Deep moist atmospheric convection in a sub‐
kilometer global simulation
Yoshiaki Miyamoto, Yoshiyuki Kajikawa, Ryuji Yoshida, Tsuyoshi Yamaura, Hisashi Yashiro, Hirofumi Tomita
(RIKEN AICS)
I.
II.
III.
Background
Experimental Settings
Methodology for detection of convection
IV. Results
V. Conclusion
戦略分野３メソ課題研究会 March 7th, 2014 @Kobe
I. BACKGROUND
※deep moist convection := ”Convection”
Byers and Braham (1949)
Convection
Convection
• Element of cloudy disturbances
• Transport heat and moisture
 Horizontal scale (x) 〜100 km
hard to explicitly solve in global models
（x〜101 – 102 km）
←cumulus parameterization
Byers and Braham (1949)
Heat & Moisture
~100 km
~100 km
2000〜
Model development ＋enhancement of computer
power => x 〜 100 km
→clouds are explicitly solved in global models
 Still coarser or comparable to obs.
Regional model (Weismann et al., 1997) : Change around Δx <= 4 km
Objective:
Reveal the dependence of the simulated convection on resolution in global
model by describing the global statistical characteristics.
Experimental design
model
NICAM (Tomita and Satoh 2004, Satoh et al. 2008)
Initial state
3‐day integrated results of 1‐step coarser resolution
SST
NCEP analysis + nudging (Reynolds weekly SST)
land
Model adjusted produced by 5 year run
Cloud physics
NSW6 (Tomita 2008)
Boundary layer turbulence
MYNN (Nakanishi and Niino 2004, Noda et al. 2008)
Surface flux
Louis (1979)
Long and short‐wave radiation
MSTRNX (Sekiguchi and Nakajima 2008)
Cumulus parameterization
‐‐
Δx
integration period（12 h）
※72 h integration before producing initial fields
Computational Cost
•
•
•
•
Nodes used: 20480 (~160000 cores)
Wall-clock time: 53 h
Sustained performance: 7~8 %
Storage: 200 TB
Δx
integration period（12 h）
※72 h integration before producing initial fields
Δ14.0 km
Snap shot of OLR
12‐h integrations
Δ7.0 km
Δ3.5 km
Δ1.7 km
Δ0.87 km
II. METHOD OF DETECTING CONVECTION
1. Detect “convective grids” by ISCCP table
2. Determine “convective core” grids
Step 1/2: Detect convective grids by ISCCP table
Δx = 14 km
(Rossow and Schiffer 1999)
Step 2/2: determine convective core grids
a) ISCCP convective grids ( )
b) Find grids ( ) at which all the surrounding 8 grids satisfy the ISCCP condition
c) Estimate horizontal gradient of vertical velocity averaged vertically in the troposphere
d) Convective grids ( ) := where vertically aved w is larger than those at surrounding 8 grids
example（Δx=3.5 km）
Δx = 3.5 km
ISCCP convective grid
Convection core grid
w（troposphere mean）
CI = 0.1 m s‐1
w = 0.1 m s‐1
Δx = 14 km
ISCCP convective grid
Convection core grid
w（troposphere mean）
CI = 0.1 m s‐1
w = 0.1 m s‐1
Δx = 7 km
Δx = 1.7 km
III. RESULTS
Δ0.87 km
Composited structure of convection (GL13)
Convection core grid
※transform the coordinate into the cylindrical around the core grid
mean of all the detected convection
symmetric around the x axis
Composite of convection (vertical velocity)
Δ14.0 km
Δ7.0 km
Δ3.5 km
Δx ≧ 3.5 km:
– Convection is represented at 1
grid
– Little dependence on resolution
Δx ≦ 1.7 km:
– Convection is represented at
multiple grids
– Intensify w/ resolution
Δ1.7 km
Δ0.87 km
※transform the coordinate into the cylindrical around the core grid
mean of all the detected convection
symmetric around the x axis
X axis is normalized by resolution
Number and distance of convection
number
10
0.25
frequency/total number
(a) number of convection
5
(b) distance between convection
14.0
7.0
3.5
1.7
0.8
0.2
0.15
10
4
0.1
0.05
10
3
0.8
1.7
3.5
 (km)
7
14
0
0
Δx ≧ 3.5 km:
– number: increase by factor of 4
– distance: 4 grids
Δx ≦ 1.7 km:
– number: decrease in increasing rate
– distance: 5 grids
4 5
number of gird
10
Summary
Global simulation with
a sub-kilometer
resolution
Finding
 Convection features (structure, number, distance)
change between Δ3.5 km  Δ1.7 km
- Δx should be 2.0〜3.0 km to resolve convection in global models
3.5 km
1.7 km
Thank you very much for your attention!
Miyamoto, Y., Y. Kajikawa, R. Yoshida, T. Yamaura, H. Yashiro and H. Tomita, 2013:
Deep moist atmospheric convection in a sub-kilometer global simulation, Geophysical
Research Letters, 40, 4922-4927.
Special thanks to Drs. H. Miura, S. Iga, S. Nishizawa, M. Satoh, our colleagues, and two anonymous reviewers for fruitful discussions. The authors are grateful to researchers and technical experts at RIKEN and FUJITSU for their kind help. The simulations were performed using the K computer at the RIKEN Advanced Institute for Computer Science. SUPPLEMENT
What is the general characteristics of convection?
Isolated convection
– Element of amospheric cloudy disturbances
– Transport of heat/moisture
What is the general characteristics of convection?

•
Jorgensen and LeMone (1989): 50% of
convection (core) has horizontal scale less
than 1 km
Resolution dependence
 Weismann et al. (1997): dependence of squall line
(Klemp and Wilhelmson (1979) cloud model)
- Characteristics changes Δx less than and
equal to 4 km
15
コアの直径 (km)
•
1
存在頻度 (積算)
Jorgensen&LeMone (1989)
http://callofduty.wikia.com/wiki/File:Cumulonimbus.jpg
Is there any threshold of
resolution in realistic conditions?
w’’
w’u’
qr
Weismann et al. (1997)
Model（NICAM, Tomita & Satoh 2004, Satoh et al. 2008）
Global



cloud-system resolving model
Icosahedral grid
nohydrostatic DC
explicit cloud expression:
Miura et al. (2007)
21
50
40
30
20
10
0
30
area icp cutoff (GL08-12 t=201208250600)
14
0.1
0.08
0.06
0.04
0
1
2
3
4
5
0
0.02
km
a re a (
解析範囲：
130—190E, ‐15—15N
0.8
1.7 3.5 7
Delta x (km)
km )
a re a ( 1 0
2)
1 0 -6
面積
a re a ( 1 0 -6 k m 2 )
14
30
area con cutoff (GL08-12 t=201208250600)
0.8
1.7 3.5 7
Delta x (km)
14
30
area icp cutoff (GL08-12 t=201208250600)
0.8
1.7 3.5 7
Delta x (km)
allave
allave
解析範囲：
130—190E, ‐15—15N
300
250
200
150
100
50
0
0.8
1.7 3.5 7
Delta x (km)
14
30
areal ave. (con) of z-aved mass flux (GL08-12 t=201208250
400
conave
350 conave
s -1 )
-2
m a s s f lu x ( x 1 0 3 k g m
8
6
4
2
0
0.8
1.7 3.5 7
Delta x (km)
14
30
m a s s f lu x ( x 1 0 3 k g m
-2
s -1 )
150
100
50
0
icpave
icpave
0.8
1.7 3.5 7
Delta x (km)
14
30
areal ave. (all) of z-aved mass flux (GL08-12 t=2012082506
areal ave. (icp) of z-aved mass flux (GL08-12 t=2012082506
10
200
面積平均質量フラックス
s -1 )
-2
m a s s f lu x ( x 1 0 3 k g m
allave
allave
解析範囲：
130—190E, ‐15—15N
p r e c ip it a t io n ( x 1 0 3 m m h - 1 )
0.15
0.1
0.05
0
0.8
1.7 3.5 7
Delta x (km)
14
30
8
6
4
2
0
conave
conave
0.8
1.7 3.5 7
Delta x (km)
14
30
areal ave. (icp) of rain flux (GL08-13 t=201208250600)
10
p r e c ip it a t io n ( x 1 0 3 m m h - 1 )
4
3
2
1
0
icpave
icpave
0.8
1.7 3.5 7
Delta x (km)
14
30
areal ave. (all) of rain flux (GL08-13 t=201208250600) areal ave. (icp) of rain flux (GL08-13 t=201208250600)
0.2
5
面積平均降水量
p r e c ip it a t io n ( x 1 0 m m h )
解析範囲：
130—190E, ‐15—15N
15
10
5
0
contot
contot
0.8
1.7 3.5 7
Delta x (km)
14
30
rea-integrated (con) z-aved mass flux (GL08-13 t=20120825
20
m a s s f lu x ( x 1 0 3 k g s - 1 )
240
220
200
180
160
140
120
100
alltot
alltot
0.8
1.7 3.5 7
Delta x (km)
14
30
m a s s f lu x ( x 1 0 3 k g s - 1 )
240
220
200
180
160
140
120
100
icptot
icptot
0.8
1.7 3.5 7
Delta x (km)
14
30
area-integrated (all) z-aved mass flux (GL08-13 t=201208250
area-integrated (icp) z-aved mass flux (GL08-13 t=20120825
面積積分質量フラックス
m a s s f lu x ( x 1 0 3 k g s - 1 )
面積積分降水量
alltot
alltot
解析範囲：
130—190E, ‐15—15N
m )
p r e c ip it a t io n ( x 1 0 m m h
contot
contot
0.8
1.7 3.5 7
Delta x (km)
14
30
area-integrated (con) of rain flux (GL08-13 t=20120825060
0.2
0.15
0.1
0
0.05
2)
4
3
2
1
0
0.8
1.7 3.5 7
Delta x (km)
14
30
p r e c ip it a t io n (
x103
mm
h -1
m
4
3
2
1
0
icptot
icptot
0.8
1.7 3.5 7
Delta x (km)
14
30
area-integrated (all) of rain flux (GL08-13 t=20120825060area-integrated (icp) of rain flux (GL08-13 t=20120825060
5
5
p r e c ip it a t io n ( x 1 0 3 m m h - 1 m 2 )
解析範囲：
130—190E, ‐15—15N
240
230
220
210
200
0.8
1.7 3.5 7
Delta x (km)
14
30
areal ave. (all) OLR (GL08-13 t=201208250600)
250
all
all
-2 )
O L R (W m
-2 )
O L R (W m
面積平均OLR
-2 )
O L R (W m
con
con
14
30
areal ave. (con) OLR (GL08-13 t=201208250600)
150
140
130
120
110
100
0.8
1.7 3.5 7
Delta x (km)
icp
icp
14
30
areal ave. (icp) OLR (GL08-13 t=201208250600)
150
140
130
120
110
100
0.8
1.7 3.5 7
Delta x (km)
東西風速
降水量
海面更正気圧
東西風速・降水量・海面更正気圧
の緯度分布
• 各解像度間に大きな差無し
• 解析値・観測値との顕著な差
無し
Skamarock (2004)
Effective resolution （~ 6-7Δx）：
それより小さい空間スケールの現象が、モデルで計算される運動エネル
ギースペクトルが-5/3則から外れる解像度
•
実現象で対流の存在する間隔 < 6-7Δx
– モデルでは現象と同様の間隔を再現できない
→Effective resolution以上で、且つ、実現象に最も近
いスケール（=6-7Δx）に最頻値が出現
•
実現象で対流の存在する間隔 > 6-7Δx
– モデルで対流間の距離を解像可能