A comparison of different versions of the Stochastically Perturbed Parametrization Tendencies (SPPT) scheme

Abstract

The Stochastically Perturbed Parametrization Tendencies (SPPT) approach is widely used in ensemble prediction systems (EPSs) to represent model error. In this paper two different versions of SPPT, independent SPPT (iSPPT) and parametrization based SPPT (pSPPT), are implemented and tested in the Convection-permitting – Limited Area Ensemble Forecasting system (C‑LAEF) developed at ZAMG (Zentralanstalt für Meteorologie und Geodynamik). iSPPT and pSPPT are both very flexible approaches in which the uncertainties of different physical parametrizations are considered by applying individually adapted stochastic patterns. In iSPPT, the uncertainties from the physical parametrizations are treated independently, while in pSPPT an interaction between the uncertainties of the different parametrization schemes are sustained. Both schemes are evaluated over one summer and one winter month and compared to a reference C‑LAEF ensemble without any stochastic physics. Verification of upper-air and surface variables reveals an increased ensemble spread for all variables when using stochastic physics. Probabilistic scores show that pSPPT is able to outperform iSPPT, particularly near the surface. Precipitation verification on the grid confirms the additional skill of pSPPT especially in convective situations in summer.

Keywords

• ensemble
• stochastic physics
• perturbations
• comparison