A5. Objective optimisation of empirical model parameters

Climate models are subject to structural uncertainty resulting from poorly constrained parameters of their parameterized physical processes. These parameters are usually tuned using expert judgment in a way that more often than not lacks planning, an established method and a clear aim. Recently, a promising objective optimization method has been developed and successfully tested on a European regional domain with the COSMO-CLM model (Bellprat et al., 2012a and b). The method has also been recently adapted at Ouranos within the same model for the North American continent. The experience in North America has shown to be more complex than that in Western Europe. First, because the larger internal variability affecting North American simulations weakens the capacity of the objective function to distinguish between outputs of differently perturbed simulations. Second, because variables that seem the most sensitive to perturbation are not the same in both continents, those related to snow for example playing a larger role in North America.

The aim of this project is to adapt this methodology to the CRCM5, to revise the objective criteria to use in North America, to establish the merit of spectral nudging in helping – through the reduction of internal variability – to obtain a better tuning (e.g. Separovic et al., 2012a), and to define the more sensitive variables of interest, putting emphasis in the exploration of those related to the soil.