Limitations of an Eddy-Covariance System in Measuring Low Ammonia Fluxes

Green manuring of legume crops can improve soil fertility and sustainability. To evaluate its agronomic and environmental effectiveness, gaseous losses of ammonia (NH3) in the surface layer need to be quantified by direct measurements in the field. However, the application of the eddy-covariance technique to atmospheric NH3 is challenging: its high reactivity, water solubility, and low background concentrations all hinder the response time of closed-path sensors for fast measurements of NH3 concentration. Ammonia emissions following green manuring were measured for 21 days using a flux system equipped with a fast-pulsed quantum-cascade tunable-infrared-laser spectrometer. The noisy cross-covariance function for this configuration indicates flux measurements are close to the limit of detection; the low signal-to-noise ratio further increases the uncertainties, introducing a mirroring effect on the fluxes, which results in the rapid alternation between emission and deposition, within the limit of detection (around 13 and 20 ng m(-2) s(-1), at the 95 and 99% confidence limits, respectively). An evaluation of the measurement errors is presented, focussing on three technical aspects of the eddy-covariance system: (1) time lag, (2) random error, and (3) limit of detection. The NH3 fluxes measured by the spectrometer are close to its limit of detection, with a random error of the same order as the flux.