1 / 110

S teering Committee M eeting

S teering Committee M eeting. SPM Slides (dmiitrii.mironov@dwd.de). 1-2 March 201 8 , Offenbach am Main, Germany. 10. Overview of current status of WGs, PPs, PTs (70 min). WG1, KENDA-O WG2, CDIC, CELO WG3a, T2(RC)2, ConSAT4 WG3b, CALMO-MAX, TERRA Nova, AEVUS, SAINT

Download Presentation

S teering Committee M eeting

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Steering Committee Meeting SPM Slides (dmiitrii.mironov@dwd.de) 1-2 March 2018, Offenbach am Main, Germany

  2. 10. Overview of current status of WGs, PPs, PTs (70 min) • WG1, KENDA-O • WG2, CDIC, CELO • WG3a, T2(RC)2, ConSAT4 • WG3b, CALMO-MAX, TERRA Nova, AEVUS, SAINT • WG4 current activities • WG5, INSPECT • WG6, CEL-ACCEL, POMPA (extension) • WG7, SPRED, CIAO

  3. WG1, KENDA-O Christoph Schraff

  4. Status Report for KENDA-O • Task 1: further development of LETKF scheme (conventional obs)  MCH: 2-year position on additive covariance inflation / regional B matrix: Claire Merker, visit at DWD in Jan.  COMET: KENDA slightly worse than COMET-LETKF, continued investigation  tests (at 2.2 km at COMET, ARPAE; parallel suite with additive infl. at MCH, …) etc.  understanding discrepancies betw. DWD and MeteoSwiss KENDA results (most critical issue, see next slides) • Task 2: extended use of observations  Mode-S operational at DWD (since 4 Oct. 2017)  work on other obs types ongoing • Task 3: lower boundary: soil moisture analysis using satellite soil moisture data  soil moisture data assimilation tests in parallel suite • Task 4: adaptation to ICON-LAM (re. C2I)  schedule set up, main milestones:  12/18: consolidated ICON-LAM LETKF;  03/19: 3DVAR + regional B-matrix + EnVar technically available;  07/19: parallel suite at DWD; risk: A. Rhodin will leave DWD end 03/18)

  5. Critical issue in KENDA-O: discrepancies between MeteoSwiss & DWD KENDA verif. results MeteoSwissanalysis verification DWD verification + 0 h COSMO-1  nudging COSMO-E  KENDA COSMO-7  nudging Winter 2016 KENDA analysis worse KENDA analysis not worse (without additive inflation)  run experiment at DWD : comparison KENDA vs. Nudging for Dec. 2016 (winter, extended low stratus periods) DWD setup (KENDA, ICON-LBC, obs (no Mode-S)), but on Swiss COSMO-E domain • DWD - standard verification: similar results as with COSMO-DE, but … 5

  6. Investigation of discrepancies between MeteoSwiss & DWD KENDA Reason for differences betw. the 2 plots: LETKF f.g. check rejects too many obs near strong inversions  (not optimal for data assimilation) if MEC is applied to ekf files written by LETKF, these obs are not used in the verification  verification tends to be blind (to differences betw. Nudging and LETKF) near strong inversions and underestimates errors particularly of analyses temperature RMSE DWD std. verif: MEC based on DWD ekf files  with LETKF first guess check KENDA forecast slightly better 1 – 27 Dec 2016 MCH-style verif: MEC based on (Swiss)cdfin files  no LETKF first guess check RMSE KENDA forecast slightly better above 850 hPa, slightly worse near surface 1 – 26 Dec 2016 KENDA analysis much worse than nudging and than in DWD stdverif at low levels 6

  7. Investigation of discrepancies between MeteoSwiss & DWD KENDA 6 – 24 h forecasts: radiosonde verification RH T wind speed wind dir. MEC based on Swiss cdfin files  no LETKF first guess check 1 – 27 Dec 2016 MEC based on DWD cdfin files  no LETKF first guess check KENDA vs. nudging 1 – 31 Dec 2016 MEC based on DWD ekf files  with LETKF first guess check 1 – 27 Dec 2016 • the way that MEC is applied, i.e. that verification is done, has rather little effect on forecast scores 7

  8. Investigation of discrepancies between MeteoSwiss & DWD KENDA • performance differences between KENDA and nudging similar on Swiss domain as on COSMO-DE domain (e.g. low stratus, surface verif, standard radiosonde verif) • COSMO (cdfin-based MEC) first guess check rejects almost no data • LETKF first guess check rejects about 5% for T, RH and about 2.5% for wind, particularly near inversions (and in stratosphere)  too many good obs are rejected • discrepancies in upper-air analysis scores at MCH and DWD are (apparently) mainly due to different quality control in verification, not due to difference in analysis and forecast performance of KENDA as a result of different model domains, ensemble LBC’s, data input, etc. • the problem, i.e. the differences between the verification results, are understood • solution:  improve model, eliminate systematic model errors (long term, WG3)  refine first guess check in LETKF analysis (first implementation and test already done)

  9. WG2, CDIC, CELO Michale Baldauf

  10. PP CELO Work done: • Migration with EULAG dynamical core (compressible) to the most recent version of COSMO 5.4h • Tuning of physical parameterizations of the CE model based on deliverables from the CALMO project • Preliminary work on implementation of tailored restart subroutine in the CE for Task 1: “implementation/coupling of ICON physics … … advice from SMC about the physics package required.” SMC recommendation: from the experiences at DWD, we cannot recommend the new common COSMO-ICON-package! M. Baldauf (DWD)

  11. PP EX-CELO Zbigniew Piotrowski delivered an updatedversionoftheproject plan (newestversionfrom 15 Feb. 2018) forapprovalby STC Dmitriiposesthefollowingquestions: how non-vanishing wind at the surface is reconciled with the non-slip condition for the velocitynon-slip condition is replaced by the transfer scheme what is meant by "explicitly solve fluxes at the 0 m level“ what "adaptation of ICON physics to fully exploit 0 m level" specifically involves. These itemshadbeendiscussedduringthe last SMC meeting. M. Baldauf (DWD)

  12. PP CDIC Task 1: Idealizedtests sometraceradvectiontestsperformed on COSMO side (see ‚New Bott scheme‘) These traceradvectiontests will beaccessible (via src_artifdata.f90)in thenext COSMO version. However, thesetests still must betransferredtothe ICON-LAM side. Task 5: Suitability of ICON dynamical core for other applications than NWP … First feedback from CLM community received.Larger recognised problem: open model top; instead relaxation analogous toCOSMO might be required M. Baldauf (DWD)

  13. The new Bott (2010) tracer advection scheme W. Schneider, A. Bott (Univ. Bonn)M. Baldauf (DWD) A closerinspectionofthecodedeliveredbyUniv. Bonn (W. Schneider)showstwoissuesthatshouldbesolved: • a lotoffieldcopyingisused in an interfaceprogram (looksinefficient)also thiscodeversioniswritten in a rather ‚non COSMO like‘ style decisiontorewritethecodecompletely • a moreseriousissue: thecodedoes not parallelizecorrectlyfor CFL>1!This is not just a lack ofexchangestatements but a design problemin theadvectionscheme (asdescribed in Bott (2010))! 3 proc.s: 1 proc.: 1 GP with CFL>1 not correctlyparallelized o o ooooooooooo ? oooo Bott (2010) proposal: ? M. Baldauf (DWD)

  14. Summary: • Toavoidtheseproblemsandbeyondthistoincreaseefficiency, MB developed a local time-steppingmethodfortheBott (2010) scheme • The reinvented/reimplementednew Bott scheme(i.e. Bott (2010) + local time-stepping) passes all idealisedadvectiontests;in particularmassconservationviolationlooks ‚least pathological‘ • Verificationof a hindcastrunduringsummer: quite neutral scores in comparisonwiththecurrentscheme(nonumericalstabilityproblemsvisible) • Efficiency gainwiththenewlocal time splittingmethod:the pure advectionschemeis~30% fasterthanthecurrentone a whole COSMO-D2 runis~5.5% faster! • Code readyfor COSMO 5.6 (goodchancetogo operational with CD2) M. Baldauf (DWD)

  15. Higher order scheme A. Will, J. Ogaja (Univ. Cottbus) (no progress on DWD side …) Next steps: • Transfer the code from version 5.0 to 5.5 ( M. Baldauf) • Run this new version on NUMEX ( M. Baldauf) • Deliver documentation (COSMO Sci. Doc. part I, possibly alsoa COSMO-TR (?)) ( A. Will) • expected availability for the official code version: ~June 2017~June 2018 (will not likely go operational within CD2) M. Baldauf (DWD)

  16. WG3a, T2(RC)2, ConSAT4 Matthias Raschendorfer

  17. PP T2(RC)2 Highlights (Harel Muskatel, 23.02.2108) • Intensive running on ECMWF computers: tuning the new radiation scheme parameters (CALMO); runs are finished; analyzis is ongoing (some results will be presented by P. Khain at ICCARUS 2018). • Monte-Carlo spectral integration: finished and tested (talk by H. Muskatel at ICCARUS 2018). • Ice and water droplets new optical properties implemented are into ICON-RRTM (same was done for COSMO); parametrizations were calculated by U. Blahak and H. Muskatel; implementation is performed by S. Gruber; testing is needed but it depends on the progress with ICON-LAM in IMS (Russia), not enough (CPU) power in Israel to run global tests. • ICON-ART aerosols are implemented into COSMO radiation scheme (almost finished); most of the coding work is finished (by D. Rieger and H. Muskatel). • Verifications is performed Chubarova et al. (talk at ICCARUS 2018).

  18. Along the aim of COSMO-SP to consider missing interactions: Mass fraction and number concentration of • cloud constituents, • precipitation and • passive tracers (aerosols) Optical properties: • optical thickness, • single scattering albedo, • asymmetry factor and • delta-transmission function Increased set of diagnostic or prognostic variables: modified by SGS-processes feed-back with SGS-processes PP T2(RC)2 Radiation transport Cloud-microphysics Local parameterizations: Parameterizations of source terms integrated in GS parameterizations: Parameterizations of SGS processes interaction STIC Turbulence Separation Circulations PT ConSAT Surface and Soil-processes

  19. Testing & Tuning of Revised Cloud Radiation Coupling T2(RC)2PP: Status & Highlights since CGM in Israel Project leader: Harel Muskatel (IMS) Pavel Khain (IMS), Alon Shtivelman (IMS), Oliver Fuhrer (MCH),Xavier Lapillonne (MCH), Gdaly Rivin (RHM), Natalia Chubarova (RHM), Marina Shatunova (RHM), Alexey Poliukov (RHM), Alexander Krisanov (RHM) Ulrich Blahak (DWD), Daniel Rieger (DWD), Matthias Raschendorfer (DWD), Simon Gruber(KIT)

  20. The highlights/news are: • Phase II accepted! • Intensive running on ECMWF computers (ongoing): • tuning the 30 new radiation scheme parameters (CALMO). • Fist set of runs is finished needs now to be analyzed. • Implementation of Monte-Carlo spectral integration into COSMO-radiation: • finished and sanity-tested • New optical properties for ice- and water-droplets implemented in ICON-RRTM: • same as what we did in COSMO. • The parametrizations were calculated by Uli. B and Harel M. • Implementation done by Simon Gruber • ICON-ART aerosols: • Implemented in COSMO radiation • Adaptation of INT2LM code for interpolation of ICON-ART fields

  21. Implementation of CAMS aerosols into COSMO radiation scheme: • Optical properties adapted from ECMWF to COSMO 8-band radiation • INT2LM and COSMO prepared to use mixing ratios of 5 different aerosol-species taken from MACC • Coding, test-runs and documentation ready • Comparisons with different aerosol-data concluded (showed at CGM) • Inter-comparison with accurate measurements in cloudy conditions -> phase 2 • Implementation of CAMS prognostic aerosols into COSMO microphysics: • Introducing the Segal&Khain-method to define cloud number concentration -> to be continued in Phase II • Implementation into ice-nucleation scheme -> starting in Phase II • Case studies, Inter-comparison and documentation -> Phase II

  22. Further tasks for Phase II: • Standard-Verification with new cloud-radiation coupling including different aerosol input • Updating official COSMO version with the recent T2(RC2)-development • New combined scheme for SGS clouds • Investigation, whether components of the shallow convection parameterization by Boeing can be used for cloud-cover parameterization • New development of shallow convection, if necessary • Implementation of a proper combination with turbulent saturation-adjustment (“statistical” cloud scheme) • Verification of revised combined scheme for SGC clouds against ground-base and satellite data

  23. General and common WG3-Task “Consolidation of the Surface-to-Atmosphere (ConSAT): according to a dynamically adapted list of actions being the base of past and (maybe) future PTs Current topic: Reformulation of surface-processes with respect to roughness-effects and numerical stability in TERRA and TURBTRAN Current contributors: Matthias Raschendorfer (DWD) Günther Zängl (DWD), (Jan-Peter S., Jürgen H.) COSMO Matthias Raschendorfer Offenbach 2018

  24. Main lessen from previous ConSAT tasks: • Background-diffusion (BD), introduced, e.g., by minimal diff.-coeff., is a substitute of missing STIC terms (acting due to non-turbulent heterogeneity). • If BD is applied in the BL even above a homogeneous surface, it partly destroys the stability reduction of SAT-velocity. • Modifications in the description of the turbulent Prandtl-layer can hardly correct the main sources of current model-errors of the diurnal cycle of near surface variables! • The process description of surface processes provides by far the largest potential for improvement!! COSMO Matthias Raschendorfer Offenbach 2018

  25. Case study: 23.06.2016 COSMO-DE with lateral boundaries from ICON-EU • only for rather smooth surfaces; applied filter • almost saturated soil due to long standing rain period before • almost no clouds due to high pressure situation; + applied filter TD_2m T_2m still much too moist in the afternoon still much too cold during day-time nocturnal warm bias removed but perhaps a new nocturnal dry bias direct analysis of T_2m and TD_2m operational configuration revised TURBDIFF imported from ICON Geneve 2017 Matthias Raschendorfer

  26. Case study: 23.06.2016 COSMO-DE with lateral boundaries from ICON-EU • only for rather smooth surfaces; applied filter • almost saturated soil due to long standing rain period before • almost no clouds due to high pressure situation; + applied filter TD_2m T_2m removed cold bias in the afternoon! almost perfect double wave! amplitude slightly overestimated still some problems during heating of the cover (stable stratification within the roughness layer) revised TURBDIFF imported from ICON + new decoupled surface cover: revised TURBDIFF imported from ICON direct analysis of T_2m and TD_2m Geneve 2017 Matthias Raschendorfer

  27. Another lessen from previous ConSAT tasks: • Direct process-parameterization can be substituted by empirical (statistical) hyper-parameterizations, trying to derive missing relations by diagnostics and sensitivity-experiments (trial&error). • What Günther has rather successfully done in ICON, e.g., related to • Background-diffusion: to be substituted by STIC-terms • Roughness-effects for the snow-cover: to be substituted in the framework of a full R-layer description COSMO Matthias Raschendorfer Offenbach 2018

  28. Illustration of a realistic surface concept: : indicator of the lowermost full atm. level may carry liquid or frozen interception water : area and indicator of total surface : area and indicator of the cover-surface (surface of all R-elements) R-layer : indicator of the snow-cover surface snow-cover : area and indicator of the bare-soil surface : indicator of the uppermost full soil level : indicator of the snow-cover surface • Critical properties: : indicator of all snow-free surfaces • R-elements are substantial, semi-transparent and thermally loosely coupled to B. • Their Temperature T_c may be different from T_b, which is distinctive from T_b1. • They may contain liquid and frozen interception water. • They can shade B from radiation. • Snow-free parts of S are mainly parts of C. • R-elements vanish within an increasing snow-cover. • Distribution of snow cover alters aerodynamic R-length (z0) and short-wave albedo. • Coexistence of liquid water and ice possible. • Smooth transition due to variance of T_sf or T_sn. Matthias Raschendorfer Offenbach 2018

  29. Illustration of the current idealized surface concept: increasing with snow-water level : indicator of the lowermost full atm. level only liquid interception water : area and indicator of total surface : area and indicator of the cover-surface (surface of all R-elements) R-layer : indicator of the snow-cover surface : area and indicator of the bare-soil surface : indicator of the uppermost full soil level : indicator of the snow-cover surface • Current implementation: : indicator of all snow-free surfaces • Total land-use surface is a topographic enlargement of the horiz. ground by the factor SAI • thermally coupled with the soil like the bare-soil surface • with fractions of distinctive sf-evaporation: B, (wetted , transpirating or sealed) C • Only one sf surface-temperature T_sf=T_c=T_b=T_b1. • R-elements carry only liquid interception water. • Mass (heat-capacity) of R-elements is neglected. • R-elements do not cover (shade) the bare-soil surface B, as they are rather a part of it. • R-layer is treated like a pure additional transport-resistance for sensible and latent heat. • Snow and rime are treated together and are concentrated in a single cover. • Snow follows surface-structure (no additional effect on z0). Matthias Raschendorfer Offenbach 2018

  30. Current state in operational ICON-TERRA: • “dirty” fixes related to snow at a rough surface implemented by Günther Z.: • Melting snow: f_sf represents mainly the surface C of sf R-elements: • Snow is artificially “pushed together”. • f_sn<1, although snow may totally cover the soil-surface B. • Mean surface temperature T_s may become > 0°C, what was missing before! • But: Artificial bare-soil evaporation with T_sf>0°C at the fraction f_sn*B needs to be restricted by potential snow-evaporation! • Dynamic sn and sf sub-tiles prevent from a too fast melting of snow. • Frosty snow: sf R-elements can’t be treated similarly: • T_sf would become too warm as C is thermally not yet decoupled from B. • Calculating a reduced albedo of “dirty” snow dependent on snow-age and z0. • f_sn may be as large as 1, although the surface C is sf. • “Dirty” snow (polluted by sf R-elements) gets only a bit warmer than pure snow. • But: Snow-evaporation is too strong and needs to be artificially restricted! • Rime: No longer treated as a contribution to snow: • Frozen R-elements can evaporate potentially even without a snow-cover • But:Enthalpy by freezing or melting of interception water not yet considered and interception of fresh snow not yet included! Offenbach 2018 Matthias Raschendorfer

  31. Plant-evaporation: reduction during late afternoon: • Introduction of a evaporation hysteresis due to restricted water supply through the stems • New implementation needs to reduce these fixesas far as possiblebutnot more than still required!! • Our beleave: • Most of these indirect “tricks” can be substituted via the full R-layer implementation!! COSMO Matthias Raschendorfer Offenbach 2018

  32. Main aims with respect to an improved process-description: • Better representation of effects caused by surface roughness also in combination with snow and interception-water • Semi-transparent and loosely coupled roughness cover: • Shading of the soil-surface B by the R-cover C (e.g. plants). • Different surface temperatures for pure soil (T_b) and R-elements (T_c). • Better adapted soil-evaporation and plant-transpiration. • Stronger thermal decoupling of C from B. • Improved diurnal cycle of near-surface variables. • Realistic treatment of snow and rime at a rough surface: • Allowing a sf roughness cover C above the snow-cover at B. • Allowing interception also of snow- or rime at C. • Modification of short-wave albedo dependent on snow-distribution. • Modification of roughness-structure by snow-cover. • Realization of a mean T_s > 0°C for a totally snow-covered soil • Revision of dynamic sn and sf sub-tiles. • Proper consideration of conversion-enthalpy of freezing or thawing of interception water. Matthias Raschendorfer Offenbach 2018

  33. Main aims with respect to numerical represenatation: • Improvement of numerical stability • Removal of T_s-oscillation and artificial flux-limiter: • Implicit treatment of T_sf and T_sn in near-surface heat-budgets. • Linear combination of heat-budgets of soil-layers, the roughness cover skin and the snow-cover. • Considering the implicit dependency of transfer-velocity for heat on T_s. • Implicit and positive-definite formulation of water-levels with respect to evaporation (of snow and interception water). • Removal of various artificial limitations. • Avoiding of artificial numerical sources due to implicit treatment. • Implicit formulation of freezing and melting of snow-water and soil-water. • Towards a smooth transition between water-phases. • Implicit and instantaneous integration of all sources of interception water Offenbach 2018 Matthias Raschendorfer

  34. Current state of new implementation: • Revised TERRA and TURBDIFF in ICON-branch: • Including all measures (green aims) with respect to implicit treatment of T_s • Contains also a separate equation for implicit T_sf • Various adaptions related to: • Calculation of snow-cover fraction f_sn • Formation of dynamic sub-tiles • Redistribution of variables for dynamic sub-tiles • A couple of consistency-checks performed: • Removal of various bugs mainly related to incomplete or inconsistent sub-tile redistribution • Now only slight differences to previous solution • Günther’s incomplete treatment frozen interc.-water so far deactivated! • Degree of vectorizations can be improved. • Shows already a lot from a previous test-implementation in the non-blocked COSMO-version of TERRA with has the following properties: • semi-transparent, loosely coupled and substantial R-cover treatment • but: without all snow- and interc.-water related conceptual and numerical adaptations. • Introduction of the missing extension towards the full R-cover treatment into the current ICON-branch with the revised treatment. Offenbach 2018 Matthias Raschendorfer

  35. Ad I: Resulting matrix of the extended linear system: altered • All 2 + k_soil budgets are always present (even for f_sn=0 or f_sn=1) • They are linearly coupled in the temperatures: created isc fes ifb • Can easily be tri-diagonalized by matrix-operations and solved by the standard solver • Partly reducible by parameters: degree of corrected implicit coupling of T_sn to the soil- and atm. temperatures isc: degree of considered flux-equilibrium in diagnostics of T_sf fes: degree of implicitness for effective surface fluxes used in the heat budgets ifb: Default for test: isc=1; fes=1; ifb=1 (full implicit solution active) - modified for diagnostic points

  36. Ad I: Test-grid-point Kenia (+33.71_+7.89) : • After-noon situation; tropical hot with strong radiation forcing • 3 hour ICON-global test-run (R2B6) with defaults of the new SAT/TERRA-scheme (dt=6 min) • Non-default settings only for the special grid-point: T_sf LHF_sf old-sx-cpl + Ifb=1 + itv=1 old-sx-cpl + Ifb=1 + itv=1 • Oscillations almost completely eliminated by ifb=1 + itv=1 • Similar result but a bit larger daily amplitudes ifb=1 + itv=1 + fes=1(not shown) itv=1: full consideration of implicit T_sx-dependency in atmospheric transfer velocity fes=1: full consideration of flux-equilibrium at the sf surface

  37. Most recent work going to be implemented soon into the ICON-branch: • Introducing an implicit treatment of the amount of interception water (WI) • Instantaneous integration of • liquid/frozen precipitation (Prc) • mdew/rime-fall or evaporation (± Evp) • echanical drip-off (Drp) • So far decoupled from implicit temperature equations • Introducing a mixed liquid/frozen phase of WI • Controlled by a T_sf-dependent liquid water fraction f_liq with a • smooth transition between liquid and frozen within a T_sf-range • Included into implicit treatment of T_sf: • linearizationof saturation humidity and f_liq as a function of T_sf • consideration of f_liq-dependent implicit latent heat from • freezing/melting of WI or Prc • ±Evp • Calculation of atmospheric correction-heat-fluxes compensating • discrepancies between hydrologic mass conversions of pater water-phases and implicit latent heat release, which are due to • linearization errors • missing implicit coupling between T_sf and WI • Not yet considered T-sf-dependent flux-corrections in radiation scheme quadratic equation for WI: • positive-definite • capacity-limited Offenbach 2018 Matthias Raschendorfer

  38. WG3b, CALMO-MAX, TERRA Nova, AEVUS, SAINT Jean-Marie Bettems

  39. COSMO WG3bStatus Jean-Marie Bettems / MeteoSwiss Offenbach, COSMO StC, March 2018

  40. WG3b highlights Summary of WG3b activities and links to related documents http://www.cosmo-model.org/content/tasks/workGroups/wg3b/default.htm

  41. PP CALMO-MAX • Calibration of unconfined model parameters • Antigoni Voudouri / HNMS, 06.2017 – 09.2019 • Consolidate results of PP CALMO, provide community tool • Strong interest of Prof C. Schaer / ETHZ group • This method is used at ETHZ to periodically calibrate COSMO-CLM • Similar method used by Chineese group with WRF • Up to 30% improvements of precipitation scores for moonsoon case over Beijing • Paper available in BAMS / May 2017

  42. PP CALMO-MAX • Two and a half days workshop at Athens beginning of January • Omar / ETHZ (original development), Yoav / IMS, Edoardo / CIRA, Euripides, Antigoni, Flora, DImitra / HNMS, Jean-Marie / MCH • Very useful discussions, minutes have been distributed • Good example of COSMO collaboration and knowledge transfer • Task 2.1 on track (COSMO-1 calibration on Daint / CSCS) • Calibrating 6 parameters of production configuration of COSMO-1 • With improvement of meta-model and performance function • In depth verification should definitely show the usefulness of the method for improving the forecast quality

  43. PP CALMO-MAX • Meta-model • On COSMO web and on GitHub in public repository • Common development between IMS and ETHZ • Octave version coming ( ECMWF) • Extension of meta-model and performance function (TBE) • Consider model internal variability to filter noise • Use 6h instead of 24h accumulated precipitation (daily cycle) • Add precipitation FSS to constraint precipitation spatial structures • Add 2m humidity constraint (to avoid over fitting 2m temperature) • Add sunshine duration constraint

  44. PP CALMO-MAX • Computing cost of the method • Running calibration in hindcast mode significantly reduces the cost of the method (and simplify the experimental setting). • It is possible to fit the MM with the minimum number of simulations,namely 2*N + N*(N-1)/2 + 1 for N parameters. • If the soil memory is not an issue, sampling a full year with representative days will considerably reduce the cost of the method; otherwise a full year is most probably required. • Impact of calibrating with a reduced domain size will be evaluated. • Case study: C1 calibration for 6 parameters: Use one year hindcast with 0.5 time domain extension:calibration requires about 8 years operational configuration equivalentIs it much? … similar to MOS or EPS calibration requirements…Less than 2 weeks time on a machine with 1000 GPU’s

  45. PP CALMO-MAX • Suggestion fordocumentationofmodeltuningparameters • Table on COSMO web, ideallyfilledupduringdevelopmentphase • Includingparametersimplementedashardcodedvalues! • Withshortdescription, default, minimumandmaximumvalues • Includinginformationaboutmodelsensitivity(summer/winter, different targetareas) • Preliminarystepsdone in CALMO / CALMO-MAX and in WG7 • Coordinationworkshopplanned in Athens in Spring 2018 • Shouldbecome a permanent taskof COSMO • All unconfinedmodelparametersshouldbedocumented in namelist • But thesenamelistsshouldbeonlyvisibletoexperts

  46. PT TERRA Nova / MSc Verena / PhD Daniel • PT TERRA Nova, 09.2016 – 06.2018, Y. Ziv / IMS MsC Verena, 12.2017 – 05.2018, Prof. Seneviratne / ETHZ • document TERRA performance, compare with CLM performance • compare v5.0 / v5.05 conservative / v5.05 aggresive / CLM • Simulations with TERRA v5.0 (EU, RU @ 7km) and CLM (EU @ 7km) performed and being analyzed (MCH tool for standard verification, additional verification). • Additional simulations with latest TERRA and on Eastern Mediteranean domain are planned, but only @ 7km. • PhD Daniel Regenass, 01.2018 – 12.2020, Prof. Schär / ETHZ • first step is to test Linda Schlemmer developments in NWP mode(topo dependent water table  strong positive effect on climate simulations) • next tasks still open, but goal is to address MCH specific shortcomings • common meeting with all stake holders took place at ETHZ on Jan. 16 • use TERRA Nova test bed

  47. PT AEVUS • Urban parameterization for operational NWP • Paola Mercogliano / CIRA, 09.2017 – 12.2018 • based on Hendrik Wouters bulk model • Code base is COSMO 5.04g, plus the latest URB development by Hendrik (TERRA-URB v2.3), including all known bug fixes developed for the climate version • Code base is ready (thanks to Uli S) • Sanity check of this new release is beeing perfomed by the PT team • AEVUS meetingICCARUS / Wednesday  February 28th (chair Paola Mercogliano)

  48. PT SAINT • Validate and update the multi-layer snow model to make it production ready • Sascha Bellaire / SLF, 07.2017 – 06.2019 • full support of SLF (Prof. Michael Lehning) • Code base is COSMO 5.04g, plus the latest SAT development by Matthias (in particular the implicit formulation of the near surface heat budget) • Code base is ready (thanks to Uli S.), code has been analysed by Sascha who is now evaluating different options ( discussion with DWD colleagues) • Martin Koehler will prepare an environment at ECMWF for tests with global ICON (but not before autumn 2018) • SAINT meetingICCARUS / Thursday  March 1st  (chair Sasha Bellaire)With participation of Matthias, Ekaterina …

  49. SNOWE and snow analysis • SNOWE defined as COSMO software by StC • full featured snow analysis package, incl. snow density • developed and maintained at RHM • code and documentation available on COSMO sitehttp://www.cosmo-model.org/content/support/software/default.htm • used for production at RHM • Snow analysis for COSMO – Status & plan ICCARUS / Thursday  March 1st   (chair Juergen Helmert)

  50. EXTPAR • In its March 2017 meeting the COSMO StC has nominated Katie Osterried, working at ETHZ for C2SM, as Source Code Administrator • The official source code is available in a private repository in the C2SM organization on GitHub : https://github.com/C2SM-RCM/extpar(latest release is v4.0) • Automatic testing using DWD and MCH configurations is implemented at CSCS (using Jenkins tool, all raw data available at CSCS) • Currently different versions of the code exist at DWD and at MPI  • Currently GRIB output is not working correctly (but NetCDF is ok)

More Related