Title: ICON The next generation global model at DWD and MPI-M Current development status and selected results of idealized tests G
1ICON The next generation global model at DWD
and MPI-MCurrent development status
and selected results
of idealized testsGünther Zängl
and the ICON development team
2The ICON-Project main goals
- Centralize Know-how in the field of global
modelling at DWD and the Max-Planck-Institute
(MPI-M) in Hamburg. - Develop a non-hydrostatic global model with
static local zooming option (ICON ICOsahedral
Non-hydrostatic http//www.icon.enes.org/). - At DWD Replace global model GME and regional
model COSMO-EU by ICON with a high-resolution
window over Europe. Establish a library of
scale-adaptive physical parameterization schemes
(to be used in ICON and COSMO-DE). - At MPI-M Use ICON as dynamical core of an Earth
System Model (COSMOS) replace regional climate
model REMO. Develop an ocean model based on ICON
grid structures and operators. - DWD and MPI-M Contribute to operational seasonal
prediction in the framework of the Multi-Model
Seasonal Prediction System EURO-SIP at ECMWF).
3Requirements for next generation global models
- Applicability on a wide range of scales in space
and time? seamless prediction - (Static) mesh refinement and limited area model
(LAM) option - Scale adaptive physical parameterizations
- Conservation of mass (chemistry, convection
resolving), energy? - Scalability and efficiency on massively parallel
computer systems with more than 10,000 cores - Operators of at least 2nd order accuracy
4Current project teams at DWD and MPI-M
- D. Majewski Project leader DWD
- (till 05/2010)
- G. Zängl Project leader DWD
(since 06/2010)
static local mesh refinement,
parallelization, optimization, numerics - H. Asensio external parameters
- M. Baldauf NH-equation set
- K. Fröhlich physics parameterizations
- D. Liermann post processing, preprocessing
IFS2ICON - D. Reinert advection schemes
- P. Ripodas test cases, power spectra
- B. Ritter physics parameterizations
- M. Köhler physics parameterizations
- U. Schättler software design
- MetBw
- T. Reinhardt physics parameterizations
- M. Giorgetta Project leader MPI-M
- M. Esch software maintenance
- A. Gassmann NH-equations, numerics
- P. Korn ocean model
- L. Kornblueh software design, hpc
- L. Linardakis parallelization, grid generators
- S. Lorenz ocean model
- C. Mosley regionalization
- T. Raddatz external parameters
- F. Rauser adjoint version of the SWM
- W. Sauf Automated testing (Buildbot)
- U. Schulzweida external post processing (CDO)
- H. Wan 3D hydrostatic model version
External S. Reich, University of Potsdam Time
stepping schemes R. Johanni MPI-Parallelization
5Present development status
- Consolidated version of hydrostatic dynamical
core with option to switch between triangles and
hexagons as primal grid - Nonhydrostatic dynamical core for triangles and
hexagons basic testing and efficiency
optimization finished - Two-way nesting for triangles as primal grid,
capability for multiple nests per nesting level
one-way nesting mode and limited-area mode are
also available - Positive-definite tracer advection scheme (Miura)
with 2nd-order accuracy, 3rd-order PPM for
vertical advection 3rd-order horizontal in
testing/optimization phase - OpenMP and MPI parallelization, blocking
(selectable inner loop length) for optimal
vectorization or cache use (depending on
architecture) - Technical preparations for physics coupling
available so far, grid-scale cloud microphysics,
saturation adjustment and convection are included - Dynamical core of ocean model currently under
revision
6Horizontal grid
7Horizontal grid
Primary (Delaunay, triangles) and dual grid
(Voronoi, hexagons/pentagons)
8Static mesh refinement (two-way nesting)
latitude-longitude windows
circular windows
9Grid structure (schematic view)
Triangles are used as primal cells Mass points
are in the circumcenter Velocity is defined at
the edge midpoints
Red cells refer to refined domain Boundary
interpolation is needed from parent to child mass
points and velocity points
10Nonhydrostatic equation system (triangular
version)
vn,w normal/vertical velocity component K
horizontal kinetic energy ? vertical vorticity
component ? density ?v Virtual potential
temperature ? Exner function
11Numerical implementation
- Momentum equation Rotational form for horizontal
momentum advection (2D Lamb transformation),
advective form for vertical advection,
conservative advective form for vertical wind
equation - Flux form for continuity equation and
thermodynamic equation Miura scheme (centered
differences) for horizontal (vertical) flux
reconstruction - implicit treatment of vertically propagating
sound waves, but explicit time-integration in the
horizontal (at sound wave time step not
split-explicit) - Two-time-level Adams-Bashforth-Moulton time
stepping scheme - Mass conservation and tracer mass consistency
12Implementation of two-way nesting
- Flow sequence 1 time step in parent domain,
interpolation of lateral boundary
fields/tendencies, 2 time steps in refined
domain, feedback - Boundary interpolation of scalars (dynamical and
tracers) - RBF reconstruction of 2D gradient at cell
center - Linear extrapolation of full fields and
tendencies to child cell points - Boundary interpolation of velocity tendencies
- RBF reconstruction of 2D vector at vertices
- Use to extrapolated to child edges
lying on parent edge - Direct RBF reconstruction of velocity
tendencies at inner child edges - Weak second-order boundary diffusion for velocity
13Implementation of two-way nesting
- Feedback
- Dynamical variables bilinear interpolation of
time increments from child cells / main child
edges to parent cells / edges - Additive mass-conservation correction for
density - Tracers bilinear interpolation of full fields
from child cells to parent cells, multiplicative
mass-conservation correction - Bilinear feedback is inverse operation of
gradient-based interpolation - For numerical stability, velocity feedback
overlaps by one edge row with the interpolation
zone - Density and (virtual) potential temperature are
used for boundary interpolation / feedback,
rhotheta and Exner function are rediagnosed
14One-way nesting and other options
- One-way nesting option Feedback is turned off,
but Davies nudging is performed near the nest
boundaries (width and relaxation coefficients can
be chosen via namelist variables) - One-way and two-way nested grids can be
arbitrarily combined - An arbitrary number of nested domains per nesting
level is allowed - Multiple nested domains at the same nesting level
can be combined into a logical domain to reduce
parallelization overhead (exception one-way and
two-way nested grids have to be assigned to
different logical domains) - An option to run computationally expensive
physics parameterizations at reduced resolution
is in preparation
15- Idealized tests
- Purpose Validation of correctness of
numerical implementation, assessment of
convergence properties and numerical stability - Jablonowski-Williamson baroclinic wave test
- Modified Jablonowski-Williamson baroclinic wave
test with moisture and cloud microphysics
parameterization - Mountain-induced Rossby wave test
- Tracer advection test Convergence study for
advection of a tracer cloud in quasi-uniform flow
16Development of baroclinic waves
- Baroclinic wave case of Jablonowski-Williamson
(2008) test suite - Nonhydrostatic dynamical core
- Basic state geostrophically and hydrostatically
balanced flow with very strong baroclinicity
small initial perturbation in wind field - Disturbance evolves very slowly during the first
6 days, explosive cyclogenesis starts around day
8 - Grid resolutions 140 km and 70 km, 35 vertical
levels - Results are shown after 10 days
location of nest
17Vertical velocity at 1.8 km AGL on day 10
70 km
140 km
140 km, nested
18Baroclinic wave test with moisture
- Modified baroclinic wave case of
Jablonowski-Williamson (2008) test suite with
moisture and Seifert-Beheng (2001) cloud
microphysics parameterization (one-moment
version QC, QI, QR, QS) - Initial moisture field RH70 below 700 hPa, 60
between 500 and 700 hPa, 25 above 500 hPa QV
max. 17.5 g/kg to limit convective instability in
tropics - Transport schemes for moisture variables
- Horizontal Miura 2nd order with flux limiter
- Vertical 3rd-order PPM with slope limiter
- Grid resolutions 70 km and 35 km, 35 vertical
levels - Results are shown after 14 days
19Temperature at lowest model level on day 14
70 km
35 km
70 km, nested
nest, 35 km
20QV (g/kg) at 1.8 km AGL on day 14
70 km
35 km
70 km, nested
nest, 35 km
21Accumulated precipitation (mm WE) after 14 days
70 km
35 km
70 km, nested
nest, 35 km
22Rossby wave generation by a large-scale mountain
- Mountain-induced Rossby-wave case of
Jablonowski-Williamson test suite - Nonhydrostatic dynamical core
- Basic state isothermal atmosphere, zonal flow
with max. 20 m/s - Standard setup with 2000-m high circular mountain
at 30N/90E - High-resolution runs 35 km mesh size 35
levels - Coarse-resolution runs 140 km
- Nested runs 140 km globally, double nesting to
35 km over mountain - Results are shown after 20 days
23Vorticity (1/s) at surface level on day 20
high-resolution (35 km)
nested (140-km domain)
coarse-resolution (140 km)
24Vorticity at surface level on day 20 (mountain
region)
high-resolution
nested (35-km domain)
coarse-resolution
25Horizontal wind at surface level (barbs),
vertical wind at 2.5 km AGL on day 20
(colours)
high-resolution
nested (35-km domain)
coarse-resolution
26Solid body rotation test case
- Uniform flow along northeast direction
- Initial scalar field is a cosine bell centered at
the equator - After 12 days of model integration, cosine bell
reaches its initial position - Analytic solution at every time step initial
condition
Error norms (l1, l2, l8) are calculated after one
complete revolution for different resolutions
27Linear vs. quadratic reconstruction
quadratic
L1 0.53764E-01 L2 0.45283E-01
- 140 km res.
- c0.2
- flux limiter
linear
L1 0.61346E-01 L2 0.51185E-01
28Convergence rates
Quadratic reconstruction
Linear reconstruction
29Summary
- The nonhydrostatic dynamical core has been
thoroughly tested and compares well with
reference solutions from a spectral model it
appears to have good stability properties over
steep topography - Two-way nesting also works numerically stable
over long times (tested for integration times up
to 100 days) and exhibits only very small
disturbances along the nest boundaries - State-of-the-art transport schemes have been
implemented for tracer advection further
investigations will be made to determine the
optimal compromise between accuracy and
computational cost - Now the focus will be directed towards completing
physics coupling, incorporating real external
parameter data, I/O parallelization, using real
NWP analysis data as input, data assimilation,
30- Thank you for your attention!