Modelling the effects of climate change on soil heterotrophic respiration in forest ecosystems. Empirical versus Mechanistic approach

The soil stores more carbon than the whole atmosphere and vegetation together, about 1950 Pg, as well as an amount of nitrogen 20 times higher than the quantity stored in the standing vegetation of either forests or cultivations. Furthermore, soil exchanges several greenhouse gases which entrap the longwavelength radiation, causing an increase of the global mean air temperature. Hence, soil plays a key role in the mitigation of climate change by terrestrial ecosystems, especially forests, which cover about 4 million hectares on the planet, equal to 30.6% of the lands. The carbon and nitrogen compounds stored into the soil and exchanged with the atmosphere are regulated by the soil biogeochemical processes. Some of them are expected to be more affected by climate change - in particular the processes under the temperature control - such as the soil organic matter decomposition and heterotrophic respiration. However, studying the temperature effect on these key processes is complex, both under present-day and even more on the long-term (up to 100 years), because of many other physical, chemical and biological factors involved. To manage this complexity, models are fundamental, even if they are based on different approaches, reflecting different assumptions. For example, most models simulate an exponential decomposition rate-soil temperature relationship that never reaches an optimum, even if the laboratory and field experiments show an acclimation of the process, that is, a decrease of the decay rate after an optimum temperature, always explained as the enzyme denaturation. A recently developed theory - the Macromolecular Rate Theory (MMRT) - explains through a thermodynamic point of view the reason why the acclimation occurs at definitely lower temperatures than those of enzyme denaturation, which can be registered under the expected climate change, but also under current climate. In the present work, two different approaches to simulate the temperature effect on the heterotrophic respiration (Rhet) have been compared: the classical empirical Exponential Function (EF) and the MMRT. Both approaches have been implemented in the conventional scheme introduced by the CENTURY model. The work aimed to understand if, and to what extent, simulating the acclimation of the process at high, or increasing, temperatures implies some relevant differences in the Rhet estimates, especially on the long-term projections. The model has been ran on two contrasting forest ecosystems and with different climate conditions, a temperate European beech forest located in Germany (Hainich) and a tropical forest located in Central Africa (Ankasa, Ghana), at different modeling time scales (from daily to monthly), both under current and climate change scenarios (2006-2099). Moreover, sensitivity and uncertainty analyses have been 6 carried out to detect the parameters to which the model is more sensitive and to quantify the uncertainty in the Rhet estimates. Thus, the work aimed also to understand if the implementation of the MMRT can reduce the uncertainty, compared to the EF, in the model estimates. The results show that the incorporation of the acclimation in a conventional scheme of the soil biogeochemical cycles - despite the application of a more complex mechanistic approach than the classical empirical EF - conversely to the initial hypothesis, does not imply relevant differences in the Rhet simulation by the two approaches. Indeed, the MMRT does not improve the simulation of the monthly Rhet fluxes under current climate scenario, with a difference of the correlation coefficient (rPea) between EF and MMRT equal to just 0.004 at Hainich and 0.0013 at Ankasa. Under climate change scenario, the relevant differences are detected only for the warmest Representative Concentration Pathway (RCP8.5) and only if the result is scaled on the entire surface of the analyzed forests, with a difference in the Rhet simulated by the two approaches on the whole period 2006-2099 equal to 95?103 tons C at Hainich and 307?103 tons C at Ankasa. Furthermore, the MMRT is more uncertain than EF both under present-day and climate change scenarios. The results achieved in the present work put in doubt the possibility to simulate the Rhet acclimation at increasing soil temperature by a mechanistic approach - the MMRT - using the ,,conventional" scheme of soil C and N cycles. This point is crucial to have reliable model predictions of the CO2 fluxes from the soil under changing climate a