Deutscher Wetterdienst GB Forschung und Entwicklung Modifications to

Deutscher Wetterdienst GB Forschung und Entwicklung Modifications to the COSMO-Model Cumulus Parameterisation Scheme (Tiedtke 1989): Implementation and Testing Dimitrii Mironov and Axel Seifert Deutscher Wetterdienst, Offenbach am Main, Germany COSMO General Meeting 2008, Krakow -1 -

Revision of the T 89 mass-flux scheme Motivation: • Once in a while, COSMO-EU simulates very high precipitation amounts that are unrealistic, e. g. 200 -400 mm in 24 h, making the COSMO-EU forecasts less useful for a number of applications, e. g. for flood prediction • Orographic precipitation is primarily on the windward side of mountains, very little in the lee • In general, a better coupling of convection, grid-scale microphysics and radiation is desirable COSMO General Meeting 2008, Krakow -2 -

Revision of the T 89 mass-flux scheme (cont’d) Modifications of the Tiedtke (1989) scheme: • Co-existence of cloud water and cloud ice (water-ice mixed phase) over a certain temperature range (see next slide) is allowed during the saturation adjustment calculations within the T 89 scheme (current version of COSMO-T 89: cloud condensate is either water or ice, depending on the temperature being above/below the freezing point); this change slows down the T 89 scheme • Detrainment of convective cloud condensate (most importantly, cloud ice) which is passed to the other COSMO-model physics and dynamics in the form of tendencies of qc and qi (current version of COSMO-T 89: detrained convective cloud condensate instantaneously evaporates); this change improves coupling of the T 89 scheme with the other COSMO-model schemes • 80% environment air relative humidity threshold for evaporation of convective rain (current version of COSMO-T 89: 95% threshold); this change partially counteracts the above mixed-phase change by increasing the surface precipitation amount All these modifications have long since been operational in the ECMWF IFS COSMO General Meeting 2008, Krakow -3 -

Mixed phase clouds in the Tiedtke scheme The modified T 89 scheme computes the saturation water vapour pressure over the (convective) cloud water-cloud ice mixed phase by interpolating between the saturation pressures over liquid water and over ice, using temperature as an interpolation variable. Ratio of ice water content (IWC) and total water content (TWC) for convective updrafts in COSMO-DE (w > 10 m/s, TWC > 1 g/m 3) taken from a single forecast of a squall line case (20. 07. 2007 00 UTC run). Lines show possible transitions from pure liquid to pure ice phase • The linear function used in the first experiments resulted in too weak convection • The quadratic function taken from the ECMWF IFS is now used in the COSMO model COSMO General Meeting 2008, Krakow -4 -

COSMO-EU Routine: mean=1. 49, max=243. 91 Total Precipitation 12 May 2008 COSMO-EU Parallel Routine: mean=1. 36, max=112. 58 COSMO General Meeting 2008, Krakow -5 -

COSMO-EU Routine: mean=1. 56, max=147. 87 Total Precipitation May 2008 Boundary Effects COSMO-EU Parallel Routine: mean=1. 46, max=193. 49 COSMO General Meeting 2008, Krakow -6 -

Example 11 June 2007 observation avg 3. 1 2. 8 mm mm max 60 44 mm mm control avg 3. 9 mm max 274 mm new scheme avg 1. 9 mm max 27 mm accumulated precipitation in mm New scheme much better, but does not work that good all the time. COSMO General Meeting 2008, Krakow -7 -

Example 27 June 2007 observation avg 2. 0 mm max 33 mm control new scheme avg 2. 4 mm max 29 mm avg 2. 3 mm max 38 mm accumulated precipitation in mm An old problem solved? Only for weak orographic precip. COSMO General Meeting 2008, Krakow -8 -

Monthly accumulation June 2007 observation 105 mm control 128 mm (+21%) new scheme 116 mm (+11 %) • Obvious problems in control. Local overprediction, too strong orographic enhancement and bias of 21%. • New convection scheme clearly improves the model, but does not solve all problems. COSMO General Meeting 2008, Krakow -9 -

Montly accumulation July 2007 observation 120 mm control 131 mm (+9%) new scheme 120 mm (+0 %) • Confirms the good results for June 2007. COSMO General Meeting 2008, Krakow - 10 -

Some QPF scores: Juni/Juli 2007 • Verification against 24 h-REGNIE precipitation estimate at each model grid point. • Bootstrap hypotheses test that confirms that the model change is statistically significant, i. e. both models version are indeed different. • FBI clearly improved, TSS und ETS somewhat worse for low threshold (both scores are not independent from FBI). Slight improvement for high thresholds. COSMO General Meeting 2008, Krakow - 11 -

Low level clouds BIAS, . . . : 00 UTC (‘LMQ-Domain’) Exp 6467(equivalent to COSMO-EU Routine) vs. Exp 6468 (modified T 89 scheme) slightly improved High level clouds slightly deteriorated COSMO General Meeting 2008, Krakow - 12 -

Deutscher Wetterdienst GB Forschung und Entwicklung Conclusions • The modified Tiedtke (1989) scheme helps improve the precipitation forecasts with the COSMO model • The unrealistic high precipitation amounts are reduced • The orographic precipitation structures are improved • The FBI (frequency bias) is reduced and so are the monthly accumulations • QPF scores, such as TSS and ETS, can be slightly worse due to the reduced FBI • Prediction of low level clouds is slightly improved • Prediction of high level clouds is slightly deteriorated COSMO General Meeting 2008, Krakow - 13 -