An intercomparison of five ammonia measurement techniques

Abstract

An intercomparison has been conducted among five instruments which measure gas phase ammonia at low concentration in the atmosphere: (1) a photofragmentation/laser‐induced fluorescence (PF/LIF) instrument; (2) a molybdenum oxide annular denuder sampling/chemiluminescence detection (MOADS) technique; (3) a tungsten oxide denuder sampling/chemiluminescence detection (DARE) system; (4) a citric acid coated denuder sampling/ion chromatographic analysis (CAD/IC) method; and (5) an oxalic acid coated filter pack sampling/colorimetric analysis (FP/COL) method. Mixtures of ammonia in air at flow rates of 1800 (STP) L/min and concentrations from 0.1 to 14 parts per billion by volume (ppbv) with the addition of possible interferants (CH3NH2, CH3CN, NO, NO2, O3, and H2O) were provided for simultaneous tests. In addition, the five instruments made simultaneous ambient air measurements both from a common manifold and from their separate inlets located at a common height above the ground. Several conclusions were reached: (1) No artifacts or interferences were conclusively established for any of the techniques, although CH3NH2 may interfere in the MOADS system. (2) Measurements from the PF/LIF and the CAD/IC methods agreed well with the prepared mixtures over the full range of ammonia concentrations. The high specificity and time resolution (l min) of the PF/LIF results allowed data from this technique to be used as a basis set for comparisons. (3) For fog‐free conditions, ambient measurements from all of the instruments generally agreed to within a factor of 2 for ammonia levels above 0.5 ppbv. The CAD/IC and PF/LIF instruments agreed to within 15% on average for all ambient data. (4) The slope from linear regression analysis of separate inlet ambient air measurements indicated that the DARE data agreed with those from the PF/LIF system to within 7%. The linear regression line intercept was 216 parts per trillion by volume (pptv), which may indicate a positive interference in the DARE data, but the DARE data were closer to the PF/LIF data (50–100 pptv higher) at the lowest ambient ammonia levels. (5) The FP/COL method measured about 66% of the ammonia as determined by the PF/LIF technique and measured even lower fractional levels in the prepared samples. These low values indicate a loss of ammonia, possibly on the Teflon prefilter, under the conditions of this study (cold temperatures and generally low relative humidity). (6) The MOADS ambient air data were about 64% of the ammonia that was observed by the PF/LIF method for levels greater than about 1 ppbv, and below this level it overestimated ammonia. Problems with the MOADS calibrations and inlets may have been responsible. (7) Ambient air data from the one period of fog formation indicated that ambient ammonia was primarily partitioned into the condensed phase, leaving the interstitial air greatly depleted. Volatilization of absorbed ammonia from water droplets entrained in the sampled air appeared to influence the MOADS system. This did not appear to affect the FP/COL, CAD/IC, or PF/LIF results as a result of fast sample flows and/or operation at ambient temperatures.