Effect of chamber closure time on soil CH4 and CO2 flux estimation by linear and non-linear model application

Soils are key components of the global carbon cycle, releasing CO₂ due to plant respiration and microbial decomposition processes and consuming or releasing CH₄ depending on the dominance of methanotrophic or methanogenic activity. Gas flux measurements have been, and still are, widely employed to improve our understanding of greenhouse gas budgets as well as the processes and mechanisms regulating them. Thus, accurate estimation of flux rates and dynamics is important.

While there’s a multitude of techniques available for greenhouse gas flux measurements, non-steady state chambers are commonly used. They are however, like other chambers and flux measurement techniques in general, prone to measurement artefacts and biases. When a non-steady state chamber is deployed on top of bare soil, the concentration gradient between the soil and the atmosphere inside the chamber is artificially altered, leading to non-linear gas concentration increases (CO₂ and CH₄) or decreases (CH₄) inside the chamber, even when chamber closure times are short. Whether the true flux rate can still be approximated using linear regression by keeping the chamber closure time short has been discussed for decades and non-linear models rooted in diffusion theory have been developed to account for the non-linearity of the concentration change (e.g., the Hutchinson-Mosier model and the non-steady state diffusive flux estimator (NDFE)). Nonetheless, only few studies have empirically evaluated the effect of chamber closure time on soil flux estimation by linear and non-linear model application, especially using high frequency data.

Using >3 000 forest soil CO₂ and CH₄ flux measurements collected with a high frequency and precision trace gas concentration analyzer (LI-7810 from LI-COR^®), we evaluated the effect of sequentially increasing the time period used for linear and non-linear (Hutchinson-Mosier, (1981)) model fitting, up to a total of 300 seconds. Initial results show that using less time for non-linear model fitting results in higher release estimates for CO₂ and higher consumption estimates for CH₄ compared to when using the full 300 seconds. We also found that flux estimates from linear and non-linear models converged when decreasing the time period used for the linear fit and increasing the time period used for nonlinear fit, indicating that linear models can provide accurate flux estimates when the time period used for the linear fit is kept short. Our results have implications not only for robust estimation of flux rates, but also for field work and flux measurement logistics and planning.