ME-3036

Perubahan Iklim

CMIP5
(Coupled Model Intercomparison Project Phase 5)
+ CMIP5
(Coupled Model Intercomparison Project Phase 5)

A framework for coordinated climate change experiments for

assessment in the IPCC AR5

CMIP5 provide a multi-model context for

1) assessing the mechanisms responsible for model

differences in poorly understood feedbacks associated
with the carbon cycle and with clouds,

2) examining climate “predictability” and exploring the

ability of models to predict climate on decadal time scales,

3) determining why similarly forced models produce a range

of responses.
http://cmip-pcmdi.llnl.gov/
https://esgf-node.llnl.gov/search/cmip5/
+

The relationship of CMIP5 (which is organized by WGCM) to

organizations established to coordinate climate research activities
internationally and to the IPIPCC, the modeling centers, and the
climate research community
+
CMIP5 Objectives
To:
 address outstanding scientific questions arising from IPCC AR4
 improve understanding of climate, i.e. to understand some of the
factors responsible for differences in model projections, including
quantifying some key feedbacks such as those involving clouds
and the carbon cycle

 evaluate how realistic the models are in simulating the recent past,

 provide projections of future climate change on two time scales,

near term (out to about 2035) and long term (out to 2100 and
beyond), and
Not intended to be comprehensive – other experiments will emerge
along the way
+ CMIP5 Experiments
Two types of climate change modeling experiments:

1) long-term (century time scale), usually started from

multicentury preindustrial control (quasi equilibrium)
integrations

2) near-term (10–30 yr), also called decadal prediction

experiments, initialized with observed ocean and sea ice
conditions

Both are integrated using atmosphere–ocean global climate

models (AOGCMs), the “standard” models used in previous CMIP
phases.

The AOGCMs respond to specified, time-varying

concentrations of various atmospheric constituents (e.g.,
greenhouse gases) and include an interactive representation of
the atmosphere, ocean, land, and sea ice.

For the long-term simulations, some of the AOGCMs will be

coupled to biogeochemical components that account for the
important fluxes of carbon between the ocean, atmosphere, and
terrestrial biosphere carbon reservoirs [Earth system models
(ESMs)],
+ Climate Change Scenarios
 The CMIP5 projections of climate change are driven by concentration or
emission scenarios called RCPs (Radiative Concentration Pathaways)

 SRES (Special Report on Emissions Scenarios) did not include policy

intervention, while the RCPs are mitigation scenarios that assume policy
actions will be taken to achieve certain emission targets.

 RCPs have been formulated based on a range of projections of future

population growth, technological development, and societal responses.

 The labels for the RCPs provide a rough estimate of the radiative forcing
in the year 2100 (relative to preindustrial conditions).

 For example, the radiative forcing in RCP8.5 increases throughout the

twenty-first century before reaching a level of about 8.5 W m−2 at the end
of the century. In addition to this “high” scenario, there are two
intermediate scenarios, RCP4.5 and RCP6, and a low so-called peak-and-
decay scenario, RCP2.6, in which radiative forcing reaches a maximum
near the middle of the twenty-first century before decreasing to an
eventual nominal level of 2.6 W m−2.

Radiative forcing: difference of incoming solar radiation absorbed by the Earth and
energy radiated back to space.
+
+

Radiative forcing: difference of incoming solar radiation absorbed by the Earth and
energy radiated back to space.
+ CMIP5 Experiments
Two types of climate change modeling experiments:

1) long-term (century time scale), usually started from

multicentury preindustrial control (quasi equilibrium)
integrations

2) near-term (10–30 yr), also called decadal prediction

experiments, initialized with observed ocean and sea ice
conditions

Both are integrated using atmosphere–ocean global climate

models (AOGCMs), the “standard” models used in previous CMIP
phases.

The AOGCMs respond to specified, time-varying

concentrations of various atmospheric constituents (e.g.,
greenhouse gases) and include an interactive representation of
the atmosphere, ocean, land, and sea ice.

For the long-term simulations, some of the AOGCMs will be

coupled to biogeochemical components that account for the
important fluxes of carbon between the ocean, atmosphere, and
terrestrial biosphere carbon reservoirs [Earth system models
(ESMs)],
+ CMIP5 Design Summary

• All models perform CORE experiments (basis for intercomparison)

• Models perform Tier 1 and Tier 2 experiments as interests and
resources dictate: these explore specific aspects of model forcing,
response, and process
+ Near term (decadal) experiments

Assess model skill on

time scales where initial
state may exert some
influence

At least 3
ensemble
members
for each
run
Assess model skill on
time scales where GHG
forcing expected to
exert some influence
+ Long term experiments
Pre-industrial control
(ca. 1850) and 20th
century, forced by
concentrations and by
emissions
Future scenarios (RCPs)
forced by
concentrations and by
emissions

“Diagnostic” runs to
assess transient climate
response, equilibrium
climate sensitivity

Green font indicates experiment performed

only by models with carbon cycle
representation
+ Long term experiments, cont’d
Partial-forcing and
single-forcing factor
experiments to quantify
C-climate feedbacks

Additional RCPs and

multi-century
simulations
And many other
simulations…

“Diagnostic” runs to
assess C-climate
feedbacks and
allowable (fossil fuel +
land use) emissions

Green font indicates experiment performed

only by models with carbon cycle
representation
All models:

Control,
historical,
and paleo

Future
scenarios
(RCPs)

Diagnostic
simulations
(feedbacks)

Attribution
runs (single
and multi-
factor)
+
Simulations only performed by
ESMs…

Forced by fossil fuel emissions and land use changes,

as opposed to concentrations
+
+ Data access:
https://esgf-
node.llnl.gov/projects
/cmip5/
+ CMIP Phase-6 (Scheduled 2021)
+
+
What is SSPs?
• SSPs are socio-economic pathways that consist of quantitative and qualitative
elements that pertain to society, energy, economy and environment.
• They are reference pathways in that they contain no climate-related policy
interventions or impacts. SSPs are envisioned in a two-dimensional ‘challenges
space’, with the degree of challenges to mitigation on one axis and the degree of
challenges to adaptation on the other.
+ Future RCP CO emissions scenarios featured in CMIP5 and their
2
CMIP6 counterparts, as well as historical CO2 emissions (in black).

https://www.carbonbrief.org/cmip6-the-next-generation-of-climate-models-explained
+
+ Climate Projections:
“Raw” Model Outputs
Sample grid point
( 2.5o x 1.67o Res.)

IPSL-CM5A-MR
RCP 4.5
+ Climate Projections:
Downscaled Rainfall Projections

Statistical
Downscaling

10 ensemble
members per
model

Area averaged
over BS
watershed
+ Coordinated Regional Climate
Downscaling Experiment
(CORDEX)
Global partnership on
regional climate
downscaling efforts, both
dynamical and empirical-
statistical downscaling
experiment.

Consists of 14 domains (a
region for which the
regional downscaling is
taking place)

Indonesia is part of South

East Asia CORDEX
Domain
CORDEX (14 domains)

East Asia
25 km x 25
South Asia km
25 km x 25
50 km x 50 Southeast
km
km
Asia
 Thailand

Philippines Lao PDR

Indonesia Cambodia

Malaysia Vietnam

UK Australia Sweden South Korea Hong Germany Japan

Kong
+

Website:
http://www.ukm.my/
seaclid-cordex/

http://www.rucore
.ru.ac.th/Home
Downscaling of CMIP5 climate projection :
+ ideal: dynamical downscaled
Dynamical
Downscaling

Bias correction
works well in
reducing spurious
signal from
original model
output
3. Downscaling of CMIP5 climate projection :
alternative: dynamical downscaled CMIP5 data are not (or insufficiently)
available
Statistical Downscaling
+

A Review of Downscaling Methods for Climate Change Projections, AARC Report.

CIESSIN, 2014.
http://www.ciesin.org/documents/Downscaling_CLEARED_000.pdf
+
EMPIRICAL-STATISTICAL
DOWNSCALING (ESD)
Motivation:
Problems with general circulation models (GCMs)on
- Coarse resolutions. Impact studies need finer
resolution for planning
- Grid parameterization schemes not okay for all
regions
- Important feedback of local factor (topography,
geography, etc.) on local and global climate is
missing

and no (or limited) computational resources.

y, geography, etc.) on local and global climate is missing
+
EMPIRICAL-STATISTICAL
DOWNSCALING (ESD)Verification/
Skill Estimates Operational
Large scale
X HINDCAST HINDCAST FORECAST

YDEV=F(XDEV) YVER=F(XVER) YOP=F(XOP)

Local Scale
Y OBSERVATION OBS ?

DEVELOPMENT PERIOD IMPLEMENTATION PERIOD

F: Transfer Function/Calibration/Correction Factor

+
Challenges to statistical downscaling as
used in climate change impacts
prediction
• It is regionally-dependent
• Will the relationship between predictor large-scale
variable and local-scale variable remain the same under
climate change?
• Karl et al. (1990) suggests that atmospheric extremes are
already included in any large enough sample of daily
observations used for training the statistical model
• Busuioc et al. (1999) suggests that a good agreement
between the GCM simulations of future climate and the
statistical model predictions might be indicative of a good
statistical model for using under climate change
+
Simple Bias Correction
http://ccafs-climate.org/bias_correction/
+
Change Factor (Delta Method)
http://ccafs-climate.org/bias_correction/
+ Quantile Based Bias Correction (1)
Fukami and Inomata (2012)
+ Quantile Based Bias Correction (2)
Fukami and Inomata (2012)
 Bjerknes’ 1904 Manifesto

To predict future states of the atmosphere.

We need:

1. A sufficiently accurate knowledge of

the initial state of the atmosphere

2. A sufficiently accurate knowledge of the

laws of physics governing its behaviour.
+ Determinism, Chaos, dan Predictability
 Konsep Determinism: perfect prediction is possible when we have
knowledge of all necessary “condition”

 Konsep Chaos: small difference in the initial state cause huge

difference later, even in the deterministic non-linear system.

 Pengetahuan kita tidak akan pernah lengkap, sehingga:

perfect forecast is imposible!

How well can we predict?  Predictability

+ Uncertainty (1)

State of the art Numerical

Weather Prediction

• Large, sophisticated
computer models that
represent the physics of
the atmosphere,
• deterministic: they do not
represent uncertainty
• But inescapably uncertain, for two
reasons:
• they are not complete and true
representations of the governing
physics
SIMPLIFICATION
• Dynamical chaos: the time evolution
(of the atmosphere) depends very
sensitively on the initial conditions
of the system
INITIAL CONDITION PROBLEM
+
Uncertainty (2)
 Quantifying uncertainty  Probability Distribution

Lower uncertainty

Higher uncertainty
 + Ensemble Prediction System (EPS)
Konsep Prediksi Ensemble
Schematic illustration in ensemble forecasting.
For non-linear f(x)

Average of Single
all evolution of
member average
initial value

• The heavy line represents the evolution of the single best analysis of the initial state of the
atmosphere, corresponding to the more traditional single deterministic forecast.
• The dashed lines represent the evolution of individual ensemble members.
• The ellipse in which they originate represents the probability distribution of initial
atmospheric states, which are very close to each other.
• At the intermediate projection, all the ensemble members are still reasonably similar.
• By the time of the final projection some of the ensemble members have undergone a regime
change, and represent qualitatively different flows.
• Any of the ensemble members, including the solid line, are plausible trajectories for the
evolution of the real atmosphere, and there is no way of knowing in advance which will
represent the real atmosphere most closely
Ensemble Prediction System (EPS)
+
Membangun Ensemble
 Pada dasarnya terdapat dua tipe
prosedur untuk menghasilkan
ensemble:
 Perturbasi di nilai initial condition 
initial condition problem
 Perturbasi pada skema
parameterisasi model  model
simplification problem

 Kombinasi dari keduanya juga

dapat dilakukan untuk
membentuk Multi-model
ensemble

 Atau altenatif lain yang lebih

simple adalah time-lag ensemble Chung Lu, et. Al (2006)
+ Produk Ensemble Prediction System (EPS)
Ensemble mean dan ensemble spread

Guidelines on Ensemble Prediction

Systems and Forecasting
Document WMO-No. 1091
World Meteorological Organization, 2012

Mean (black contours) and spread (colour shading) for PMSL forecast (T
+ 72)
+ Ensemble Climate Projection
http://etccdi.pacificclimate.org/list_27_indices.shtml

ETCCDI Extreme Indices

CDD – Maximum length of dry spell,
maximum number of consecutive days with RR
< 1mm; Indicator of drought

R50mm -- seasonal count of days when

PRCP≥ 50mm

RX1day -- Monthly maximum 1-day

precipitation

Zhang et al. (2011)

 PDFs of CDD Anomalies (TRMM vs.
Model) (Baseline)

SEACLID/CORDEX Southeast Asia simulations

reasonably captured the distribution of CDD for
historical period (1985-2005)
 PDF of anomalous CDD for historical vs. End of
+ Century (Southern SUMATRA)

The PDF of CDD distribution tended to be shifted to

the right for RCP4.5 & 8.5
IPCC AR5 SYR 2014

Sustainable Development
 Result and Applications:
+ Future Climate Analysis
• Some climate analysis using
downscaled precipitation
projection: changes on seasonal
rainfall
 Result and Applications:
+ Changes on extreme rainfall occurrence
BASELINE
2001-2010

FUTURE PROJECTION

• Extreme rainfall on baseline

period (1981-2010): >60
mm/day. Relative probability of
that event was calculated for the
last 10 years of baseline period
(1991-2000)
• Probability of INCREASING
extreme rainfall occurrence
was estimated from 30 member
of future projection.
 Result and Applications:
+ Change on Landslide Hazard
BASELINE FUTURE PROJECTION
1981-2010
2021-2030 2031-2040

• A trigger of landslide: 2 days- HAZARD LEVEL

run wet spell that precedes a
2021-2030 2031-2040
more than 60 mm/day rainfall
 based on historical record
• Average frequency of that
event was calculated over
BASELINE period
• Probability of INCREASING
frequency was estimated for
FUTURE PROJECTION
• HAZARD LEVEL was
estimated from the probability
map
 Result and Applications:
+ Change on Drought Hazard
BASELINE FUTURE PROJECTION
1981-2010 2011-2040

• On baseline • On future
period, a modified period,
Drought Hazard probability of
Index (Shahid and DHI increase
Behrawan, 2008) atleast one
was calculated level relative
based on relative to baseline level
frequency of 15 was calculated
days-run dry spell from ensemble
during rainy season future
(DJF), 4 consecutive projection.
Extreme Dry SPI-3
and SPI6.