As part of this study, special emphasis was put on the upgrade of the OLC monitoring site and the inclusion of a highly sensitive ambient CO analyzer. This analyzer is based on the measurement principle of gas filter correlation, non-dispersive infrared absorption (model TEI 48C-TL with a hand-selected PbSe detector and optimally matching preamplifier, and gold-plated absorption cell mirrors), and has been modified here by improving the control of the absorption cell temperature, and by introducing a zero trap (ZT) of 0.5 % Pd on alumina catalyst bed (type E221 P/D, Degussa Corp.) controlled at 180 oC. If ambient air was led through the ZT, the ambient CO was quantitatively removed by oxidation to CO2. By use of a remotely controlled zero valve (ZV), the analyzer was frequently switched into zero modes. The analyzer’s flow diagram is shown in Figure 11. The sensitivity of the analyzer was automatically determined at least once per day by introducing a CO standard cal gas dynamically mixed with ultra-pure zero air to levels between 500 and 4000 ppbv via the 2-way 3-port valve CV, upstream of the inlet Teflon membrane filter.
Figure 11: Schematic of the modified CO analyzer with added needle valve NV, solenoid valves CV and ZV for periodic calibration and background (zero) determination, respectively, and the CO zero trap ZT consisting of a catalyst that quantitatively oxidizes CO to CO2 at 180 °C.
Prior to implementation, the analyzer was set up for operation and testing at the School’s mobile air quality research laboratory (MAQREL) stationed at the University of Georgia Experiment Station, Bledsoe Farm near Griffin, GA. Figures 12, 13 and 14 show results from this test phase.
Figure 12: Analog signal output (1 min averages) of the modified CO analyzer, periodically subject to zero and standard addition (cal) checks aside from regular ambient air measurements.
The top panel in Fig. 12 shows the direct influence of the detector’s operating temperature (represented by “rack T”) on the baseline signal, i.e. the lower the temperature the higher the instrument’s signal background. The bottom panel is an expanded view showing the analyzer’s response to zero and cal gas modes, which is well below one minute. The 1 min data were reduced by interpolating between the zero signals, subtracting the interpolated baseline from the ambient measurement signal, and applying (multiplying) the interpolated sensitivity values in ppbv/Vdc, determined by the delta-increase of the signal due to the standard addition to the ambient air sample. Note, that at OLC, the standard gas was added to pure (CO-free) zero air. The CO mixing ratios were then compared on the basis of time, in terms of response to real changes in ambient concentrations and potential instrument drift, see e.g. Fig. 13, and in form of a linear regression of a scatter plot in Fig. 14.
Figure 13: Ambient CO mixing ratios measured by the newly modified CO analyzer (black) compared to the MAQREL operated reference analyzer (brown).
Figure 14: Regression of the new analyzer’s converted signal against the reference CO mixing ratio. The agreement is near unity at a correlation coefficient of 0.9.
The sensitivities of both analyzers were very similar (496 versus 499 ppbv/Vdc), with an experimental precision of ~2 % at 500 ppbv, however the lower detection limits (DL) expressed as 2-sigma variability of the 1 min zero averages during one hour seemed different. The higher DL of the new analyzer (4.6 vs. 2.4 ppbv) reflects its sensitivity to more noise due to the less tight temperature control of its absorption cell. The accuracy of the analyzer can be estimated as the RMS error of uncertainties in the calibration tank concentration (2 %), the mass flow controllers (4 % each MFC), the background variation (4 %), and potential inaccuracies from interpolation of the measured ambient CO during span checks (10 %). Thus, the total uncertainty in the CO measurement is estimated at ±12 % for the entire measuring range of 0 - 5000 ppbv.