Title: On attempts to supersede multiplicative noise to represent model uncertainties in NWP

Author: Martin Leutbecher (ECMWF)
Sarah-Jane Lock (ECMWF)
Pirkka Ollinaho (ECMWF)
Peter Bechtold (ECMWF)
Anton Beljaars (ECMWF)
Richard Forbes (ECMWF)
Robin Hogan (ECMWF)
Irina Sandu (ECMWF)

In order to obtain a reliable ensemble forecasting system, model uncertainties need to be adequately represented in the forecast model. The uncertainties associated with parametrized physical processes are dependent on the atmospheric state. For instance, in a convectively active region the uncertainty will have different structure and amplitude than in a quieter anticyclonic situation. Multiplicative noise is a simple way of achieving state-dependent stochastic perturbations. This approach has been used at ECMWF for over 15 years. The original formulation (termed stochastic physics at the time) has been refined to use a spatially and temporally continuous random pattern. While the revised scheme is now called Stochastically Perturbed Parametrization Tendency scheme (SPPT), it still uses the basic approach of multiplicative noise, i.e. the perturbation is equal to the unperturbed total physics tendency times a random number.

The formulation of SPPT makes it difficult to differentiate the degree of uncertainty between different physical processes. Furthermore, the sampled pdf may be too degenerate, i.e. the associated covariances are too low rank. Lastly and perhaps most importantly, the physical consistency of the perturbations is difficult to achieve. Therefore, new ways are being explored to integrate the representation of uncertainties into the existing physical parametrizations at ECMWF. The poster will report on recent work at ECMWF. Results will be presented based on experiments with the global medium-range ensemble.

GMAO Head: Steven Pawson Global Modeling and Assimilation Office NASA Goddard Space Flight Center

Curator: Nikki Privé Last Updated: Feb 9 2015