480 likes | 607 Views
HIRLAM-6, development since last time. Strategy - ALADIN - MF - collaboration Data assimilation, 3D/4D-VAR, surface Observation Usage Parameterisation – turbulence and convection Surface and radiation Physics coupling - boundary conditions Meso-scale modelling EPS
E N D
HIRLAM-6, development since last time • Strategy - ALADIN - MF - collaboration • Data assimilation, 3D/4D-VAR, surface • Observation Usage • Parameterisation – • turbulence and convection • Surface and radiation • Physics coupling - boundary conditions • Meso-scale modelling • EPS • Regular Cycle with the Reference (FMI)
HIRLAM-6 Memorandum of Understanding • Targets • achieve highest possible accuracy for severe weather and of wind, precipitation and temperature • develop 3D/4D-VAR further and its use of non-conventional data • maintain the regular analysis/forecasting cycle • continue development of synoptic model 10-20 km • develop meso-scale non-hydrostatic operational model with suitable physical parameterisation • Overhaul of complete System • develop methods for probabilistic forecasting • continue development of verification methods
HIRLAM strategy - synoptic • Synoptic model, 10-20 km, every 6 hours -> 2 (3) days, 4D-VAR and satellite data over a (fairly) large area • provides comprehensive set of forecast parameters for applications and driving other models • boundary conditions and tight coupling to meso-scale model • covers window between ECMWF forecasts - more recent observations and boundaries (frames)
HIRLAM strategy - meso-scale • Meso-scale data assimilation and model , 2-3 km non-hydrostatic model +3-12 (24 h) • physics for 2km, explicit convection • turbulence and radiation non-local (later, ~ 1 km ) • rapid update cycle, vast amount of regional data available, conv/non-conv, reflectivity, precipitation .. • 4D-VAR /3D-VAR FGAT - if in short time - spinup? • Boundary field impact, transparent boundary conditions !
HIRLAM research profile • Physics interfaces - combinations • HIRLAM physics / AROME physics • Synoptic physics HIRLAM/ALARO • Synoptic 4D-VAR - migrate to ALARO • Meso-scale 4D-VAR • Meso-scale basis functions - Jb - • Observations - radar winds, surface, refl. Cloud, • Large scale coupling - spectral - extension zone • Meso-scale validation • Probabilities with EPS and physical perturbations • Surface modelling and assimilation (SST)
HIRLAM meso-scale group • Learning - set up of ALADIN - climate - coupling • DMI-SMHI-FMI-INM - • Set up of domain(s) • Physics interface - temporary - general HIRLAM and AROME • First experiments • Coupling with HIRLAM outer model
Data assimilation -3D-VAR • 3D-VAR background constraint Jb : • (xb - H(y))TB-1 (xb - H(y)) , sigma-b, horizontal variation, new structure functions • => Background check, analysis increments • Analytical balance (enh) ->statistical balance
3D-VAR (cont) • FGAT - First Guess at Appropriate Time
4D-VAR Data Assimilation • Adjoints of semi-Lagrangian spectral model • Multi-incremental minimisation - low resolution • Optimisations of transforms • > significant gain in economy, feasible for operations
4D-VAR single obs 3 Dec 99 06-12 3 Dec 06 ->3 Dec 12 3 Dec 06
4D-VAR argument • Optimal solution in time including all information • Iterativ method enabels non-linear operators - • possible in 3D too, but : • Non-linear analysis can transfer a vortex • The model analyses non-observed quantaties • Possible to use integrated observations • Enables high time resolution of data and time sequence can be utilised - e.g. radar • Model generated structure functions • necessary for meso-scale
4D-VAR Estimated cost of SL incremental 4D-VAR Estimated computer requirements of SL incremental 4D-VAR
4D-VAR activity now • Jc DFI - control of noise - NNMI in iterations • Optimisation • Multi-incremental and real trials • 120 - 45 km minimisation, 22 - 17 km fcs • about 1 hour for very large area
Analysis of surface parameters • OI SST and Ice analysis • Ocean Sea Ice SAF data - • New OI snow analysis ready for implementation • QC and bias correction (due to height differences) • Tuning of 2m T och RH analysis (statistics) Old New
New Snow analysis • SSM/I will help – LAND SAF data -
Observation Usage • Conventional data • radiosonde launch times • radiosonde drift • comparing observation availability • Remote sensing data • AMSU-A • AMSU-B • QuikScat • Radar doppler winds • GPS ZTD • WINDPROFILER
Reference case GPS included Radar 20020712_06 (analysis time)
HIRLAM EWP feasibility study
Forecast Model - parameterisation • Turbulence (CBR TKE-l) • Much attention to stable case - more mixing at high stability - modified - cut - smooth Ri >1 • Increased roughness - vegetational - orografical • Direction of surface stress vector • => filling of lows, reduce 10 m wind • Moist conservative and moist stability version • effect of condensation on stability
Turning of stress and smooth mixing (Tijm, 2004)
Snow scheme in ISBA main modifications to original code: • Only new snow scheme on fractions 3 and 4 and now 5 • Force-restore formulation replaced by heat conduction • Heat capacity of uppermost layer replaced by 1 cm • moist soil. • A second soil layer (7.2 cm) • Forest area decreased so that at least 10% of area • is low-vegetation • At present (temporarily!) no soil freezing • Forest tile, being developed - canopy snow and ground
ISBA: snow covering parts of fractions 3 and 4 snow in beginning of timestep Snow change • Features of the snow scheme: • move the snow from fractions 3 and 4 to fraction 6 every timestep • one layer of the snow, with a thermally active layer < 15 cm • water in the snow, which can refreeze • varying albedo and density • mirroring of temperature profile in the ground to assure correct memory Thermally active layer Ts snow T snow Ts 3 and 4 Ts2 3 and 4 Td 3 and 4 Ts2 snow Tdsnow mixing of T in soil between timesteps Tclim
Soil moisture adapts in assimilation to different vegetation types
Radiation and snow cover • Soil Freezing - implemented • esat for ground <0 for ice implemented • esat over water and ice following K-I Ivarsson • distribution water - ice in clouds to be consistent - large effect on emissivity - implemented • radiation for sloping ground calculated - for HR
Convection - condensation • Kain-Fritsch Rash-Kristjanson • extensive tests and verification at 22 km • better humidity • 11 km indicates better results • Expensive, and very much so, on vector systems • Possible vectorised version
Model dynamics and embedding • Coupling between SL advection and physics • Semi-Lagrangian mods for orography (T eq.) • Boundary relaxation (Host orography, interp.) • Development of transparent boundary conditions • Incremental Digital Filter Initialisisation • Ensemble forecasts with HIRLAM • Verification methods - meso-scale - Workshop • Climate system developments • System - upgrades - Reference test - RCR • Communication - HeXNeT - RCR monitoring
Transparent LBC progress • 2D-shallow water model - several results • 3D-simplest 2 layer baroclinic • 3D-multilevel Z - • eigenvalues - Laplace transform • demonstrated • 3D-mulitlevel eta - to be done • Spectral LAM - extension zone - programming ?
New HR rotated climate data sets 0.025 0.0125
Conclusions • Systematic near surface errors adressed and worked on • turbulence, surface scheme, radiation-clouds • New orientation towards Meso-scale • Collaboration with ALADIN • 4D-VAR for synoptic scales • More remote sensing • Lateral Boundary conditions developing - necessary • Monitoring and quality of Reference system
DMI Jan/ Feb 2003 Bias corrected
SMHI HIRLAM - 11 km -> HR-FAR
SMHI HIRLAM - Dec -> HR-FAR