Convection‐permitting ensembles: Challenges related to their design and use

• Published in: Quarterly journal of the Royal Meteorological Society 2019-09, Vol.145 (S1), p.90-106
• Main Authors: Frogner, Inger‐Lise, Singleton, Andrew T., Køltzow, Morten Ø., Andrae, Ulf
• Format: Article
• Language: English
• Subjects: added value; Convection; convection‐permitting; CPEPS; Design; ensemble prediction; EPS; MEPS; Precipitation; scale dependency; Statistical methods; Weather forecasting; Winter
• ISSN: 0035-9009; 1477-870X
• DOI: 10.1002/qj.3525

Challenges related to the design and use of a convection‐permitting ensemble (CPEPS) are discussed. In particular the scale‐dependent predictability of precipitation and the use of a CPEPS as well as its potential added value over global ensemble prediction systems (EPS) are investigated. Forecasts of precipitation from the operational CPEPS in Finland, Norway and Sweden (MEPS) are used for the investigations. It is found that predictability for scales smaller than ∼60 km is lost rapidly within the first 6 h of the forecast with the smallest predictable scale growing more slowly to ∼100 km over the following 18–24 h. However, there is large case‐to‐case variability and the ensemble perturbations fail to become fully saturated, especially in winter, suggesting a weakness in the design of the ensemble. The added value of CPEPS over deterministic forecasts and coarser resolution EPSs is discussed with summary statistics and case‐studies. It is shown that the added value varies between seasons and lead times. For precipitation there is an added value for both severe precipitation events and for precipitation/no precipitation decisions. The added value is higher in summer compared to winter and for shorter lead times compared to longer lead times. Added value: Brier Skill Score of IFSENS ensemble with IFSENS control as reference forecast (black), MEPS ensemble with MEPS control as reference forecast (blue), MEPS control with IFSENS control as reference (light blue) and MEPS ensemble with IFSENS ensemble as reference forecasts (red). Challenges related to the design and use of CPEPSs are discussed; the predictability for scales smaller than ∼60 km is lost rapidly within the first 6 h with the smallest predictable scale growing more slowly to ∼100 km over the following 18–24 h. For precipitation there is added value of CPEPS over deterministic forecasts and coarser resolution EPSs for precipitation events, although the added value is higher in summer compared to winter and for shorter lead times compared to longer lead times.