COSMO WG3-WG5 workshop 9 March 2005, Langen

WG3-WG5 workshop, , Langen 1 Agenda for joint WG3-WG5 workshop (1) WP 5.6.2: “Verification of new schemes from WGs 1-3” 09: :30 Marco Arpagaus and Matthias Raschendorfer: Introduction to WP : :15 Francis Schubiger, Pirmin Kaufmann, Dominique Ruffieux, and Matteo Buzzi: “A first attempt to WP 5.6.2” 10: :45 Discussion, next steps, i.e., 1. Who should run the test cases, WGs 1-3 or WG5? (all) 2. Test cases for WP 5.6.2: defined by WGs 1-3 or WG5 (or both)? (all) 3. What can be improved? – First experience from WG5 members … (all) 4. What could be the next (WG3) candidates for WP 5.6.2? (Arpagaus) 5. … 10: :15 coffee break

WG3-WG5 workshop, , Langen 2 Agenda for joint WG3-WG5 workshop (2) WP 5.5.1: “Validation of near-surface boundary layer processes” 11:15 -11:45 Gerd Vogel and Felix Ament: “Impact of soil moisture on near surface atmospheric layers” Other results from WG5 11: :15 Chiara Marsigli: “Analysis of LM behaviour in different configurations on some test cases” 12: :45 Ulrich Damrath: ”Some systematic LM-errors” 12: :45 lunch break 13: :00 WG5 workshop

WG3-WG5 workshop, , Langen 3 “Verification of new schemes from WGs 1-3” Aim(s) of WP A first attempt … TKE scheme: operational verification results Experiments at ECMWF: exp001 & exp002 In-depth verification by WG5 members Discussion 1. Who should run the test cases, WGs 1-3 or WG5? (all) 2. Test cases for WP 5.6.2: defined by WGs 1-3 or WG5 (or both)? (all) 3. What can be improved? – First experience from WG5 members … (all) 4. What could be the next (WG3) candidates for WP 5.6.2? (Arpagaus) 5. …

WG3-WG5 workshop, , Langen 4 Aim(s) of WP Already done: Establish first WP in WP5 (“Verification and Case Studies”) concerning case studies … As soon as possible: Try to bring WG5 results closer to the needs of WG3 (as well as WGs 1 & 2), i.e., get some (WG3- independant) results based on conditional verifications of test cases, test periods, and operational model suites. Christmas 2006 (?): Provide a common and easy-to-use tool that allows WG 1-3 members to do conditional verifications themselves --- the tool should ideally be provided, maintained, and used by WG5.  a first attempt to WP 5.6.2: a simple (sensitivity-) experiment …

WG3-WG5 workshop, , Langen 5 TKE scheme: operational verification results (1) LM: model has a cold bias in the PBL

WG3-WG5 workshop, , Langen 6 TKE scheme: operational verification results (2) LM: model has a cold bias in the PBL

WG3-WG5 workshop, , Langen 7 TKE scheme: operational verification results (3) aLMo test-suite (old [OPR] vs TKE scheme [TKE]; Sep & Dec): TKE scheme has a cold bias in the PBL (vs OPR (& obs))

WG3-WG5 workshop, , Langen 8 TKE scheme: operational verification results (-) The model seems to have a cold (LM; aLMo test-suite; aLMo with old scheme generally too cold) and moist (relative humidity! - LM; aLMo with old scheme generally too moist) bias in the PBL, [especially during winter (aLMo only, CNL3, p. 107)] and at night (LM, CNL4 p. 98)

WG3-WG5 workshop, , Langen 9 TKE scheme: operational verification results (-) Hence, the experiments are aimed to have an impact on the temperature and (specific) humidity structure of the PBL Note that this also means that you should primarily look at the PBL structure, and not just at the diagnosed parameters at 2m/10m.

WG3-WG5 workshop, , Langen 10 Experiments at ECMWF: exp001 & exp002 (1) A cold (and moist) model bias first of all is a hint for too much evaporation in the model. In this case, for cloud free conditions, the model would tend to systematically have too small heating rates within the boundary layer during day time, connected with too much moistening. On the other hand, during winter there is only little forcing of the boundary layer during day time and cooling during the night may be more important. Thus, a too cold boundary layer may also be a hint for too strong heat fluxes during the night from the atmosphere towards the soil. In this case the model would tend to have too strong cooling rates during clear sky night conditions.

WG3-WG5 workshop, , Langen 11 Experiments at ECMWF: exp001 & exp002 (2) Large domain (461 x 389) encompassing most of Europe IBCs and LBCs from ECMWF analyses LM version 3.12, 7 km mesh size, 45 vertical levels Exp001: most namelist parameters identical to operational setting at DWD Exp002: same as exp001 except for pat_len = 0. and ztk_hmin = ztk_mmin = 0. (see below) 72 hours forecast starting 12 Dec UTC High pressure system (SLP and 500 hPa) over central Europe; fairly stable stratification for large parts of the domain, some low level clouds over northern Europe Full model level output (GRIB) stored on ECFS at ECMWF; extraction of data with fieldextra

WG3-WG5 workshop, , Langen 12 Experiments at ECMWF: exp001 & exp002 (3)

WG3-WG5 workshop, , Langen 13 Experiments at ECMWF: exp001 & exp002 (4) It should be possible to check both possibilities (see above) by comparing the evolution of boundary layer profiles (‘diurnal cycle’) of temperature and specific humidity (dew point temperature) in exp001 with those of the corresponding analysis (or with observations directly). Exp002 is a run with decreased vertical mixing during stable conditions (pat_len = 0. (instead of 500.) switches off the circulation term in the prognostic TKE equation; additionally, ztk_hmin = ztk_mmin = 0. (instead of 1.) sets the minimal diffusion coefficients for scalars (heat, moisture) and momentum to zero). Assuming that too much mixing during the night plays an important role, exp002 should have a reduced cold bias.

WG3-WG5 workshop, , Langen 14 Experiments at ECMWF: exp001 & exp002 (5) Conditional verification: look for cloud free situations (at least low and mid-level clouds; obs & model) look for a stably stratified PBL (obs & model) possibly compare the above with 'opposite' situations (i.e., cloudy situations, unstably stratified PBL) This would allow to see whether there is too much vertical exchange/diffusion in case of a stably stratified PBL (and with a correctly forecasted cloud cover). Additionally, differences over land/sea would also be interesting (the amount of low level clouds over sea may be a potential problem in exp002).

WG3-WG5 workshop, , Langen 15 In-depth verification by WG5 members Francis Schubiger, Pirmin Kaufmann, Dominique Ruffieux, and Matteo Buzzi: “A first attempt to WP 5.6.2”

WG3-WG5 workshop, , Langen 16 poor mans conditional verification (1)

WG3-WG5 workshop, , Langen 17 poor mans conditional verification (2)

WG3-WG5 workshop, , Langen 18 poor mans conditional verification (3)

WG3-WG5 workshop, , Langen 19 poor mans conditional verification (4)

WG3-WG5 workshop, , Langen 20 poor mans conditional verification (5)

WG3-WG5 workshop, , Langen 21 Discussion Discussion of results … … and: 1. Who should run the test cases, WGs 1-3 or WG5? (all) 2. Test cases for WP 5.6.2: defined by WGs 1-3 or WG5 (or both)? (all) 3. What can be improved? – First experience from WG5 members … (all) 4. What could be the next (WG3) candidates for WP 5.6.2? (Arpagaus) 5. …

WG3-WG5 workshop, , Langen 22 Who should run the test cases? WGs 1-3: To be sure that the correct model version and namelists are used (PK, FS) To be sure that everyone looks at the same data WG 5 members individually: To be sure to have the right model output (MM) To be sure that the cases are of interest to the WG5 member (CM)  WG3 favours the 2nd option.

WG3-WG5 workshop, , Langen 23 Test cases: defined by WGs 1-3 or WG5? WGs 1-3: Because a heavy precipitation case may not be very suitable to test clear-sky radiation issues … WG 5: To be sure that the cases are of interest to the WG5 member (CM) ‘smallest common multiple’ - run some cases defined by WGs 1-3 and some more of particular interest to WG5: members More work … ‘largest common divider’ - run the cases WGs 1-3 select from the WG5-defined list: There may be none …

WG3-WG5 workshop, , Langen 24 What can be improved? First experience from WG5 members …

WG3-WG5 workshop, , Langen 25 The next candidates for WP 5.6.2? WG3: another sensitivity-experiment for the TKE scheme (?) 3d turbulence (WP 3.9.2; 2.8km mesh size) Kain-Fritsch scheme (WP 3.3.1) WGs 1 & 2: ???