Watanabe et al. 2014

SOUSEI Program International WS, November 25th, 2014
Understanding and attributing climate
changes in the last decade
Masahiro Watanabe
AORI, University of Tokyo
Contributors: H Shiogama, Y Imada, H
Tatebe, M Mori, Y Kamae, R K Lestari, T
Ogura, M Ishii, M Kimoto and team MIROC
Research targets in Theme A
Prediction and diagnosis of imminent global climate change
 Near-term predictability (incl. ENSO predictability)
 Attribution & mechanisms of past climate changes
 Event attribution (EA)
 Clouds and climate sensitivity  T Ogura
Climate model development for CMIP6 and innovative
system for coupled data assimilation  M Ishii
Timeline of the MIROC group
FY2012
2013
2015
2014
2016
2018
2017
post SOUSEI?
SOUSEI program
Exps using CMIP5 version models
CMIP6 model development
current (no longer fast) machine
next (hopefully faster) machine
DECK
CMIP6
MIPs
* First CMIP6 experiments may be started in early 2016
ENSO & near-term prediction
MIROC5-CGCM (T85L40+1degL44) CHFP
• Anomaly assimilation for ocean T, S
• Initialization for atmos with NCEP reanalysis
• 1979-present, 8 members, 4 times a year
Feb
Prediction skill for Nino3.4 SST anomaly
Aug
Plans for update (resolution, initialisation)
May
Nov
Imada et al. (2014 submitted)
MIROC5-AGCM (T85L40) C20C
• 4 types of experiments
• 1946-2013, 10 members
• 2010-2013 100 members
• Definition of natural SST/sea ice
SST/sea ice
Attribution experiments
EA2013
Peterson et al. (2014 BAMS)
Radiative forcing
Historical PI fixed
Historical
ALL
NST
Natural*
ALF
NAT
Attribution experiments
Research Issues
EA
• 2010 Russian heat wave (Watanabe et al. 2013)
• 2010 drought in Amazon (Shiogama et al. 2013)
• 2012 heavy rain in Japan (Imada et al. 2013)
• 2013 heat wave in USA (Shiogama et al. 2014)
• 2013 heat wave in Japan (Imada et al. 2014)
Attribution of long-term changes
• Global warming hiatus (Watanabe et al. 2014)
• Recent Eurasian cold winters (Mori et al. 2014)
• NH heat waves (Kamae et al. 2014)
• Increasing biomass burning in Sumatra (Lestari et al. 2014)
Global warming hiatus
Internally generated variability?
 Excess energy to TOA for 2001-2010 (Loeb et al. 2012 Nature Geo; Allan et al. 2014 GRL)
 Intensification of ocean heat uptake (Meehl et al. 2011 Nature CC; Watanabe et al. 2013 GRL)
 Associated with the Pacific decadal variability (Meehl and Teng 2012 GRL;
Global-mean sea-level change (mm)
Global-mean surface temperature (℃)
Meehl et al. 2013 JC; Kosaka & Xie 2013 Nature; England et al. 2014 Nature CC;
Watanabe et al. 2014 Nature CC; Trenberth et al. 2014 Nature CC; Meehl et al. 2014 Nature CC)
Trenberth et al. (2014)
UK Met Office (2013)
Global warming hiatus
Internally generated variability?
 Excess energy to TOA for 2001-2010 (Loeb et al. 2012 Nature Geo; Allan et al. 2014 GRL)
 Intensification of ocean heat uptake (Meehl et al. 2011 Nature CC; Watanabe et al. 2013 GRL)
Long-term global warming
TOA radiative energy imbalance
(Net downward)
Atmosphere
Surface
(warmed)
Subsurface
Hiatus
Net TOA imbalance for 2001-2010=0.5∼0.6 Wm2
(still excess energy to the system)
Atmosphere
Surface
(not warmed)
64%
29%
Deep ocean
cf IPCC (2013)
More heat transferred to deep layer
Hiatus reproduced in MIROC-CGCM
Partial wind overriding experiments
MIROC5.2 (T85L40) 5-member ensembles for 1958-2012
* ASYM-H: Tropical (30S-30N) t anomaly replaced with JRA55 reanalysis
* ASYM-C: As in ASYM-H but with external forcing fixed at 1850
SAT change from 1990-1999 to 2001-2012
Observations
ASYM-H
Watanabe et al. (2014, Nature CC)
Tropical ocean wind stress anomalies are sufficient
to reproduce surface temperature anomaly pattern in the hiatus
Hiatus reproduced in MIROC-CGCM
Partial wind overriding experiments
Decadal wind stress variability substantially contributes to the
warming acceleration in the 1980-90s and hiatus in the 2000s
CMIP models
Obs
Decomposition of decadal-mean
SAT changes
ASYM-H
ASYM-C diff.
ASYM-H
(tropical tDecade
anomaly DTINT
replaced 1980s
with reanalysis)
+0.11K
1990s
ASYM-C
(ASYM-H
+
2000s
external forcing
fixed at 1850)
DTINT/DTALL
47%
+0.13K 38%
-0.11K
27%
Watanabe et al. (2014, Nature CC)
Substantial contribution of the internal decadal variations
Fractional contribution decreases as rising signal of anthropogenic warming
Surface air temperature anomaly [℃]
Seasonality in warming trends
Courtesy of Y Kamae
 Hiatus seen year round over oceans
 Hiatus only in DJF over land  What has happened in warm seasons?
Attribution of increase in NH heat waves
MIROC5 ALL runs reproduce increasing f
Contribution to heatwave frequency (25-50N)
f = fADIR + fASST + fNAT
fADIR
fNAT
fASST
38%
43%
19%
Long-term 40%
15%
45%
Decadal
Combined analysis of ALL, NAT, NST ensembles
 Considerable contribution of direct radiative
Effect of GHGs to increasing heat wave frequency.
Kamae et al. (2014 GRL)
Hiatus and East Asian summer
Given the negative PDO accompanies hiatus ―
 Higher SST in the western Pacific  Stronger subtropical high?
 Higher SST in the North Pacific  Amplifies warming response?
What we observed
in 2013 hot summer
High
Warm gets warmer, cold gets colder
SAT anomaly over Japan
Temperature anomaly (℃)
monthly anomaly for 1999-2012
Warm seasons
Cold seasons
1960
1980
2000
Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov
Urabe & Maeda (2014 SOLA)
Why cold winters?
Due to warm western Pacific (hiatus)
& warm Eurasia?
Why cold winters occur in the last decade?
EOF2 = WACE
Decadal-mean winter (DJF) SAT diff
from 1994-2003 to 2004-2013
• cold winters
• warm winters
EOF1 = AO
Mori et al. (2014, Nature Geo)
• EOF2 = Warm Arctic and Cold Eurasia (WACE)
• Eurasian winter SAT explained by a combination
between the AO and WACE
• The WACE index is highly correlated with BarentsKara sea-ice anomalies (r=-0.81)
Arctic sea-ice reduction favors cold Eurasia
 MIROC4-AGCM (T106L56) 100-member ensembles with
 Low sea-ice condition in Barents-Kara Sea (LICE)
 High sea-ice condition in Barents-Kara Sea (HICE)
 Phase of the WACE pattern (=EOF2) depends on sea-ice anomaly
EOF1=AO
31%
20%
PDFs of PCs
probability (%)
MIROC4-AGCM
ERA Interim
– LICE
– HICE
23%
17%
probability (%)
EOF2=WACE
model-PC1
– LICE
– HICE
model-PC2
Mori et al. (2014, Nature Geo)
Sea-ice driven signal against noise
SAT difference (˚C)
Ensemble mean SAT difference (LICE-HICE) over central Eurasia
sampling size of ensemble member
Ensemble mean WACE index for LICE & HICE
mean PC2
LICE
HICE
sampling size of ensemble member
minimum number for
robust signal detection
Robust circulation response
to sea-ice anomaly with only
≥ 20 members,  weak
SN ratio for SAT
 AO acts as noise
Mori et al. (2014, Nature Geo)
Summary
 Supported by the SOUSEI program,
MIROC group has actively worked on
attribution of recent climate changes
and weather events as well as climate
prediction
 The outcomes have been/will be
beneficial to
 Physical understanding of changing
climate
 Outreach to public
 The
above activity
willclimate
be
 Collaboration
with
services
continued/expanded towards CMIP6,
hopefully funded by the post-SOUSEI
backup
Global warming hiatus
Hypotheses
Externally forced response?
 Weakening of solar activity (Kaufmann et al. 2011 PNAS)
 Increase of the stratospheric aerosols (Solomon et al. 2011 Science)
 Accumulated effect of minor volcanoes (Schmidt et al. 2014 Nature Geo)
Internally generated variability?
 Climate ‘noise’ (Easterling and Wehner 2009 GRL)
 Intensification of ocean heat uptake
(Meehl et al. 2011 Nature CC; Watanabe et al. 2013 GRL)
 Associated with the Pacific decadal variability
(Meehl and Teng 2012 GRL; Meehl et al. 2013 JC)
Unclassified
 Decrease of the stratospheric water vapor (Solomon et al. 2010 Science)
Energy is still in
Weakness of the hypothesis: Externally forced hiatus
Top of Atmosphere (TOA) radiation budget by Clouds and the
Earth's Radiant Energy System (CERES) indicates net storage of
0.5±0.43 W/m2 into the climate system for 2001-2010
Loeb et al. (2012 Nature Geo)
Weilicki et al. (1996 BAMS)
 Where is the missing energy?
Ocean interior is warming
Continuous rise of global-mean sea level
UK Met Office (2013)
Intensified Pacific trades and heat uptake
Partial assimilation experiments using CGCM
* Eastern Pacific cooling  hiatus (Kosaka and Xie 2013 Nature)
* Intensified Pacific trades in the 2000s
 hiatus w/ increasing ocean heat uptake (England et al. 2014 Nature CC)
Model response to 1992-2011 wind trend
England et al. (2014)
Partial Wind Overriding historical experiments
MIROC5.2 (T85L40) 5-member ensembles for 1958-2012
Excess energy to the system 
* ASYM-H: Tropical (30S-30N) t anomaly replaced with JRA55 reanalysis
* ASYM-C: As in ASYM-H but with external forcing fixed at 1850
2001-2010 average:
0.64±0.26 W/m2 (ASYM-H)
0.5±0.43 W/m2 (Loeb et al. 2012 Nature Geo)
0.62±0.43 W/m2 (Allan et al. 2014 GRL)
TOA radiative imbalance
ASYM-H
Watanabe et al. (2014, Nature CC)
Attribution of hiatus
Contribution of natural internal variability & external forced
component to the global warming acceleration & hiatus
Substantial contribution of the internal decadal variations
Fractional contribution decreases as rising signal of anthropogenic warming
Decomposition of decadal-mean
SAT changes
ASYM-H
Decade
1980s
1990s
2000s
ASYM-C diff.
DTINT DTINT/DTALL
+0.11K 47%
+0.13K 38%
-0.11K
27%
Watanabe et al. (2014, Nature CC)
Attribution of hiatus
Distinct structure of ocean warming
Zonal-mean temperature change from 1990-1999 to 2001-2012
Obs
(Ishii)
DTALL
=ASYM-H
DTINT
More heat uptake
in the subtropics by the
wind-driven circulation
DTEXT
More heat uptake
in northern high latitudes
(mostly North Atlantic)
Watanabe et al. (2014, Nature CC)
Introduction
In the recent decades . . .
 北極域で温暖化が顕著 (北極温暖化増幅)
 ユーラシアの中緯度域で寒冷化
The 10-year mean winter (DJF) SAT difference
between 1994-2003 and 2004-2013.
 再解析データを用いた研究
北極海における急速な海氷の減少が
ユーラシアの寒冬を強制していることを示唆
(e.g. Liu et al. 2012; Tang et al. 2013)
 モデリング研究
ロバストな大気応答は未だに得られていない
(e.g. Honda et al. 2009; Peings & Magnusdottir 2014)
非常に活発な冬の大気の内部変動が、海氷の減少
に対する大気応答の検出を難しくしている。
疑問
最近の寒冬の増加は海氷減少の応答?自然変動?
 AGCMを用いた大規模アンサンブルシミュレーションによっ
て、冬に減少が著しいバレンツ・カラ海の海氷減少が、ユー
ラシアの中緯度域に低温偏差を強制し得ることを示す。
 地表気温偏差に対する大気の内部変動の寄与と、海氷の
変動に対する大気応答の寄与とに分離し、定量化する。
National Snow & Ice Data Center
Model : MIROC4-AGCM (T106L56)
Period
: 9月〜3月
GHG
: 2011–2012 (RCP4.5) 状態
Ensemble : 100 メンバー
SST
: 月別気候値
(HadISST, 1979-2000)
sea-ice concent. (%)
Experimental Design
バレンツ・カラ海の海氷密接度 (9月)
high-ice years
low-ice years
 2種類のアンサンブル実験: low-ice (LICE) 実験と high-ice (HICE) 実験
 バレンンツ・カラ海で海氷が多かった10年と少なかった10年でコンポジットした
海氷密接度場を、 それぞれ LICE と HICE 実験で用いた。
LICE–HICE (SON)
LICE–HICE (DJF)
change in recent decade (DJF)
バレンツ・
カラ海
sea-ice concentration (%)
Observed and simulated change in SAT, SLP and Z500 (DJF)
Difference of composite fields (LICE-HICE)
Z500 (m)
ERA
Interim
Z500 (m)
AGCM
ensemble
mean
near-surface air temperature (˚C)
モデルのアンサンブル平均は、観測の特徴をよく再現している。
stippling indicates 95% statistical confidence
Two leading modes governing winter SAT anomalies (OBS)
EOF1 (31%)
SAT & SLP(cont.) & V10
AO
r = 0.85
EOF2 (23%)
decrease
WACE
r = -0.81
increase
 EOF2 は、“Warm Arctic and Cold Eurasia” (WACE) パターン。
 PC2 は、バレンツ・カラ海の海氷偏差と強い相関(r = -0.81)。
 バレンツ・カラ海の海氷変動に付随する地表気温の変動はWACEで表現され、
それは北極振動とは独立で個別の変動である。
Two leading modes governing winter SAT anomalies (OBS)
EOF1 (31%)
SAT & SLP(cont.) & V10
r = 0.85
EOF2 (23%)
decrease
r = -0.81
increase
 ユーラシア中央部における寒冬は、北極振動(内部変動)とWACE(海氷減少による応答)
の組み合わせで説明される。
 近年の寒冬の頻度の増加には、WACE の寄与が大きいことが示唆される(WACEは有意
な正のトレンドを示すので)。
Two leading modes governing winter SAT anomalies (AGCM)
ERA Interim
EOF1 (31%)
AGCM
EOF1 (20%)
AO
EOF2 (23%)
EOF2 (17%)
WACE
 200メンバーのアンサンブルシミュレーションの卓越変動は、観測のそれをよく再現する。
 WACEが卓越変動であるか否かは、海氷の変化に依存しない!(intrinsic mode)
 しかしながら、WACEの出現頻度は海氷偏差に依存する。
Scatter plots of PC scores
ERA Interim
AGCM
contour: PDF of
PCs for both LICE
and HICE
• cold winters
• warm winters
difference
between LICE
and HICE PDFs
 LICE の寒冬 (blue •) の頻度は、HICE の寒冬 (blue x ) の頻度よりも明らかに多い。
逆もまたしかり (red marks; • and x)。
 実際に、LICE の 14.4% のメンバーが、ユーラシア大陸中央部で寒冬を示し、それは
HICE の 6.2% よりも 2倍以上大きい。
* The cold and warm winters over central Eurasia (60E–120E, 40N–60N) are defined by averaged
SAT anomalies above and below one standard deviation, respectively.
MIROC in CMIP5
Who are we
 Joint modeling group among
AORI Univ of Tokyo, National Institute for
Environmental Studies (NIES), and JAMSTEC
Model lineup
 CGCM
MIROC5 (T85L40+1deg)
MIROC4h (T213L56+0.25deg)
 ESM
MIROC-ESM (T42L80+1deg)
MIROC-ESM-CHEM (T42L80+1deg)
MIROC in CMIP6
Support for modeling activity
 Program for Risk Information on Climate Change
(‘SOUSEI’ program), 2012-2016, 5000K USD
Model lineup
 CGCM
MIROC5 (T85L40+1deg)MIROC6-CGCM (T85L56+1deg)
MIROC4h (T213L56+0.25deg)
 ESM
MIROC-ESM (T42L80+1deg)MIROC6-ESM (T85L80+1deg)
MIROC-ESM-CHEM (T42L80+1deg)
 NICAM AGCM (7km/14km/28km)
Attribution of hiatus
Wind stress variability responsible for the global-mean SAT change
10-member AMIP runs w/ and w/o anthropogenic warming (AMIP-H & AMIP-C)
reveal the decadal t variability is approximately independent of the global warming
PC timeseries
t EOF1
reanalysis
Global-mean SAT
t EOF1
AMIP-H
Watanabe et al. (2014, Nature CC)
Strengthening of heat uptake efficiency
Discrepancy in k between ‘observations’ and CMIP5
Simplified energy balance
Assuming small Td and steady Ts,
approximate balance will be
Heat uptake efficiency k may be estimated
using obs DT, DF and GCM-based range of
feedback parameters l
What’s the process making differences in k?
Watanabe et al. (2013 GRL)
Pacific Decadal Oscillation (PDO)
Dominant natural decadal variability in
the Pacific atmosphere-ocean system
PDO tends to be in
its negative phase
during 2000-2012
Trenberth et al. (2014 Nature CC)
Global-mean surface temperature anomaly [℃]
Record September 2014
Linear trend = 0.61 ℃/century
+0.34 ℃ (wrt 1981-2010)
In reality the record high temp continues from June !
http://www.data.jma.go.jp/cpdinfo/temp/sep_wld.html
Global-mean surface temperature anomaly [℃]
Record October 2014
Linear trend = 0.63 ℃/century
+0.34 ℃ (wrt 1981-2010)
In reality the record high temp continues from June !
http://www.data.jma.go.jp/cpdinfo/temp/oct_wld.html
Has hiatus started changing?
http://www.data.jma.go.jp/cpdinfo/temp/oct_wld.html
Decadal forecast suggests the end of hiatus?
….. But skill lacks in the tropical Pacific
Operational forecast by Hadley Centre
5-95% range
of the forecast
from 2014
Nov2008-Oct2013
surface temperature
anomaly
http://www.metoffice.gov.uk
Simulated ocean warming
Zonal-mean ocean
temperature anomaly
for 2001-2010
Observations
(Ishii data)
MIROC5
(11-member
historical runs)
Warming overestimated, but the structure not too bad
Watanabe et al. (2013 GRL)
Dominant internal variability
2001-2010 SST anomalies in MIROC5
For ith member (i=1,N)
 Ensemble mean does not show the hiatus
 There are members simulating hiatus
 How to extract the structure associated with
the hiatus in the model?
Ti  T  Ti
Thiatus
‘Hiatus’ SST pattern (regression across members)
T 

SATg
Obs SST anomaly (2000-2012)
Watanabe et al. (2013 GRL)
Anthropogenic SST changes
Estimated from CMIP AOGCM attribution experiments
SST trends 1946-2012
Fraction of interannual/decadal
components to linear trends
MIROC3.2
(10members)
CMIP5 MME
Watanabe et al. (2014, Nature CC)
Attribution of hiatus
Does wind stress trend (ie weakening of the Walker cell) play a role?
-NO. Global mean SAT time series is virtually unchanged w/o trend
Watanabe et al. (2014, Nature CC)
Attribution of hiatus
Wind stress variability responsible for ENSO = EOF2
10-member AMIP runs w/ and w/o anthropogenic warming reveal
the interannual t variability is almost independent of the global warming
Watanabe et al. (2014, Nature CC)