ICON. The Icosahedral Nonhydrostatic model: Formulation of the dynamical core and physics-dynamics coupling

Transcription

1 ICON The Icosahedral Nonhydrostatic model: Formulation of the dynamical core and physics-dynamics coupling Günther Zängl and the ICON deelopment team PDEs on the sphere 2012

2 Outline Introduction: Main goals of the ICON project The dynamical core and physics-dynamics coupling Selected results: from dycore tests to NWP applications Summary and conclusions

3 ICON ICON = ICOsahedral Nonhydrostatic model Joint deelopment project of DWD and Max-Planck-Institute for Meteorology for the next-generation global NWP and climate modeling system Nonhydrostatic dynamical core on an icosahedral-triangular C-grid; coupled with (almost) full set of physics parameterizations Two-way nesting with capability for multiple nests per nesting leel; ertical nesting, one-way nesting mode and limited-area mode are also aailable

4 ICON Primary deelopment goals Better conseration properties (air mass, mass of trace gases and moisture, consistent transport of tracers) Grid nesting in order to replace both GME (global forecast model, mesh size 20 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 down to mesh sizes that require a nonhydrostatic dynamical core Scalability and efficiency on massiely parallel computer architectures with O(10 4 +) cores At MPI-M: Deelop an ocean model based on ICON grid structures and operators; Use limited-area mode of ICON to replace regional climate model REMO. Later in this decade: participate in the seasonal prediction project EURO-SIP

5 0 ) ( 0 ) ( ) ( pd n n pd n t n t t g z c z w w w w t w n c z w n K f t n,w: normal/ertical elocity component : density : Virtual potential temperature K: horizontal kinetic energy : ertical orticity component : Exner function blue: independent prognostic ariables Nonhydrostatic equation system (dry adiabatic)

6 Numerical implementation (dynamical core) Two-time-leel predictor-corrector time stepping scheme; for efficiency reasons, not all terms are ealuated in both sub-steps For thermodynamic ariables: Miura 2 nd -order upwind scheme (centered differences) for horizontal (ertical) flux reconstruction; 5-point aeraged elocity to achiee (nearly) second-order accuracy for diergence implicit treatment of ertically propagating sound waes, but explicit time-integration in the horizontal (at sound wae time step; not split-explicit); larger time step (usually 4x or 5x) for tracer adection / fast physics For numerical conenience, the thermodynamic equation is reformulated to an equation for Exner pressure Numerical filter: fourth-order diergence damping

7 Numerical implementation (tracer adection) Finite-olume tracer adection scheme (Miura) with 2 nd -order and 3 rd -order accuracy for horizontal adection; extension for CFL alues slightly larger than 1 aailable 2 nd -order MUSCL and 3 rd -order PPM for ertical adection with extension to CFL alues much larger than 1 (partial-flux method) Monotonous and positie-definite flux limiters Option to turn off adection of cloud and precipitation ariables (and moisture physics) in the stratosphere Option for (QV) substepping in the stratosphere

8 Numerical implementation (physics-dynamics coupling) Fast-physics processes: incremental update in the sequence: saturation adjustment, turbulence, cloud microphysics, saturation adjustment, surface coupling Slow-physics processes (conection, cloud coer diagnosis, radiation, orographic blocking, sub-grid-scale graity waes): tendencies are added to the right-hand side of the elocity and Exner pressure equation Diabatic heating rates related to phase changes and radiation are consistently treated at constant olume

9 Special discretization of horizontal pressure gradient (apart from conentional method; Zängl 2012, MWR) Precompute for each edge (elocity) point at leel the grid layers into which the edge point would fall in the two adjacent cells A S dashed lines: main leels pink: edge (elocity) points blue: cell (mass) points

10 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 2 2 ) ( 2 1 ) ( ~ c e p c e c c c z z z c g z z z Note: the quadratic term has been approximated using the hydrostatic equation to aoid computing a second deriatie Treatment at slope points where the surface is intersected: ) ( 2 A S A p A S z z x c g x x

11 Selected experiments and results Highly idealized tests with an isolated steep mountain, mesh size 300 m: atmosphere-at-rest and generation of nonhydrostatic graity waes Jablonowski-Williamson baroclinic wae test with/without grid nesting DCMIP tropical cyclone test with/without grid nesting Real-case tests with interpolated IFS analysis data

12 atmosphere-at-rest test, isothermal atmosphere, results at t = 6h ertical wind speed (m/s), potential temperature (contour interal 4 K) circular Gaussian mountain, e-folding width 2 km, height: 3.0 km (left), 7.0 km (right) maximum slope: 1.27 (52 ) / 2.97 (71 )

13 ambient wind speed 10 m/s, isothermal atmosphere, results at t = 6h ertical (left) / horizontal (right) wind speed (m/s), potential temperature (contour interal 4 K) circular Gaussian mountain, e-folding width 2 km, height: 7.0 km maximum slope: 2.97 (71 )

14 ambient wind speed 25 m/s, isothermal atmosphere, results at t = 6h ertical (left) / horizontal (right) wind speed (m/s), potential temperature (contour interal 4 K) circular Gaussian mountain, e-folding width 2 km, height: 7.0 km maximum slope: 2.97 (71 )

15 ambient wind speed 7.5 m/s, multi-layer atmosphere, results at t = 6h ertical (left) / horizontal (right) wind speed (m/s), potential temperature (contour interal 2 K) 3D Schär mountain, height: 4.0 km, peak-to-peak distance 4.0 km maximum slope: 2.73 (70 )

16 Jablonowski-Wiliamson test, surface pressure (Pa) after 10 days 160 km 80 km 40 km

17 Jablonowski-Wiliamson test, ertical wind at 1.5 km AGL (m/s) after 10 days 160 km 80 km 40 km

18 Jablonowski-Wiliamson test, surface pressure (Pa) after 10 days 160 km 80 km 160/80 km, two-way nesting

19 DCMIP tropical cyclone test with NWP physics schemes, eolution oer 12 days Absolute horizontal wind speed (m/s) Left: single domain, 56 km; right: two-way nesting, 56 km / 28 km

20 Real-case forecasts initialized with interpolated IFS analyses: mean sea-leel pressure 2 Jan 2012, 00 UTC h 18 Jun 2012, 00 UTC h

21 Real-case forecasts initialized with interpolated IFS analyses: temperature at 10 m AGL 2 Jan 2012, 00 UTC h 18 Jun 2012, 00 UTC h

22 Real-case forecasts initialized with interpolated IFS analyses: 7-day accumulated precipitation 2 Jan 2012, 00 UTC h 18 Jun 2012, 00 UTC h

23 WMO standard erification against IFS analysis: 500 hpa geopotential, NH blue: GME 40 km with IFS analysis, red: ICON 40 km with IFS analysis

24 WMO standard erification against IFS analysis: 500 hpa geopotential, SH blue: GME 40 km with IFS analysis, red: ICON 40 km with IFS analysis

25 Summary and conclusions The dynamical core of ICON combines efficiency, high numerical stability and improed conseration properties The two-way nesting induces ery weak disturbances, supports ertical nesting and a limited-area mode Forecast quality with full physics coupling is comparable with the operational GME een though systematic testing and tuning is only in its initial phase Next major step: coupling with data assimilation Visit also the ICON posters by Reinert et al. and Ripodas et al.