Quantifying biases in open-path eddy covariance CO2 flux measurements caused by spectroscopic effects in broadband non-dispersive infrared gas analyzers

Open-path eddy covariance systems, based on broad-band non-dispersive infrared (NDIR) gas analyzers, are widely used for CO2 and H2O flux measurements in remote locations around the world, because of their low power consumption, fast response, and reliable operation. Nevertheless, agreement between open- and closed-path CO2 fluxes has limited inter-site comparability, especially in cold or non-growing seasons and low-flux environments, where physiologically unreasonable CO2 uptake is often observed by the open-path systems. A possible explanation is sensor-surface heating from internal-electronics power dissipation and solar radiation, which causes unaccounted gas density changes in the optical path. Fast-response thermometers, co-located with the gas analyzer, have been used to correct these effects. However, the fragility of the thermometers has prevented the wide adoption of this approach. A challenge for the open-path sensor design is that in-situ air temperature affects not only the gas density but also the broadened half-width and intensity of the spectral absorption lines. We hypothesize that fast air temperature fluctuations in the optical path of the gas analyzer can change the amount of absorbed light and cause errors in the CO2 concentration measurement. Because of the natural covariance of sensible and CO2 fluxes, such errors are well correlated with the vertical wind and can potentially propagate into flux calculations. We used spectral-line parameters, obtained from the high-resolution transmission molecular spectroscopic database (HITRAN), to evaluate the temperature effects on the integrated absorption spectra of CO2-air-mixtures across the 4.2 to 4.3 μm infrared active region utilized by NDIR analyzers. Results show that air temperature strongly influences absorption, and if not properly corrected, potentially introduces biases in the CO2 concentration measurements. Strong lines exhibit Doppler broadening, where the line peak and width decline with increasing temperatures, causing underestimation of CO2 concentration. Weak lines exhibit the opposite behavior. Based on our simulations, optimizing the optical filter passband can balance these opposing effects and greatly reduce the temperature dependence. In practice, manufacturing tolerances, shifts in the center wavelength, and the temperature sensitivity of the optical filters prevent complete elimination of the temperature-line broadening. A 13-nanometer shift in the filter pass band can introduce a 0.008 mmol m-3 K-1 underestimation in the CO2 concentration, which is a 0.67 μmol m-2 s-1 systematic error in CO2 flux per 100 watts of sensible heat flux.