300 likes | 479 Views
HARMONIE a common effort of HIRLAM, ALADIN and LACE on high resolution modelling. Jeanette Onvlee COSMO General Meeting Cracow, 20080916. Outline. Background and scope of the HARMONIE cooperation HARMONIE mesoscale modelling developments; status and plans
E N D
HARMONIEa common effort of HIRLAM, ALADIN and LACE onhigh resolution modelling Jeanette Onvlee COSMO General Meeting Cracow, 20080916
Outline • Background and scope of the HARMONIE cooperation • HARMONIE mesoscale modelling developments; status and plans • Ensemble forecasting activities and plans • Other developments
History • Background: Wish to join forces to meet the challenges of mesoscale modelling • Common ambitions • Play to each partner’s strengths • Good experiences in previous research cooperation • Goals of cooperation: full code cooperation on common mesoscale model in IFS/Arpege code framework, joint development of short-range ensemble forecasting system • A new name: • HARMONIE: HIRLAM-ALADIN Research on Mesoscale Operational NWP In Euromed • Common activities: • joint scientific/strategic planning • setup of ALADIN code-based systems in HIRLAM institutes • joint meetings and working/training weeks • from “division of labour” increasingly to interacting, mixed research teams • ECMWF increasingly also involved (common stakes in IFS system, research cooperation in physics, dynamics and EPS)
Present status of HARMONIE mesoscale • Data assimilation: • 3D-VAR/FGAT • use of obs: basically everything which ECMWF assimilates is available; in practice: SYNOP/SHIP/BUOY, TEMP, AMDAR, AMSU-A/B, GEO/AMVs, wind profiler, radar winds • Var-QC • Forecast model • dynamics: NH-ALADIN (spectral, SLSI) • Upper air model physics: three flavours • AROME/ EDMF, explicit deep convection, ECMWF radiation • ALARO/ Bougeault/TKE, 3MT, ALADIN radiation • HIRLAM/ CBR, KFB, RK/CAM3, Saavijarvi radiation (baseline only) • Surface model/DA: • Externalized scheme SURFEX: ISBA soil, canopy/forest, snow, urban, lake • OI assimilation scheme • INCA analysis system for nowcasting, validation • Monitoring, validation and verification tools
The Arome-France model config Arome-France current domain Aladin-France domain
AROME’s resolved convection : a deep change for bench forecasters’ expertise job Arome Radar observation Aladin The ‘application side’ of the ‘double penalty’ syndrome for verification: details of AROME bring good information about the structure of the field but they might be more misleading about the timing-position than their ALADIN counterparts at the latter’s scale
Some other HARMONIE suites... NRT suites with AROME/ALARO by almost all HIRLAM partners and some more (Hungary...) Operational suites with ALARO/3MT in several LACE countries In total ~ 13 systems, typically 2.5km, 40 levels
Data assimilation: algorithms • 3D-VAR: • Different ways of blending in info from larger scale 4D-VAR • LACE: implementation of 3D-VAR operationally in all LACE countries • Experimentation with rapid update cycling • 4D-VAR in preparation; • Flow dependency by wavelets, ensemble assimilation; • Take in concepts from HIRLAM 4D-VAR experience: Jdfi, Jk, multiple outer loops, ... • Experimentation with hybrid variational / ensemble assimilation techniques - ETKF, Lorenc • Surface: replace OI with EKF scheme • Initialization with scale-selective DFI
Data assimilation: use of observations • Use of observations: Increase range of observations to be assimilated: • HIRLAM: Comprehensive observation impact studies (CIS) • Now on synoptic scale (AMSU-A/B over sea/land/ice, GEO and MODIS AMV, scatterometer over Atlantic/Arctic, initial results positive; starting on convection exps with SEVIRI, radar winds, GPS) • Transfer to mesoscale, focus on summer convection and radar • High-resolution sat data: e.g. SEVIRI, IASI, ADM; cloud– and land-contaminated data, varBC, tuning of obs error stats and impact studies • Radar winds/precip, GPS, BUFR TEMP • Radar processing inhomogeneity, data exchange, beam blocking critical issues • Surface: screen level parameters, (scatterometer) soil moisture, SST/sea ice, lake, snow, snow on ice
Assimilation of radar winds: Neutral to slightly positive scores in AROME-France: RADAR CNTRL Psurf Precip DD FF Hu T R0
Forecast model: Dynamics • NH core validation and comparison with hydrostatic: LACE, ECMWF • Vertical Finite Element formulation • Variable map factor for use of SISL over large domains • Improved quality of LBC • Nesting experiments with various configurations • More mass-conserving (less diffusive) SL interpolators (also for coupling to chemistry)
Forecast model: upper air physics • AROME: • Highest priority: solve presently remaining problems: overestimate of severe convection, extreme precipitation, negative wind bias over steep orography, minimum temp under very stable conditions • Physics-dynamics interactions • Impact of additional deep convection parametrization (either within EDMF or separate) • 3D-turbulence scheme • Tuning of microphysics • Representation of orographic roughness • Stable boundary layer modifications to turbulence scheme • Surface: compare AROME/HIRLAM snow/forest schemes, add snow on ice parametrization • ALARO/3MT: • Now operational in several LACE institutes • Comparison with AROME on 5km scale ongoing in HIRLAM • Convergence between various physics options? Different coding strategies make common interfacing at low level in source code difficult.
AROME performance : low-level scores • Objective scores of AROME-France using French automatic surface obs network (hourly data every ~30km) • Beats ALADIN-France in most respects 10m windspeed Scores over France for 5-18 February 2008 (Arome in pink Aladin in blue) forecast range (h) 2m Temperature 2m Humidity 2nd AROME training course, Lisbon, March 2008 forecast range (h) forecast range (h)
Δx=9.0 km (2x) Δx=4.5 km (2x) Δx=2.3 km (3x) A0 with 3MT => A0 without 3MT => Observed precipitations => ‘Resolved’ convection => 3MT’s sampling of the ‘grey-zone’ (ALARO-0)
Forecast model: other aspects • Numerical efficiency and portability: profiling studies • Diagnostics, validation and verification: • Exchange started, but should be intensified • HIRLAM: routine monitoring/verification system, increasingly enhance with appropriate mesoscale diagnostics tools (incl from COSMO) • ALADIN/LACE: Catching up with the state of the art • Exchange with other consortia within SRNWP/ Verification project Programme. • System aspects: • HIRLAM/LACE: Explore common script system? • SRNWP/Interoperability programme
Ensemble forecasting activities Existing real-time LAM EPS systems within HIRLAM/ALADIN/LACE: • LAEF: • Downscaling of Arpege, plus breeding, ETKF, physics perturbations; Eur area + part Atlantic, +72h, 20km res, 16 members • NORLAMEPS: • downscaling of ECMWF TEPS, N.Atl/Eur area, +72h, 12km resolution, 21 members • SREPS • 5 regional models nested in 4 global models, with SLAF; NAtl/Eur area, +72h, 27km resolution, 80 members Under construction: GLAMEPS (others to be replaced by/integrated into this??) • Downscaling of EUROTEPS, plus HIRLAM/ALADIN SV’s, ETKF, perturbations in physics and surface, ... • NAtl/Eur area, target resolution 10km, 40 layers, 50-60 members
Ensemble forecasting activities LAEF: • New breeding/blending/multi-physics system offers improvements over pure downscaling version => to be implemented operationally soon • 1st + 2d moment calibration of pdf promising • Optimize operational setup through e.g. introduction of SMS • Target resolution: 10km • Evaluate outcome of Beijing 2008
Verification for a two-month period: June-August 2007: Clear improvement of extended LAEF system Impact of bias correction and 2d moment calibration
GLAMEPS Joint multi-model EPS system for HIRLAM & ALADIN. Present status: • GLAMEPS prototype setup (version 0) implemented at ECMWF. • HIRLAM components: EUROTEPS, physics perturbations, HIRLAM (forcing) SVs, ETKF, surface perturbations • ALADIN components: breeding, blending, multi-physics, ETKF, ALADIN SV • Calibration: BMA for 1st / 2d moment of pdf, Gaussian and non-Gaussian parameters • Ongoing: configuration tests for prototype distributed GLAMEPS system (v1: March 2009). Continue parallel experimentation in laboratory • Continue pursuit of Eur. LAMEPS coopera- tion in TIGGE-LAM
GLAMEPS Model domain (for HIRLAM components) and EUROTEPS targeting areas for Northern, Middle and Southern Europe
Other (HIRLAM) developments (1)Coupling with atmospheric chemistry Activities: • Make HIRLAM output better suited as input to ACT models (postprocessing, vert resolution in BL, …) • Dynamics: better mass-conserving properties: • more accurate interpolators • mass conserving SL scheme of Kaas et al. • ENVIRO-HIRLAM coupled system being installed as HIRLAM chemistry branch at DMI To be done: • include aerosols in microphysics • include desired physics options where necessary (e.g. radiation scheme) • Start study of chemistry-cloud feedbacks
-GEMS/TNO -EMEP CAC-Aerosol Dynamics Modal model Log-normal modes: nuclei, accumulation, coarse Moment equations: Intra-modal coagulation, Intra-modal coagulation, condensation WRF-CHEM Gas-phase chemistry: RADM, RACM, CBMZ Aerosol dynamics: MOSAIC, SORGAM Photolysis: Madronich Cloud chemistry Convection Deposition Plumerise Current version of DMI-ENVIRO-HIRLAM modelling systems, showing the components of a forecast Min. number of advected quantities: RADM 2 gas-phase chemistry, no cloud chemistry, CAC aerosol dynamics ~ 50 Max. number of advected quatities: Carbon bond IV gas-phase chemistry, cloud chemistry, 8-bin Mosaic aerosol dynamics ~ 250
Other developments • HIRLAM: Coupling with ocean model • HIRLAM/ALADIN: Simplified model version for academia • HIRLAM/ALADIN: Models as tools for regional climate modelling: • Update of HIRLAM as “climate branch” • Push use of ALADIN as tool for regional projections • Definition and setup of climate branch for HARMONIE
RCR 00 + 6h AROME 00 + 6h Radar SMHI H22 conv. prec SMHI H22 strat.prec.
Downscaling EPS with HIRLAMRKKF - cloud scheme: verif. at 2007/08/22 12utc +42-48 +18-24
Aladin Rapid Update Cycle (Ald-Hun) RMSE difference Expected extra from 3-hourly cycling: - more SYNOPs (Ps only) - more AMDAR - more AMV - more Wind Profiler - smaller error in the innovation vector for ATOVS (due to more frequent analysis) Preliminary results: - improvement for all fields (see figures on the left where red shades indicate that 3h cycling is better than 6h cycling) Next: - diagnose spin-up in the 3h backgroundforecast U RHU