Carbon dioxide exchanges in an alpine tundra ecosystem (Gran Paradiso National Park, Italy): A comparison of results from different measurement and modelling approaches

Mountain ecosystems are particularly sensitive to the effects of climate change and require a detailed understanding of how their components respond to the varying environmental conditions. However, the knowledge of crucial biogeochemical variables, such as carbon dioxide (CO2) fluxes in the Critical Zone of high-altitude grassland and Alpine tundra, is still patchy. In this framework, two methods are typically adopted to measure CO2 exchanges between the atmosphere and the underlying ecosystem: eddy covariance and accumulation chambers. Potential drawbacks affect both approaches, also depending on the scale at which studies are performed. On the one hand, eddy covariance method provides large spatial coverage, despite requiring specific statistical assumptions that are not always valid in remote environments. In addition, the eddy covariance method allows measuring only the CO2 net ecosystem exchange (NEE), while the separation of CO2 emission (ecosystem respiration, ER) and uptake (gross primary production, GPP), needed for understanding ecosystem functioning and its possible future changes, requires partitioning procedures based on regressive models. On the other hand, accumulation chambers allow for direct measurements of both NEE and ER but are usually limited to sampling small areas and the chambers themselves could potentially disturb plant-soil dynamics. The present study investigates such issues by combining three complementary methods (i.e., eddy covariance, portable and automated accumulation chambers) to measure summer CO2 fluxes at the soil-vegetation-atmosphere interface in a heavily instrumented high-altitude Alpine tundra site. While eddy covariance and portable chamber measurements provide consistent estimates of the NEE, a significant discrepancy between the results of the different methodologies was found for daytime ER. In particular, the eddy covariance daytime and nighttime partitioning methods failed in representing ER at the ecosystem scale. The discrepancy between modelled and measured ER suggests that different processes are involved during day and night, which are not fully represented in models. These aspects would require further explorations to achieve site-specific assessments of daytime ER.