Representing atmospheric model uncertainties: Applications in seasonal forecasts with CNRM-CM

To account for atmospheric model uncertainties in the seasonal forecasting system based on CNRM-CM (Voldoire et al. 2013), two stochastic perturbation methods have been introduced in the ARPEGE-Climate atmospheric model, namely:

• stochastic dynamics (Batté and Déqué, 2016): prognostic variables (T, q, ψ) are perturbed with random simultaneous corrections (δX) of model errors estimated by nudged seasonal runs over the hindcast period;
• SPPT (Palmer et al. 2009): multiplicative physical parameterization tendencies perturbations using a random spectral pattern generator. In ARPEGE-Climate only u,v tendencies are perturbed.

We present here separate assessments of the impact of these methods on seasonal forecast quality.