• English 
  • Spanish 

Is it possible to separate model calibration and evaluation? A multi-physics study with modern radiation and soil datasets

Over the last years, a large effort has been put in developing Regional Climate Models capable to accurately represent the physical processes behind regional-scale climate and climate change. However, the evaluation process is still facing important challenges. Climate models are complex programs that manage many variables and are adjusted with many parameters, some of which are difficult or impossible to measure (Mauritsen et al. 2012). In most evaluation studies, only a few variables are used, being precipitation and temperature the most popular ones. Furthermore, there is not a clear enough discrimination between the calibration and evaluation processes. This can lead to reduce the bias balancing out errors, instead of obtaining a better physical realism.
In this context, multi-physics ensembles (MPEs) appear as an interesting methodology. These ensembles are built by perturbing a model changing the physical parameterizations used to represent unresolved phenomena (e.g., microphysics, cumulus, etc). In this work, our goal is to show how a multi-variable analysis of an MPE can be used to improve the understanding of the physical realism of a model, and of the sources of uncertainty. With this aim, a MPE is analysed looking not only P&T, but also radiation fluxes, cloud cover, soil moisture and albedo.