ArticlePDF Available

The WCRP CMIP3 multimodel dataset - A new era in climate change research


Abstract and Figures

A coordinated set of global coupled climate model [atmosphere-ocean general circulation model (AOGCM)] experiments for twentieth- and twenty-first-century climate, as well as several climate change commitment and other experiments, was run by 16 modeling groups from 11 countries with 23 models for assessment in the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (AR4). Since the assessment was completed, output from another model has been added to the dataset, so the participation is now 17 groups from 12 countries with 24 models. This effort, as well as the subsequent analysis phase, was organized by the World Climate Research Programme (WCRP) Climate Variability and Predictability (CLIVAR) Working Group on Coupled Models (WGCM) Climate Simulation Panel, and constitutes the third phase of the Coupled Model Intercomparison Project (CMIP3). The dataset is called the WCRP CMIP3 multimodel dataset, and represents the largest and most comprehensive international global coupled climate model experiment and multimodel analysis effort ever attempted. As of March 2007, the Program for Climate Model Diagnostics and Intercomparison (PCMDI) has collected, archived, and served roughly 32 TB of model data. With oversight from the panel, the multimodel data were made openly available from PCMDI for analysis and academic applications. Over 171 TB of data had been downloaded among the more than 1000 registered users to date. Over 200 journal articles, based in part on the dataset, have been published so far. Though initially aimed at the IPCC AR4, this unique and valuable resource will continue to be maintained for at least the next several years. Never before has such an extensive set of climate model simulations been made available to the international climate science community for study. The ready access to the multimodel dataset opens up these types of model analyses to researchers, including students, who previously could not obtain state-of-the-art climate model output, and thus represents a new era in climate change research. As a direct consequence, these ongoing studies are increasing the body of knowledge regarding the understanding of how the climate system currently works, and how it may change in the future.
Content may be subject to copyright.
T he history of climate change
modeling was first characterized
in the 1980s by a number of
distinct groups developing, running,
and analyzing model output from their
own models with little opportunity
for anyone outside of those groups to
have access to the model data. This
was partly a consequence of relatively
primitive computer networking and
data transfer capabilities, along with
the daunting task of collecting and
storing such large amounts X
Open access to
an unprecedented,
comprehensive coordinated
set of global coupled climate
model experiments for twentieth
and twenty-first century climate
and other experiments is changing
the way researchers and
students analyze and learn
about climate.
A New Era in Climate Change Research
1384 SEPTEMBER 2007
of model data (Meehl 1995). Starting in the mid-
1990s, a World Climate Research Programme
(WCRP) committee [now named the WCRP/Climate
Vari abilit y a nd P re dic tability (CL IVAR) Work ing
Group on Coupled Models (WGCM)] organized the
first global coupled climate model intercomparison
exercise whereby modeling groups performed control
runs and idealized 1% yr–1 CO2 increase experiments
(Meehl et al. 1997). A subset of model data was then
collected and archived at the Program for Climate
Model Diagnosis and Intercomparison (PCMDI) and
made available to researchers outside the modeling
groups. Subsequently there were several additional
phases of the Coupled Model Intercomparison Project
(CMIP), termed CMIP2 and CMIP2+ (Meehl et al.
2000, 2005b; Covey et al. 2003). The latter marked
the first time that every field from each model com-
ponent (atmosphere, ocean, land, and sea ice) from
the control and 1% CO2 increase experiments was
collected and made available for analysis. However,
only output from the control runs and 1% CO2 ex-
periments were collected because those represented
the most scientifically straightforward response of
the climate system to an unambiguous change in
external forcing. Limitations in data transfer and
storage still restricted the collection of output from
the early climate change scenario experiments [e.g.,
experiments using the IS92a scenario as described
in the Intergovernmental Panel on Climate Change
(IPCC) Second Assessment Report; Kattenberg et al.
1996]. It was recognized that such an exercise would
ce rta in ly be us eful at s ome s tag e to o pen up t he ou tput
of state-of-the-art climate change scenario experi-
ments for analysis by the wider community.
During t he lead-up to the IPCC Third Assessment
Report (TAR) in the late 1990s, a set of emission
scenarios for twenty-first-century climate was
produced and documented in the Special Report on
Em ission Scena rio s (ty pic ally refe rred to as the SRE S
emission scenarios; Nakicenovic et al. 2000). The
climate modeling community was asked to perform
experiments with these scenarios for assessment
in the TAR. The late date and the large number of
scenarios (numbering about 30 at the time) dictated
that only two (A2 and B2) could be run by a limited
number of groups that had the wherewithal to per-
form such experiments with the associated consider-
able computing requirements on such short notice.
There was litt le time to analyze these data, and only a
few fields were collected and assessed by the authors of
the TAR to illustrate possible future climate changes
(Cubasch et al. 2001; Giorgi et al. 2001). Subsequently,
output from some of these experiments was collected
by the IPCC Data Distribution Centre in Hamburg,
Germany (,
and made available to the climate change impacts
community. But, this still amounted to only a few
models and experiments, and was aimed at a limited
segment of the climate science community.
As planning for the IPCC Fourth Assessment Report
(AR4) commenced in 2003, the climate modeling
community, as represented at the international level
by WGCM, recognized that this process had to be
better organized and carefully coordinated. Not only
must there be more lead time for the modeling groups
to be able to marshal improved model versions and
the requisite computing resources to participate, but
there should also be time and capability for the model
data to be analyzed by a larger group of researchers.
In this way, it was desired that more studies based on
these model experiments could be performed by more
scientists in time for the AR4, thus providing a better
assessment of the state of human knowledge on climate
variability and climate change from the models.
MULTIMODEL DATASET. In consulating with
the IPCC Working Group 1 cochairs, in late 2003
WGCM embarked on a process to coordinate a set
of experiments covering many aspects of climate
variability and change that could be performed by as
many modeling groups as possible with state-of-the-
art global coupled climate models [sometimes referred
to as atmosphere–ocean general circulation models
(AOGCMs)]. The model data were then collected and
made available for analysis (Meehl et al. 2004, 2005b).
However, a crucial part of this effort was to archive
AFFILIATIONS : MEEHL—National Center for Atmospheric
Research,* Boulder, Colorado; COVE Y AND TAYL OR—Program
for Climate Model Diagnosis and Intercomparison, Livermore,
California; DELWORTH AND STOU FFERGeophysical Fluid Dynamics
Labor atory, Princeton, New Jersey; LAT IF Leibniz-Institut fuer
Meereswissenschaften, Kiel, Germany; MCAVAN EYBureau of
Meteorology, Research Centre, Melbourne, Australia; MITCHELL
Hadley Centre, Exeter, United Kingdom
*The National Center for Atmospheric Research is sponsored by
the National Science Foundation
CORRESPONDING AUTHOR: Gerald A. Meehl, National Center
for Atmospheric Research, P.O. Box 3000, Boulder, CO 80307
E-mail: meehl
The abstract for this article can be found in this issue, following the
table of contents.
In fin al form 19 March 2007
©2007 American Meteorological Society
and actually organize the data so that they were readily
available to the international climate science com-
munity for analysis. PCMDI agreed to take on this
considerable challenge, which was destined to be the
third phase of CMIP, or CMIP3. PCMDI’s role proved
to be crucial in CMIP3, the largest international global
coupled climate model experiment and multimodel
analysis effort ever attempted.
The list of experiments included the following
(single realizations were acceptable, but modeling
groups were encouraged to run multimember
1) Twentieth-century simulation to year 2000 (prefer-
able starting from pre-industrial conditions in the
late 1800s) with anthropogenic and natural forc-
ings as modeling groups deemed appropriate;
2) Climate change experiment: Twenty-first-century
simulation with SRES B1 (low forcing, i.e., CO2
concentration about 550 ppm by 2100) from 2000
to 2100;
3) Climate change experiment: Twenty-first-century
climate change simulation with SRES A1B
(medium forcing, i.e., CO2 concentration of about
700 ppm by 2100) from 2000 to 2100;
4) Climate change experiment: Twenty-first-century
simulation with SR ES A2 (high forcing, i.e., CO2
concentration about 820 ppm by 2100) from 2000
to 2100;
5) Climate change commitment experiment: Fix
all concentrations at year 2000 values and run to
2100 (CO2 ~ 360 ppm);
6) Climate change commitment experiment: Fix all
concentrat ions at ye ar 210 0 va lues for B1 a nd run
to 2200 (CO2 ~ 550 ppm);
7) Climate change commitment experiment: Fix all
concentrations at year 2100 values for A1B and
run to 2200 (CO2 ~ 700 ppm);
8) Idealized forcing and stabilization experiment:
1% yr–1 CO2 increase to doubling at year 70 with
corresponding control run, and an additional
150 yr with CO2 fixed at 2 × CO2;
9) Idealized forcing and stabilization run: 1% yr–1
CO2 increase run to quadrupling with an addi-
tional 150 yr with CO2 fixed at 4 × CO2,
10) 100-yr (minimum) control run with all forcings
held constant encompassing same time period as
in 1 above;
11) Climate sensitivity experiment: Instantaneously
double CO2 and run to equilibrium with atmo-
sphere coupled to a nondynamic slab ocean [also
as input to the Cloud Forcing Model Intercom-
parison Project (CFMIP)];
12) Extend one A1B and B1 climate change commit-
ment experiment simulation to 2300.
A fundamental part of the earlier phases of CMIP
(described above) was the idealized 1% yr–1 CO2
increase experiments, so those were also retained
in the CMIP3 list above as standard calibration
runs to better intercompare the coupled models
responses. These experiments were also necessary
to calculate the transient climate response (TCR),
defined as the globally averaged surface air tem-
perature increase at the time of CO2 doubling in
a 1% yr–1 compound CO2 increase experiment, a
standard metric to assess the coupled transient
response. Equilibrium climate sensitivity, another
standard metric for comparing model responses,
was also obtained from the atmosphere coupled to
the nondynamic slab ocean equilibrium 2 × CO2
An extensive list of fields was requested to be
supplied to PCMDI by the modeling groups. The
volume of model data was so large that, in the
international context, conventional online data
transfer mechanisms became impractical. Therefore,
modeling groups were sent hard disks and asked
to copy their model data onto the disks in netCDF
format and then mail the disks to PCMDI where the
model data were downloaded and cataloged.
To provide an idea of the model outputs that were
collected, we summarize here in general terms the
types of model variables furnished by the modeling
groups. For a full list of fields that were requested
with detailed descriptions of the variables, see
High-priority f ields are as follows (a few examples
of each are given in parentheses; note there were
additional low-priority fields requested as well that
are not listed here):
Mon thly mea n 2D a tmo sphe re or lan d su rfa ce d ata
(e.g., surface temperature, precipitation, sea level
pressure, soil moisture);
Time-independent 2D land surface data (e.g.,
orography, land area fraction);
Monthly mean 3D atmosphere data (e.g., air
temperature, winds, geopotential heights);
Monthly mean 1D ocean data (e.g., northward
ocean heat transport);
Monthly mean 2D ocean data (e.g., ocean meridi-
onal overturning streamfunction);
Monthly mean 0D or 2D ocean or sea ice data (e.g.,
sea surface height, sea level, sea ice fraction, sea ice
1386 SEPTEMBER 2007
Time-independent 2D ocean data (e.g., ocean
bottom topography);
Monthly mean 3D ocean data (e.g., temperature,
salinity, ocean currents);
Daily mean 2D atmosphere data (e.g., surface
air temperature, precipitation, sea level pressure,
winds, surface energy balance components);
Daily mean 3D atmosphere data (e.g., air tempera-
ture, winds);
Three hourly 2D atmosphere data (e.g., surface
air temperature, precipitation, sea level pressure,
winds, surface energy balance components);
Extremes indexes (calculated from daily data,
five temperature-related indices, five precipita-
tion-related indices), from Frich et al. (2002, their
Table 1).
The modeling groups proceeded to complete as many
of the experi ments as they could manage during 2004.
By early 2005, a total of 16 modeling groups from 11
countries participated with 23 models. Considerable
resources (human and computing) were devoted to
this project. PCMDI collected and archived more
than 30 TB of model data by that time (www-pcmdi. Subsequently, another
modeling group has contributed data to CMIP3, so
that the WCRP CMIP3 multimodel dataset now
consists of 17 modeling groups from 12 countries
and 24 models. Figure 1a shows the current break-
down of models and experiments in the WCRP
CMIP3 multimodel dataset, and Fig. 1b indicates the
ensemble members that were submitted by the groups
for each experiment. As shown in Fig. 1, monthly
means were generally collected, with some daily
data and even some 6-hourly data generated to drive
regional models and for other applications.
PHASE OF CMIP3. The WGCM Climate Simula-
tion Panel (G. Meehl, chair, C. Covey, T. Delworth,
M. Latif, B. McAvaney, J. Mitchell, and R. Stouffer)
coordinated the collection of the model data, and
then undertook organizing the analysis phase in
2004 when sufficient model data had been archived
to allow the initiation of analysis projects. Several
announcements were made first by e-mail, and then
in generally read publications that would reach the
climate science community (e.g., Meehl et al. 2004).
Since the schedule would be tight for analyses to
be done and submitted for publication in time to
be assessed for the AR4, it was decided to hold a
workshop in early 2005 where preliminary results
from the analyses could be presented. By late 2004
nearly 300 scientists had registered to have access
to the multimodel dataset, but it was unclear how
many would actually have completed enough work
to present results at the workshop.
Meanwhile, to encourage participation of
U.S. scientist s, the U.S. Climate Variabilit y and
Predictability program (CLIVAR) made a significant
contribution by coordinating the Coupled Model
Evaluation Project (CMEP) that resulted in
multiagency funding for 21 analysis projects (www.
Results from analyses of the multimodel data-
set were presented by 125 scientists from all over
the world at the workshop that was convened and
or gan ized b y U.S . CL IVAR and WGCM and hos ted by
the International Pacific Research Center (University
of Hawaii) on 1–4 March 2005 (http://ipcc-wgl.ucar.
edu/meeting/CMSAW/). These results were intended
to feed directly into the AR4 process. To be assessed
as par t of t he A R4, it was intend ed t hat p apers shou ld
be submitted to peer-reviewed journals by late
spring 2004. Many of the participants at the Hawaii
workshop, as well as a number of others, ended up
submitting nearly 200 papers for assessment. This
was judged to be a considerable success, given the
tight time frame and the fact that most scientists
performed these analyses without additional funding
or resources over and above what they already had in
place (the exception being the CMEP investigators).
Si nce t hen , add iti onal pa pers have b een prepared and
submitted, and from those submitted papers over 200
have already appeared in the peer-reviewed literature,
with many more either in the review process or in
preparation (
DATASET. Though it is beyond the scope of this
short summary article to provide a comprehensive
review of all the analyses published to date, we
choose here to select a few illustrative examples to
provide an idea of the types of analyses that have
been performed. Figure 2 shows globally averaged
surface air temperature time series from the experi-
ments compiled directly from the archived model
data (
The numbers in the figure indicate how many
models completed each phase of the experiments in
ti me to be assesse d in the IPCC AR4 (the nu mber of
ensemble members for each experiment and model
is shown separately in Fig. 1b). The shading is ±one
standard deviation of the intermodel variability.
This figure depicts the largest number of AOGCMs
that have ever b een assemble d to s imulat e twent iet h-
and twenty-first-century climate and climate change
commitment. Results show that the widely quoted
observed twentieth-century warming of about 0.6°C
(e.g., Trenberth et al. 2007) is well simulated by the
FIG. 1. (a) Summar y of climate model experiments per formed with AOGCMs in the multimodel archive.
Colored fields indicate that some but not necessarily all variables of the specific data type (separated by
climate system component and time interval) have been archived at PCMDI. Where different shadings
are given in the legend, the color indicates whether single or multiple ensemble members are available.
(b) Number of ensemble members performed for each experiment and each scenario. Details on the
scenarios, variables, and models can be found at the PCMDI Web page ( /about _
ipcc.php). Note that some of the ensemble members using the CCSM3 were run on the Earth Simulator
in Japan in collaboration with the Central Research Institute of Electric Power Industry (CRIEPI).
1388 SEPTEMBER 2007
models. For the twenty-first century (computed
as the difference of mean temperature for years
2090–99 minus 1980–99), the models show an aver-
age warming of 1.8°, 2.8°, and 3.4°C for the low (B1),
medium (A1B), a nd high (A2) forcing scenarios,
respectively. For the commitment experiments, by
2100 the climate system warms by about an addi-
tional 0.6°C after concentrations are stabilized in
2000 while, for the other two commitment experi-
ments, by 2200 there is about another half-degree
warming over and above what occurred by 2100 in
B1 and A1B, respectively.
The availability of such a large number of models
provides considerable opportunity to explore model
simulation capability of va rious aspects of twentieth-
century climate (e.g., see publications listed at www. One example in
Fig. 3 shows the first and second EOFs of Antarctic
sea ice concentration from observations (Figs. 3a,b),
and also for a number of models’ simulations of
sea ice concentration. [Note that EOFs depict the
principal spatial patterns of variability (see, e.g.,
Kutzbach 1967)]. Though each model has its own
characteristic sea ice variability pattern, all show a
dipole with negative values in the Atlantic sector,
and positive values in the Pacific sector (Holland
and Raphael 2006). This type of quantification of
model simu lat ion c apabil ity of w hat we ha ve alre ady
observed provides a baseline for the degree of confi-
dence we can place in the models and how they may
simulate future changes. In this case, the overall
agreement in the basic pattern of variability between
the models and the observations builds confidence
that sea ice variability in a future warmer climate
can be usefully studied.
The CMIP3 multimodel dataset has also been
used to help understand climate changes that have
already been observed during the twentieth century.
For example, model results for the twentieth century
have been analyzed, in concert with additional single
forcing datasets from some of the models, to show
that the signature from large volcanic eruptions, such
as Krakatoa in the late nineteenth century, persist
and are manifested by reduced ocean heat content
for decades after the event. This offsets, to a certain
extent, the positive radiative forcing and associated
warming that would otherwise have occurred due to
increasing greenhouse gases in the early twentieth
century (e.g., Delworth et al. 2005; Gleckler et al.
Another way that the CMIP3 multimodel dataset
has been useful in interpreting observed climate
variability and trends over the latter part of the
twentieth century is demonstrated in a comparison
of different time scales of tropospheric and surface
temperature variability from the multimodel simu-
lations of twentieth-century climate to observed
quantities from satellites, radiosondes, and surface
weather stations (Santer et al. 2005). Figures 4a
and 4b show the relationship of variability on the
monthly time scale between globally averaged
surface temperature (x axis) and weighted estimates
of tropospheric temperature variability (y axis). The
colored symbols are results from 49 separate realiza-
tion s from 19 AO GCMs from t he mul timodel data set
for twentieth-century climate that included combi-
nations of anthropogenic and natural forcings. The
black symbols denote different observed radiosonde
and satellite datasets paired with two observed
surface temperature datasets. Figures 4c and 4d are
the same as Figs. 4a and 4b, but for trends from 1979
to 1999 in the models and observations. Note that in
all panels, the observations fall along a regression
line relating the model results, with the location of a
particular model realization on that regression line
depending mostly on simulated El Niño amplitude
in the models. The regression line lies above the
black line (which has a slope of 1.0), indicating that
FIG. 2. Multimodel means of surface warming for the
twent y-first centur y for the scenarios A2 , A1B, and B1,
and corresponding twentieth-century simulations. Values
beyond 2100 are for the climate change commitment
experiments that stabilized concentrations at year
2100 values for B1 and A1B. Linear trends from the
corresponding control runs have been removed from
these time series. Lines show the multimodel means,
and shading denotes the ± 1 std dev intermodel range.
Discontinuities between different periods have no physi-
cal meaning due to the fact that the number of models
run for a given scenario is different for each period and
scenario, as indicated by the numbers given for each
phase and scenario in the bottom part of the panel.
there is enhancement of the magnitude of tem-
perature variability in the troposphere compared
to the surface in both the models and observations.
However, Figs. 4c and 4d show the relationship
among trends has less agreement between models
and observations. Therefore, either there are dif-
ferent physics operating at monthly and trend
time scales in the observations (whereby there can
somehow be good agreement on the monthly time
scale and less agreement on the trend time scale), or
this result points to the difficulties of constructing
accurate small trends with disparate observed data
with associated discontinuities in observing systems
over time.
Regarding changes of future climate a question
that is frequently asked is, What will El Niño do in
the future? Several ana lyses of the multimodel dataset
have been performed to address that question (e.g.,
Guilyardi 2006; Meehl et al. 2006; Merryfield 2006),
and Fig. 5 summarizes results from one such study
by van Oldenborgh et al. (2005). This figure attempts
to address the question of what future amplitude of
El Niño events could be in a future warmer climate
depicted in the multimodel dataset. Figure 5 clearly
shows a wide range of possible future behaviors
across the various models, agreeing with the other
studies cited above that there is no clear indication
from the models regarding future changes in El Niño
amplitude. This model dependence is the result of
several factors, not least of which is that no two
observed El Niño events are alike, and that different
models capture various aspects of the mechanisms
thought to produce El Niño events.
In several of the El Niño studies cited above, the
authors attempted to subselect models that more
faithfully simulated various metrics that applied to
FIG. 3. (a) First two EOFs from observed winter sea ice concentration (1979–99) scaled by the std dev
of the corresponding PC time series. Contour interval is 5%, 0 contour omitted, and negative values
are shaded. (b) First EOF of winter sea ice concentration from AOGCM simulations of the twentieth
centur y from 1960 to 1999 using linearly detrended data (Holland and Rafael 2006) .
FIG. 5. Relative change in
El Niño magnitude (first
EOF of detrended month-
ly SST in the region 10°S–
10° N, 12E–9W) in the
CMIP3 multimodel dataset.
The more reliable models,
as defined by their ability to
simulate several different
El Niño metrics in the current
climate, are dark red (after
van Oldenborgh et al. 2005).
1390 SEPTEMBER 2007
FIG. 4. (a), (b) Information is provided on amplification of the monthly time scale surface temperature
variability in two weighted tropospheric temperature products as defined in Santer et al. (2005). (c),
(d) Same as in (a), (b), but depicting the relation between decadal time-scale trends at the surface
and in the troposphere. The colored symbols in each panel indicate realizations from 49 ensemble
members for twentieth-century climate simulations from 19 AOGCMs from the multimodel archive.
The fitted regression lines (in red) are based on model data only. The black lines denote a slope of
1.0. Values above the black lines indicate tropospheric enhancement, and values below the black
line indicate tropospheric damping of surface temperature changes. Black symbols indicate results
from separate radiosonde and satellite MSU data paired with two surface temperature datasets.
The blue shading in (c) and (d) defines the region of simultaneous surface warming and tropospheric
cooling. Results are for the deep Tropics (20°N to 20°S), and are more fully described in Santer et al.
FIG. 6. Evolution of the Atlantic MOC as defined by
the maximum overturning at 24°N for the period
1900–2100 using 21 realizations of the response to
the A1B emissions scenario from nine AOGCMs. The
MOC simulations with a skill score larger than one are
solid lines; those from models with a smaller skill score
are dashed. The weighted ensemble mean is shown by
the thick black curve together with the weighted std
dev (thin black lines) . Observational estimates of the
circulation at 24°N [15.75 ± 1.6 Sv (1 Sv = 106 m3 s–1);
Ganachaud and Wunsch 2000 ; Lumpkin and Speer
2003] at the end of the last centur y are shown as the
red cross centered at year 1989. (top) The weighted
(solid) and unweighted (dashed) std devs (from
Schmittner et al. 2005).
observed El Niño phenomena. In another variation
on that technique, Schmittner et al. (2005) studied
possible future changes of the ocean meridional over-
turning circulation (MOC) in the Atlantic by more
heavily weighting models that accurately simulated
certain hydrographic properties and observation-
based circulation estimates (Fig. 6). Using 28 simula-
tions from 9 different AOGCMs from the multimodel
dataset, S chm itt ner et a l. (2005) were able to come u p
with a best estimate of projected MOC behavior using
the weighted model results. Their results indicate a
gradual projected reduction in the amplitude of the
MOC over the course of the twenty-first century
for the A1B emission scenario, finally amounting
to a weakening of 25% (±25%) by the year 2100. No
model shows a sudden shutdown of the MOC during
the twenty-first century. These results agree with an
assessment of a larger number of models from the
WCRP CMIP3 multimodel dataset in the IPCC AR4
(Meehl et al. 2007).
As noted above, modeling groups were asked to
calculate and submit 10 indexes of extreme weather
and climate events outlined by Frich et al. (2002).
Five of the indexes related to temperature, and five
to precipitation. In all, nine of the modeling groups
completed those calculations. Tebaldi et al. (2006)
analyzed results from those data, and Fig. 7 shows
results from two of the precipitation indices in terms
of global averages and geographical changes for the
end of the twenty-first century for the A1B scenario
from nine models. Precipitation intensity increases
almost everywhere (for a given event more precipita-
tion occurs in the future), but dry days (number of
days in between precipitation events) also increase
in some areas. This seems counterintuitive, but in
some regions, particularly where circulation and
other climate changes are associated with reduced
average precipitation (e.g., Meehl et al. 2005a), there
is a longer time period between precipitation events,
but when it does rain it rains harder.
Finally, the sheer number of AOGCMs contrib-
uting to the WCRP CMIP3 multimodel dataset has
allowed some of the f irst calculations of probabilistic
climate change information. For example, using
the techniques outlined in Furrer et al. (2007b),
Figs. 8a and 8b from Furrer et al. (2007a) show, for
21 models from the multimodel dataset for the A1B
scenario, seasonal [December–February (DJF) and
June–August (JJA)] values of temperature increases
with an 80% chance of occurrence by the end of the
twenty-first century. Conversely, Figs. 8c and 8d
show contours of probabilities of the occurrence of at
least a 2°C warming for the two seasons. The results
in Fig. 8 were obtained using a technique employed
in Furrer et al. (2007a) wherein probability density
func tions (PDFs) of temperat ure cha nge at each grid
point are computed from the multimodel dataset.
This is done by first calculating the temperature
differences from each member of the multimodel
ensemble, averaged for A1B for 2080–99 minus
1980–99 for DJF and JJA, and regressing those
differences upon basis functions, that is, a series of
fields that are chosen as starting points to explain
the possible common large-scale patterns of the
climate change signal. A statistical model is then
formulated through a hierarchical Bayes framework,
and a Markov chain Monte Carlo calculation then
estimates the true coefficients of the regression and
the uncertainty around them, plus estimates of the
Weighting by the relative agreement among the
models [such as that used in another technique that
produces probabilistic climate change information
by region by Tebaldi et al. (2004)] is not assumed in
this method. By recombining the coefficients with
the basis functions, an estimate is derived of the true
FIG. 7. Changes in extremes based on multimodel simulations from nine global coupled climate mod-
els, from Tebaldi et al. (2006). (a) Globally averaged changes in precipitation intensity (defined as the
annual total precipitation divided by the number of wet days) for a low (B1), middle (A1B), and high
(A2) forcing scenarios. (b) Changes of spatial patterns of precipitation intensity based on simulations
between two 20-yr means (2080–99 minus 1980–99) for the A1B scenario. (c) Globally averaged changes
in dry days (defined as the annual maximum number of consecutive dry days). (d) Changes of spatial
patterns of dry days based on simulations between two 20-yr means (2080 –99 minus 1980–99) for the
A1B scenario. Solid lines in (a) and (c) are the 10-yr smoothed multimodel ensemble means; the envelope
indicates the ensemble mean standard deviation. Stippling in (b) and (d) denote areas where at least 5
of the 9 models concur in determining that the change is statistically significant. Extremes indices are
calculated following Frich et al. (2002) and are shown for land points only. Each model’s time series has
been centered around its 1980–99 average and normalized (rescaled) by its std dev computed (after
detrending) over the period 1960–2099 ; then the models were aggregated into an ensemble average,
both at the global average and at the grid-box level (units are std devs).
1392 SEPTEMBER 2007
climate change field and of the uncertainty around
it. Probability density functions of the temperature
change are then derived for each grid point over the
entire globe, and represent the joint probability of a
given warming at each grid point.
This and the other studies shown above are only
a few examples of the many more published results
from the analyses of the multimodel dataset that
can be seen online at
CONCLUSIONS. An unprecedented interna-
tional effort to run a coordinated set of twentieth- and
twenty-first-century climate simulations, as well as
several climate change commitment experiments, was
organized by the WCRP/CLIVAR WGCM for assess-
ment in the IPCC AR4. Model data were collected,
archived, and made available to the international
climate science community by PCMDI. This is the
first time such a large set of AOGCM climate change
simulations has been made openly available for
analysis. As such, it represents a new era in climate
science research whereby researchers and students can
obtain permission to access and analyze the AOGCM
data. Such an open process has allowed hundreds
of scientists from around the world, many students,
FIG. 8. Probabilistic climate change results from 21 AOGCMs, 2080–99 compared to 1980–99, for the A1B
scenario, converted to a common 5° lat–lon grid: (a) DJF and (b) JJA values of temperature increase with an
80% chance of occurrence by the end of the twenty-first century. Also shown are contours of probabilities of
the occurrence of at least a 2°C warming for (c) DJF and (d) JJA (Furrer et al. 2007a).
and researchers from developing countries, who had
never before had such an opportunity, to analyze the
model data and make significant contributions not
only to the IPCC AR4, but to human knowledge of the
workings of climate variability and climate change.
This unique and valuable multimodel dataset will be
maintained at PCMDI and overseen by the WGCM
Climate Simulation Panel for at least the next several
years. It will serve as a resource for climate science
that promises to change the way students, developing
country scientists, and experienced climate scientists
perform analyses and learn about the climate system.
For instructions regarding how to obtain access the
multimodel dataset, see
ACKNOWLEDGMENTS. We acknowled ge the
internat ional modeling groups for providing their data for
analysis, the Program for Climate Model Diagnosis and
Intercomparison (PCMDI) for collecting and archiving
the model data, the WCRP/CLIVAR Working Group on
Coupled Models (WGCM), a nd their Coupled Model Inter-
compari son Project (CMIP) and Cl imate Simulation Pa nels
for organi zing the model data a nalysis act ivity, and the IPC C
WG1 TSU for technical support. The CMI P3 Mult i-Model
Data Archive at Lawrence Livermore National Laborator y
is supported by the Office of Science, U.S. Department
of Energy. Portions of this study were supported by the
Office of Science (BER), U.S. Department of Energy,
Cooperative Agreement No. DE-FC02-97ER62402, and
the National Science Foundation. The National Center
for Atmospheric Research is sponsored by the National
Science Foundation.
Covey, C., K. M. AchutaRao, U. Cubasch, P. Jones, S. J.
La mbe rt, M. E . Ma nn, T. J. Phill ips, an d K. E. Taylo r,
2003: An overview of results from the Coupled Model
Intercomparison Project. Global Planet. Change, 37,
Cubasch, U., and Coauthors, 2001: Projections of
future climate change. Climate Change 2001: The
Scientific Basis, J. T. Hou ght on et al., E ds. , Ca mbr idge
Universit y Press, 525–582.
Delworth, T. L., V. Ramaswamy, and G. L. Stenchikov,
2005: The impact of aerosols on simulated ocean
temperature and heat content in the 20th century.
Geophys. Res. Lett., 32, L24709, doi:10.1029/
Frich, P., L. V. Alexander, P. Della-Marta, B. Gleason,
M. Haylock, A. M. G. Klein Tank, and T. Pearson,
2002: Observed coherent changes in climatic ex-
tremes during the second half of the twentieth
century. Climate Res., 19, 193–212.
Furrer, R., R. Knutti, S. Sain, D. Nychka, and G. A.
Meehl, 2007a: Spatial patterns of probabilistic tem-
perature change projections from a multivariate
Bayesian analysis of climate model output. Geophys.
Res. Lett., 34, L06711, doi:10.1029/2006GL027754.
, S. R. Sain, D. Nychka, and G. A. Meehl, 2007b:
Multivariate Bayesian analysis of atmosphere–ocean
general circulation models. Environ. Ecol. Stat., in
Ganachaud, A., and C. Wunsch, 2000; Improved
estimates of global ocean circulation, heat transport
and mixing from hydrographic data. Nature, 408,
Giorgi, F., and Coauthors, 2001: Regional climate infor-
mation—Evaluat ion and projections. Climate Change
2001: The Scientific Basis, J. T. Houghton et al., Eds.,
Cambridge University Press, 583–638.
Gleckler, P. J., T. M. L. Wigley, B. D. Santer, J. M. Gregory,
K. AchutaRao, a nd K. E. Taylor, 2006 : Volcanoes and
climate: Krakatoa’s signature persists in the ocean.
Nature, 439, 675.
Guilyardi, E., 2006: El Niño–mean state–seasonal cycle
interactions in a multi-model ensemble. Climate
Dyn., 26, 329–348.
Holland, M. M., and M. N. Raphael, 2006: Twentieth
century simulation of the Southern Hemisphere
climate in coupled models. Part II: Sea ice condit ions
and variability. Climate Dyn., 26, 229–245.
Kattenberg, A., and Coauthors, 1996 : Climate models—
Projections of future climate. Climate Change 1995:
The Science of Climate Change, J. T. Houg hton et al.,
Eds., Cambridge University Press, 285–358.
Kutzbach, J., 1967: Empirical eigenvectors of sea level
pressure, surface temperature and precipitation
complexes over North America. J. Appl. Meteor., 6,
Lumpkin, R., and K. Speer, 2003: Large-scale vertical
and horizontal circulation in the North Atlantic
Ocean. J. Phys. Oceanogr., 33, 1902–1920.
Meehl, G. A., 1995: Global coupled general circulation
models. Bull. Amer. Meteor. Soc., 76, 951–957.
, G. J. Boer, C. Covey, M. Latif, and R. J. Stouffer,
1997: Intercomparison makes for a better climate
model. Eos, Trans. Amer. Geophys. Union, 78,
445–446, 451.
, and
, 2000: The Coupled
Model Intercomparison Project (CMIP). Bull. Amer.
Meteor. Soc., 81, 313 –318.
, C. Covey, M. Latif, B. McAvaney, J. F. B. Mitchell,
and R. Stouffer, 2004: Soliciting participation in
climate model analyses leading to IPCC Fourth
Assessment Report. Eos, Trans. Amer. Geophys.
Union, 85, 274.
, J. M. Arblaster, and C. Tebaldi, 2005a: Under-
standing future patterns of precipitation extremes
in climate model simulations. Geophys. Res. Lett.,
32, L18719, doi:10.1029/2005GL023680.
, C. Covey, B. McAvaney, M. Latif, and R. J. Stouffer,
2005b: Overview of the Coupled Model Intercompari-
son Project. Bull. Amer. Meteor. Soc., 86, 89–93.
, H. Teng, and G. W. Branstator, 2006: Future
changes of El Nino in two global coupled climate
models. Climate Dyn., 26, 549–566.
, and Coauthors, 2007: Global climate projections.
Climate Change 2007: The Physical Science Basis,
Cambridge University Press, in press.
Merryfield, W., 2006: Changes to ENSO under CO2
doubling in the IPCC AR4 coupled climate models.
J. Climate, 19, 4009–4027.
Nakicenovic, N., and Coauthors, 2000: IPCC Special
Report on Emissions Scenarios. Cambridge University
Press, 599 pp.
Santer, B. D., and Coauthors, 2005: Amplification of
surface temperature trends and variability in the
tropical atmosphere. Science, 309, 1551–1556.
Schmittner, A., M. Latif, and B. Schneider, 2005: Model
projections of the North Atlantic thermohaline
circulation for the 21st century assessed by observa-
tions. Geophys. Res. Lett., 32, L23710, doi:10.1029/
Tebaldi, C., L. O. Mearns, D. Nychka, and R. L.
Smith, 2004: Regional probabilities of precipitation
change: A Bayesian analysis of multimodel simula-
tions. Geophys. Res. Lett., 31, L24213, doi:10.1029/
, J. M. Arblaster, K. Hayhoe, and G. A. Meehl,
2006: Going to the extremes: An intercomparison
of model-simulated historical and future changes in
extreme events. Climatic Change, 79, 185–211.
Trenberth, K. E., and Coauthors, 2007: Observations:
Surface and atmospheric climate change. Climate
Change 2007: The Physical Science Basis, Cambridge
University Press, in press.
van Oldenborgh, G., S. Philip, and M. Collins, 2005:
El Niño in a cha nging climate: A multi-model study.
Ocean Sci., 1, 81–95.
1394 SEPTEMBER 2007
... General Circulation Model (GCM) simulations as part of Coupled Model Intercomparison Project (CMIP) are essential for providing quantitative climate projections for the twenty-first century [5]. The fourth Intergovernmental Panel on Climate Change (IPCC) assessment report was created using the CMIP phase 3 (CMIP3) GCM simulations [6]. The fifth assessment report (AR5) of the IPCC was created using the CMIP5 [7]. ...
Full-text available
In the coming years, climate change is predicted to impact irrigation water demand considerably, particularly in semi-arid regions. The aim of this research is to investigate the expected adverse impacts of climate change on water irrigation management in Saudi Arabia. We focus on the influence of climate change on irrigation water requirements in the Al Quassim (97,408 ha) region. Different climate models were used for the intermediate emission SSP2-4.5 and the high emission SSP5-8.5 Coupled Model Intercomparison Project Phase 6 (CMIP6) scenarios. The FAO-CROPWAT 8.0 model was used to calculate reference evapotranspiration (ETo) using weather data from 13 stations from 1991 to 2020 and for both the SSP2-4.5 and SSP5-8.5 scenarios for the 2040s, 2060s, 2080s, and 2100s. The findings indicated that, for the 2100s, the SSP2-4.5 and SSP5-8.5 scenarios forecast annual average ETo increases of 0.35 mm/d (6%) and 0.7 mm/d (12.0%), respectively. Net irrigation water requirement (NIWR) and growth of irrigation water requirement (GIWR) for the main crops in the Al Quassim region were assessed for the current, SSP2-4.5, and SSP5-8.5 scenarios. For SSP5-8.5, the GIWR for the 2040s, 2060s, 2080s, and 2100s are expected to increase by 2.7, 6.5, 8.5, and 12.4%, respectively, compared to the current scenario (1584.7 million m3). As a result, there will be higher deficits in 2100 under SSP5-8.5 for major crops, with deficits of 15.1%, 10.7%, 8.3%, 13.9%, and 10.7% in the crop areas of wheat, clover, maize, other vegetables, and dates, respectively. Optimal irrigation planning, crop pattern selection, and modern irrigation technologies, combined with the proposed agroecological zones and NIWR values, can support water resources management. The findings can assist managers and policymakers in better identifying adaptation strategies for areas with similar climates.
... Biasutti et al. (2012) note a 30 % decrease specific to spring and summer rainfall. Girvetz et al. (2009) andMeehl et al. (2007) project a decrease in Puerto Rico's annual precipitation, somewhere between 10 % and 30 %. Magrin et al. (2007), from the Fourth Assessment Report of the IPCC, project Latin America's annual precipitation to change any-where from −40 % to +10 %. Cashman et al. (2010) find that the Caribbean will exhibit drying trends throughout the spring, summer, and fall; however, they project increased precipitation during the winter months, while Angeles et al. (2007) project a wetter climate for the Caribbean through 2050, driven heavily by increased rainfall during the rainy season, most likely a result of rising sea-surface temperatures which bolster large storm systems, potentially leading to an increase in the scale of extreme precipitation events. ...
Full-text available
General circulation model (GCM) outputs are a primary source of information for climate change impact assessments. However, raw GCM data rarely are used directly for regional-scale impact assessments as they frequently contain systematic error or bias. In this article, we propose a novel extension to standard quantile mapping that allows for a continuous seasonal change in bias magnitude using localized regression. Our primary goal is to examine the efficacy of this tool in the context of larger statistical downscaling efforts on the tropical island of Puerto Rico, where localized downscaling can be particularly challenging. Along the way, we utilize a multivariate infilling algorithm to estimate missing data within an incomplete climate data network spanning Puerto Rico. Next, we apply a combination of multivariate downscaling methods to generate in situ climate projections at 23 locations across Puerto Rico from three general circulation models in two carbon emission scenarios: RCP4.5 and RCP8.5. Finally, our bias-correction methods are applied to these downscaled GCM climate projections. These bias-correction methods allow GCM bias to vary as a function of a user-defined season (here, Julian day). Bias is estimated using a continuous curve rather than a moving window or monthly breaks. Results from the selected ensemble agree that Puerto Rico will continue to warm through the coming century. Under the RCP4.5 forcing scenario, our methods indicate that the dry season will have increased rainfall, while the early and late rainfall seasons will likely have a decline in total rainfall. Our methods applied to the RCP8.5 forcing scenario favor a wetter climate for Puerto Rico, driven by an increase in the frequency of high-magnitude rainfall events during Puerto Rico's early rainfall season (April to July) as well as its late rainfall season (August to November).
... In the present paper, we reconsider four precipitation ECs which were originally identified using previous generations of model ensembles (CMIP3 and CMIP5, Meehl et al. 2007), and analyze to what extent they are still verified in CMIP6-including a comparison with CMIP5, checking both their existence in the new ensemble and their robustness as suggested by Simpson et al. (2021). The majority of the ECs considered in this study do not pass this test, confirming that the identification of Emergent Constraint mechanisms for hydrological variables requires extensive validation and often does not survive in the latest generation of models. ...
Full-text available
An Emergent Constraint (EC) is a physically-explainable relationship between model simulations of a past climate variable (predictor) and projections of a future climate variable (predictand). If a significant correlation exists between the predictand and the predictor, observations of the latter can be used to constrain model projections of the former and to narrow their uncertainties. In the present study, the EC technique has been applied to the analysis of precipitation, one of the variables most affected by model uncertainties and still insufficiently analysed in the context of ECs, particularly for the recent CMIP6 model ensemble. The main challenge in determining an EC is establishing if the relationship found is physically meaningful and robust to the composition of the model ensemble. Four precipitation ECs already documented in the literature and so far tested only with CMIP3/CMIP5, three of them involving the analysis of extreme precipitation, have been reconsidered in this paper. Their existence and robustness are evaluated using different subsets of CMIP5 and CMIP6 models, verifying if the EC is still present in the most recent ensemble and assessing its sensitivity to the detailed ensemble composition. Most ECs considered do not pass this test: we found one EC not to be robust in both CMIP5 and CMIP6, other two exist and are robust in CMIP5 but not in CMIP6, and only one is verified and is robust in both model ensembles.
... Tejavath et al. (2019) also reported the weakening and westward shifting of the Walker circulation in the twentieth century. The weakening is thought to be linked partly to global warming, and partly with increase in the frequency of El Niño and a natural reduction in the frequency of La Niña (Meehl et al. 2007;Power and Smith 2007;Power and Kociuba 2011;Vecchi et al. 2006;Vecchi and Wittenberg 2010). ...
Full-text available
We observe significant changes in the ENSO-Indian summer monsoon rainfall (ISMR) relationship in past three multidecadal epochs (early epoch: 1931-1960, middle epoch: 1961-1990, and recent epoch: 1991-2015) based on consistent correlation. The rainfall during early epoch was above normal, however, in other two epochs it was relatively dry. The ENSO-ISMR relationship and its changes are mostly pronounced in the monsoon core zone. During the early epoch, the effect of La Niña was more dominant by accounting higher number of La Niña events with higher values of Niño 3.4 index, which can be attributed to the surplus rainfall over most regions of India. The rainfall during the consecutive two multidecadal epochs is below normal and this may be attributed to the strong El Niño events. The dryness in the northeast region cannot be linked with ENSO events as it has insignificant relationship with the rainfall over this region. The weakening of Low Level Jet during the recent epoch can be a reason to reduce monsoon rainfall during recent periods. It is evident that the reduced moisture transport also accounts for the reduced precipitation. The changes in equatorial Walker and regional Hadley circulations during the three epochs show the same mode of variability and in phase with the changes in rainfall and other circulation related parameters. This study emphasizes the basin mode warming of world oceans and the increased frequency of El Nino events during the recent epochs linked to the climate change and its impact of ISMR.
Full-text available
This study evaluates the ability of several Regional Climate Models (RCMs) to simulate rainfall patterns in the South Caucasus region. In total, 8 RCM simulations were assessed against the CRU observational database over different domains, among them two from the Coordinated Regional Climate Downscaling Experiment (CORDEX). Seasonal climatology, annual rainfall cycles and interannual variability in RCM outputs were estimated for 8 homogeneous sub-regions against several observational datasets. Different metrics covering from monthly and seasonal to annual time scales are analyzed over the region of interest. The results confirm the distinct capabilities of climate models in capturing the local features of the climatic conditions of the South Caucasus region. At the same time, the analysis shows significant deviations in individual models depending on the sub-region and season; however, the ensemble mean is in better agreement with observations than individual models. Overall, the analysis presented here demonstrates that, the multi-model ensemble mean adequately simulates rainfall in the South Caucasus and, therefore, it can be used to assess future climate predictions for the region. This work promotes the selection of RCM runs with reasonable performance in the South Caucasus region, from which, for the first time, a high-resolution bias-adjusted climate database can be generated for future risk assessment and impact studies.
Utilizing historical records is limited in estimating the full distribution of plausible Tropical Cyclone (TC) risk due to insufficient samples. To alleviate this limitation, this study presents a semi‐physical TC rainfall model to produce spatially and temporally resolved TC rainfall sequences for a proper TC risk assessment. The model incorporates a statistical‐based track model that follows the Markov renewal process to generate synthetic TC tracks for any number of years, with a physics‐based model that acknowledges the interaction between TCs and the environment to generate tropical cyclone rainfall. After simulating the TC rainfall data, we then fit it to a probability distribution function to explore the spatially heterogeneous risk induced by landfalling TCs. The procedure is applied in South Korea as a case study to conduct a country‐scale‐based vulnerability evaluation from damaging TC impacts. As a result, we show that the proposed TC model is promising in producing sufficient numbers of TC tracks that do not only follow the spatial distribution of the historical data but also reveal hidden paths that could be utilized in examining scenarios outside what have been historically observed. Results also demonstrate that the model is suitable for properly estimating the amount of TC‐induced rainfall across various locations over the study area. The generated TC rainfall sequences allow us to reliably characterize higher return periods that are often underestimated or overlooked when the observed data is employed due to sampling limitations. Moreover, our model can properly capture the distribution of rainfall extremes that follow a heterogeneous pattern throughout the study area and vary per return period. Overall, our results show that the proposed approach is effective in providing sufficient TC sequences that could help in improving TC risk assessment.
Full-text available
Simulation of future climate changes, especially temperature and rainfall, is critical for water resource management, disaster mitigation, and agricultural development. Based on the category-wise indicator method, two preferred Global Climate Models (GCMs) for the Ishikari River basin (IRB), the socio-economic center of Hokkaido, Japan, were examined from the newly released Coupled Model Intercomparison Project Phase 6 (CMIP6). Climatic variables (maximum/minimum temperature and precipitation) were projected by the Statistical DownScaling Model (SDSM) under all shared socioeconomic pathway-representative concentration pathway (SSP-RCP) scenarios (SSP1-1.9, SSP1-2.6, SSP2-4.5, SSP3-7.0, SSP4-3.4, SSP4-6.0, SSP5-3.4OS, and SSP5-8.5) in two phases: 2040–2069 (2040s) and 2070–2099 (2070s), with the period of 1985–2014 as the baseline. Predictors of SDSM were derived from CMIP6 GCMs and the reanalysis dataset NOAA-CIRES-DOE 20th Century Reanalysis V3 (20CRv3). Results showed that CMIP6 GCMs had a significant correlation with temperature measurements, but could not represent precipitation features in the IRB. The constructed SDSM could capture the characteristics of temperature and precipitation during the calibration (1985–1999) and validation (2000–2014) phases, respectively. The selected GCMs (MIROC6 and MRI-ESM-2.0) generated higher temperature and less rainfall in the forthcoming phases. The SSP-RCP scenarios had an apparent influence on temperature and precipitation. High-emission scenarios (i.e., SSP5-8.5) would project a higher temperature and lower rainfall than the low-emission scenarios (e.g., SSP1-1.9). Spatial–temporal analysis indicated that the northern part of the IRB is more likely to become warmer with heavier precipitation than the southern part in the future. Higher temperature and lower rainfall were projected throughout the late twenty-first century (2070s) than the mid-century (2040s) in the IRB. The findings of this study could be further used to predict the hydrological cycle and assess the ecosystem's sustainability.
Full-text available
Projections of changes in climate extremes are critical to assessing the potential impacts of climate change on human and natural systems. Modeling advances now provide the opportunity of utilizing global general circulation models (GCMs) for projections of extreme temperature and precipitation indicators. We analyze historical and future simulations of ten such indicators as derived from an ensemble of 9 GCMs contributing to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC-AR4), under a range of emissions scenarios. Our focus is on the consensus from the GCM ensemble, in terms of direction and significance of the changes, at the global average and geographical scale. The climate extremes described by the ten indices range from heat-wave frequency to frost-day occurrence, from dry-spell length to heavy rainfall amounts. Historical trends generally agree with previous observational studies, providing a basic sense of reliability for the GCM simulations. Individual model projections for the 21st century across the three scenarios examined are in agreement in showing greater temperature extremes consistent with a warmer climate. For any specific temperature index, minor differences appear in the spatial distribution of the changes across models and across scenarios, while substantial differences appear in the relative magnitude of the trends under different emissions rates. Depictions of a wetter world and greater precipitation intensity emerge unequivocally in the global averages of most of the precipitation indices. However, consensus and significance are less strong when regional patterns are considered. This analysis provides a first overview of projected changes in climate extremes from the IPCC-AR4 model ensemble, and has significant implications with regard to climate projections for impact assessments.
Full-text available
Projections of changes in climate extremes are critical to assessing the potential impacts of climate change on human and natural systems. Modeling advances now provide the opportunity of utilizing global general circulation models (GCMs) for projections of extreme temperature and precipitation indicators. We analyze historical and future simulations of ten such indicators as derived from an ensemble of 9 GCMs contributing to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC-AR4), under a range of emissions scenarios. Our focus is on the consensus from the GCM ensemble, in terms of direction and significance of the changes, at the global average and geographical scale. The climate extremes described by the ten indices range from heat-wave frequency to frost-day occurrence, from dry-spell length to heavy rainfall amounts. Historical trends generally agree with previous observational studies, providing a basic sense of reliability for the GCM simulations. Individual model projections for the 21st century across the three scenarios examined are in agreement in showing greater temperature extremes consistent with a warmer climate. For any specific temperature index, minor differences appear in the spatial distribution of the changes across models and across scenarios, while substantial differences appear in the relative magnitude of the trends under different emissions rates. Depictions of a wetter world and greater precipitation intensity emerge unequivocally in the global averages of most of the precipitation indices. However, consensus and significance are less strong when regional patterns are considered. This analysis provides a first overview of projected changes in climate extremes from the IPCC-AR4 model ensemble, and has significant implications with regard to climate projections for impact assessments.
Full-text available
We present probabilistic projections for spatial patterns of future temperature change using a multivariate Bayesian analysis. The methodology is applied to the output from 21 global coupled climate models used for the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. The statistical technique is based on the assumption that spatial patterns of climate change can be separated into a large scale signal related to the true forced climate change and a small scale signal due to model bias and variability. The different scales are represented via dimension reduction techniques in a hierarchical Bayesian model. Posterior probabilities are obtained with a Markov chain Monte Carlo simulation. We show that with 66% (90%) probability 79% (48%) of the land areas warm by more than 2°C by the end of the century for the SRES A1B scenario.
Most climate models predict a weakening of the North Atlantic thermohaline circulation (THC) for the 21st century when forced by increased levels of greenhouse gas concentrations. The model spread, however, is rather large, even when the forcing scenario is identical, indicating a large uncertainty in the response to forcing. In order to reduce the model uncertainties a weighting procedure is applied that considers the skill of each model in simulating the climatological hydrographic properties. This procedure yields an `estimate' for the evolution of the North Atlantic THC during the 21st century by taking into account a measure of model quality when computing the ensemble mean. The analysis predicts a gradual weakening of the North Atlantic THC by about 30% until 2100.
The combined representation of fields of three climatic variables with empirical orthogonal functions, herein referred to as eigenvectors, is discussed. The eigenvectors are derived from measurements of monthly mean sea-level pressure, surface temperature and precipitation at 23 points in North America for 25 Januarys. Selected eigenvectors of the individual climatic variables are presented; however, the major part of the paper is devoted to the presentation of eigenvectors consisting of combinations of three climatic variables. Empirical eigenvectors derived from fields of two or more meteorological variables have been used in statistical prediction models, but none of the studies to date displayed examples of these eigenvectors or discussed the internal consistency of the combined representations. In this paper it is shown that the structure of the covariances between the three climatic variables,as portrayed by the combined representations, is consistent with synoptic experience. This result i...