mellor yamada モデルの拡張 - jamstec › hpci-sp › info › meso_sg2014 › pdf › ... ·...
TRANSCRIPT
1/14
Terra Incognita となる解像度のためのMellor‐Yamada モデルの拡張
An extension of Mellor‐Yamada model to apply for the resolution ofTerra Incognita
伊藤純至(東大AORI/気象研)新野宏 (東大AORI)中西幹郎 (防衛大)Chin‐Hoh Moeng (NCAR)
2/14
1. “Terra Incognita” (Grey Zone) problem in PBL turbulence modeling
LES’s subgrid‐scale
1‐D PBL model
f
×
Spectrum density
Frequency 100 s1 Day4 Days
Resolution of conventional numerical model (≧5km) Horizontal resolution near
L (〜1km‐100m): Terra Incognita problem
Spectrum of wind near surface
Size of energy containing eddy L
※ Terra Incognita: 未知の領域PBL: Planetaty Boundary Layer 大気境界層
3/14
ObjectivesDevelop a new PBL model for resolution of Terra Incognita(TI)•Design based on LES reference
e.g. Honnert et al. (2012); Shin and Hong (2013); Kitamura@MRI (Submitted)
•Test run of a new model with horizontal resolution of TIe.g. Ramachandran et al. (2011); Boutle et al. (2014)
→ extension of Mellor‐Yamada’s PBL scheme used in JMANHM etc.
An extension of UK model’s non‐local PBL scheme
4/14
2. LES Reference• Basic equation: Boussinesq approximation, Dry, Smagorinsky‐
Lilly’s SGS• Domain: 18km×18km×4.5km• Resolution: 25m• Lateral Bound.: Doubly periodic• Bottom Bound.: Bulk method (Momentum) / Heat flux
Q=0.2K×m/s• Geostropic winds: V=5 m/s• Δt: 0.2 S
Vertical velocity in middle of CBL18km
0 km
CBL (Convective Boundary Layer) grows
Sensitivity on Q and V (not shown)
5/14
Horizontal filter
Grid valuesHorizontal filter of size ΔH=3×25=75m
Obtain
Larger ΔH
f (xi ) f (xi )G(xi xi ) d xi G(xi xi ) 1/h ( xi xi h )
0 ( xi xi h )
ΔH=5×25=125m
Top‐hat filter:
Only wAverage wVariation w2
Triplet w3
Grid scale motionsSub‐filter TKETurbulent transport
Associated with
should be modeled for resolution ΔH
6/14
Scaling and TKE budget
etui
exi
ijSij g0
f3 xi
ujeuj e 10
uj puj p Production
Free convection scaling (Deardroff 1970) is operated on results with “*” • Length: height of convective
boundary layer h• Velocity : Convective velocity w*
• Heat flux : Surface flux Q
MY model solves prognostic equation of TKE(乱流運動エネルギー):
Advection Turbulent transport Dissipation
Vertical profiles ofover whole x‐y cross section
ΔH*→∞: Meso‐scale
limit
Range 0.5 < ΔH* (≡ΔH/h) <2 correspond to TI scale
7/14
3. Design of model: Estimated terms in TKE budget for various ΔH
*
0
0.2
0.4
0.6
0.8
1
1.2
-0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Hei
ght z
*
e* budget
H*=3.16
1.270.800.33
0
0.2
0.4
0.6
0.8
1
1.2
-0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1H
eigh
t z*
e* budget
H*=3.16
1.270.800.33
Vertical Profiles of terms in TKE budget (horizontal average)
q3
L
① Dissipation:
[Solid] Turbulenttransportsmaller for small ΔH
*
Dissipation(almost constant)
[dotted] Advection(resolved)
Production(resolved)
② Turbulent transport (T.T.): T.T. qLh
exi
Assumed down‐gradient form
→ q: Turbulent velocity ∝ TKE
8/14
① Horizontally averaged Dissipation length <Lε>
Estimated <Lε> for various ΔH*
Lε of Meso‐scale limit (e.g. Nakanishi and Niino,2009)
“Partition function” of TKE
(Honnert et al. 2012)can prescribe <Lε> for TI scale asLε×F(ΔH
*)
9/14
② Diffusion lengthTurbulent transport=‐Ld q∂TKE/∂xi
Horizontal diffusion Vertical diffusion
Slope correspond to Ld
!Neglect counter‐gradient diffusion plotted in 2nd and 4thquadrants
ΔH*=1.27
ΔH*=0.33
To reduce Ld(ΔH*) as
Ld(∞)×F(ΔH*) seem
reasonable To rationalize the neglect of counter‐gradient diffusion → A posteriori test
10/14
4. A posteriori test
• Replace Smagorinsky’s model in LES to“TI model” | MY model (Nakanishi and Niino 2009)
• Grid number: 12×12×150• Horizontal resolution dx=1.5km | 1km | 500m
Typical “TI scale” ↑• Vertical resolution: dz=25m• Environments are the same for LES reference to compare
11/14
Non‐dimensional horizontal resolution dx*(≡ dx/h)
0
0.5
1
1.5
2
2.5
3
1 2 3 4 5 6 7 8 9 10
dx*
Time (h)
TI dx=1500mTI dx=1000m
TI dx=500m
TI scale
LES scale
1.5 km
1 km
500 m
12/14
0
0.1
0.2
0.3
0.4
0.5
0 2 4 6 8 10 12
<w2 >/
w*2
Time (h)
MYNN =500mTIB =500m
@z=0.5h
0
0.1
0.2
0.3
0.4
0.5
0 2 4 6 8 10 12
<w2 >/
w*2
Time (h)
MYNN =1000mTIB =1000m
Strength of resolved convection
“TI Model” reproduce resolved convection in accord with LES reference
LES, Observation<w*2> 〜0.4
MY
TI model
MY
TI model
w*2
w*2
dx=1km dx=500m
+ : Suggested by LES reference
13/14
Conclusion 1/2
• Design of “TI model” based on LES→ Almost completed (To be sumitted to BLM?)
• Implement in a numerical weather prediction (e.g. JMANHM, WRF) would be very easy↑ only reduce length scale in MY model
• We will not peruse further refinement (3rd order closure? stochastic forcing?)
14/14
Conclusion 2/2
• Current JMANHM with MY model maybehave qualitatively similar manner to “TI model” (伊藤ら,2013年春気象学会), but it is unjustified
• Seeking verification by JMANHM with “TI model” ← Karman vortex shading was not good…
Simulated cloud
by JMANHM with a LFM configuration (dx=2km)
15/14
Local LεLε without horizontal average @ ΔH
*=0.8, z*=0.5
16/14
鉛直速度w@5h,z=0.5h
dx=1km
dx=500m
x (km) x (km)
y (km)
y (km)
鉛直速度w(m/s)