44
Mesoscale impacts of explicit numerical diffusion in a convection-permitting model Wolfgang Langhans, Jürg Schmidli, Christoph Schär Institute for Atmospheric and Climate Science, ETH Zurich March 3, 2010 Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 1 / 21

Mesoscale impacts of explicit numerical diffusion in ... - IAC Wiki

Embed Size (px)

Citation preview

Mesoscale impacts of explicit numericaldiffusion in a convection-permitting model

Wolfgang Langhans, Jürg Schmidli, Christoph Schär

Institute for Atmospheric and Climate Science, ETH Zurich

March 3, 2010

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 1 / 21

Outline

1 Introduction

2 COSMO setup

3 Impact of explicit diffusionCloud structuresAlpine heat budgetSpectral analysis

4 The origin of small-scale energy

5 Conclusions

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 2 / 21

Introduction (1)

dt= · · · − ν

(∂4χ

∂x4 +∂4χ

∂y4

)4th-order diffusion operatoralong model surfaces

with χ = u, v , w , p′, T ′, qv , qc , qi

2 3 4 5 6 7 80

0.02

0.04

0.06

0.08

0.1

0.12

0.14

n = λ/∆ x

Dam

ping

/ tim

e st

ep

Explicit diffusive filter response

0.0608

0.032

0.02048

c4=1.2E−04 s−1

c4=6.7E−05 s−1

c4=4.2E−05 s−1

hd_coeff_χ=0.75

hd_coeff_χ=0.40

hd_coeff_χ=0.25

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 3 / 21

Introduction (2)

Upstream-biased advection → no explicit numerical filtersneeded (Wicker and Skamarock 2002)Still it is convenient to apply explicit numerical filters (Knievelet al. 2007)Effective resolution depends on filtering of shortwavelengths (Skamarock 2004) → Importance for CRMIdealized studies show strong influence of numerical andsub-grid turbulent filtering at kilometer-scales (Takemi andRotunno 2003; Bryan 2005)

Objectives and MotivationInvestigate kilometer-scale real-case simulations usingexplicit diffusionIs the bulk Alpine heat budget sensitive to explicit diffusion?How’s diffusion of specific prognostic variables related toconvective precipitation?

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 4 / 21

Introduction (2)

Upstream-biased advection → no explicit numerical filtersneeded (Wicker and Skamarock 2002)Still it is convenient to apply explicit numerical filters (Knievelet al. 2007)Effective resolution depends on filtering of shortwavelengths (Skamarock 2004) → Importance for CRMIdealized studies show strong influence of numerical andsub-grid turbulent filtering at kilometer-scales (Takemi andRotunno 2003; Bryan 2005)

Objectives and MotivationInvestigate kilometer-scale real-case simulations usingexplicit diffusionIs the bulk Alpine heat budget sensitive to explicit diffusion?How’s diffusion of specific prognostic variables related toconvective precipitation?

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 4 / 21

Introduction (2)

Upstream-biased advection → no explicit numerical filtersneeded (Wicker and Skamarock 2002)Still it is convenient to apply explicit numerical filters (Knievelet al. 2007)Effective resolution depends on filtering of shortwavelengths (Skamarock 2004) → Importance for CRMIdealized studies show strong influence of numerical andsub-grid turbulent filtering at kilometer-scales (Takemi andRotunno 2003; Bryan 2005)

Objectives and MotivationInvestigate kilometer-scale real-case simulations usingexplicit diffusionIs the bulk Alpine heat budget sensitive to explicit diffusion?How’s diffusion of specific prognostic variables related toconvective precipitation?

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 4 / 21

Introduction (2)

Upstream-biased advection → no explicit numerical filtersneeded (Wicker and Skamarock 2002)Still it is convenient to apply explicit numerical filters (Knievelet al. 2007)Effective resolution depends on filtering of shortwavelengths (Skamarock 2004) → Importance for CRMIdealized studies show strong influence of numerical andsub-grid turbulent filtering at kilometer-scales (Takemi andRotunno 2003; Bryan 2005)

Objectives and MotivationInvestigate kilometer-scale real-case simulations usingexplicit diffusionIs the bulk Alpine heat budget sensitive to explicit diffusion?How’s diffusion of specific prognostic variables related toconvective precipitation?

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 4 / 21

Introduction (2)

Upstream-biased advection → no explicit numerical filtersneeded (Wicker and Skamarock 2002)Still it is convenient to apply explicit numerical filters (Knievelet al. 2007)Effective resolution depends on filtering of shortwavelengths (Skamarock 2004) → Importance for CRMIdealized studies show strong influence of numerical andsub-grid turbulent filtering at kilometer-scales (Takemi andRotunno 2003; Bryan 2005)

Objectives and MotivationInvestigate kilometer-scale real-case simulations usingexplicit diffusionIs the bulk Alpine heat budget sensitive to explicit diffusion?How’s diffusion of specific prognostic variables related toconvective precipitation?

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 4 / 21

Introduction (2)

Upstream-biased advection → no explicit numerical filtersneeded (Wicker and Skamarock 2002)Still it is convenient to apply explicit numerical filters (Knievelet al. 2007)Effective resolution depends on filtering of shortwavelengths (Skamarock 2004) → Importance for CRMIdealized studies show strong influence of numerical andsub-grid turbulent filtering at kilometer-scales (Takemi andRotunno 2003; Bryan 2005)

Objectives and MotivationInvestigate kilometer-scale real-case simulations usingexplicit diffusionIs the bulk Alpine heat budget sensitive to explicit diffusion?How’s diffusion of specific prognostic variables related toconvective precipitation?

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 4 / 21

Introduction (2)

Upstream-biased advection → no explicit numerical filtersneeded (Wicker and Skamarock 2002)Still it is convenient to apply explicit numerical filters (Knievelet al. 2007)Effective resolution depends on filtering of shortwavelengths (Skamarock 2004) → Importance for CRMIdealized studies show strong influence of numerical andsub-grid turbulent filtering at kilometer-scales (Takemi andRotunno 2003; Bryan 2005)

Objectives and MotivationInvestigate kilometer-scale real-case simulations usingexplicit diffusionIs the bulk Alpine heat budget sensitive to explicit diffusion?How’s diffusion of specific prognostic variables related toconvective precipitation?

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 4 / 21

COSMO setup

Version: 4.3 & 4.8Dynamics:

split-explicit RK-3 scheme (Wickerand Skamarock, 2002)

p’, T’ dynamics

5th-order advection, Bott 2nd-order qxadvection

Monotonic 4th-order diffusion operator

Physics:

prognostic TKE-based 1D turbulencescheme (λ∞ = 100 m),

no cumulus scheme, TERRA_ML

one-moment graupel scheme

Large Alpine domain:

501× 451× 45

∆ϕ = ∆λ = 0.02◦, ∆t = 30 s

IC/BC:

ECMWF operational analysis

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 5 / 21

COSMO setup

Version: 4.3 & 4.8Dynamics:

split-explicit RK-3 scheme (Wickerand Skamarock, 2002)

p’, T’ dynamics

5th-order advection, Bott 2nd-order qxadvection

Monotonic 4th-order diffusion operator

Physics:

prognostic TKE-based 1D turbulencescheme (λ∞ = 100 m),

no cumulus scheme, TERRA_ML

one-moment graupel scheme

Large Alpine domain:

501× 451× 45

∆ϕ = ∆λ = 0.02◦, ∆t = 30 s

IC/BC:

ECMWF operational analysis

σ of 500 hPa geopotential height11.-20. July 2006

(gpdm)

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 5 / 21

COSMO budget diagnosis

What? Extracts model tendencies

∂θ

∂t= −ADV +

Lv

cpdSl +

Ls

cpdSf + MT + Qr + MHD

Saas-Fee

=∂∂t

TADV + HD + MIC + RAD + TUR = TOT

Leadtime →

Tend

ency→

O. Fuhrer (MeteoSwiss)

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 6 / 21

Cloud structures (4 km MSL)No Diffusion

16U

TC

uvwpt0.75

17U

TC

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 7 / 21

Cloud structures (4 km MSL)No Diffusion

16U

TC

uvwpt0.75

17U

TC

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 7 / 21

Cloud structures (vertical slice)11 UTC 13 UTC

15 UTC 17 UTC

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 8 / 21

Potential temperature tendencies (13 UTC)3d Advection Vertical advection

Microphysics Turbulence

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 9 / 21

Diurnal cycle of the heat budget

Net Advection

Turbulence Microphysics

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 10 / 21

. . . and the impact of explicit diffusion

Net Advection

Turbulence Microphysics

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 11 / 21

Diurnal cycle of precipitation

Switzerland Alps

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 12 / 21

Diurnal cycle of w-spectra no diffusion

4 km

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 13 / 21

Diurnal cycle of w-spectra with diffusion

4 km

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 14 / 21

Spectral damping

model

cloud scale

Alpine scaleMeso scale

4 km

model

cloud scale

Alpine scaleMeso scale

8 km

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 15 / 21

Diffusion applied to specific variables

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 16 / 21

Diffusion applied to specific variables

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 16 / 21

Diffusion applied to specific variables

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 16 / 21

Diffusion applied to specific variables

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 16 / 21

Diffusion applied to specific variables

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 16 / 21

Diffusion applied to specific variables

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 16 / 21

Diffusion applied to specific variables

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 16 / 21

Diffusion applied to specific variables

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 16 / 21

Diffusion applied to specific variables

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 16 / 21

The origin of small-scale energy (1)

Diffusion on discrete levels

Modification of net heating

All levels

< 1 km AGL

> 1 km AGL

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 17 / 21

The origin of small-scale energy (2)

Moisture advection:Bott 2nd, Semi-Lagrangian,

van Leer

Modification of heating

Net

Advection

Latent heating

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 18 / 21

Conclusions

“Dilemma”: Strong explicit diffusion results in→ Less energy at poorly resolved scales

→ Reduced convective growth→ Modified impact on mesoscale

Diffusion of qv ,p’,w little influence, diffusion of u,v,T’ largeinfluenceOrigin of grid-scale perturbations at lowest model levelsPartly numerically generated grid-scale noiseSolutions (besides avoiding these scales):

Always a compromise with the current setup (e.g. weakdiffusion of momentum only)Monotonic schemes

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 19 / 21

Conclusions

“Dilemma”: Strong explicit diffusion results in→ Less energy at poorly resolved scales

→ Reduced convective growth→ Modified impact on mesoscale

Diffusion of qv ,p’,w little influence, diffusion of u,v,T’ largeinfluenceOrigin of grid-scale perturbations at lowest model levelsPartly numerically generated grid-scale noiseSolutions (besides avoiding these scales):

Always a compromise with the current setup (e.g. weakdiffusion of momentum only)Monotonic schemes

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 19 / 21

Conclusions

“Dilemma”: Strong explicit diffusion results in→ Less energy at poorly resolved scales

→ Reduced convective growth→ Modified impact on mesoscale

Diffusion of qv ,p’,w little influence, diffusion of u,v,T’ largeinfluenceOrigin of grid-scale perturbations at lowest model levelsPartly numerically generated grid-scale noiseSolutions (besides avoiding these scales):

Always a compromise with the current setup (e.g. weakdiffusion of momentum only)Monotonic schemes

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 19 / 21

Conclusions

“Dilemma”: Strong explicit diffusion results in→ Less energy at poorly resolved scales

→ Reduced convective growth→ Modified impact on mesoscale

Diffusion of qv ,p’,w little influence, diffusion of u,v,T’ largeinfluenceOrigin of grid-scale perturbations at lowest model levelsPartly numerically generated grid-scale noiseSolutions (besides avoiding these scales):

Always a compromise with the current setup (e.g. weakdiffusion of momentum only)Monotonic schemes

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 19 / 21

Conclusions

“Dilemma”: Strong explicit diffusion results in→ Less energy at poorly resolved scales

→ Reduced convective growth→ Modified impact on mesoscale

Diffusion of qv ,p’,w little influence, diffusion of u,v,T’ largeinfluenceOrigin of grid-scale perturbations at lowest model levelsPartly numerically generated grid-scale noiseSolutions (besides avoiding these scales):

Always a compromise with the current setup (e.g. weakdiffusion of momentum only)Monotonic schemes

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 19 / 21

Thanks for your attention

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 20 / 21

Reduction of preciptiation

RUN TOTCH PEAKCH TOTALPS PEAKALPSnone -00.58 -28.83uvwpt0.75 -37.94 -55.76 -37.26 -32.27uv0.75 -30.23 -51.08 -27.32 -24.13p0.75 -13.88 -17.97 +00.91 -01.25t0.75 -27.52 -43.59 -29.06 -27.56q0.75 -04.04 -34.17 -06.14 -04.17uvw0.4 -17.93 -46.14 -20.22 -15.86uvw0.25 -12.23 -35.32 -16.06 -14.21w0.75 -05.04 -32.24 -06.52 -08.37

(%)

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 21 / 21

Reduction of preciptiation

RUN TOTCH PEAKCH TOTALPS PEAKALPSnone -00.58 -28.83uvwpt0.75 -37.94 -55.76 -37.26 -32.27uv0.75 -30.23 -51.08 -27.32 -24.13p0.75 -13.88 -17.97 +00.91 -01.25t0.75 -27.52 -43.59 -29.06 -27.56q0.75 -04.04 -34.17 -06.14 -04.17uvw0.4 -17.93 -46.14 -20.22 -15.86uvw0.25 -12.23 -35.32 -16.06 -14.21w0.75 -05.04 -32.24 -06.52 -08.37

(%)

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 21 / 21

Reduction of preciptiation

RUN TOTCH PEAKCH TOTALPS PEAKALPSnone -00.58 -28.83uvwpt0.75 -37.94 -55.76 -37.26 -32.27uv0.75 -30.23 -51.08 -27.32 -24.13p0.75 -13.88 -17.97 +00.91 -01.25t0.75 -27.52 -43.59 -29.06 -27.56q0.75 -04.04 -34.17 -06.14 -04.17uvw0.4 -17.93 -46.14 -20.22 -15.86uvw0.25 -12.23 -35.32 -16.06 -14.21w0.75 -05.04 -32.24 -06.52 -08.37

(%)

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 21 / 21

Reduction of preciptiation

RUN TOTCH PEAKCH TOTALPS PEAKALPSnone -00.58 -28.83uvwpt0.75 -37.94 -55.76 -37.26 -32.27uv0.75 -30.23 -51.08 -27.32 -24.13p0.75 -13.88 -17.97 +00.91 -01.25t0.75 -27.52 -43.59 -29.06 -27.56q0.75 -04.04 -34.17 -06.14 -04.17uvw0.4 -17.93 -46.14 -20.22 -15.86uvw0.25 -12.23 -35.32 -16.06 -14.21w0.75 -05.04 -32.24 -06.52 -08.37

(%)

Wolfgang Langhans COSMO/CLM User Seminar 2010 March 3, 2010 21 / 21