NSRL - Atmospheric CO2

Development of a high precision atmospheric ¹⁴CO₂ measurement capability
¹⁴C is a nearly ideal inverse tracer for fossil-fuel derived CO2 in the atmosphere- in which it has zero abundance due to radioactive decay. This is in strong contrast to high activity levels in the ambient atmosphere, biosphere, and upper ocean maintained by natural ¹⁴C production and (transiently) above-ground nuclear weapons testing. On the D scale, D ¹⁴C of fossil-fuel derived CO2 is -1000 ‰, and the present day atmosphere is about 100 ‰. Simple mass balance considerations suggest that a ¹⁴CO2 measurement precision of 3 ‰ will deliver a detection capability for recently-added fossil fuel in ambient, well-mixed air of ~1 ppm (i.e., 1100 ‰ /375 ppm). For the more realistic case of a Lagrangian sampling scheme in which an air mass is sampled at times t1 and t2, Monte Carlo error analysis indicates that for a ¹⁴C measurement precision of 3 ‰ the detection capability for the change in fossil-fuel CO2 content (Cff) will be ~ 1.6 ppm. Figure 1 shows this in contrast to uncertainties for the CO method [Novelli et al., 1998], which increase with Cff. CO can thus be seen as a potentially precise fossil fuel emission detector, while ¹⁴CO2 is the more accurate. They will be most useful if used together.

Figure 1. Comparison of uncertainty when using CO (thin lines) or D¹⁴C (thick lines) to determine recently-added fossil-fuel derived CO2 (Cff). For D¹⁴C, the lines refer to measurement uncertainties of 4‰ (small dashes), 3‰ (solid), and 2‰ (long dashes). For CO, lines refer to uncertainty in the CO:CO2 emissions ratio (R) of 50% (small dashes), 33% (solid), and 25% (long dashes). In this analysis, we assumed the only uncertainty in [CO2] and D¹⁴CO2 at t1 and at t2 was the measurement uncertainty. Note that this analysis does not include any biases in the CO method that arise from neglecting OH consumption, biomass burning emissions, or CO production from hydrocarbon oxidation.

We have recently developed a high precision preparation and measurement capability for CO2 extracted from 3-6L parent samples of whole air. Figure 2 shows results for replicate extraction and graphitization aliquots taken from a large tank of whole air collected from NOAA/CMDL's NIWOT Ridge Site and measured at the UCI-AMS facility (http://www.ess.uci.edu/ams/). Individual instrumental measurement precisions were 1.7 to 1.9‰ and the total variance (i.e., including contributions to variance from extraction, graphitization and measurement) amongst all the measurements was 1.8‰ at 1 sigma. The results include one 3-sigma outlier, which if rejected indicates a residual variance of 1.34‰ (at 1-sigma). Total variances for measurement standards and Niwot tank test samples were comparable, suggesting that the extraction procedure has not contributed measurably to the total variance.

Figure 2. D¹⁴C results for replicate extraction and graphitization aliquots of Niwot Tank air. Single aliquots of CO2 were extracted in amounts sufficient for preparation of 2 separately graphitized targets. Samples were extracted and prepared at INSTAAR and measured at UC-I.

Extraction of CO₂ from whole air

CO2 is extracted from air samples by a simple cryogenic purification procedure depicted schematically in Figure 3. A collection flask (or flasks) is fitted to the evacuated line and flask air flowed through it at a rate set by a mass flow controller. Water is first separated from the air gas mixture by freezing in a water trap at -80°C. CO2 is then frozen into the CO2 trap with liquid nitrogen at -196°C and remaining non-condensable components are removed by the vacuum pump. Once all of the air in the flask has been flowed through the system, the CO2 trap is isolated and the CO2 is allowed to sublime. The total amount collected is measured and transferred to a pyrex tube at the flame-off ports. Similar procedures are used at NOAA-CMDL for quantification of CO2 mixing ratios [Zhao et al., 1997] and at INSTAAR's Stable Isotope Lab for associated high-precision ¹³CO2 measurements [Trolier et al., 1996].

Figure 3: Schematic of NSRL atmospheric CO2 extraction line (see "Lab Tour" for actual photo).

Ion statistical constraints on measurement precision and sample size

Conventional methods of radiocarbon analysis rely on precise counting of beta particles emitted during decay of ¹⁴C back to stable ¹⁴N, where the mean life of ¹⁴C is 8267 years. In contrast, AMS provides for direct counting of ¹⁴C atoms. Sensitivity of the AMS method is thus ~4 orders of magnitude greater than for decay counting and, for comparable levels of precision, sample size requirements are equivalently reduced. This development makes possible the application of high precision ¹⁴C measurement to relatively small (3-6 L) samples of whole air.

In all cases the statistical limit on precision (s) is determined by the number of ¹⁴C counts (n), including those for associated measurement standards, such that;

(1)

and, more specifically;

(2)

where sstdand ssa are the statistical errors for the standard and sample, respectively, (from equation 1) and F is the respective measured Fraction modern..Figure 4 shows the relationship between the number of counts obtained (for both the unknown and standard) and the statistical limit on precision, for samples at ambient activity. For AMS systems displaying measurement stability (i.e. for which the relative efficiency of obtaining ¹⁴C counts and ¹²C or ¹³C currents is stable over the course of analyzing the large number of samples and standards comprising a single AMS run), statistics for several accompanying measurement standards can be pooled. This lowers the statistical limit on precision by reducing substantially the first squared term in equation 2, as shown in the figure for the example of pooling 6 standards. We have generally sought to obtain >500,000 counts for authentic samples and standards. This counting goal leaves some headroom for non-statistical (mainly preparative) contributions to the long term measurement variance.

Figure 4: Dependence of statistical limit on precision (sF)on number of ¹⁴C ion counts for both the sample and standard(s).

From simple ion statistics (i.e. isotope abundance, and ionization and transmission efficiencies in the accelerator, as in equation 3) we can determine that a 3L sample of parent air at ambient mixing ratio and activity would deliver ~1,200,000 ¹⁴C counts if it were possible to "count to the last atom". Experience has shown, however, that ionization generally falters as graphite sample targets ablate and heat up in the sputter source. Thus, parent sample size requirements may often exceed those based on typical operational efficiency factors.

(3)

where,

V_sa = sample volume in L

V₀= normal volume of perfect gas (22.4 L/mol)

M_CO2= ambient molar CO₂ mixing ratio (380ppmv)

R₁₄= standard ¹⁴C:¹²C abundance ratio (1.176x10^-12)

F_m = ambient Fraction modern (~1.1)

f_I= ionization efficiency factor (~0.1)

f_tr= transmission efficiency factor (~0.35)

References cited:

Trolier, M., J.W.C. White, P.P. Tans, K.A. Massarie, and P.A. Gemery, Monitoring the isotopic composition of atmospheric CO2: Measurements from the NOAA global air sampling network, Journal of Geophysical Research, 101 (D20), 25,897-25,916, 1996.

Zhao, C.L., P.P. Tans, and K.W. Thoning, A high precision manometric system for absolute calibrations of CO2 in dry air, Journal of Geophysical Research, 102 (D 5), 5885, 1997.