Evaluation of automatic ammonia monitors for application in an air quality monitoring network

Six automatic ammonia analyzers were studied both in the field and in an environmental test chamber, on their suitability for the Netherlands National Air Quality Monitoring Network (NAQMN). The instruments studied were a−1 continuous-flow denuder, a−1 WO3-coated thermodenuder, a−1 V2O5-coated thermodenuder, a commercial Differential Optical Absorption Spectroscopy (DOAS monitor, and a−1 chemiluminescence NOx monitor with NH3 converter. In the environmental test chamber functional parameters such as range, precision and detection limit, were determined, while the field study was mainly addressed to operational and technical performance of the instruments. The continuous-flow denuder met all the functional requirements and appears to be a−1 suitable ammonia analyzer for various purposes, for instance determination of deposition fluxes, but it must be technically modified (e.g. by using lower liquid flow rates) to meet the requirements for operation in a−1 network. The W03 thermodenuder might be suitable provided that some improvements, such as the use of an “autorange” NOx monitor, are made. The other instruments were rejected as network monitors since they did not meet two or more requirements and cannot be sufficiently improved. The converter method is only useful when high NH3 (and low NH a) concentrations are expected, such as in stacks or stables, because of interference by NH4+ and differences between the efficiencies of the Mo and the Nt converter after exposure to high concentrations of N-components. The V2O5 thermodenuder is too inaccurate, has a−1 limited range and does not sample continuously, which may yield non-representative values because ambient NH3 concentrations can vary rapidly, especially in areas with many sources. The photoacoustic monitor and DOAS needs too frequent attention. The noise level of the DOAS system appeared to be too high: 2 μg m−3 for hourly average concentrations. The photoacoustic monitor suffered from many problems leaving only 1% useful data. The scarce results showed large discrepancies compared with the other methods.