Estimation of ecosystem respiration and its components by means of stable isotopes and improved closed-chamber methods

Abstract

Ecosystem respiration (Reco) is the second largest flux of CO2 between the biosphere and the atmosphere. It consists of several components, such as plant respiration and soil respiration (Rsoil), each of which may respond differently to abiotic factors, and thus to global climate change. Rsoil, which is the largest component of Reco, is often quantified by the closed-chamber method, where automated chambers can provide information on Rsoil on a high temporal scale. Although it is a widely used method, some methodological biases are still not fully understood. One emergent issue is the overestimation of closed-chamber fluxes at low atmospheric turbulence. Thus, this potential bias needs to be quantified, and methods need to be developed, to yield correct estimates of Rsoil. Apart from correct quantification of the flux of CO2 from Rsoil, the isotopic composition of C in CO2 (or δ13C) can reveal important information on the partitioning of Rsoil into autotrophic and heterotrophic respiration. Traditionally, measurements of δ13C have been performed by isotope-ratio mass spectrometry, limiting the applicability to low frequency manual measurements. However, recent advances in laser spectroscopy have allowed for real-time measurements of δ13C, thereby providing new ways to investigate the CO2 fluxes of natural ecosystems at a high temporal scale. This PhD thesis had three main aims that were all addressed experimentally in a Danish beech forest: The first main aim was to quantify the effect of overestimation of automated closed-chamber soil CO2 fluxes due to low atmospheric turbulence, and to develop methods to account for this effect. The second main aim was to quantify the individual components of Reco at an annual, seasonal and diel time scale, and the third main aim was to combine an isotope quantum cascade laser with an automated closed-chamber system to yield high temporal δ13C of chamber-based ecosystem CO2 fluxes. To address the first main aim, we measured Rsoil hourly for one year by automated closed-chambers. The data showed a clear diel pattern of Rsoil across all seasons with highest rates during nighttime. However, further analysis showed a clear negative relationship between measured flux rates and atmospheric turbulence measured as friction velocity (u∗) above the canopy, suggesting that the measured Rsoil was overestimated at low atmospheric turbulence. Filtering out data at low u∗ values removed, or even inverted, the observed diel pattern, such that the highest fluxes were now observed during daytime, and also led to a substantial decrease of 21 %, depending on u∗ threshold value, in the estimated annual Rsoil. By installing fans to produce continuous turbulent mixing of air around the soil chambers, we tested the hypothesis that overestimation of soil CO2 fluxes during low u∗ can be eliminated if proper mixing of air is ensured, and indeed the use of fans removed the overestimation of Rsoil during low u∗. To address the second main aim, total Reco was measured by the eddy covariance method and the components of tree stem respiration (Rstem), heterotrophic Rsoil from trenched soil, heterotrophic and autotrophic Rsoil from intact soil, and coarse root respiration (Rroot) were measured every two hours by automated closed-chambers for one year. We found that the contribution of Rstem to total Reco varied across the year, by only accounting for 6 % of Reco during winter and 16 % during summer. In contrast, Rsoil showed a fairly similar contribution to Reco during winter, spring and summer of 52, 45 and 49 %, respectively, while the contribution increased to 79 % during autumn. By using the trenching method, we found that autotrophic Rsoil accounted for 34 % of Rsoil during summer. Diel Rstem and Rroot showed a clear pattern during summer with the highest respiration seen around 13:00-15:00 CET for Rstem, and the highest respiration seen from 9:00-15:00 for Rroot. In contrast, Rsoil showed the lowest respiration during daytime with no clear difference in the diel pattern between the intact and trenched soil plots. Finally, we calculated the annual Rsoil for different transects at the site, and found that annual Rsoil estimated from a previously used transect at the site was underestimated by 20 %, due to Rsoil of the transect not being representative for the spatial heterogeneity of Rsoil at the site. To address the third main aim, an Aerodyne quantum cascade laser for CO2 isotopes was combined with a LI-8100A/8150 automatic closed-chamber system to yield the δ13C of CO2 during automated chamber measurements. The δ13C of the respired CO2 for each chamber measurement was determined by the Keeling plot methodology. We found that the δ13C measured by the laser was influenced by the water vapour and CO2 concentration of the sample air. However, we quantified these dependencies, and implemented a correction method to yield precise measurements of δ13C. The corrections increased the average δ13C determined from the Keeling plots by 2.1 and 3.4 ‰ for the water vapour and the CO2 concentration dependence corrections, respectively. The system was used during a two month campaign where we measured δ13C every two hours from intact soil, trenched soil, tree stems and coarse roots. The results revealed an average δ13C of -29.8, -29.7, -30.2 and -32.6 ‰ for the intact soil plots, the trenched soil plots, the stem plots and the coarse root plots, respectively. Taken together, the work presented in this PhD thesis shows that periods with low atmospheric turbulence can provide a significant source of error in Rsoil rates estimated by the closed-chamber techniques and that erroneous data must be filtered out to obtain unbiased diel patterns, accurate relationships to biotic and abiotic factors, and before estimating Rsoil fluxes over longer time scales. The work also shows that artificial turbulent air mixing may provide a method to overcome the issue with overestimated fluxes, allowing for measurements even at low atmospheric turbulence. Furthermore, the results show that a quantum cascade laser can successfully be combined with an automated closed-chamber system to yield δ13C of ecosystem CO2 fluxes at a high temporal scale, but also that the measured δ13C is highly influenced by water vapour and CO2 concentration, why a calibration procedure, as developed in this study, is crucial to yield precise measurements of δ13C.