Revised parameter settings in the COSMO-DE boundary layer (TKE) scheme Deutscher Wetterdienst GB Forschung und Entwicklung Revised parameter settings in the COSMO-DE boundary layer (TKE) scheme Axel Seifert Deutscher Wetterdienst, Germany
Modifications of PBL scheme Blackadars parameterization of the turbulent length scale: ECMWF/IFS: lturlen = 375 m COSMO model: lturlen = 500 m New COSMO-DE: lturlen = 150 m In the olde COSMO-DE setup sub-grid cloudiness started early, compared to the original Sommeria and Deardorff (1977) scheme. The revised settings use the linearization as originally suggested by SD77
Fraction Skill Score A scale dependent precipitation skill score for convective-scale NWP models Δ = 1: 0 vs 1 → bad forecast on 1Δ-scale Δ = 3: 4/9 vs 3/9 → OK forecast on 3Δ-scale Δ = 5: 6/25 vs 6/25 → perfect forecast on 5Δ-scale Roberts and Lean (2005)
Fraction Skill Score In the following we chose 21x21 grid points, i.e., 60x60 km2. Roberts and Lean (2005)
Precip and QPF-Scores 01. Juni 2008 06-18 UTC Weak large-scale forcing by quasi-stationary occlusion zone
Precip and QPF-Scores 15. Mai 2008 06-18 UTC Slowly moving warm front over Rheinland-Pfalz/Hessen/Sachsen
Precip and QPF-Scores 03. Juni 2008 06-18 UTC Occlusion zone in the Elbe region with several convergence lines
Precip and QPF-Scores 07. Juni 2008 06-18 UTC Weak slowly moving cold front (?)
Total scores from 1 May to 9 June 2008 06-18 UTC Significantly more strong (convective) precipitation events. Slight improvement of ETS for thresholds larger than 5 mm / 12 h Much better FSS especially for high thresholds
Total scores from 1 May to 9 June 2008 06-18 UTC und 18-06 UTC! Including day and night in the statistics gives a less clear signal!
Hovmöller diagramms or precipiation rate 1 May to 9 June 2008 In the operational COSMO-DE almost no diurnal cycle, the afternoon maximum is very weak. Clear improvement in the model behavior using the modified PBL parameters. Reduced nocturnal drizzle, better afternoon maximum
Monthly precipitation accumulations form May 2008 The total accumulated precipitation is robust. Both variants of COSMO- DE give a good monthly budget.
Deutscher Wetterdienst GB Forschung und Entwicklung Conclusions Changing the PBL parameters (turlen, SGS cloud) has a strong effect on the initiation of convection and the diurnal cycle of precipiation during summer. For COSMO-DE this resolves the notorious problem of the underestimation of convective activity. Therefore the modification was made operational at DWD on 10 Sep 2008. The modifications have also a strong effect on the temperature profile and 2m-temperature. Both increase, making the profile more unstable. This might introduce a positive bias in the temperature forecasts. More research on improved PBL schemes is necessary (UTCS). The work presented here shows that the mixing length, and the moist turbulence assumption are important and need special attention.
Was bewirkt das physikalisch? Beide Änderungen bewirken generell einen reduzierten vertikalen Austausch in der Grenzschicht: Die Längenskala geht insbesondere in die Dissipationsrate ein: Kleinere Längenskala → mehr Dissipation → weniger TKE→ weniger Austausch Die Änderungen in der Feuchtturbulenz reduzieren die subskaligen Wolken: Weniger Wolken → reduzierte TKE-Produktion → weniger TKE → weniger Austausch Weniger vertikale Durchmischung führt dann zu größerer vertikalen Gradienten, d.h. erhöhter lokaler Instabilität, was dann die skalige Konvektionsauslösung erleichtert oder ermöglicht. Insbesondere in der Nacht kann eine zu starke Durchmischung zu skaligen Grenzschichtwolken führen, die sogar Niederschlag produzieren, und dann am Vormittag die Ausbildung eine konvektiven Grenzschicht verhindern oder verlangsamen.
Precip and QPF-Scores 14. Mai 2008 06-18 UTC
Precip and QPF-Scores 18. Mai 2008 18-06 UTC
Precip and QPF-Scores 02. Juni 2008 06-18 UTC Wetterlage: Okkulsionszone im Bereich Elbe, Konvergenzlinie in Bereich Rheinland- Pfalz/Baden-Württemberg
Precip and QPF-Scores 31. Mai 2008 06-18 UTC
Precip and QPF-Scores 29. Mai 2008 18-06 UTC
Zeitpunkt der max. Niederschlagsrate 03. Juni 2008 06-18 UTC
Precip and QPF-Scores 03. Juni 2008 18-06 UTC
Case study 9 June 2007: Prä-frontale Konvergenzline führt zu starker Konvektion: Radar turlen=500m, clc=0.75 turlen=250m, clc=0.50 turlen=150m, clc=0.50 turlen=150m, clc=0.25 turlen=150m, SD77
Hovmöller diagrams of precipitation rate from 1 June to 21 June 2007 Radar turlen=500m, clc=0.75 turlen=250m, clc=0.50 turlen=150m, clc=0.50 turlen=150m, clc=0.25 turlen=150m, SD77
Hovmöller diagramms of precipitation frequency 1 May to 9 June 2008 Confirms the findings seen in the precip rate diagramms.
Hovmöller diagrams of precipitation frequency Hovmöller diagrams of precipitation frequency from 1 June to 21 June 2007 Radar turlen=500m, clc=0.75 turlen=250m, clc=0.50 turlen=150m, clc=0.50 turlen=150m, clc=0.25 turlen=150m, SD77
Standard SYNOP-Verifikation: 00 UTC Lauf Standard SYNOP-Verifikation: 00 UTC Lauf (Achtung: Hier Experiment für Juni 2007) 2m-Temperatur und -Feuchte zeigen ein deutliche Änderung.
Akkumulierter Niederschlag und QPF-Scores 13. Mai 2008 06-18 UTC