Mid Holocene dynamic vegetation highlights unavoidable climate feedbacks

Green Sahara and a northern limit of forest in the northern hemisphere are key characteristics of the differences between the mid Holocene and present-day climate. However, the strength of vegetation feedback and the ability of state-of-the-art climate model to properly represent it still an issue. A reason is that vegetation lies at the critical zone between land and atmosphere. Its variations depend on interconnected factors such as light, energy, water and carbon and, in turn, affect climate and environmental factors. These interconnexions makes it difficult to disentangle the factors that affect the representation of vegetation in a fully interactive model. Dynamical vegetation introduces additional degrees of freedom in climate simulations, so that a model that produces reasonable results when vegetation is prescribed might not be able to properly reproduce the full coupled system, when feedbacks that are not dominant when the system is constraint induce first order cascading effects in coupled mode. Here we investigate the climate-vegetation feedback in mid-Holocene and pre-industrial simulation with the IPSL climate models using 3 different settings of the dynamical vegetation that combining differences in the choice of representation of photosynthesis, bare soil evaporation and parameters defining the vegetation competition and distribution. We show that whatever the set up the major differences expected between the mid-Holocene and preindustrial climates remains similar, but the realisms of the simulated climate can be very different due to cascading climate-vegetation feedbacks that trigger vegetation growth and snow-ice-temperature-soil feedbacks.  Interestingly, with this IPSLCM6 version of the IPSL model (Boucher et al., 2020) all the mid-Holocene simulations produce vegetation in the Sahara-Sahel region compatible with the green Sahara period, but the representation of boreal forests is strongly affected by the different vegetation modeling choices.