Preliminary Testing of Stochastic Physics Schemes for the Next Global Ensemble Forecast System (GEFS) Implementation

One of the forecast uncertainties in numerical weather prediction is associated with the fact that model is imperfect. To account for the deficiency of the model, in ensemble forecast application, stochastic perturbation is added to model variables to represent the uncertainty of the model. In 2010, STTP (Stochastic Total Tendency Perturbation) was first introduced in GEFS, and it is still used in the latest GEFSv11 implementation (2015). More recently, an alternate suite of methods has been developed and tested in GEFS that more specifically targets uncertain model components and is more aligned with those used in other state-of-the-art ensemble systems. This suite includes three different components, SPPT, SHUM and SKEB, that combined improve model performance in terms of error-spread relationship, especially in the tropics. As this approach is expected to be used in the next GEFS (v12) implementation, comprehensive tests are needed to verify its validity in improving ensemble forecasts. In addition, the FV3 in NEMS framework will also replace the GSM as the dynamical model for GEFSv12. The characteristics of FV3 are remarkably different from GSM, and so are the applications of stochastic physics schemes in the model. In this study, we conduct a few experiments to evaluate the model performance with this new stochastic scheme. The results from NEMS-GSM-based GEFS and NEMS-FV3-based GEFS are both included in this study.