008 at Galesville and 18 Aug–30 Nov 2007 at Kewanee are believed to be associated with errors in the wavelength-tracking algorithms that were corrected by the manufacturer in late 2007.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00063.1

The long-term stability of the WS-CRDS instruments is demonstrated by the residuals for NOAA/ESRL field standards using only the initial, predeployment calibration (Fig. 3). The residuals typically change slowly and do not, in general, have a unidirectional trend over the duration of the deployment. Large deviations in the residuals on 17–30 November 2007 and 2 November–3 December 2008 at Galesville and 18 August–30 November 2007 at Kewanee are believed to be associated with errors in the wavelength-tracking algorithms that were corrected by the manufacturer in late 2007.

After applying the empirical correction for HDO effects, the residuals averaged over the 30-month field deployment are 0.03 ±0.20, 0.16 ±0.16, 0.54 ± 0.40, 0.29 ±0.09, and −0.03 ±0.21 ppm for the Centerville, Galesville, Kewanee, Mead, and Round Lake sites, respectively (Fig. 3). At Kewanee, the calibration shifted by 0.3 ppm upon deployment, contributing to its comparatively large mean residual. Excluding Kewanee, 2σ of the magnitude of the uncorrected drifts are less than 0.38 ppm. By the end of the 30-month deployment, the instruments had drifted by 0.2–0.8 ppm from the initial laboratory calibrations. The measured drifts without field calibrations, while large compared to the WMO target accuracy of 0.1 ppm (WMO 2009), are very small compared to the alternative NDIR systems, which typically drift by 0.3 ppm day−1 (Stephens et al. 2011). Even in situations for which accuracy better than 1 ppm is not required, however, field calibrations are still recommended as a means of testing for software and hardware problems.

An estimate of the calibrated residuals is given by the magnitude of the median difference between the residuals of the two references gases and is 0.02 ±0.14, 0.07 ±0.13, 0.04 ±0.15, 0.17 ±0.07, and 0.17 ±0.07 ppm for Centerville, Galesville, Kewanee, Mead, and Round Lake, respectively; 95% (2σ) of the points are within 0.20, 0.23, 0.22, 0.32, and 0.38 ppm, respectively.

d. Uncertainty resulting from analyzer drift

Because the tower data are corrected using the 20-h residuals from the NOAA/ESRL field standards, the uncertainty resulting from analyzer drift is related to the change between the field calibrations. For example, if the residual is 0.5 ppm on day 1 and 0.6 ppm on day 2, assuming the instrument drifts steadily in between the measurements, an estimate of the error resulting from drift between calibrations is the difference between the two, or 0.1 ppm. For these data we simply applied the field calibration as a zero shift to the tower data for that day. Probability distributions of the uncertainty resulting from analyzer drift are shown in Fig. 4. The instrument at Mead (CADS07) had significantly less noise than the other instruments; 1σ of the differences between subsequent quasi-daily residuals are less than 0.06 ppm, and 2σ of the differences are less than 0.36 ppm.

Fig. 4.
Fig. 4.

Probability distributions of the difference between each field calibration and the subsequent field calibration performed 20 h later, as an estimate for the uncertainty resulting from analyzer drift. Shown are probability distributions for each of the five instruments (for the higher of the two field standards). The bin size is 0.02 ppm and the largest bin contains all of the points larger than 0.3 ppm.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00063.1

e. Comparison of NOAA/ESRL NDIR and PSU WS-CRDS measurements of CO2 mixing ratio at West Branch

In part to assess the range of isotopic values measured in the region, the Kewanee instrument (CADS05) was, upon the conclusion of the MCI, relocated to the NOAA/ESRL tall tower site at West Branch. The software was also updated to enable the instrument to measure both HDO and H2O. Shown in Fig. 5 are the CO2, H2O, and isotopic ratio between HDO and H2O. Here the isotopic ratio of HDO to H2O (δD) is defined in the typical way as
eq3
where R is the ratio of HDO to H2O, and Rstandard is the international standard reference ratio [Vienna Standard Mean Ocean Water (VSMOW)]. Previous in situ measurements of δD in water vapor have been limited, but recently WS-CRDS and related off-axis integrated cavity output spectroscopy have been evaluated for this use (e.g., Gupta et al. 2009; Wang et al. 2009; Worden et al. 2010). Although Gupta et al. (2009), for example, calibrated δD every few hours with known liquid water standards, the current data were not field calibrated because their primary purpose was to correct the CO2 mixing ratios. The initial laboratory calibration was performed by measuring water vapor from a room temperature liquid water bubbler containing −62‰ liquid water. A shift of −100 ±20‰ was applied to this calibration because of the known shift in δD upon the evaporation from liquid to vapor (Cappa et al. 2003). Although this shift is an estimate, it affects only the reported δD, not the CO2. The resulting δD shown in Fig. 5c varies widely, between −280‰ in the winter and −40‰ in the summer, and is highly correlated with the H2O. These values of δD are reasonably consistent with values typically seen in other studies (e.g., Lawrence et al. 2004). The empirical relationship between the water vapor measured on an H2O line (HH2O) and that measured on an HDO line and converted to H2O (HHDO) is found to be (Fig. 6a)
eq4
Fig. 5.
Fig. 5.

The (a) CO2 mixing ratio measured at West Branch for January–August 2010 at 99 m AGL; (b) H2O mixing ratio; (c) HDO/H2O isotopic ratio, where the horizontal line indicates the value of isotopic ratio of the laboratory calibration (−162‰). Beyond the initial laboratory calibration on a single instrument, the water measurements are uncalibrated. (d) Corrections to the reported water vapor and (e) carbon dioxide resulting from the empirical HDO correction are unique to this dataset.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00063.1

Fig. 6.
Fig. 6.

(a) The random residual resulting from noise in the relationship between HDO and H2O. (b) CO2 error resulting from noise in the relationship between HDO and H2O.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00063.1

This relationship was applied as part of the conversion from CO2wet to CO2dry described in section 2c. This empirical relationship derived directly from the observations captures the tendency of HDO to be enhanced at high levels of humidity, which, if uncorrected, would cause a tendency toward overestimation of the water vapor concentration at high humidity levels. A shift of +100‰ in δD, as an extreme example, results in 10% error in the reported water vapor, leading to a CO2 correction of 1.4 ppm at 400 ppm CO2 and 3% H2O. The corrections to the reported water vapor and carbon dioxide resulting from the empirical HDO correction unique to this dataset are shown in Figs. 5d,e. Throughout the deployment, 95% (2σ) of the δD corrections to the hourly data are less than 0.4 ppm in magnitude. A random residual resulting from noise in the relationship between HDO and H2O remains (Fig. 6a); 95% of the residuals induce an error of less than 0.3 ppm CO2, using the relationship 0.012CO2wetH (Fig. 6b). While the empirical correction reduces the average magnitude of error by only a small amount, prior to the correction the error is a seasonally dependent bias, whereas afterward it is a random error.

We compare the CO2 mixing ratio data after applying the isotopic ratio correction for January–September 2010. The NOAA/ESRL NDIR and PSU WS-CRDS data are independently quality controlled. The flow rate of the NOAA/ESRL measurement is ~8 L min−1, while that of the PSU WS-CRDS systems is ~200 mL min−1. Five-minute samples are used for both despite this difference in flow rate. The hourly daytime-only difference (PSU − NOAA) measurements at West Branch are shown in Fig. 7a. While the median difference (−0.12 ppm) is near the WMO recommended limit of 0.1 ppm for CO2 (WMO 2009), the standard deviation is large, at 1.37 ppm. On a daily time scale (daytime only), the median difference is −0.13 ±0.63 ppm (Fig. 7b).

Fig. 7.
Fig. 7.

(a) Daytime-only hourly differences between NOAA/ESRL NDIR and PSU WS-CRDS measurements of CO2 mixing ratio for January–August 2010 at West Branch at 99 m AGL. (b) As in (a), but on a daily time scale. (c) NOAA/ESRL daytime-only hourly total uncertainty estimates (×) and daily mean total uncertainty (·). Note that three to five points in each subplot are outside of the range.

Citation: Journal of Atmospheric and Oceanic Technology 29, 3; 10.1175/JTECH-D-11-00063.1

The total uncertainty includes atmospheric variability as well as measurement uncertainty and scale uncertainty (0.07 ppm). Over the duration of the comparison, the median of the hourly daytime-only NOAA/ESRL total uncertainty estimates is 0.13 ±0.68 ppm (Fig. 7c), with no clear seasonal dependence. Over the comparison as a whole, the median difference between the measurements is about the same as the median NOAA/ESRL uncertainty estimate.

The difference between PSU and NOAA/ESRL measurements tends to be larger during the growing season, when there are large deviations in CO2 mixing ratio and higher humidity levels. Considering only 1600–1700 LST, in order to sample when the atmosphere is well mixed and stable with time, the median of the differences is −0.33 ±0.83 ppm for July–August.

There are several factors that may contribute to the differences between the NDIR and WS-CRDS measurements at West Branch. There was a large difference in flow rates between the two instruments—approximately a factor of 50—compounding small errors in timing. Although the lag of maximum correlation changed by about 15 min from the beginning of the comparison to the end, for the sake of simplicity a single time shift was applied to the entire dataset to account for the difference in timing between the two systems. The MCI region is largely agricultural, with the local region surrounding West Branch being dominated by corn (Miles et al. 2012). Large local fluxes cause large deviations in CO2, accentuating the small differences in timing and affecting the comparison of the two measurements.

For comparison, the 1992–99 NOAA–Commonwealth Scientific and Industrial Research Organisation (CSIRO) flask air intercomparison experiment indicated differences of 0.2 ±0.2 ppm for biweekly measurements at Cape Grim (Masarie et al. 2001), Australia. With predominant winds transporting well-mixed air over the Southern Ocean to the station, the range of CO2 values was much smaller (<5 ppm seasonal drawdown) than that of the MCI region, reducing the effects of timing differences compared to the current study.

4. Conclusions

The WS-CRDS instruments deployed during the Mid-Continent Intensive measured both carbon dioxide and water vapor, and used the water vapor measurement to correct the CO2 for water vapor interference effects, including dilution and overlapping absorption bands. The H2O measurement in this early version of the WS-CRDS was made at a HDO line, however, necessitating corrections during postprocessing; drying the air sample (e.g., via Nafion dryers) would have simplified data postprocessing, although we do not feel it is necessary in the current WS-CRDS instrument measuring CO2 and CH4 (G2301). If the sample is dried using Nafion dryers, then we plumb the system such that the calibration air also passes through the dryer. This both equilibrates the sample and calibration gas moisture levels to a similar value and provides a test for leaks in the drying system. If applying the water vapor correction to the carbon dioxide value, then it is recommended to perform periodic tests using moistened calibration gas to ensure data quality. Quasi-daily (20 h) measurements of field standards show that long-term instrument drift is less than 0.38 ppm (2σ) over a 30-month period. After applying a field calibration in which known tanks are sampled every 20 h, residuals from known values are, depending on the site, from 0.02 ±0.14 to 0.17 ±0.07 ppm. Finally, a comparison with NOAA/ESRL measurements at West Branch for 8 months shows a median daytime-only difference of −0.13 ±0.63 ppm on a daily time scale.

Other techniques to continuously measure CO2 at remote, tower-based locations, including the two-cell NDIR-based method employed at the NOAA/ESRL towers and the single-cell NDIR-based method (e.g., http://www.raccoon.ucar.edu and amerifluxco2.psu.edu), are not commercially available as a completed unit. In addition, both require frequent calibration (on hourly instead of daily orders). The use of the WS-CRDS systems simplifies instrument development and deployment (i.e., there is no need to construct the system or use as many field calibration standards), but the up-front cost is higher. Stephens et al. (2011) points out that the benefits of laser-based instruments such as WS-CRDS may often outweigh their greater initial cost because of their better stability, more moderate calibration requirements, and stable ancillary H2O measurement, thereby enabling a moist air sample measurement.

The commercialization of WS-CRDS has made practical CO2 measurement networks of unprecedented density (Lauvaux et al. 2011). Including two NOAA tall towers at West Branch, Iowa, and Park Falls, Wisconsin; The Pennsylvania State University/University of Missouri Ameriflux tower at Ozarks, Missouri; and the University of Minnesota Rosemount tower, there were nine towers within a 500 km × 800 km domain during the MCI.

Acknowledgments

The authors gratefully acknowledge tower personnel (K. Hosman, F. Jenkins, D. Vandiver, D. McSorely, R. Book, R. Dodds, and R. Cook) for allowing tower access and for on-site help; without their cooperation this work would not have been possible. This research was sponsored by the U.S. Department of Energy Office of Science TCP Program (DE-FG02-06ER64315) and the U.S. Department of Commerce NOAA/Office of Global Programs (NA08OAR4310533). Mid-Continental Intensive tower data are available online (http://ring2.psu.edu).

REFERENCES

  • Allan, D., 1966: Statistics of atomic frequency standards. Proc. IEEE, 54, 221230.

  • Bakwin, P. S., Tans P. P. , Hurst D. F. , and Zhao C. , 1998: Measurements of carbon dioxide on very tall towers: Results of the NOAA/CMDL program. Tellus, 50B, 401415.

    • Search Google Scholar
    • Export Citation
  • Busch, K. W., and Busch M. A. , 1997: Cavity Ringdown Spectroscopy: An Ultratrace Absorption Measurement Technique. American Chemical Society Symposium Series, Vol. 720, Oxford Press, 269 pp.

    • Search Google Scholar
    • Export Citation
  • Cappa, C. D., Hendricks M. B. , DePaolo D. J. , and Cohen R. C. , 2003: Isotopic fractionation of water during evaporation. J. Geophys. Res., 108, 4525, doi:10.1029/2003JD003597.

    • Search Google Scholar
    • Export Citation
  • Chen, H., and Coauthors, 2010: High-accuracy continuous airborne measurements of greenhouse gases (CO2 and CH4) using the cavity ring-down spectroscopy technique. Atmos. Meas. Tech., 3, 375386.

    • Search Google Scholar
    • Export Citation
  • Crosson, E. R., 2008: A cavity ring-down analyzer for measuring atmospheric levels of methane, carbon dioxide, and water vapor. Appl. Phys., 92B, 403408.

    • Search Google Scholar
    • Export Citation
  • Denning, A. S., and Coauthors, 2005: Science implementation strategy for the North American carbon program. Prepared for the Carbon Cycle Science Steering Group and the Interagency Working Group On Carbon, 83 pp. [Available online at http://www.globalcarbonproject.org/global/pdf/NACP_ImplementationPlanJune04.pdf.]

    • Search Google Scholar
    • Export Citation
  • Gupta, P., Noone D. , Galewsky J. , Sweeney C. , and Vaughn B. H. , 2009: Demonstration of high-precision continuous measurements of water vapor isotopologues in laboratory and remote field deployments using wavelength-scanned cavity ring-down spectroscopy (WS-CRDS) technology. Rapid Commun. Mass Spectrom., 23, 25342542, doi:10.1002/rcm.4100.

    • Search Google Scholar
    • Export Citation
  • Lauvaux, T., and Coauthors, 2011: Constraining the CO2 budget of the corn belt: Exploring uncertainties from the assumptions in a mesoscale inverse system. Atmos. Chem. Phys. Discuss., 11, 20 85520 898, doi:10.5194/acpd-11-20855-2011.

    • Search Google Scholar
    • Export Citation
  • Lawrence, J. R., and Coauthors, 2004: Stable isotopic composition of water vapor in the tropics. J. Geophys. Res., 109, D06115, doi:10.1029/2003JD004046.

    • Search Google Scholar
    • Export Citation
  • Masarie, K. A., and Coauthors, 2001: NOAA/CSIRO Flask Air Intercomparison Experiment: A strategy for directly assessing consistency among atmospheric measurements made by independent laboratories. J. Geophys. Res., 106 (D17), 20 44520 464.

    • Search Google Scholar
    • Export Citation
  • Miles, N. L., Richardson S. J. , Davis K. J. , Lauvaux T. , Andrews A. E. , West T. O. , Bandaru V. , and Crosson E. R. , 2012: Large amplitude spatial and temporal gradients in atmospheric boundary layer CO2 mole fractions detected with a tower-based network in the U.S. upper Midwest. J. Geophys. Res., 117, G01019, doi:10.1029/2011JG001781.

    • Search Google Scholar
    • Export Citation
  • National Research Council, 2010: Verifying Greenhouse Gas Emissions: Methods to Support International Climate Agreements. National Academies Press, 124 pp.

    • Search Google Scholar
    • Export Citation
  • Ogle, S., and Coauthors, 2006: Science plan: Mid-Continent Intensive Campaign of the North American Carbon Program. U.S. Global Change Research Program, North American Carbon Program, 11 pp. [Available online at http://www.nacarbon.org/nacp/documents/NACP_MCI_SciPlan_8-06.pdf.]

    • Search Google Scholar
    • Export Citation
  • Stephens, B. B., Miles N. L. , Richardson S. J. , Watt A. S. , and Davis K. J. , 2011: Atmospheric CO2 monitoring with single-cell NDIR-based analyzers. Atmos. Meas. Tech. Discuss., 4, 43254355, doi:10.5194/amtd-4-4325-2011.

    • Search Google Scholar
    • Export Citation
  • Wang, L., Caylor K. K. , and Dragoni D. , 2009: On the calibration of continuous high-precision δ18O and δ2H measurements using an off-axis integrated cavity output spectrometer. Rapid Commun. Mass Spectrom., 23, 530536, doi:10.1002/rcm.3905.

    • Search Google Scholar
    • Export Citation
  • WMO, 2009: 14th WMO/IAEA meeting of experts on carbon dioxide, other greenhouse gases and related tracers measurement techniques. WMO TD 1487, 163 pp. [Available online at http://www.wmo.int/pages/prog/arep/gaw/documents/revised_SEPT_2009_GAW_186_TD_No_1487_CD.pdf.]

    • Search Google Scholar
    • Export Citation
  • Wofsy, S. C., and Harriss R. C. , 2002: The North American Carbon Program (NACP). NACP Committee of the U.S. Carbon Cycle Science Program Rep., 62 pp. [Available online at http://www.isse.ucar.edu/nacp/nacp.pdf.]

    • Search Google Scholar
    • Export Citation
  • Worden, J. K., and Coauthors, 2010: Estimate of bias in Aura TES HDO/H2O profiles from comparison of TES and in situ HDO/H2O measurements at the Mauna Loa Observatory. Atmos. Chem. Phys. Discuss., 10, 25 35525 388, doi:10.5194/acpd-10-25355-2010.

    • Search Google Scholar
    • Export Citation
  • Zhao, C., Bakwin P. S. , and Tans P. P. , 1997: A design for unattended monitoring of trace gases on a tall tower. J. Atmos. Oceanic Technol., 14, 11391145.

    • Search Google Scholar
    • Export Citation
Save
  • Allan, D., 1966: Statistics of atomic frequency standards. Proc. IEEE, 54, 221230.

  • Bakwin, P. S., Tans P. P. , Hurst D. F. , and Zhao C. , 1998: Measurements of carbon dioxide on very tall towers: Results of the NOAA/CMDL program. Tellus, 50B, 401415.

    • Search Google Scholar
    • Export Citation
  • Busch, K. W., and Busch M. A. , 1997: Cavity Ringdown Spectroscopy: An Ultratrace Absorption Measurement Technique. American Chemical Society Symposium Series, Vol. 720, Oxford Press, 269 pp.

    • Search Google Scholar
    • Export Citation
  • Cappa, C. D., Hendricks M. B. , DePaolo D. J. , and Cohen R. C. , 2003: Isotopic fractionation of water during evaporation. J. Geophys. Res., 108, 4525, doi:10.1029/2003JD003597.

    • Search Google Scholar
    • Export Citation
  • Chen, H., and Coauthors, 2010: High-accuracy continuous airborne measurements of greenhouse gases (CO2 and CH4) using the cavity ring-down spectroscopy technique. Atmos. Meas. Tech., 3, 375386.

    • Search Google Scholar
    • Export Citation
  • Crosson, E. R., 2008: A cavity ring-down analyzer for measuring atmospheric levels of methane, carbon dioxide, and water vapor. Appl. Phys., 92B, 403408.

    • Search Google Scholar
    • Export Citation
  • Denning, A. S., and Coauthors, 2005: Science implementation strategy for the North American carbon program. Prepared for the Carbon Cycle Science Steering Group and the Interagency Working Group On Carbon, 83 pp. [Available online at http://www.globalcarbonproject.org/global/pdf/NACP_ImplementationPlanJune04.pdf.]

    • Search Google Scholar
    • Export Citation
  • Gupta, P., Noone D. , Galewsky J. , Sweeney C. , and Vaughn B. H. , 2009: Demonstration of high-precision continuous measurements of water vapor isotopologues in laboratory and remote field deployments using wavelength-scanned cavity ring-down spectroscopy (WS-CRDS) technology. Rapid Commun. Mass Spectrom., 23, 25342542, doi:10.1002/rcm.4100.

    • Search Google Scholar
    • Export Citation
  • Lauvaux, T., and Coauthors, 2011: Constraining the CO2 budget of the corn belt: Exploring uncertainties from the assumptions in a mesoscale inverse system. Atmos. Chem. Phys. Discuss., 11, 20 85520 898, doi:10.5194/acpd-11-20855-2011.

    • Search Google Scholar
    • Export Citation
  • Lawrence, J. R., and Coauthors, 2004: Stable isotopic composition of water vapor in the tropics. J. Geophys. Res., 109, D06115, doi:10.1029/2003JD004046.

    • Search Google Scholar
    • Export Citation
  • Masarie, K. A., and Coauthors, 2001: NOAA/CSIRO Flask Air Intercomparison Experiment: A strategy for directly assessing consistency among atmospheric measurements made by independent laboratories. J. Geophys. Res., 106 (D17), 20 44520 464.

    • Search Google Scholar
    • Export Citation
  • Miles, N. L., Richardson S. J. , Davis K. J. , Lauvaux T. , Andrews A. E. , West T. O. , Bandaru V. , and Crosson E. R. , 2012: Large amplitude spatial and temporal gradients in atmospheric boundary layer CO2 mole fractions detected with a tower-based network in the U.S. upper Midwest. J. Geophys. Res., 117, G01019, doi:10.1029/2011JG001781.

    • Search Google Scholar
    • Export Citation
  • National Research Council, 2010: Verifying Greenhouse Gas Emissions: Methods to Support International Climate Agreements. National Academies Press, 124 pp.

    • Search Google Scholar
    • Export Citation
  • Ogle, S., and Coauthors, 2006: Science plan: Mid-Continent Intensive Campaign of the North American Carbon Program. U.S. Global Change Research Program, North American Carbon Program, 11 pp. [Available online at http://www.nacarbon.org/nacp/documents/NACP_MCI_SciPlan_8-06.pdf.]

    • Search Google Scholar
    • Export Citation
  • Stephens, B. B., Miles N. L. , Richardson S. J. , Watt A. S. , and Davis K. J. , 2011: Atmospheric CO2 monitoring with single-cell NDIR-based analyzers. Atmos. Meas. Tech. Discuss., 4, 43254355, doi:10.5194/amtd-4-4325-2011.

    • Search Google Scholar
    • Export Citation
  • Wang, L., Caylor K. K. , and Dragoni D. , 2009: On the calibration of continuous high-precision δ18O and δ2H measurements using an off-axis integrated cavity output spectrometer. Rapid Commun. Mass Spectrom., 23, 530536, doi:10.1002/rcm.3905.

    • Search Google Scholar
    • Export Citation
  • WMO, 2009: 14th WMO/IAEA meeting of experts on carbon dioxide, other greenhouse gases and related tracers measurement techniques. WMO TD 1487, 163 pp. [Available online at http://www.wmo.int/pages/prog/arep/gaw/documents/revised_SEPT_2009_GAW_186_TD_No_1487_CD.pdf.]

    • Search Google Scholar
    • Export Citation
  • Wofsy, S. C., and Harriss R. C. , 2002: The North American Carbon Program (NACP). NACP Committee of the U.S. Carbon Cycle Science Program Rep., 62 pp. [Available online at http://www.isse.ucar.edu/nacp/nacp.pdf.]

    • Search Google Scholar
    • Export Citation
  • Worden, J. K., and Coauthors, 2010: Estimate of bias in Aura TES HDO/H2O profiles from comparison of TES and in situ HDO/H2O measurements at the Mauna Loa Observatory. Atmos. Chem. Phys. Discuss., 10, 25 35525 388, doi:10.5194/acpd-10-25355-2010.

    • Search Google Scholar
    • Export Citation
  • Zhao, C., Bakwin P. S. , and Tans P. P. , 1997: A design for unattended monitoring of trace gases on a tall tower. J. Atmos. Oceanic Technol., 14, 11391145.

    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    The CO2 mixing ratio cumulative difference from the mean of the three WS-CRDS systems available for testing. All units sampled from a 4-l buffer volume during a 3-day laboratory test. No further calibrations were performed beyond the initial characterization. Results shown are 1-min averages.

  • Fig. 2.

    Differences between calibrated results and known values (measured − known) for each PSU site during a round-robin test completed in February 2008. Tanks 2 and 4 were not tested at Round Lake and unequilibrated tests are not shown. The CO2 levels are 338.81, 369.39, 401.68, and 431.78 ppm for tanks 1, 2, 3, and 4, respectively.

  • Fig. 3.

    Residuals (measured − known) for the higher of two NOAA/ESRL field standards sampled every 20 h at (a) Centerville, (b) Galesville, (c) Kewanee, (d) Mead, and (e) Round Lake. These data are used to apply quasi-daily field calibrations. Beginning in June 2009, Nafion dryers humidified the field standards, and the residual values prior to empirical correction are indicated (gray symbols). Large deviations in the residuals on 17–30 Nov 2007 and 2 Nov–3 Dec 2008 at Galesville and 18 Aug–30 Nov 2007 at Kewanee are believed to be associated with errors in the wavelength-tracking algorithms that were corrected by the manufacturer in late 2007.

  • Fig. 4.

    Probability distributions of the difference between each field calibration and the subsequent field calibration performed 20 h later, as an estimate for the uncertainty resulting from analyzer drift. Shown are probability distributions for each of the five instruments (for the higher of the two field standards). The bin size is 0.02 ppm and the largest bin contains all of the points larger than 0.3 ppm.

  • Fig. 5.

    The (a) CO2 mixing ratio measured at West Branch for January–August 2010 at 99 m AGL; (b) H2O mixing ratio; (c) HDO/H2O isotopic ratio, where the horizontal line indicates the value of isotopic ratio of the laboratory calibration (−162‰). Beyond the initial laboratory calibration on a single instrument, the water measurements are uncalibrated. (d) Corrections to the reported water vapor and (e) carbon dioxide resulting from the empirical HDO correction are unique to this dataset.

  • Fig. 6.

    (a) The random residual resulting from noise in the relationship between HDO and H2O. (b) CO2 error resulting from noise in the relationship between HDO and H2O.

  • Fig. 7.

    (a) Daytime-only hourly differences between NOAA/ESRL NDIR and PSU WS-CRDS measurements of CO2 mixing ratio for January–August 2010 at West Branch at 99 m AGL. (b) As in (a), but on a daily time scale. (c) NOAA/ESRL daytime-only hourly total uncertainty estimates (×) and daily mean total uncertainty (·). Note that three to five points in each subplot are outside of the range.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 806 224 17
PDF Downloads 446 96 3