Optimizing the closure period for improved accuracy of chamber-based greenhouse gas flux estimates

Non-steady-state chambers are often used for greenhouse gas flux measurements, and while there are recommendations on how long to keep the chamber closed, it is less investigated to what extent the length of the chamber closure period affects the estimated flux rates and which closure periods may provide the most accurate linear and non-linear flux estimates. Previous studies have shown that the closure of non-steady-state chambers induces a non-linear concentration development inside the chamber, even across short chamber closure periods, and that both linear and non-linear flux estimates are impacted by the chamber closure period itself. Based on 3,159 individual soil CO₂ and CH₄ flux measurements, we analyzed how linear regression and Hutchinson and Mosier (1981) modeled flux estimates are affected by the length of the chamber closure period by increasing it in increments of 30 sec, with a minimum and maximum chamber closure period of 60 and 300 sec, respectively. Across all detected flux measurements, the effect of chamber closure period length varied between 1.4–8.0 % for linear regression estimates and between 0.4–17.8 % for Hutchinson–Mosier estimates, and the largest effect sizes were observed in high flux regions. While both linear regression and Hutchinson–Mosier based estimates decreased as the chamber closure period increased, we observed a clear convergence of flux estimates as shorter and longer chamber closure periods were used for linear regression and Hutchinson–Mosier based estimation, respectively. This suggests using closure periods as short as possible for linear regression flux estimation or ensuring long-enough closure periods to observe a stabilization of Hutchinson–Mosier flux estimates over time. This analysis was based on soil flux measurements, but because the perturbation of the concentration gradient is related to the non-steady-state chamber technique rather than the measured ecosystem component, our results have implications for all flux measurements conducted with non-steady-state chambers.