9:10-9:30 Introductions of research groups and themes.

9:30-10:40 State-of-the-art GCMs, examples from the IPCC AR4

M. Kimoto: An overview of Japanese contributions to AR4
J. Slingo: Facing up to the demands of resolution, complexity and uncertainty in Earth System Modelling: Is there a choice?
K. Maruyama: Global warming projections for IPCC AR4 and CRIEPI's
future perspective

10:40-11:00 break

11:00-12:40 High resolution GCMs, weather and climate

T. Sakamoto: Impact of horizontal resolution on the Kuroshio's representation
L. Shaffrey: First Results from the HiGEM Project
W. Ofuchi: Predictability of Weekly to Decadal Variability and Its Implication to Earth System Modeling
K. Takahashi: High resolution coupled non-hydrostatic atmosphere-ocean
 simulations on the Earth Simulator

      
12:40-14:00 Lunch

14:00-15:40 Integrative science: water and climate

T. Anderson: Representing climate feedbacks in biogeochemical ocean GCMs
T. Suzuki: Projection of future sea level and its variability in a climate model:
ocean processes and Greenland and Antarctic ice-melt contributions
M. Roberts: High-resolution ocean climate modelling
     

15:40-16:00 break
 
16:00-17:40 Integrative science: carbon, ecosystems and climate

M. Kawamiya: Development of an integrated earth system model on the Earth Simulator
D. Tsumune: Future plan of ocean carbon cycle model for the assessment of carbon sequestration in the North Pacific
P.L. Vidale: A review of carbon cycle research with Hadley Centre models
      

18:30 Reception at the Guest House


Day 2

9:00-10:40 Integrative science: chemistry, aerosols and climate

K. Sudo: Chemistry-Aerosol modelling in the FRCGC Earth System Model
T. Nozawa: Impact of aerosols on the climate change in the 20th century
M. Sanderson: Missing chemistry-climate interactions in GCM simulations
M. Woodage: The impact of dust aerosols on the simulation of current day climate

      
10:40-11:00 break

11:00-12:40 Using past climates to understand the Earth system

      
12:40-14:00 Lunch

14:00-15:40 Understanding climate sensitivity and feedbacks: the promises of ESM

T. Ogura: Climate sensitivity of CCSR/NIES/FRCGC AGCM with different cloud modelling assumptions
B. Collins: The next generation of global environmental models at the Hadley Centre
      

15:40-16:00 break

16:00-17:40 HPC and ESM


      
18:30 End of the workshop

1. M. Kimoto, State-of-the-art + AR4

M. Kimoto: An overview of Japanese contributions to AR4

2. J. Slingo, State-of-the-art + AR4

Facing up to the demands of resolution, complexity and uncertainty in Earth System Modelling: Is there a choice?

J. Slingo, NCAS CGAM, Univ. of Reading, UK

The recent Pan-WCRP workshop on monsoon modelling identified a number of key issues, all of which constitute generic model problems and place limitations on our ability to simulate the climate system. This presentation will summarise these problems, using illustrative examples from the current IPCC AR4 models. Recommendations for future research priorities will be discussed along with the implications for the competing demands on computing resources posed by resolution, complexity and uncertainty.


3. K. Maruyama, State-of-the-art + AR4

Global warming projections for IPCC AR4 and CRIEPI's future perspective

Koki Maruyama and Yoshikatsu Yoshida
Central Research Institute of Electric Power Industry

For the IPCC AR4, CRIEPI conducted multi-century three-member
ensemble global warming projection experiments using the
Community Climate System Model Version 3 to investigate the
long-term response of climate system to stabilization levels
of green house gases (GHGs). The scenario experiments carried
out consist of 20th century historical simulation and future
scenario experiments up till year 2450. The future scenarios
include IPCC SRES scenarios, A1B and B1, for the 21st century,
the green house gases (GHGs) stabilization and overshoot
scenarios beyond the 21st century.

From the obtained projections, implications to the world energy
policy are (1) GHGs stabilization level under the A1B
stabilization scenario may not meet the goal of UNFCCC because
the sea ice will vanish in the Arctic region which might
correspond to the dangerous anthropogenic interference with
climate system: (2) GHGs stabilization level under the B1
stabilization scenario might be one of the target levels.
However, an appropriate level of carbon dioxide concentration
in the atmosphere should be decided from a clear criterion,
which still remains unknown: (3) A further research of overshoot
scenarios should be pursued because it is expected to be useful
for risk managements to cope with low and late emission
reduction of carbon dioxide in the world.

In addition to the above implications, one of the important
directions in future plan is to explore the feasibility of
zero-emission world. For this purpose, we are going to
contribute to develop an earth system model where terrestrial
and ocean carbon cycle, dynamic vegetation, and atmospheric
chemistry are incorporated.


4. T. Sakamoto, High-resolution GCMs, weather and climate

Impact of horizontal resolution on the Kuroshio's representation

Takashi T. SAKAMOTO (1), Tatsuo SUZUKI(1), Seita EMORI (2,1),
         Hiroyasu HASUMI (3), and Akimasa SUMI (3)

  1: Frontier Research Center for Global Change (FRCGC), JAMSTEC
  2: National Institute for Environmental Studies (NIES)
  3: Center for Climate System Research (CCSR), the Univ. of Tokyo


  A climate model called MIROC3.2 developed by CCSR, NIES, and FRCGC has
two setups for high-resolution (hereafter, MIROC3.2_hi) and medium-
resolution (hereafter, MIROC3.2_med). The atmospheric component of
MIROC3.2_hi (MIROC3.2_med) is a T106 (T42) spectral model and the oceanic
component consists of an OGCM of 0.28deg x 0.19deg (1.4deg x 0.56--1.4deg)
horizontal resolution.
  In MIROC3.2_hi, the Kuroshio, that is a western boundary current of
subtropical ocean gyre in the North Pacific, is realistically represented,
but in MIROC3.2_med, it is not because its horizontal resolution is
insufficient to represent a narrow ocean current.
  We investigate that how the state of the Kuroshio is changed when
atmospheric component of MIROC3.2_med (MIROC3.2_hi) and oceanic component
of MIROC3.2_hi (MIROC3.2_med) are coupled. We call it MIROC3.2_mAhO
(MIROC3.2_hAmO). We will show preliminary results about the differences
of the Kuroshio in the four setups of MIROC3.2 in the presentation.


5. Shaffrey, high resolution

First Results from the HiGEM Project


HiGEM is a national UK programme in high resolution modelling of the
global environment. In a three year programme (2004-2006), the new
Hadley Centre climate model HadGEM1 has been increased in resolution
to 1 degree in the atmosphere and 1/3 degrees in the ocean. The aim is
to advance the fidelity of simulations of the global environment, and
to improve our understanding of climate variability on timescales from
days to centuries.

Results will be presented from the control integration of the HiGEM
model with the focus on the small-scale interactions between
components in the climate system. Examples include the modulation of
atmospheric windstress by oceanic tropical instability waves and
small-scale ocean-atmosphere coupling in the mid-latitudes.



6. W. Ofuchi, high resolution + weather

Predictability of Weekly to Decadal Variability and Its
Implication to Earth System Modeling

Wataru Ohfuchi, Hideharu Sasaki, Takeshi Enomoto, Nobumasa
Komori and Akira Kuwano-Yoshida (with inputs from many
collaborators)

The Earth Simulator Center
Japan Agency for Marine-Earth Science and Technology

We have been conducting "hindcast" experiments with
high-resolution atmospheric and oceanic general
circulation models (GCMs) for weekly to decadal time
scale. Also we have been testing a coupled GCM and are
developing prediction system for seasonal to interannual
time scale. This talk briefly describes the results so
far, and discusses their implications to some issues on
predictability and verifiability of Earth system models.
Some key questions are;

1) Can we rely on prediction or "projection" of Earth
system models without good predictability of seasonal to
interannual variability?

2) Are we fully utilizing available state-of-art
observations in order to verify our models and initialize
predictions?


7. K. Takahashi, high-resolution + weather

High resolution coupled non-hydrostatic atmosphere-ocean simulations on the Earth Simulator
K. Takahashi, Earth Simulator Center, Yokohama, Japan


 Global/Regional coupled non-hydrostatic atmosphere-ocean
 simulation code has been developed in the Earth Simulator Center
 since FY2003. The simulation code was designed in order to simulate
 atmosphere-ocean interactions with high resolution. In the coupled
 model, high performance computation has been performed due to
 advanced parallel techniques of each component and coupling
 schemes to be used on the Earth Simulator.

 Several validation results and impacts of high resolution
 coupled simulations will be presented in this talk. Preliminary
 results of high resolution coupled simulations will show that
 small scale interactions between atmosphere and ocean will
 play important roles. Near future work under collaboration
 of MOU will be also introduced.


8. T. Anderson, Water+Climate
Representing climate feedbacks in biogeochemical ocean GCMs.

Thomas R. Anderson, National Oceanography Centre Southampton, Waterfront Camputs, Southampton SO15 5AQ, UK (presenting author)
Ian J. Totterdell, Met. Office Hadley Centre, FitzRoy Road, Exeter EX1 3PB, UK

Simple nutrient-phytoplankton-zooplankton-detritus (NPZD) models have for many years been used to represent the marine ecosystem in ocean biogeochemical general circulation models (GCMs). Climate feedbacks may however be linked to the ecology of individual plankton groups rather than the bulk assemblage. Recent developments in ecosystem modelling have therefore seen the explicit representation of plankton functional types (PFTs), such as diatoms or coccolithophores, in models. Adding complexity does however present difficulties, particularly if the ecology of the organisms is not well understood, model results are sensitive to parameter values, and there are only sparse data for validation. Also, a model that can adequately simulate current ecosystems may not show the correct response to climate change, and this may be a more serious problem for more complex models. A grand challenge for ecosystem modellers is to formulate a model that is sufficiently complex to adequately capture system feedbacks of interest, yet sufficiently simple to be amenable to parameterisation. We will identify the major climate feedbacks associated with marine biogeochemistry and discuss how these may be represented in ecosystem models to be included in GCMs.

9. T. Suzuki, water and climate

Projection of future sea level and its variability in a climate model:
ocean processes and Greenland and Antarctic ice-melt contributions

Tatsuo Suzuki, Hiroyasu Hasumi, Takashi T. Sakamoto, Teruyuki
Nishimura, Ayako Abe-Ouchi, Tomonori Segawa, Naosuke Okada, Akira
Oka, and Seita Emori

Using a climate model, we projected future sea level and its
variability based on two scenarios for 21st century greenhouse gas
emission. The globally averaged sea level rise attributable to the
steric contribution was 23 and 30 cm for the two scenarios. The results
of the high-resolution model (MIROC3.2_hi) and a medium-resolution
version (MIROC3.2_med) of the same model for global and local sea level
change agreed well in spite of the different climate sensitivity, the
A1B run induced global warming of about 4.0°C in MIROC3.2_hi and 3.4°C
in MIROC3.2_med at the end of the 21st century.  It is because the
total heat flux into the ocean is similar in the both models during the
21st century, though the net heat flux into ocean in MIROC3.2_hi is
larger than that in MIROC3.2_med during the early 21st century.The
upper ocean in MIROC3.2_hi also warms up more than that in
MIROC3.2_med.The distribution of sea level changes during 21st century
predicted in MIROC3.2_hi also resembled that in MIROC3.2_med on a large
scale. However, MIROC3.2_hi presented more detailed ocean structure
changes under global warming. The changes in the ocean structure
affected not only the spatial distribution of sea level rise, but also
changes in local sea level variability. These distributions are
strongly influenced by changes in the wind stress forcing under global
warming. However, the common future is not established in
high-resolution models.


10. M. Roberts, water and climate

High-resolution ocean climate modelling
M. Roberts
Hadley Centre, Exeter, UK

Coupled climate models have, until recently, been generally rather
coarse in resolution and hence unable to represent small scale features
of the atmosphere or ocean. Increasing the resolution of one component,
as was done with the HadCEM model at the Hadley Centre (1/3 degree ocean
but 2.5x3.75 degree atmosphere) significantly improved the ocean
simulation, but seemed to have little impact on the coupled climate.

Observations and modelling studies over the last few years suggest quite
significant coupling between atmosphere and ocean at small scales, as
suggested by studies of Tropical Instability Waves in the equatorial
Pacific (an oceanic process which causes changes to atmospheric winds,
moisture and even perhaps interacts with the ITCZ). Studies have also
been made suggesting direct coupling over major boundary currents as
well, such as the Kuroshio and North Atlantic Current. Clearly, in order
to represent such features in models, one needs high resolution and
relatively long integration timescales to study the long term
interaction.

Initial results from the HiGEM model will be presented, which has a
resolution of 1/3 degree in the ocean and 1.25x0.83 degrees in the
atmosphere. Coupling effects in the tropical Pacific will be presented,
and compared to the standard HadGEM1 model, which is the IPCC AR4 model
used by the Hadley Centre.


11. Kawamiya, Integrative science: carbon

Development of an integrated earth system model on the Earth Simulator

M. Kawamiya and T. Matsuno
(Frontier Research Center for Global Change / JAMSTEC)

Frontier Research Center for Global Change (FRCGC) launched in FY 2002 a project to develop an integrated earth system model that operates on the Earth Simulator, in collaboration with the Center for Climate System Research of the University of Tokyo, National Institute of Environmental Studies. The project aims at development of a model where biological and chemical processes important for the global environment are included to interact with climate changes. The model is developed by adding individual component models to atmospheric and oceanic general circulation models (GCMs). The component models are terrestrial and oceanic carbon cycle models and an atmospheric chemistry model. Improvements of the physical climate model are required in order to extend the model top to the middle atmosphere. Preliminary results with fully-coupled climate-carbon cycle model show a significant positive feedback between climate change and carbon cycle, while another preceding model exhibits an even stronger feedback. Other foci of the project include: experiments with the atmospheric chemistry component model, which demonstrate that impact of climate change on other green house gases such as tropospheric ozone and methane could be significant; examination of resolution-dependence of momentum transfer to the stratosphere by gravity waves using high resolution Atmospheric GCMs which explicitly resolve gravity waves.


12. D. Tsumune, Integrative science: carbon

Future plan of ocean carbon cycle model for the assessment of carbon sequestration in the North Pacific

Daisuke Tsumune
Environment Science Research Laboratory
Central Research Institute of Electric Power Industry (CRIEPI)
Abiko, Chiba, JAPAN


We plan to assess the ocean carbon sequestration mainly in the North Pacific near Japan. Ocean carbon sequestration is still one of important future options to stabilize the carbon dioxide concentration in the atmosphere. Ocean general circulation models assessed the effect and environmental impact of ocean sequestration. These results include the uncertainties both in the carbon cycle and ocean circulation process.
At first, we assess the ocean circulation in the North Pacific in considering with inter-annual variability. Chlorofluorocarbon (CFCs) concentrations and ideal age are simulated in a global integration of the Parallel Ocean Program (POP) with realistic surface forcing derived from the NCEP/NCAR reanalysis (1958-2000). The present study focuses on the subtropical and subpolar thermocline of the North Pacific due to the relatively good observational coverage and the well documented decadal timescale variability of the circulation for that region.
Calculated CFC distribution is generally in good agreement with observation by WOCE. On the other hand, there are some discrepancies around the North Pacific Intermediate Water.  Tracer ventilation ages are computed from the simulation using the CFC partial pressure (pCFC). Biases arise in pCFC ages due to the interaction of mixing with the non-linear CFC atmospheric time-histories, and the model CFC ages significantly underestimate ideal age in the mid- to lower-thermocline and in the shadow zone of the eastern tropics and subtropics. Variability in pCFC ages is well correlated with variability of ideal age on interannual timescales, though the amplitude of the pCFC variability may be substantially less due to the offsets between the tracer and ideal ages and their spatial gradients.

13. P.L. Vidale, Integrative science: carbon

A review of carbon cycle research with Hadley Centre models
P.L. Vidale, NCAS CGAM, Univ. of Reading, UK
C. Jones and M. Warnier, UKMO Hadley Centre, Exeter, UK


One of the major goals of UJCC is to construct an Earth System Model capable of simulating the global carbon cycle, in order to address the still existing uncertainties in the magnitude of carbon-climate feedbacks (Cox  et al., 2000; Friedlingstein et al., 2001, Friedlingstein et al.,
2003, Jones et al., 2003 , Zeng et al., 2004b). Our development work is based on existing Hadley Centre model components, e.g. as in HadCM3LC (Cox et al., 2001), which have been used in the past to assess different aspects of the carbon cycle. As an example application we describe results from the first phase of the C4MIP intercomparison project, in which only the atmospheric and terrestrial biosphere submodels were coupled together.
First, we present an overview of the simulated climate and global mean changes in the terrestrial biosphere. We then compare the model with point-flux measurements from CarboEurope tower data (Aubinet et al., 2000; Valentini et al., 2003) and a brief analysis of the interannual variability. Important uncertainties about the regional distribution of sources and sinks still remain. Finally, we present some more recent results from higher resolution simulations and discuss the value of current high resolution products from remote sensing in integrating the information from surface observational networks.


14. K. Sudo, Integrative science: chemistry, aerosols and climate

Chemistry-Aerosol modelling in the FRCGC Earth System Model

Kengo Sudo1, Masayuki Takigawa1, Tatsuya Nagashima2, Masaaki, Takahashi3
1Frontier Research Center for Global Change (/JAMSTEC), 3173-25 Showa-machi, Kanazawa-ku Yokohama 236-0001, Japan
TEL:81-45-778-5722, FAX:81-45-778-5496, EMAIL:kengo@jamstec.go.jp
2National Institute for Environmental Studies, Tsukuba, Japan.
3Center for Climate System Research, University of Tokyo, Chiba, Japan.

In the framework of Earth System modelling (ESM) at the Frontier Research Center for Global Change (FRCGC), we have been developing a chemistry-aerosol coupled climate model. In this talk, we will introduce our chemistry-aerosol component in the FRCGC ESM which includes detailed chemistry and aerosol processes in the troposphere and stratosphere. In the context of “chemistry-climate interaction”, we also discuss the impact of climate change on ozone distribution and related chemistry in our past and future simulations with a chemistry coupled climate model CHASER[1,2] which is used in our ESM. Interaction between atmospheric chemistry and climate is being recognized as an important issue on future prediction of climate and global environment. To investigate the roles of climate change in changing processes of ozone distribution and related chemistry in the stratosphere/troposphere is important as a first step toward understanding of chemistry climate interaction. Our past (reproduction) simulation of ozone distribution (with the SSTs prescribed by the HadiSST data in the CHASER model) suggests that the past-present climate change as represented by a ~0.6K rise in global mean surface air temperature has a significant effect on tropospheric ozone budgets due to water vapor increase and enhanced ozone transport from the stratosphere. In our future simulation following the SRES scenarios, we obtain much different ozone distributions for different emission scenarios during 2000 to 2100, but very similar climate change effects on ozone: reduction in lower tropospheric ozone by 10-20% due to water vapor changes and increases in upper tropospheric ozone by 5-15% associated with enhanced ozone transport from the stratosphere as suggested by [2]. Future climate change also has a significant impact on methane (CH4). The global CH4 trend calculated in our “control” run, only with emission changes, basically reflects CH4 emissions specified with each scenario. This emission-induced CH4 change, however, appears to be reduced significantly by climate change in our “warming” run (with warming as well as emission changes) for each scenario due to increases in water vapor and temperatures associated with tropospheric warming which enhance the CH4 loss reaction with OH radical. It should be noted that although magnitude of the projected climate change is much dependent on scenario, our simulation shows a CH4 reduction by ~20% due to climate change in 2100 for all three scenarios. The CHASER experiments for the IPCC-AR4 related projects (Air Quality-Climate) will be also discussed.

References
[1]Sudo, K., M. Takahashi, J. Kurokawa, and H. Akimoto, J. Geophys. Res., 107, 10.1029/2001JD001113, 2002a.
[2]Sudo, K., M. Takahashi, and H. Akimoto, Geophysical Research Letters., 30, 24,2256, doi:10.1029/2003GL018526, 2003.


15. T. Nozawa, Integrative science: chemistry, aerosols and climate

Impact of aerosols on the climate change in the 20th century

Authors: Toru Nozawa, Tatsuya Nagashima, Tokuta Yokohata, Hideo Shiogama, Toshihiko Takemura, Simon A. Crooks, and the K-1 Japan project team

Abstract: Transient climate change simulations in the 20th century are performed to diagnose relative importance of various forcing agents as well as to enhance the model’s reliability. We consider a set of “complete” forcings that consists of solar and volcanic forcings for natural contributions, and well-mixed GHGs, ozone, sulfate and carbonaceous aerosols, and land-use forcings for anthropogenic contributions. Our model can reproduce observed surface air temperature (SAT) change with a set of “ complete” forcings. Geographical distributions of the SAT trends also show quite good agreement with observations. The observed warming in the early 20th century results from the natural contributions; combined effects of the solar and volcanic forcings. Warming in the latter half of the 20th century, on the other hand, results from the anthropogenic contributions; aerosol cooling due to human activity offsets about 30% of the GHG warming. The consideration of carbonaceous aerosols as an agent of anthropogenic forcing is one of the main features of our simulation. Although simulated global mean SAT does not show much difference between the cases with and without an increase in the carbonaceous aerosols emission, geographical distribution of the SAT trend in the middle of the 20th century were properly reproduced in the simulation with an increase in carbonaceous aerosols emission.


16. M. Sanderson , aerosols+climate

Missing chemistry-climate interactions in GCM simulations
M. Sanderson
Hadley Centre, Exeter, UK


Many climate simulations use levels of carbon dioxide, methane, ozone
and other trace gases that are either fixed at certain levels, or change
with time in a prescribed manner.  The impact of climate and chemistry
on these trace gases is often not included.  However, the changing
climate is likely to have a significant impact on the chemistry of these
trace gases.  A warmer climate means that many chemical reactions will
proceed more quickly.  The amount of water vapour in a warmer atmosphere
will also be larger than the present day, which will increase the rate
of destruction of ozone, and correspondingly increase the production
rate of the hydroxyl radical, OH.  The chemical destruction of most
trace gases begins with attack by the OH radical, so their lifetimes
will be shorter, reducing their radiative effects (e.g. methane).  The
emission of some highly reactive natural hydrocarbons (e.g., isoprene)
may also increase in a warmer climate.  This presentation will give a
brief overview of some of the missing chemistry-climate interactions
from GCM simulations that could be included by coupling the STOCHEM
chemistry model to the Hadley Centres climate model.


17. Woodage, 'Integrative science: chemistry, aerosols and climate'

The impact of dust aerosols on the simulation of current day climate
M. Woodage, CGAM, UK

Natural and anthropogenic aerosols have an significant impact on the earth's
radiation budget and it is therefore important to include their effects in earth
system models. It is well known that changes in aerosol emissions due to man's
activities are an important component of climate change experiments, but they
can also have significant shorter-term regional impacts due to their spacial
inhomogeneity and complex radiative properties. Fully interactive modelling of
aerosols within GCMs is challenging due to the fact that many of the processes
are poorly understood, observations of emissions and aerosol loadings are sparse,
and the transport and aerosol chemistry codes tend to be are computationally
expensive. The work presented here will focus on some experiments carried out
including dust aerosol in the UK_HiGEM model run on the NERC HPCx computer as
part of the HiGEM model development project.

18. JC Hargreaves, Using past climates to understand the Earth system

Efficiently constraining climate sensitivity with ensembles of paleoclimate simulations

JC Hargreaves, JD Annan, R Ohgaito, A. Abe-Ouchi, S. Emori

We use a recently developed efficient probabilistic estimation technique to estimate the sensitivity of the Earth's temperature to a doubling of atmospheric carbon dioxide. The method is based on the ensemble Kalman filter, which we apply to the CCSR/NIES/FRCGC AGCM (the atmospheric component of MIROC3.2). We attempt to validate the results using hindcasts of the Last Glacial Maximum and conclude that our ensembles are probably biased towards too high a climate sensitivity. Within the framework of our single model experiment we show that climate sensitivity of more than 6C is hard to reconcile with the paleoclimate record.


19. JD Annan, Using past climates to understand the Earth system

The implications of using observational evidence to constrain climate sensitivity.

JD Annan and JC Hargreaves

Recently several independent estimates of climate sensitivity have been obtained from various historical climate events. These are generally referred to as observational constraints on climate sensitivity and are much less subjective than the expert assessment originally presented in the 1979 Charney report and subsequently adopted by the IPCC.

In this talk I will discuss the conclusions that can be drawn from these estimates of climate sensitivity based on a range of observational constraints, and demonstrate how these constraints can be applied to improve the next generation of earth system models.

20. T. Ogura, Understanding climate sensitivity and feedbacks: the promises of ESM

Climate sensitivity of CCSR/NIES/FRCGC AGCM with different cloud modelling assumptions
T. Ogura et al., NIES

Two versions of CCSR/NIES/FRCGC atmospheric general circulation model are
coupled to a slab ocean model, which yield different values of equilibrium climate sensitivity of 6.2K and 4.1K. The difference between the two versions is attributed to the different treatment of cloud microphysics, which affect the cloud radiative forcing response to CO2 increase, especially in the southern middle latitudes. Results of cloud water budget analysis suggest that this cloud forcing response is dependent on the control cloud ice distribution and the temperature range where ice cloud is replaced by liquid cloud in response to temperature rise. The obtained results underline the importance of cloud ice modelling for increasing our confidence in the projected climate change.

21. B Collins, ESM

The next generation of global environmental models at the Hadley Centre
B. Collins
Hadley Centre, Exeter, UK

This talk describes the next Hadley Centre climate model HadGEM2. This
model will fully live up to its description as a Global Environmental
Model by including many of the Earth System processes important for
predicting climate change.

Climate modelling has progressed tremendously since its earliest days.
Originally only the atmosphere was considered, then it was realised that
the land surface and oceans needed to be included.  The third IPCC
report (in 2001) showed that it was necessary to include representations
of both aerosols and greenhouses gases in order to replicate the 20th
Century temperature record. In 2000 Cox et al. showed that including the
carbon cycle in climate models could dramatically change the predicted
response of the HadCM3 model to anthropogenic forcing, from 4.0K to 5.5K
by the year 2100. This highlighted the importance of Earth System
feedbacks in the climate system, and the necessity of including such
feedbacks in climate models in order to predict future climate change.

The concept of climate feedbacks has also raised the importance of
natural sources of greenhouse gases and aerosols. These are excluded
from the definition of radiative forcing which only includes
anthropogenic sources. However natural sources can be significantly
affected by a change in climate, either amplifying or dampening the
climate change signal. They therefore need to be taken into account to
the same extent as anthropogenic sources.

The HadGEM2 model will include representations of land and ocean
biogeochemistry, anthropogenic and natural aerosols, and atmospheric
chemistry. This will allow a better estimate of the future climate
change, and of the uncertainties involving Earth System feedbacks.

22. R. Benshila, HPC

Improving parallelization of OPA on the Earth-Simulator.
R. Benshila, LODYC-IPSL, France

This talk will give an overview of the modifications made in the OPA model to improve its parallelization on the Earth-Silumator.
OPA parallelization is based on MPI. The first task was to find the appropriate domain decomposition along the longitude and latitudes directions. Because of the fixed dimensions of the model domain, only some specific values of the number of cpu used in the i and j directions are suitable. Next, the shape of the sub-domains will influence the performances in many (and opposite) ways. A rectangle shape with large number of points along the i direction will favor the model vectorization (even if almost all the do loops are unrolled along i and j directions). A square shape will minimize the number of points located at the boundary of the sub-domain and thus reduce the size of the data that need to be exchanged between the cpus. Finally, a domain decomposition along i and j directions needs about twice more communications than a model decomposition along the i or the j axis.
Once the appropriate domain decomposition was defined, our second task was to minimize the number of needed communications.
A trick in the convergence test of the elliptic solver of the model allowed us to suppress easily a significant number of communications for each model time-step. Next, the main part of our work focussed on cluster several communications in one unique and larger communication, in order to decrease latency time induced by the communication set up. This was done by (1) grouping sets of 2D communications in one 3D communication, (2) grouping consecutive communications involving for example zonal and meridional current or temperature and salinity. Third, by increasing the number of overlapping bands only in the elliptic solver routine (less expensive since performance is a balance between local computation and communications), we cluster the communications and perform one larger communication every N iterations of the solver instead of every iterations as it was done before.
In conclusion, OPA parallelization was greatly improved as we are now able to used about 50 nodes for a domain size of 1622 x 1082 x 101ABSTRACT MISSING

23. G. Robinsons, HPC

Feasibility of Improving UM performance by using Task Parallelism
G. Robinson, UKMO Hadley Centre, Exeter, UK

There are a number of issues with the UKMO Unified Model (UM, the foundation of all Hadley Centre climate models), which limit scaling on large number of processes. The dynamic nature of the workload and the need for synchronization causes the code to scale badly on larger numbers of processors. The workload balance is difficult to rectify since it depends not only on the static decomposition of data but on a number of dynamic factors which vary between iterations and even timesteps. Optimisation of MPI processes such as global sums is hard due to the requirement to produce identical results independent of the number of processes.

The UM HIGEM problem was run on 8 nodes, with each node hosting one MPI process and there being 1, 4 and 8 processes on a node. Best results are achieved with 4 tasks per MPI process. This run has the MPI scaling costs of an 8 process run but can provide the computational performance of 24 processes for the parts of the conjugate solver routines. The code section was chosen to be representative of the loop structures and the interaction of the MPI communications/GCOM library. It should be relatively easy to expand the task parallelism throughout the remainder of the UM considering each section and placing directives as necessary to ensure all GCOM calls are SERIAL and any non-trivially PARALLEL loops are either guided by explicit directives or remain SERIAL.

This test has only considered a small section of the code and the extension of task parallelism to the remainder of the program can be undertaken with care. This shows the ease with which task parallelism can be adopted for expensive parts of the code and the remainder left serial. On the ES there are relatively few obviously expensive routines which can be targeted and the number of routines which must be effectively made to run well with task parallelism is high. This is not the case with other system such as the HPCx IBM Power series clusters.


24. S. Wilson, HPC

Optimisation of HiGEM on scalar machines: HPCx
S. Wilson, UKMO Hadley Centre, Exeter, UK and NCAS CGAM, Reading, UK

HiGEM is the high resolution configuration of the Met Office climate model (HadGEM). There are many new considerations for running the model at higher resolutions, one of which is the model's performance when run on many processors. If a model runs slowly, or does not scale, then the amount of science researchers are able to do will be limited.

In addition to the Earth Simulator, HiGEM is also being run on the HPCx, an IBM p690+ Regatta system based in the UK. A study into the performance of the model on this system will be presented, together with the optimisations implemented from the results of this study. These optimisations give a significant speed increase. The suitability of these optimisations for a vector supercomputer (e.g. Earth Simulator) will also be discussed.

There are differences in the performance of various  aspects of HiGEM on the ES and HPCx due to different system architectures. The reasons for these differences, and what these signify for model systems design for HiGEM, and future Met Office models, will be presented.

A partnership between the Hadley Centre, the NCAS Centre for Global Atmospheric Modelling, the Earth Simulator and the Center for Climate System Research