Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
Zbigniew Piotrowski1), Bogdan Rosa, Damian Wójcik,
Michał Ziemiański
Institute of Meteorology and Water Management – National Research Institute
Towards soundproof dynamical core for COSMO model – high resolution weather prediction for the Alps
COSMO-EULAG development at the Polish hydrological and meteorological service (IMGW-PIB)
Activities started in 2009 within the „Conservative Dynamical Core”priority project of COSMO consortium(http://www.cosmo-model.org/content/tasks/pastProjects/cdc/default.htm)
Goal of the project was ``to get a dynamical core with some explicit conservation properties, the most important one being mass conservation. This should be achieved with no reduction of accuracy or efficiency. A further goal (probably connected with the former) is to improve the capability of the model to work in steep terrain.''
Within the project, IMGW with the help of MeteoSwiss and DWD took the task „... to demonstrate that the EULAG itself, and the basic ideas defining the EULAG construction, can be used as a robust and credible base for construction of future operational weather models, and especially the COSMO model.”
COSMO-EULAG operationalization (CELO) – the new priority project of COSMO for 2012-2016
After a successful accomplishment of a series of idealized and semi-realistic Alpine flow tests performed with EULAG model, see topical issue of Acta Geophysica „Modeling Atmospheric Circulations with Sound-Proof Equations”, COSMO consortium approved new priority project CELO with aim „To deliver fully operational weather prediction package” with anelastic dynamical core, without data assimilation in the scope of this project.
Current efforts concentrate on: - full integration of EULAG dynamical core (DC) with COSMO framework - consolidation and optimization of the DC forumulation, in particular proper link to parameterizations and the soil model - testcases focusing on examination of a series of meteorological scenarios with 2.2 km, 1.1 km, 0.55 km and 0.27 km horizontal resolution
Two fundamental algorithms: MPDATA advection + GCRK pressure solverTwo optional modes for integrating fluid PDEs:• Eulerian - control-volume and Lagrangian - trajectory wise integral
Optional fluid equations (nonhydrostatic):• Anelastic• Compressible/incompressible Boussinesq• Incompressible Euler/Navier-Stokes’• Fully compressible, explicit or (prototype) semi-implicit• Anelastic MHD (unique ability to reproduce MHD solar cycles !)• Anelastic for fully unstructured grid formulation
Available strategies for simulating turbulent dynamics:• Direct numerical simulation (DNS)• Large-eddy simulation, explicit and implicit (LES, ILES)Portfolio of applications from microturbulence to orographic flows (e.g Rotunno et al, 1999), evolution of sand dunes (Ortiz et al. 2009) to stellar.
EULAG model for multiscale flows
Numerical design
⇒ system implicit with respect to all dependent variables.
On grids co-located with respect to all prognostic variables, it can be inverted algebraically to produce an elliptic equation for pressure
solenoidal velocity contravariant velocity
subject to the integrability conditionBoundary conditions on
Boundary value problem is solved using nonsymmetric Krylov subspace solver - a preconditioned generalized conjugate residual GCR(k) algorithm
(Smolarkiewicz and Margolin, 1994; Smolarkiewicz et al., 2004)
Imposed on
All principal forcings are assumed to be unknown at n+1
Smolarkiewicz et al., J. Comput. Phys. 227 (2007), 'Building resolving large-eddy simulations and comparison with wind tunnel experiments'
Current state of COSMO-EULAG dynamical core
We run prototype COSMO with anelastic dynamical core on MeteoSwiss 2.2 km/1.1 km operational domain with limited set of parameterizations of COSMO (currently no tuning of parametrizations to EULAG)
We compare against Runge-Kutta dynamical core with same range of parameterizations and with the results of operational MeteoSwiss forecast
We achieve good results given the stage of the project
COSMO – EULAG is stable and robust, the results are rich in detail due to small effective diffusion (stability of the model does not depend on any numerical filtering or explicit diffusion)
ESTS
Case study: role of shallow convection
parameterization
Computational domain :• (WE x NS) = (520 x 350) / (1014 x 678) / (806 x 806) (2.2km ) / (1.1km ) / (0.55km) • standard operational COSMO-2 model of Meteo-Swiss vertical distribution of levels (61),
Gal-Chen coordinate system with top at 23588.50m, • Lateral absorber width = 45km / 23km / 11.0km• Top sponge base height = 15km
Model setup
1.1km2.2km
0.55km
10th SRNWP Workshop, Offenbach/Main, 14 May 2013
Dynamics :• Saturation adjustment is on• Numerical diffusion turned off „on” for Runge-Kutta• Semi-Lagrangian advection of moist quantities
Microphysics :• Standard COSMO microphysics parameterization including ice, rain, snow and graupel
precipitation
Radiation :• Calculated every 15 minutes (2.2km) or 6 minutes (1.1km and 0.55km) of simulation Other parameterizations :• Vertical turbulence model with surface layer fluxes• Standard COSMO surface model
Orography:• Operational Meteo-Swiss for 2.2 km, recalculated from SRTM for 1.1 and 0.55 km• standard Cosmo orography filtering was applied
Initial and boundary data for all simulations are interpolated from COSMO – 7 forecast of MeteoSwiss.
Model setup
10th SRNWP Workshop, Offenbach/Main, 14 May 2013
Synoptic map for 27.07.2012 Synoptic map for 28.07.2012
Test case
• Surface weak and shallow low-pressure system• Some (rather weak) upper-air forcing
10:00 UTC 14:00 UTC
16:00 UTC 18:00 UTC
Test case : MSG High Resolution Visible (HRV) channel
ESTS
Spatial structure of convectionComparision of Runge-Kutta and
EULAG dynamical cores with and without shallow convection
parameterization
Cloud radiance (high res.) 14:00 UTC
CE 2.2km SCP on CE 2.2km SCP off
MSG infrared band
RK eq. 2.2km SCP on RK eq. 2.2km SCP off
Cloud radiance (high res.) 14:00 UTC
MSG infrared band
CE 1.1km SCP on CE 1.1km SCP off
CE 0.5km SCP off
CE 0.5km SCP on
Summary : convective cloud pattern
●Qualitatively, while the SCP have minor impact on RK 2.2 km spatial pattern of convective clouds, it significantly influences the pattern of the CE convective clouds, for all resolutions tested (from 2.2 to 0.55 km)
●In CE, the forecasts without SCP result in less extensive spread of convective clusters which is also more in agreement with observations, for the analyzed case
●In CE with SCP, the spatial pattern of convective clouds is similar across tested resolutions, during the day
●In CE without SCP, the spatial pattern of convective clouds is similar across resolutions for early stages of the convection, for later stages the pattern differs between resolutions
Conclusions and remarks
EULAG dynamical core was tested within COSMO framework for stable and convective NWP scenarios over the Alps for 2.2 km, 1.1 km and 0.55 km resolutions, with good results,
Current coupling is simplistic (first order contributions from physics and diffusion of COSMO parameterizations and diffusion operator)
Full integration with COSMO framework in progress, systematic quantitative assessment of results is expected
'PantaRhei' project at ECMWF investigates the possibilitiesof coupling or merging EULAG numerics with IFS for futurenon-hydrostatic weather prediction
Selected challenges of merging COSMO and EULAG
Standarizing treatment of boundary conditions
Merging EULAG A-grid approach with default COSMO C-grid
Making good use of the lowest model level at the surface (0 m) in the EULAG core
Changing focus from multi to single (or limited) application – code structure, performance, etc.
Accurately coupling to COSMO physics packages
Allowing for the accurate formulation of generalized vertical coordinate
• Multidimensional positive definite advection transport algorithm (MPDATA).
Starts with iteration of upwind scheme, then applies nonlinear corrective iterations of upwind with negative diffusion
1. MOTIVATION Computation of the inverse metric coefficients
Equations are derived based on the assumption that 16 differential identities:
are satisfied. For the transformation at hand:
we are left with 10 nontrivial identities:
computational space(ordered like Cartesian grid)
x,y,z – physical coordinates, here in lat-lon-z for EULAG DC
1. MOTIVATION Metric term formulations – impact on cloud area fraction
Standard formulation of EULAG DC with hard-coded Gal-Chen coordinate transform.
New metric terms definitionindependent of Gal-Chen coordinate transform,
Default COSMO metric term formulation