04/24/23 Zängl04/24/23 Zängl

ICON

The next-generation global model for numerical weather prediction and

climate modeling of DWD and MPI-M

Current development status and the way towards preoperational testing

Günther Zängl and the ICON development team

24.04.23 Zängl24.04.23 Zängl

General overview of ICON and its current development status

Selected results of idealized mountain tests and of (a very first) real-data forecast

Further planning towards preoperational testing Summary

Outline

24.04.23 Zängl24.04.23 Zängl

ICON = ICOsahedral Nonhydrostatic model Joint development project of DWD and Max-Planck-Institute for

Meteorology for the next-generation global NWP and climate modeling system

Nonhydrostatic dynamical core for triangles and hexagons as primal grid; C-grid discretization; coupling with physics parameterizations in principle completed

Two-way nesting for triangles as primal grid, capability for multiple nests per nesting level; vertical nesting, one-way nesting mode and limited-area mode are also available

Work in progress: Thorough testing (and debugging) of physics parameterizations and physics-dynamics coupling, GRIB2 I/O, further optimization of MPI parallelization, data assimilation, ocean model on triangular grid

ICON

24.04.23 Zängl24.04.23 Zängl

Primary development goals Better conservation properties (air mass, mass of trace gases

and moisture, consistent transport of tracers; hexagonal core also conserves energy)

Grid nesting in order to replace both GME (global forecast model, mesh size 30 km) and COSMO-EU (regional model, mesh size 7 km) in the operational suite of DWD

Applicability on a wide range of scales in space and time ("seamless prediction"); therefore, the dynamical core was requested to be nonhydrostatic

Scalability and efficiency on massively parallel computer architectures with O(104+) cores

ICON

24.04.23 Zängl24.04.23 Zängl

G. Zängl Project leader DWD two-way nesting, parallelization, optimization, numerics

H. Asensio external parametersM. Baldauf NH-equation setK. Fröhlich physics parameterizationsM. Köhler physics

parameterizationsD. Liermann post processing,

preprocessing IFS2ICONF. Prill numerics, optimization,

parallelizationD. Reinert advection schemesP. Ripodas test cases, power

spectraB. Ritter physics parameterizationsA. Seifert cloud microphysicsU. Schättler software designMetBwT. Reinhardt physics parameterizations

M. Giorgetta Project leader MPI-M physics

parameterizations

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

R. Müller pre- and postprocessing

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, physics parameterizations

External: R. Johanni: MPI-Parallelization

Project teams at DWD and MPI-M

24.04.23 Zängl

The horizontal grid

24.04.23 Zängl

Primary (Delaunay, triangles) and dual grid (Voronoi, hexagons/pentagons)

The horizontal grid

24.04.23 Zängl24.04.23 Zängl

latitude-longitude windows circular windows

Grid structure with nested domains

24.04.23 Zängl24.04.23 Zängl

0)(

0)(

)(

vv

vpdnn

vpdn

tn

vt

vt

gz

czwwvwwv

tw

nc

zv

wnKvf

tv

vn,w: normal/vertical velocity component

: density

v: Virtual potential temperature

K: horizontal kinetic energy

: vertical vorticity component

: Exner function

blue: independent prognostic variables

Nonhydrostatic equation system

24.04.23 Zängl

• Two-time-level predictor-corrector time stepping scheme• 2D Lamb transformation for nonlinear momentum advection• Flux form for continuity equation and thermodynamic equation;

Miura 2nd-order upwind 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); larger time step (default 4x) for tracer advection / fast physics

• Mass conservation and tracer mass consistency

Numerical implementation

24.04.23 Zängl

• Finite-volume tracer advection scheme (Miura) with 2nd-order and 3rd-order accuracy for horizontal advection; 2nd-order MUSCL and 3rd-order PPM for vertical advection with extension to CFL values larger than 1; monotonous and positive-definite flux limiters

Special measures to improve the numerical stability over steep terrain:

• Option for truly horizontal temperature diffusion over steep slopes• Option for truly horizontal discretization of the horizontal pressure

gradient over steep slopes

Numerical implementation

• Precompute for each edge (velocity) point at level the grid layers into which the edge point would fall in the two adjacent cells

Discretization of horizontal pressure gradient (apart from conventional method)

dashed lines: main levelspink: edge (velocity) pointsblue: cell (mass) points

A

S

• Reconstruct the Exner function at the mass points using a quadratic Taylor expansion, starting from the point lying in the model layer closest to the edge point

Discretization of horizontal pressure gradient

22 )(

21)(~

cev

vpce

ccc zz

zcgzz

z

• Note: the quadratic term has been approximated using the hydrostatic equation to avoid computing a second derivative

• Treatment at slope points where the surface is intersected:

)(2 ASA

v

vpAS

zzxc

gxx

24.04.23 Zängl

• Cloud microphysics from COSMO-EU (hydci_pp)• New mass-conserving saturation adjustment scheme (U. Blahak/

A. Seifert)• Turbulence schemes from COSMO-model (M. Raschendorfer) and

ECHAM• Tiedtke-Bechtold convection scheme (from ECMWF)• COSMO cloud cover scheme; new scheme under development by

M. Köhler• Ritter-Geleyn and RRTM radiation schemes• TERRA surface scheme

Physics parameterizations

• Isolated circular Gaussian mountain, e-folding width 2000 m, various mountain heights (see subsequent slides)

• Isothermal atmosphere at rest or with u0 = 20 m/s

• Horizontal mesh size 300 m• 50 vertical levels, top at 40 km, gravity-wave damping layer starts

at 27.5 km• dry dynamical core with turbulence scheme for vertical mixing• Results are shown after 6 h of integration time

Idealized steep-mountain test

24.04.23 Zängl

vertical wind speed (m/s) potential temp. (c.i. 4 K), hor. wind speed (m/s)

mountain height: 2.25 km, isothermal atmosphere at restmaximum slope 0.95 (43.5°)

conventional pressure gradient discretization, no z-diffusion of temperature

24.04.23 Zängl

conventional pressure gradient discretization, with z-diffusion of temperature

vertical wind speed (m/s) potential temp. (c.i. 4 K), hor. wind speed (m/s)

mountain height: 2.25 km, isothermal atmosphere at restmaximum slope 0.95 (43.5°)

24.04.23 Zängl

conventional pressure gradient discretization, with z-diffusion of temperature

vertical wind speed (m/s) potential temp. (c.i. 4 K), hor. wind speed (m/s)

mountain height: 3.0 km, isothermal atmosphere at restmaximum slope 1.27 (52°)

24.04.23 Zängl

z-based pressure gradient discretization, and z-diffusion of temperature

vertical wind speed (m/s) potential temp. (c.i. 4 K), hor. wind speed (m/s)

mountain height: 3.0 km, isothermal atmosphere at restmaximum slope 1.27 (52°)

24.04.23 Zängl

vertical wind speed (m/s) potential temp. (c.i. 4 K), hor. wind speed (m/s)

mountain height: 3.75 km, isothermal atmosphere at restmaximum slope 1.59 (58°)

z-based pressure gradient discretization, and z-diffusion of temperature

24.04.23 Zängl

vertical wind speed (m/s) potential temp. (c.i. 4 K), hor. wind speed (m/s)

mountain height: 4.5 km, isothermal atmosphere at restmaximum slope 1.90 (62°)

z-based pressure gradient discretization, and z-diffusion of temperature

24.04.23 Zängl

vertical wind speed (m/s) potential temp. (c.i. 4 K), hor. wind speed (m/s)

mountain height: 3.5 km, isothermal atmosphere with u0 = 20 m/s

maximum slope 1.49 (56°)

z-based pressure gradient discretization, and z-diffusion of temperature

24.04.23 Zängl

vertical wind speed (m/s) potential temp. (c.i. 4 K), hor. wind speed (m/s)

mountain height: 4.5 km, isothermal atmosphere with u0 = 20 m/s

maximum slope 1.90 (62°)

z-based pressure gradient discretization, and z-diffusion of temperature

• Global mesh size 40 km, refinement to 20 km over Europe• Model top at 45 km, 60 levels (another expt.: 67.5 km / 90 levs)• Time step in global domain: 60s / 240s• Initialization with interpolated IFS analysis data for 01.01.2011• Total integration time: 20 days• Results shown on subsequent slides: 3-day forecast in

comparison with corresponding analysis data

Real-data test

Temperature at 3 km AGL

24.04.23 Zängl

Specific humidity at 3 km AGL

24.04.23 Zängl

Pressure at 3 km AGL

24.04.23 Zängl

24.04.23 Zängl

Further steps towards preoperational testing

• Thorough investigation of still exisiting numerical instabilities; bug fixes or improvement of algorithms

• Test suites for ICON and GME with interpolated IFS analysis data; systematic tuning of physics parameterizations and their mutual interaction (maybe also physics-dynamics coupling)

• Work on GRIB2 I/O, interpolation to lat-lon grid and pressure/ height levels, meteogram output, …

• Data assimilation

24.04.23 Zängl

Conclusions and further remarks

• We are on a good way, but there is still a lot to do…

• Efficiency and scalability: significantly better than for GME; roughly comparable to the COSMO model

• Time planning: start with preoperational testing late 2011 / early 2012; operational use (hopefully) by mid-2013