Addressing a systematic bias in carbon dioxide flux measurements with the EC150 and the IRGASON open-path gas analyzers

•CO2 flux measurement errors (IRGASON & EC150) scale with kinematic temperature flux.•Relative errors are most pronounced when true CO2 flux is small and heat flux large.•Using a fast-response air temperature to scale CO2 absorption minimizes bias.•Agreement between open- and closed-path IRGA CO2 fluxes is substantially improved.•Fast-instead of slow-response air temperature should be used to scale CO2 absorption. Across a global network of eddy covariance flux towers, two relatively new open-path infrared gas analyzers (IRGAs), the IRGASON and the EC150, are increasingly used to measure net carbon dioxide (CO2) fluxes (Fc_OP). Differences in net CO2 fluxes derived from open- and closed-path IRGAs in general remain poorly constrained. In particular, the performance of the IRGASON and the EC150 for measuring Fc_OP has not been characterized yet. These IRGAs measure CO2 absorption, which is scaled with air temperature and pressure before converting it to instantaneous CO2 density. This sensor-internal conversion is based on a slow-response thermistor air temperature measurement. Here, we test if the high-frequency temperature attenuation causes selectively systematic Fc_OP errors that scale with kinematic temperature fluxes. First, we examine the relationship between wintertime Fc_OP and kinematic temperature fluxes for eight northern ecosystems. Second, we investigate how residuals between Fc_OP and CO2 fluxes from co-located closed-path IRGAs (FC_CP) are related to kinematic temperature fluxes for three different ecosystem types (i.e., boreal forest, grassland, and irrigated cropland). We find that kinematic temperature fluxes, but not mean ambient air temperatures or CO2 flux regime, consistently determine the absolute magnitude of Fc_OP errors. This selectively systematic bias causes the most pronounced relative Fc_OP errors to occur when “true” CO2 fluxes are low and kinematic temperature fluxes are high (e.g., northern ecosystems during the winter). The smallest relative errors occur during periods with large “true” CO2 fluxes and low kinematic temperature fluxes. To address this bias, we replace the slow-response air temperature in the absorption-to-CO2 density conversion with a fast-response air temperature derived from sonic anemometer measurements. The use of the fast-response air temperature improves the agreement between half-hourly Fc_OP and FC_CP for all open- versus closed-path IRGA comparisons. Additionally, cumulative Fc_OP and Fc_CP sums ar