1 / 23

Michael Baldauf Deutscher Wetterdienst, Offenbach, Germany

Michael Baldauf Deutscher Wetterdienst, Offenbach, Germany. Numerical contributions to the Priority Project ‘Runge-Kutta’ COSMO General Meeting, Working Group 2 (Numerics) Bukarest, 18.09.2006. Outline Stability analysis of the p‘T‘-dynamics (PP ‚Runge-Kutta, Task 10)

Download Presentation

Michael Baldauf Deutscher Wetterdienst, Offenbach, Germany

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Michael Baldauf Deutscher Wetterdienst, Offenbach, Germany Numerical contributions to the Priority Project ‘Runge-Kutta’COSMO General Meeting, Working Group 2 (Numerics)Bukarest, 18.09.2006

  2. Outline • Stability analysis of the p‘T‘-dynamics (PP ‚Runge-Kutta, Task 10) • Vertical advection of 3. order (PP ‚Runge-Kutta‘, Task 8) • New stability criteria for the small and the big timestep • Tool for conservation properties of LM (PP ‚Runge-Kutta‘, Task 3)

  3. PP ‚Runge-Kutta‘, Task 10 • Von-Neumann-stability analysis of a 2D (horizontal + vertical) system • with advection + sound + buoyancy ( + smoothing, filtering) • constraints: • no bondaries (wave expansion in an  extended medium) • assumptions about the base state: p0=const, T0=const valid for atmospheric motions with about 2-3 km vertical extension • no orography (i.e. no metric terms) • only horizontal advection

  4. (p‘T‘-Dynamics)

  5. Tool to inspect stability properties • von-Neumann-analysis for any combination of cs, U, dT0/dz, T, t, ... : • calculate the amplification matrizes Q (4*4-matrix) • calculate the eigenvalues (use of LAPACK-EV-subroutines) • search of the biggest EV by scanning through kx x = -...+ , kz z = -  ...+ . • ‘verification’: • analytically known stability limits (advection, sound, divergence damping) are calculated correctly • physical stratification instability (i.e. N2<0) will be reproduced( combination ‘buoyancy + sound’ works)

  6. fully explicit uncond. unstable - forward-backward (Mesinger, 1977), unstaggered grid stable for Cx2+Cz2<2 neutral 4 dx, 4dz forward-backward, staggered grid stable for Cx2+Cz2<1 neutral 2 dx, 2dz forward-backw.+vertically Crank-Nic. (2,4,6=1/2) stable for Cx<1 neutral 2 dx forward-backw.+vertically Crank-Nic. (2,4,6>1/2) stable for Cx<1 damping 2 dx Sound • temporal discret.:‘generalized’ Crank-Nicholson=1: implicit, =0: explicit • spatial discret.: centered diff. Courant-numbers:

  7. Choose CN-parameters for buoyancy in p‘T‘-dynamical core of the LMK =0.6 =0.7 =0.5 (‚pure‘ Crank-Nic.) =0.8 =0.9 =1.0 (purely implicit)  choose =0.7 as the best value

  8. ‚Verification‘ of the stability analysis tool: dependency from stratification C=-0.35 C=-0.37 C=-0.38 critical value: C=-0.399  N=0 C=-0.39 C=-0.4 C=-0.395  Tool works for ‚buoyancy + sound‘ C=-0.41 C=-0.42 C=-0.45

  9. xkd

  10. Influence of Cdiv Cdiv = xkd * (cs * t/ x)2 ~0.35 Cdiv=0.025 Cdiv=0.05 Cdiv=0.075 Cdiv=0.1 Cdiv=0.15

  11. without divergence filtering with 3D- divergence filtering with 2D- (only horizontal)divergence filtering

  12. stability of the single waves for Cadv=1, Csnd=0.7 without divergence filtering with 3D- divergence filtering with 2D- (only horizontal)divergence filtering

  13. summary von-Neumann-stability analysis of a 2D (horizontal + vertical) system with advection + sound + buoyancy ( + smoothing, filtering) • without divergence damping: weak instability of long waves remains • without orography: no relevant difference between 2D- and (assumed better) 3D-Divergenzdämpfung.remark A. Gassmann: no difference in amplitude- but in phase-information • experience from LMK: there exist cases which are unstable with 2D-divergence damping (orography?)there exist cases which can be stabilized by 2D-divergence damping  implementation of 3D-div.-damping seems to be reasonable? • LMK-Testsim. mit Bryan-Fritsch-Test: nur mit 3D-Divergenzdämpfung stabil

  14. PP ‚Runge-Kutta‘, Task 8 Improved vertical advection for the dynamical variables (u, v, w, T or T‘, p‘) Motivation: explicitly resolved convection •  vertical advection has increased importance  use a scheme of higher order (compare: horizontal adv. from 2. order to 5. order in RK-scheme) •  greater w (~20 m/s)  Courant-criterium is violated implicit scheme or CNI-explicit scheme up to now: implicit (Crank-Nicholson) advektion 2. order (centered differences) new: implicit (Crank-N.) advektion 3. order  LES with a 5-banddiagonal-matrix implicit Adv. 3. Ordn. in every RK-step: computation costs ~30% of total computation time! planned: outside of the RK-scheme (splitting error?, stability with fast waves?)

  15. Implicit Vertical Advection for dynamic variables (u, v, w, T or T‘, p‘) Generalized Crank-Nicholson-advection (Dimensionless) Advection operator for centered differences 2. order (3-point-stencil):  Lin. eq. system with a tridiagonal matrix, needs ~3 N operations Motivation for a better scheme: explicitly resolved convection, higher values of w

  16. dim.less advection operator for upwind 3. order (4-point stencil) case Cj>0 ( Crowley 3. order, e.g. Tremback et al., 1987) • =1/2: unconditionally stable, damping, truncation error 3. order • >1/2: unconditionally stable, damping, trunc. error 1. order • <1/2: unstable • Lin. eq. system with a 5-band diagonal matrix • needs ~14 N operations • LMK: Subr. complete_tendencies_uvwTpp_CN3

  17. Idealized 1D advection test C=1.5 80 timesteps analytic sol. implicit 2. order implicit 3. order implicit 4. order C=2.5 48 timesteps

  18. Real case study: LMK (2.8 km resolution) ‚12.08.2004, 12UTC-run‘ implicit vertical adv. 2. order difference: 3. order - 2. order

  19. Real case study: LMK (2.8 km resolution)‚ 25.06.2005, 00UTC-run‘ implicit vertical adv. 2. order difference: 3. order - 2. order

  20. Calculation of small timestep • Use correct gridlength:x = R cos (important for bigger areas) • Use correct 2D criterion (dependency from x and y)  Subr. calc_small_timestep • To do: • tcrit is calculated with T=303 K (?) • influence of buoyancy?

  21. 2-dim. horizontal Advection • 2D-advection in RK-schemes by a simple adding of tendencies(operator splitting (e.g. corner transport upstream (CTU) method) is not possible for upstream 3., 5., ... order ) • this is limited by |Cx| + |Cy| < const.this can be proven for RK2 + upwind 3. orderit holds empirically also for RK3 + upwind 5. order • compare with the usual formulated 2-dimensional stability criterion:  Subr. check_cfl_horiz_advection

  22. PP ‚Runge-Kutta‘, Task 3 Tool to inspect conservation properties of LM balance equation for scalar  : temporal change flux divergence sources / sinks • integral over a volume (arbitrary square-stone): ready • Subr. init_integral_3D: define square-stone (in the transformed grid!), domain decomp. • Function integral_3D_total: calc. volume integral • Function integral_3D_cond: calc. vol. integral over individual processor • surface integral over the fluxes: work to do!

More Related