A storyline approach to select the CMIP6 model ensemble to be downscaled for the South America domain

The selection of the Global Climate Models (GCMs) needed to provide the initial and boundary conditions for the Regional Climate Model (RCM) downscaling is the first research question to answer for each RCM simulation in a given domain. The methodology is not unique and depends on the second research question that is typically behind any study based on dynamical downscaling, which is the physical process we want to study and which are the mechanisms that are relevant for it, or in a more general way which is the storyline of large-scale circulation we are interested in. In the South America region there are several large-scale circulation features that are relevant to describe the climate of this region together with the specific geographical characteristics and local scale processes and their interaction. In this work we will focus on two different large-scale circulation features: the South America Monsoon and the extratropical cyclones. For each of the two a criterion needs to be established to select which CMIP6 ensemble is more suitable for one or the other or for both, and to represent the uncertainty in their future evolution. To this aim we have defined and used a series of indicators derived from the comparison between the observation and the model and based on precipitation in specific regions of interest like La Plata and Amazon basin, and Southeastern Brazil region. A second set of indicators were based on the mean sea level pressure and wind at two different levels (850 hPa and 200 hPa) aimed at characterizing the main large scale circulation patterns (South American Low-Level jets, South Atlantic Subtropical High, South Pacific Subtropical High, trade winds, Bolivian High, cyclonic circulation vortex and upper-level jet stream). By mean of these indicators a sub set of CMIP6 models has been identified for each of the large-scale circulation features and the fitness for purpose of the whole ensemble evaluated. It is common to use different model climate sensitivities to represent the uncertainty in future evolution. Here we take a different approach and represent the uncertainty in terms of the influence of a small set of large-scale drivers, including both tropical and polar regions, which are known to affect the local circulation feature of interest and which optimally span the range of local circulation response uncertainty. The uncertainty in climate sensitivity is handled by expressing these dynamical storylines of change in terms of global warming levels. In this way, a small set of CMIP6 models can be chosen which is both fit for purpose and which avoids missing plausible worst-case scenarios for regional climate risk.