PANDOWAE (Predictability and Dynamics of Weather Systems in the Atlantic-European Sector) is a research group of the Deutsche Forschungsgemeinschaft. Using the Plant Craig stochastic convective parameterization in an ensemble forecast system Pieter Groenemeijer George Craig DACH 2010, Bonn
Goals and motivation study... the roles of large and small scale processes on predictability of high impact weather potential benefits of adaptive ensembles i.e. ensembles that take into account which sources of uncertainty limit predictability more and which less, depending on the weather pattern
Planned ensemble forecast system ECMWF EPS ensemble ~50 km COSMO ~7 km COSMO ~2.8 km initial conditions: introduction of variability of the large-scale flow ~ 5-10 members selected by COSMO-LEPS clustering system (Molteni et al., 2001) x 5-10 members x 5-10 members introduction of variability introduced by convection through a stochastic convective scheme introduction of boundary-layer variability
Planned ensemble forecast system ECMWF EPS ensemble ~50 km COSMO ~7 km COSMO ~2.8 km initial conditions: introduction of variability of the large-scale flow ~ 5-10 members selected by COSMO-LEPS clustering system (Molteni et al., 2001) x 5-10 members x 5-10 members introduction of variability introduced by convection through a stochastic convective scheme introduction of boundary-layer variability
Principle of Convective Parameterisation Represent effects of unresolved cumulus convection assume controlled by large-scale flow: e.g. removal of CAPE control is statistical: e.g. average precipitation over many cloud individual storms not predictable
Principle of Convective Parameterisation Represent effects of unresolved cumulus convection assume controlled by large-scale flow: e.g. removal of CAPE control is statistical: e.g. average precipitation over many cloud individual storms not predictable Convective activity within a model grid cell may be highly variable control is statistical.
Implementation of the Plant Craig scheme in COSMO 7km Large scale forcing calculated as average over a 35x35 km area (indep. of model resolution) Convective plumes are drawn from a probability density function (PDF) that describes the chance of finding a plume of a given size within a given grid cell (Plant and Craig, 2008) The PDF is derived from equilibrium statistics (Cohen and Craig, 2006).
Comparison of precipitation structure
Ensemble forecast results 10 members selected from operational ECWMF EPS forecasts (COSMO LEPS clustering; Molteni et al, 2001) Each member drives 10 COSMO simulations with the Plant Craig scheme Study the variability within group driven by 1 EPS member ( internal ) and total variability ( total ), i.e. internal + external
Ensemble forecast results 2 weakly forced summertime convection cases Vorticity and geopotential height at 500 hpa
Ensemble forecast results 3 moderately forced convection cases Vorticity and geopotential height at 500 hpa
Ensemble forecast results One winter storm case Vorticity and geopotential height at 500 hpa
Ensemble forecast results bluish: weakly forced summertime convection greenish: summertime convection with moderate forcing red: winter storm weakly forced cases moderately forced cases winter storm case
Ensemble forecast results bluish: weakly forced summertime convection greenish: summertime convection with moderate forcing red: winter storm results of precipitation smoothed over area used in the PC scheme weakly forced cases weakly forced cases moderately forced cases moderately forced cases winter storm case
Ensemble forecast results blueish: weakly forced summertime convection greenish: summertime convection with moderate forcing red: winter storm
Conclusions 1. the stochastic scheme introduces considerable amounts of variability relative to that introduced by the ECMWF ensemble, when using hourly rain accumulation as the verification metric 2. after spatial smoothing over the averaging area used in the stochastic scheme, considerable amounts of variability remain 3. the relative importance of the variability introduced by the stochastic scheme depends on the weather situation and appears to be proportional to the amount of convective precipitation it produces (see poster of Michael Würsch for implications for data assimilation)
Thank you! References: R. S. Plant and G. C. Craig, 2008: A Stochastic Parameterization for Deep Convection Based on Equilibrium Statistics. J. Atmos. Sci., 65, 87 105 Craig, G. C. and B. G. Cohen, 2006: Fluctuations in an equilibrium convective ensemble. Part I: Theoretical basis. J. Atmos. Sci., 63, 1996 2004. Molteni, F., Buizza, R., Marsigli, C., Montani, A., Nerozzi, F., and Paccagnella, T.: A strategy for High Resolution Ensemble Prediction, Part I: Definition of Representative Members and Global Model Experiments, Quart. J. Roy. Meteor. Soc., 127, 2069 2094, 2001.