Europe has experienced several major summer heatwaves and drought events in the past decades, which severely influenced terrestrial ecosystems. The response of soil carbon dynamics as well as the CO2 fluxes to global climate change remains one of the largest uncertainties for future climate projection. Meanwhile, the impacts of climate change on forest and grassland ecosystems have received substantial attention, but only a few studies have focused on the effects on shrubland ecosystems. Therefore, this dissertation addresses the influences of climate change on soil carbon cycling as well as CO2 fluxes in shrubland ecosystems.

In the CLIMAITE project, we found the soil carbon stocks increased ca. 19% during the long-term, multi-factorial climate experiment with elevated CO2, warming and prolonged summer drought in 2005-2013. Therefore, the research in Paper I focused on the fate and the stability of the extra stored soil carbon and the nitrogen stocks seven years after the climate change experiment was terminated. Revisiting a previous climate experiment to study the changes and recovery/resilience is a novel idea that has rarely been explored. The main findings of this study were the extra stored soil carbon from the previous eight-year CO2 fumigation has been lost again seven years after termination of the treatment and the increased soil carbon input during the original experiment stimulated the decomposition of the old soil carbon via priming, which brought the soil carbon pool back to the initial equilibrium status. Increases and decreases in topsoil nitrogen pools generally mimicked changes in soil carbon, which indicate a large transfer of nitrogen from deeper soil layers upon increasing plant demand. In summary, the soil carbon and nitrogen pools in this ecosystem are highly dynamic and may change rapidly in response to changes in major ecosystem drivers causing potentially large feedback to climate change.

In the chronosequence study (Paper II) about how different ecosystem CO2 flux components as well as the total carbon balance change over the lifecycle of Calluna vulgaris (L.), we found the ecosystem carbon balance was highly non-linear across a three-decade timescale, exhibiting a sinusoidal curvature between carbon sink/source strength and time after disturbance. This is a highly relevant observation for ecosystem models and highlights the need to take vegetation age into account when projecting ecosystem carbon balance and feedback on climate change. The carbon compensation point was observed at the age of four years, while it took seven years before the cumulative carbon loss in the period after cutting had been compensated for an equal amount of carbon uptake. And carbon payback from the ecosystem to the atmosphere started from the age of 16 years. Therefore, optimizing the vegetation management cycles could be beneficial from the perspective of maximizing the ecosystem carbon uptake and storage capacity in heathland ecosystems.

In the exploration of how six shrubland ecosystems along a European climatic gradient of temperature and precipitation respond to long-term, ecosystem-level experimental nocturnal warming and repeated growing season drought, we quantified the ecosystem-level CO2 fluxes (Paper III). Across sites, gross primary productivity (GPP) exhibited higher sensitivity than ecosystem respiration (Reco) in response to drought and warming and therefore was the dominant contributor to the resulting net ecosystem exchange (NEE). We also proposed a general conceptual framework, which explained the observed responses of GPP and Reco to drought and warming at all six sites based on the correct site-specific positioning of each site on the growing season soil moisture curves. This concept could be valuable and potentially easily applicable in future meta-analyses as a better alternative to using the Gaussen index of aridity. We believe that our findings are important for understanding the impacts of climate change on shrubland ecosystems, and the results have implications for the modelling of future projections of ecosystem CO2 fluxes.

The different sensitivities of GPP and Reco to drought, which were investigated in Paper III, were further explored in Paper IV. Here, a Danish heathland/grassland ecosystem was exposed to more severe experimental drought and GPP, Reco and RS were measured with more than 30 campaigns over two years. A novel, stepwise modelling approach was applied in order to improve the estimation of ecosystem carbon balance and its components, taking parameters of temperature, soil water content (SWC), and plant biomass into account, as well as testing if treatment-wise parameter fitting improved the model performance. Our results showed that periods of low SWC in the summer season depressed GPP more than Reco but Reco declined more than GPP at the annual scale, which indicated that seasonal rewetting mitigated the severe summer drought impacts on the annual scale and GPP has a higher resilience than Reco. Meanwhile, our model estimation indicated that soil respiration (RS) plays a relatively stable fraction of Reco in response to drought treatment. Overall, our modelling approach revealed that model parameters often changed under different treatments, which is often not considered in models used to predict ecosystem responses to future climate. How model parameters themselves change in response to climate is the key challenge to solve in order to improve our projections of ecosystem responses to future climate change.