Download The Met Office Hadley Centre climate modelling capability: the

Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
Phil. Trans. R. Soc. A (2007) 365, 2635–2657
doi:10.1098/rsta.2007.2087
Published online 30 July 2007
The Met Office Hadley Centre climate
modelling capability: the competing
requirements for improved resolution,
complexity and dealing with uncertainty
B Y V. P OPE *, S. B ROWN , R. C LARK , M. C OLLINS , W. C OLLINS ,
C. D EARDEN , J. G UNSON , G. H ARRIS , C. J ONES , A. K EEN , J. L OWE ,
M. R INGER , C. S ENIOR , S. S ITCH , M. W EBB AND S. W OODWARD
Met Office, Hadley Centre, Fitzroy Road, Exeter EX1 3PB, UK
Predictions of future climate change require complex computer models of the climate
system to represent the full range of processes and interactions that influence climate.
The Met Office Hadley Centre uses ‘families’ of models as part of the Met Office Unified
Model Framework to address different classes of problems. The HadGEM family is a
suite of state-of-the-art global environment models that are used to reduce uncertainty
and represent and predict complex feedbacks. The HadCM3 family is a suite of well
established but cheaper models that are used for multiple simulations, for example, to
quantify uncertainty or to test the impact of multiple emissions scenarios.
Keywords: HadGEM; climate models; climate change; uncertainty
1. Introduction
Useful climate predictions depend on having the most comprehensive and
accurate available models of the climate system. However, any single model
will still have limitations in its application for certain scientific questions and
it is increasingly apparent that we need a range of models to address these
ranges of applications. There are two primary reasons for this. First, there is
inherent uncertainty in predictions, which means that ensemble predictions are
needed with many model integrations. Second, technological advances have
not kept pace with scientific advances. A model that included the latest
understanding of the science at the highest resolution would require computers
of several orders of magnitude faster than today’s machines. For these reasons,
the Met Office Hadley Centre has adopted a flexible approach to climate
modelling based on model ‘families’ within which we define a suite of models
aimed at addressing different aspects of the climate prediction problem. All of
these models are flavours of the Met Office’s unified weather forecasting and
climate modelling system.
* Author for correspondence (vicky.pope@metoffice.gov.uk).
One contribution of 9 to a Theme Issue ‘Climate change and urban areas’.
2635
This journal is q 2007 The Royal Society
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
2636
V. Pope et al.
Our current modelling families are as follows.
(i) HadGEM1 (Johns et al. 2006; Martin et al. 2006; McLaren et al. 2006).
A state-of-the-art global environment model, building on, but substantially
changed from our earlier model, HadCM3.
(ii) HadCM3 (Gordon et al. 2000; Pope et al. 2000). A well-established coupled
climate model that is cheap to run on current computers (e.g. it can be run
on a PC, www.climateprediction.net).
Specific applications for which we have recently used these models include
the following.
(i) Earth system feedbacks (see §3 for details). These have been incorporated
separately into coupled ocean–atmosphere models in both families,
HadGEM1 and HadCM3. Work has been started in incorporating these
in-line into our next model family, HadGEM2, which will include our first
true Earth system model.
(ii) QUMP (Murphy et al. 2004). Quantifying uncertainty in model
predictions using ensembles of scientifically distinct versions of HadCM3
(see §4 for details).
(iii) Regional ( Wilson et al. 2006; Buonomo et al. 2007) and highresolution models (led by the NERC HiGEM project and the UK,
NERC and the Hadley Centre, Japan UJCC project). These build on
the standard global models, HadGEM1 (HiGEM) and HadCM3
(regional).
(iv) FAMOUS. A low-resolution version of HadCM3 designed to run approximately 10 times faster (Jones et al. 2005). It is well suited to long runs,
large ensembles or use on PCs and its speed has allowed it to be optimally
tuned. Results are directly traceable to HadCM3 and processes of interest
in HadCM3 can be studied further in long simulations or parameter
ensembles carried out with FAMOUS.
In addition, we have a suite of tools available to us for detailed model
evaluation. Notably, this includes a variety of observational datasets, model to
satellite approaches and increasing use of the strong links between numerical
weather prediction and climate within our unified modelling framework that
allows us to evaluate many of the ‘fast’ physical processes on short time scales
making use of real-time observations.
This paper presents some of the highlights of the latest developments in some
of these modelling systems at the Hadley Centre. Full details on each topic can
be found in the references cited. In §2, we focus on HadGEM1 and how it
compares with HadCM3. In §3, we discuss Earth system feedbacks. In §4, we
outline our approach to quantifying uncertainty. Evaluation is discussed
throughout the paper as it underpins our confidence in the model results. In
§5, we outline further improvements to the model, which have been incorporated
into the next model family HadGEM2. We also point to some key developments
for the future.
Phil. Trans. R. Soc. A (2007)
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
Hadley Centre climate modelling capability
2637
2. HadGEM1
(a ) Simulation of present-day climate
Full details of HadGEM1 are given elsewhere (see references above). Table 1
provides a summary of the key model schemes in HadGEM1 and HadCM3. A key
aspect of our strategy is for a unified model (UM), used for both numerical
weather prediction and climate modelling. It has therefore been a particular aim
in building HadGEM1 to incorporate the new semi-Lagrangian dynamical core,
which has recently been implemented in the Met Office’s operational forecast
models. In addition, continuing research into climate processes has yielded a
number of new physical parametrizations, improving the representation of these
processes and adding new functionality to the model. Furthermore, increases in
computing resources, coupled with the different behaviour of the semiLagrangian dynamics at low resolution, have motivated us to increase the
horizontal and vertical resolutions in both the atmosphere and ocean when
compared with HadCM3. Taken together, these developments mean that
HadGEM1 is very different from its predecessor.
The evaluation of HadGEM1 against observations and reanalyses indicates
that most aspects of the simulation are significantly improved compared with
HadCM3 (Martin et al. 2006). The basic model variables of temperature, winds
and moisture are improved in the free atmosphere, as is mean sea-level pressure.
These improvements can be attributed to the increased resolution and the new
dynamics and physics packages. Some of the most impressive improvements are
in the tropopause structure and the reduced surface pressure bias in the Arctic.
The transport of both water vapour and tracers is also dramatically improved.
Some aspects of variability are less well simulated however, in particular tropical
Pacific sea surface temperatures (SSTs), El Niño Southern Oscillation (ENSO)
variability and monsoon rainfall. These problems are the subject of ongoing
research and there have already been substantial improvements made to a new
model family, HadGEM2, which is in the process of being documented (see also §5).
A summary of the key improvements in physical processes in HadGEM1 is
as follows.
(i) The new advection scheme results in improved transport of water vapour
and tracers, although some compromises have to be made to ensure
conservation of tracers.
(ii) The improved representation of clouds and convection improves the
representation of clouds, water vapour and radiative properties of the
atmosphere.
(iii) Increased horizontal and vertical resolutions improve spatial and
temporal variability.
(iv) The new boundary layer scheme and surface tiling improve properties of
the boundary layer, interactions with convection and interactions
between the atmosphere and the land and ocean surfaces.
(v) The new gravity wave scheme improves tropospheric and stratospheric
gravity waves as well as large-scale winds across the globe.
(vi) The inclusion of an interactive sulphur cycle, including the indirect aerosol
effects, means that radiative forcing can be represented more realistically.
Phil. Trans. R. Soc. A (2007)
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
2638
V. Pope et al.
Table 1. Summary of the key model schemes in HadGEM1 and HadCM3.
HadCM3
HadGEM1
atmospheric grid
hydrostatic
horizontal
resolution
(atmosphere)
vertical resolution
(atmosphere)
Arakawa-B grid
yes
2.58 latitude!3.758 longitude
Arakawa-C grid
no
1.258 latitude!1.8758 longitude
19 levels; hybrid pressure;
Lorenz grid
physics–dynamics
coupling
dynamics
sequential
38 levels; hybrid height; terrain following
near bottom boundary; Charney–Phillips
grid
parallel split (slow processes), sequential
split (fast processes) (Dubal et al. 2004)
semi-Lagrangian advection, conservative
monotone treatment of tracers; semiimplicit time integration (Staniforth et al.
2003; Davies et al. 2005)
Edwards & Slingo (1996) and Cusack et al.
(1999a)
non-local mixing scheme for unstable
boundary layers (Lock et al. 2000); local
Richardson number scheme for stable
boundary layers (Smith 1990, 1993)
mixed-phase scheme including prognostic
ice content; solves physical equations for
microphysical processes using particle
size information (Wilson & Ballard 1999)
revised scheme including diagnosed deep
and shallow convection; new thermodynamic closures at lifting condensation
level; new CMT parametrization based
on flux–gradient relationships; parametrized entrainment/detrainment rates for
shallow convection; based on ideas from
Grant & Brown (1999) and Grant (2001);
convective anvil scheme
(Gregory 1999)
gravity wave drag scheme with low-level
flow blocking (Webster et al. 2003)
derived from global land-based one kilometre base elevation (GLOBE) dataset
at 1 0 resolution
MOSES-II (Essery et al. 2001); nine surface
tile types plus coastal tiling; seasonally
varying vegetation (Lawrence & Slingo
2004)
Smith (1990); parametrized RH-crit
(Cusack et al. 1999b); vertical gradient
area cloud scheme (Smith et al. 1999)
radiation
boundary layer
Eulerian advection, splitexplicit time integration
(Cullen 1993)
Edwards & Slingo (1996) and
Cusack et al. (1999a)
local Richardson number
mixing scheme (Smith 1990,
1993)
microphysics
Senior & Mitchell (1993);
evaporation of precipitation
as in Gregory (1995)
convection
mass flux scheme (Gregory &
Rowntree 1990); convective
downdraughts (Gregory &
Allen 1991); convective
momentum transport
(Gregory et al. 1997)
gravity wave drag
Gregory et al. (1998)
orography
derived from US Navy 10’
dataset
hydrology
MOSES-I (Cox et al. 1999)
clouds
Smith (1990); prescribed
critical relative humidity for
cloud formation (RH-crit)
(Continued.)
Phil. Trans. R. Soc. A (2007)
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
Hadley Centre climate modelling capability
2639
Table 1. (Continued.)
HadCM3
river routing
aerosols
horizontal resolution (ocean)
vertical resolution
(ocean)
ocean model
HadGEM1
simple basin aggregate output embedded 18!18 river transport/hydrology
instantaneously to ocean
submodel (Oki & Sud 1998)
interactive sulphate (anthro- interactive sulphate, sea salt, black carbon
and biomass burning aerosol schemes;
pogenic sources only, direct
direct/indirect radiative forcing
effect only); all other aero(Jones et al. 2001; Woodage et al. 2003;
sols and effects prescribed
Roberts & Jones 2004)
1.258!1.258
18!18 except in tropics where it increases
smoothly to 1/38 in latitude at the
equator
20 levels
40 levels
Gordon et al. (2000)
HadGOM1 ( Johns et al. 2004): EVP
(elastic–viscous–plastic) sea-ice
dynamics; multiple-category ice thickness
distribution; linear free surface scheme;
high-resolution coast/bathymetry;
fourth-order advection of active tracers
(vii) A new land hydrology scheme and improved vegetation scheme
(including a seasonal cycle but not interactive vegetation) are included
to give better radiation and water balance and improved representation
of rivers and run off into the sea.
(viii) Improved sea-ice dynamics and thermodynamics improve the transport,
growth and decay of sea ice.
These improvements are described in detail by Martin et al. (2006). Here, we
have chosen to highlight two major improvements that are particularly important
in the response of the model to global warming, namely cloud and sea ice.
One of the most significant improvements in HadGEM1 has been in the
representation of clouds and cloud radiative properties (Martin et al. 2006). This
is of particular relevance to climate prediction, as clouds continue to provide the
major source of uncertainty when considering estimates of climate sensitivity
from contemporary models (Ringer et al. 2006; Soden & Held 2006; Webb et al.
2006). In HadGEM1, we have succeeded in improving the distributions of
different cloud types (low or high altitude, optically thick or thin) while at the
same time retaining a simulation of the top-of-atmosphere radiation budget,
which compares very favourably with that observed. This indicates that we have
eliminated many compensating errors associated with clouds that were present in
HadCM3, the most apparent being a tendency to generate small amounts of
extremely optically thick cloud rather than larger amounts of thinner
(‘intermediate thickness’) cloud. Figure 1 shows an example of this and
illustrates the improvement in the representation of low-level cloud, i.e. where
the cloud top is below approximately 700 hPa in the atmosphere. Recently, Bony &
Dufresne (2005) found that tropical cloud feedbacks in current coupled climate
change experiments differ most in areas of large-scale subsidence, consistent with
Phil. Trans. R. Soc. A (2007)
2640
low/thick: HadGAM1
low/thick: HadAM3
3 6 9 12 15 20
low / intermediate: ISCCP
3 6 9 12 15 20
low/intermediate: HadGAM1
3 6 9 12 15 20
low/intermediate: HadAM3
60° E 120° E 180 120° W 60° W
5 10 15 20 30 40
00
60°E 120°E 180 120°W 60°W
5 10 15 20 30 40
00
60°E 120°E 180 120°W 60°W
V. Pope et al.
60° N
50° N
40° N
30° N
20° N
10° N
EQ
10° S
20° S
30° S
40° S
50° S
60° S
0
low / thick: ISCCP
0
5 10 15 20 30 40
Figure 1. Comparison of low-level cloud simulated in HadGAM1 and HadAM3 (the atmosphere-only versions of HadGEM1 and HadCM3) with
satellite observations from the International Satellite Cloud Climatology Project (ISCCP). Low-level cloud is defined as being below 680 hPa. Cloud is
further classified in terms of optical thickness into ‘thick’ and ‘intermediate’ thickness categories (adapted from Martin et al. (2006)).
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
Phil. Trans. R. Soc. A (2007)
60° N
50° N
40° N
30° N
20° N
10° N
EQ
10° S
20° S
30° S
40° S
50° S
60° S
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
Hadley Centre climate modelling capability
2641
the hypothesis that low clouds play a key role. Webb et al. (2006) have confirmed
this by providing direct evidence of considerable low-top cloud responses in areas
which contribute most to inter-model differences in global cloud feedback and
climate sensitivity. Uncertainties in cloud feedbacks are also discussed in §4.
Another major improvement in HadGEM1 is in the representation of sea ice
(McLaren et al. 2006). The geographical ice extent generally agrees well with
observations, especially in winter, with the exception of the HadGEM1 winter ice
being too extensive in the North Pacific. In the Arctic, HadGEM1’s lead fraction is
within the observational range of HadISST and values derived from RGPS. The
seasonal cycle of ice area is improved in HadGEM relative to HadCM3 (not shown),
with the winter maximum now occurring at the correct time in both the Arctic and
the Antarctic. The spatial distribution of ice thickness in HadGEM1 is also much
improved relative to HadCM3, particularly in the Arctic (figure 2), where the
thickest ice is now banked up against the Canadian Archipelago as observed.
Sensitivity experiments suggest that the spatial pattern of ice thickness is improved
by the new sea-ice dynamics scheme, and that the magnitude of the ice thickness is
improved by resolving the sub-grid-scale ice thickness distribution.
(b ) Simulation of future climate response
The transient climate change response in the two models has been compared in
an idealized scenario in which atmospheric carbon dioxide concentrations are
increased by 1% per annum for 80 years. This scenario has previously been shown
to lead to statistically significant changes in global and regional climate for a
range of climate quantities of interest. The values of effective climate sensitivity
and total ocean heat uptake in HadGEM1 are found to be similar to those in
HadCM3 and, consequently, the global mean surface warming is also similar.
On a regional scale, more differences are evident between the two models, with
differences in patterns of the climate feedback parameter, surface warming
(figure 3) and precipitation all being evident. In the atmosphere above the surface
level and in the ocean below the surface, there are noticeable differences in the
structure of temperature change. Figure 3 shows that HadGEM1 warms more than
HadCM3 over the Arctic Ocean and northern Canada and Alaska, with differences
of 18C or more in some places. At mid-latitudes and over large areas of land,
HadGEM1 warms less than HadCM3 again by 18C or more in some places. Indeed,
HadGEM1 actually cools south of Greenland at the time of CO2 doubling, in an area
likely to be affected by changes in the thermohaline circulation.
The patterns of radiative forcing in the two models are similar (Johns et al.
2006; using slab ocean experiments), so the differences in surface warming must
be due to differences in local feedbacks, including latent heat damping. Following
Boer & Yu (2003), we have estimated local climate sensitivity and analysed the
contributing factors. Given the changes in the global cloud distribution described
above, we would expect the cloud-related feedbacks to differ between the models.
For example, HadGEM1 has a weaker short-wave cloud feedback than HadCM3
over South America and southern Africa, corresponding to the lower temperature
increases in these regions seen in figure 3. Over the Amazon, the dominance of
the strong short-wave feedback contributed to the strong carbon cycle feedback
in HadCM3 (Cox et al. 2000) and this is likely to be reduced in HadGEM1.
Differences in the feedbacks related to low-level clouds are also apparent: there is
Phil. Trans. R. Soc. A (2007)
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
2642
V. Pope et al.
(b)
(a)
0.01
1
2
3
4
5
0.01
1
2
3
4
5
Figure 2. Grid-box mean ice thickness in March (m): (a) HadGEM1 and (b) HadCM3.
a weaker negative feedback (leading to greater warming) over the southern
tropical trade cumulus regions, whereas a weaker positive feedback (and thus
reduced warming) occurs over the Pacific stratocumulus region. There are also
differences between the models’ clear-sky feedbacks. HadGEM1 has stronger
clear-sky short-wave feedback at high latitudes, particularly over the Northern
Hemisphere ocean, presumably resulting from the changes to both sea-ice
dynamics and thermodynamics and the retreat of sea ice.
Although the global mean climate sensitivities are very similar (the effective
climate sensitivity is 3.1 and 2.8 K in HadCM3 and HadGEM1, respectively),
changes in the local feedbacks due to the many differences between the models
can lead to substantially different local temperature increases due to increased
CO2. The generally improved representation of cloud in HadGEM1, for example,
should, all other things being equal, enable us to have greater confidence in this
model’s predicted cloud feedbacks. The detailed analysis provided by Johns et al.
(2006) demonstrates that, individually, the many changes between HadCM3 and
HadGEM1 lead to both increases and decreases in the globally averaged climate
sensitivity, with their cumulative effect being relatively small. Differences
between the geographical patterns of the feedbacks may arise due to interactions
between the changes or through the individual feedbacks combining nonlinearly.
3. Earth system processes
There is growing recognition that different parts of the Earth system affect one
another and both improved projections of climate change and reliable evaluation
of mitigation options require us to include these feedbacks in models. To
illustrate this, we focus on a number of Earth system components.
(a ) Carbon cycle processes
Inclusion of Earth system components such as the carbon cycle allows us to
extend our research from scientific questions such as how atmospheric composition
Phil. Trans. R. Soc. A (2007)
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
2643
Hadley Centre climate modelling capability
(a) 90° N
60° N
30° N
0
30° S
60° S
90° S
(b) 90° N
60° N
30° N
0
30° S
60° S
90° S
–5.0 –3.0 –2.0 –1.5 –1.0 – 0.2
0.2
1.0
1.5
2.0
3.0
5.0
(c) 90° N
60° N
30° N
0
30° S
60° S
90° S
60° E
0
120° E
180
120° W
–2.50 –1.50 –1.00 –0.75 –0.50 –0.10 0.10 0.50 0.75
60° W
1.00 1.50 2.50
Figure 3. 1.5 m temperature change after 80 years in model integrations in which CO2 is increased
by 1% per year. (a) HadCM3, (b) HadGEM1 and (c) HadGEM1–HadCM3.
will affect climate to more politically relevant questions such as how anthropogenic
activity will affect climate, through changes in atmospheric composition.
The land and ocean are currently taking up about half of the anthropogenic
emissions of carbon dioxide, suppressing the rate at which CO2 increases and climate
Phil. Trans. R. Soc. A (2007)
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
2644
V. Pope et al.
warms. However, this uptake of carbon is known to be sensitive to climate, so climate
change will feedback on itself by modifying the natural carbon cycle: an effect
normally excluded from general circulation model climate projections, but which
may produce significant acceleration of climate change (Cox et al. 2000).
Climate change acts to weaken both the natural terrestrial and ocean uptake of
CO2 from the atmosphere. Faster decomposition of dead plant material exceeds
increases in vegetation productivity due to higher CO2 levels, leading to release of
carbon from the world’s soils. In the oceans, warmer temperatures reduce the
chemical solubility of CO2 and also decrease the mixing of CO2 to depth due to
stratification of the surface waters. A recent study of 11 coupled climate–carbon cycle
models found unanimous agreement that climate change will reduce natural carbon
uptake, providing a positive feedback and accelerated climate change (Friedlingstein
et al. 2006). Le Quere et al. (2007) present observational evidence that recent climate
changes may already be weakening the natural carbon sink in the Southern Ocean.
(b ) Carbon cycle implications
Stabilization scenarios are receiving increasing amounts of interest both
politically and scientifically. At present, about half of anthropogenic CO2
emissions are absorbed naturally, but there is growing consensus that this
fraction will reduce due to the action of climate change on the natural carbon
cycle. Such climate–carbon cycle feedbacks will therefore influence the amount of
carbon emissions required to stabilize atmospheric CO2 levels (Jones et al. 2006).
Simulated feedbacks between the climate and the carbon cycle (figure 4) imply a
reduction of 21–33% in the integrated emissions (between 2000 and 2300) for
stabilization, with greater fractional reductions necessary at higher stabilization
concentrations. Any mitigation or stabilization policy that aims to stabilize
atmospheric CO2 levels must take into account climate–carbon cycle feedbacks
or risk significant underestimate of the action required to achieve stabilization.
(c ) Radiative forcing from ozone
After CO2 and water vapour, the most important greenhouse gases are
methane and ozone. Unlike the first two, methane and ozone react chemically
within the atmosphere. Ozone in fact is not directly emitted into the atmosphere
(except in very small quantities), but is produced by the chemical reactions of
hydrocarbons and oxides of nitrogen. These same compounds also affect the rate
at which methane is removed from the atmosphere. Thus, anthropogenic
emissions of hydrocarbons and oxides of nitrogen although not necessarily
greenhouse gases themselves will have an indirect climate forcing through their
effect on methane and ozone (e.g. Derwent et al. 2001; Collins et al. 2002).
The combined effects of anthropogenic ozone precursors can be calculated using
three-dimensional chemistry models. Results are subject to uncertainty both
globally and regionally, as seen in the recent multi-model intercomparison of Gauss
et al. (2006), which included STOCHEM results coupled to both HadAM3 and
HadGEM1. However, there is a consensus that twentieth century changes in
tropospheric ozone have contributed a positive radiative forcing of approximately
1 W mK2 in parts of the tropics and maybe a small cooling forcing in high southern
latitudes. Consideration of the interactions with other components of the Earth
Phil. Trans. R. Soc. A (2007)
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
Hadley Centre climate modelling capability
2645
cumulative CO2 emissions (GtC)
4000
3000
2000
1000
0
450
550
650
750
stabilization level (ppm)
1000
Figure 4. Cumulative emissions totals from 2000 to 2300 for the five WRE stabilization scenarios,
both with (grey lines) and without (black lines) climate–carbon cycle feedbacks as simulated by the
simple model. The lower and upper subdivisions within the bars show cumulative emissions up to
2100 and 2200, respectively.
system, such as deposition to vegetation and biogenic VOC emissions, is required in
order to simulate future tropospheric ozone levels.
(d ) Impact of climate change on ozone and methane
As well as reactive gases affecting climate, work by Johnson et al. (2001) has
shown that changes in climate can also affect the atmospheric concentrations of
reactive greenhouse gases. The rate of loss of ozone from the troposphere
increases with the concentration of water vapour. This is expected to increase in
a warmer climate. One of the by-products of this ozone loss reaction is the OH
radical. Reaction with this OH radical is the main destruction route for methane.
A compounding factor is that this destruction rate is temperature dependent,
increasing as temperature increases. The overall effect of climate change on
chemical reaction rates therefore is to increase the destruction of both methane
and ozone—a negative feedback. Figure 5 shows the impact of climate change on
methane and ozone in HadCM3. Both are expected to increase in the future, due
to increasing anthropogenic emissions; however, all other things being equal, the
rates of increase will be less than expected without allowing for climate change.
There are, however, other factors to consider. Collins et al. (2003) showed that
the influx of stratospheric ozone into the troposphere was likely to increase in the
future, due to changes in the atmospheric circulation. Sanderson et al. (2003)
predicted an increase in hydrocarbon emissions from vegetation in a warmer
climate, and Gedney et al. (2004) predicted an increase in methane emissions
from natural wetlands. Sanderson et al. (2007) showed that increasing CO2 levels
in HadGEM1 would decrease the removal of tropospheric ozone by plants due to
CO2-induced stomatal closure reducing the flux of O3 into stomatal cavities
(figure 6). The uncertainties in all these processes are sufficiently large that it is
not yet possible to conclude whether climate change will increase or decrease the
greenhouse forcing from ozone and methane.
Phil. Trans. R. Soc. A (2007)
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
2646
V. Pope et al.
(a) 4000
(b)
control
climate change
6.0
O3 (molecules × 1036)
CH4 (ppb)
3500
3000
2500
5.5
5.0
4.5
2000
4.0
1500
2000
2025
2050
2075
2000 2020 2040 2060 2080 2100
Figure 5. The impact of climate change on (a) methane and (b) ozone in HadCM3. In the control
run, the reactive chemistry evolves according to the predictions of anthropogenic emissions
following the SRES A2 scenario, but the meteorology used is appropriate for a 1990s climate. In the
climate run, both the chemistry and the climate evolve according to the A2 scenario. Data from
Johnson et al. (2001).
(e ) Impact of ozone on plant productivity
Excessive O3 concentrations can damage plant cells and therefore reduce
productivity, with the risk of decreasing crop yield (McKee & Long 2001;
Morgan et al. 2003, 2004). As an illustration of this, Sitch et al. (2005) used an
off-line vegetation model driven by changes in climate, CO2 and O3, consistent
with the IS92a emissions scenario, and compared the terrestrial carbon storage in
simulations with and without O3 physiological effects. O3 concentrations were
projected to increase by 10–20% over the most populated parts of the temperate
and tropical regions, and this was simulated to reduce terrestrial carbon storage
by up to 4 kg C mK2 in the eastern USA, Europe, north and Southeast Asia and
the tropics, a reduction of approximately 25%. When O3 effects were investigated
with and without CO2 effects, it was found that O3 damage was reduced under
higher CO2 concentrations as a result of CO2-induced stomatal closure reducing
the flux of O3 into stomatal cavities (as described above).
(f ) Ocean biogeochemistry
The production of dimethyl sulphide (DMS) by ocean phytoplankton has long
been thought to form part of a feedback process on global climate.
DMS is the major source of sulphate aerosol over the ocean, forming cloud
condensation nuclei and affecting cloud properties. The CLAW hypothesis
(Charlson et al. 1987) states that changes in oceanic DMS supply to the
Phil. Trans. R. Soc. A (2007)
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
Hadley Centre climate modelling capability
2647
change in O3 (ppb)
80° N
70° N
60° N
50° N
40° N
30° N
15° W
0
0
15° E
1
2
30°E
3
45°E
4
60°E
5
Figure 6. The impact on ozone concentrations of reduced removal by leaf stomata due to a
doubling of CO2 levels in HadGEM1. The CO2 levels are only allowed to affect the ozone removal
through stomata, they are not affecting the climate. Data from Sanderson et al. (2007).
atmosphere may cause changes to aerosols and clouds and hence global climate.
Changes in the amount of radiation reaching the ocean may feedback on the
planktonic production of DMS. Climate simulations with HadCM3 have shown
that oceanic DMS emissions do form part of a global negative feedback process
(Gunson et al. 2006).
(g ) Role of mineral dust
Mineral dust aerosol is found throughout the troposphere, where it interacts
directly with both short- and long-wave radiation. Changes in the world’s
ecosystems and climate have the potential to modify the atmospheric dust
content, providing a feedback on climate change and affecting the ecosystems
themselves. When vegetation and climate change from a coupled climate–carbon
cycle model simulation were used to drive the mineral dust scheme in the
atmosphere-only model version, HadAM3, the global mean dust load increased
by a factor of approximately 3 (Woodward et al. 2005). Such changes would be
likely to lead to impacts on marine ecosystems, where dissolved iron delivered by
atmospheric transport of dust can be a very important micronutrient. Changes in
the supply of iron to the surface ocean may form a significant climate feedback in
the future by affecting biological activity and hence ocean carbon uptake and
oceanic production of DMS. An iron-cycling version of the HadOCC ocean
carbon cycle model will be implemented in a model version in the HadGEM2
family and coupled to the atmospheric dust scheme.
(h ) Earth system models
In order to understand the overall feedbacks of chemistry and ecosystems on
climate, the relevant processes need to be included within the climate model. The
current generation of climate models have been tested with some of these
Phil. Trans. R. Soc. A (2007)
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
2648
V. Pope et al.
(a)
climate
greenhouse effect
aerosols
water cycle
greenhouse
gases
CO2, CH4
fires: soot
mineral dust
chemistry
(b)
ecosystems
climate
water usage
greenhouse effect
human
emissions
aerosols
water cycle
greenhouse
gases
human
emissions
CO2, CH4
fires: soot
mineral dust
chemistry
human
emissions
ecosystems
land-use change
Figure 7. Schematic of the coupling in the (a) current (HadGEM1) and (b) next generation
(HadGEM2) Hadley Centre climate model families. Dotted lines, off-line components; solid lines,
online components.
feedbacks individually (figure 7a). Figure 7a schematically represents the
couplings present in the Hadley Centre models that generated the results for
the IPCC Fourth Assessment Report. Aerosols are coupled interactively to the
climate model, but the chemical fields to drive the aerosols and supply reactive
greenhouse gases are coupled off-line through the supply of data files. Similarly,
the cycling of carbon through ecosystems is not included in HadGEM1, and only
in separate experiments with HadCM3LC. Figure 7b represents the next
generation ‘Earth system’ model. In this configuration, the chemistry and the
ecosystem models will be coupled interactively to the aerosols and the climate.
This will allow the important feedbacks to be quantified. Our next generation
family of models, HadGEM2, will include versions in which the main key
processes and their interactions are incorporated (figure 7b).
Phil. Trans. R. Soc. A (2007)
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
Hadley Centre climate modelling capability
2649
4. Quantifying uncertainty in predictions of climate change
While great progress has been made in recent years in building more accurate
physically based models of the climate system, uncertainties in the projections
made by the models still remain (e.g. Collins 2007). There are three principal
sources of uncertainty in climate predictions, which are as follows.
(i) Those which arise owing to unknown future emissions of greenhouse gases
and other forcing agents related to both human and natural activities.
(ii) Those due to internally generated chaotic fluctuations in the system—e.g.
the phase of the ENSO at some point many decades in the future.
(iii) Those due to the simplifying assumptions made in building climate
models. In particular, the representation of sub-grid-scale physical
processes.
The latter source of uncertainty has been assessed in the past using the rather
ad hoc approach of comparing the outputs of different climate models in
international assessment exercises (e.g. Cubasch et al. 2001). In recent years, the
Hadley Centre has been at the forefront of the development of a new method of
quantifying uncertainties in climate predictions and producing predictions in
terms of the probabilities of different outcomes (probability distribution
functions (pdfs); Murphy et al. 2004, 2007).
While the method can in principle be applied to quantify all sources of
uncertainties, the main focus of the project so far has been on the latter source,
i.e. the one arising from the representation of physical processes in models. These
sub-grid-scale processes are parametrized by means of simplified physically based
or empirical relationships between the large-scale resolved model variables and
the small-scale processes such as those associated with clouds and radiation. Such
relationships, loosely termed the ‘model physics’, are controlled by parameters
that are, in themselves, uncertain. By perturbing these model parameters, it is
possible to generate an ensemble of possible future outcomes based on different
versions of the same climate model (Murphy et al. 2004, 2007; Stainforth et al.
2005). Information from these ensembles of models may then be assembled
together with information from the real climate system to produce pdfs of
climate change for a range of different variables.
Figure 8 shows an example of the simulation of historical and future global
mean temperature change using a ‘perturbed physics’ ensemble approach (Collins
et al. 2006). It can be seen that perturbing the model parameters does result in a
relatively wide spread of outcomes on centennial time scales. While the magnitude
of global mean temperature change is not necessarily an indicator of the
magnitude of regional change and its impacts, it does provide a useful benchmark
and to that extent highlights how uncertainties in model formulation translate to
uncertainties in predictions. Perturbed physics ensembles have been analysed in
detail to show, rather like the small ensemble of the world’s climate models, that
climate feedbacks associated with clouds and radiation are the principal driver of
uncertainties in global mean temperature change (Webb et al. 2006).
Further steps are required to derive pdfs for transient regional climate change
from the perturbed physics ensembles. The 17-member ensemble of Collins et al.
(2006), with coupling to a dynamic ocean model, was limited in size by computer
Phil. Trans. R. Soc. A (2007)
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
2650
V. Pope et al.
global temperature anomaly (K)
6
4
2
0
1900
1950
2000
year
2050
2100
Figure 8. Time series of global, annual mean temperature from two perturbed physics ensembles of
HadCM3. Temperatures are expressed as anomalies with respect to a period of 1860–2000. The
black lines are curves which use the ensemble members described by Collins et al. (2006). The grey
lines are from a new ensemble with much reduced North Atlantic and Arctic SST and salinity
biases, which also sample a slightly larger range of atmospheric and surface feedbacks. (The
number of ensemble members makes it difficult to distinguish the different curves but the desire is
to illustrate the existence of ensemble spread and the increase in spread with time.) The thick black
curve shows observed record of global mean temperature change.
resources and is clearly insufficient to fully sample modelling uncertainty (29
model parameters were simultaneously varied here). For this reason, a technique
has been developed to augment ensemble size by time-scaling patterns of
equilibrium climate change obtained from much larger ensembles of perturbed
physics experiments with coupling to a simpler mixed-layer (‘slab’) ocean model
(Williams et al. 2001), and forced by pre-industrial and twice pre-industrial CO2.
In this way, we can estimate the transient regional response for a larger fraction
of parameter space in a cost-effective way (Harris et al. 2006). Figure 9 shows
‘plumes’ of evolving uncertainty in change in two key variables over the
historical and future period under a single emissions scenario, with 245 timescaled equilibrium simulations here contributing to the spread in response.
Although this is a relatively large ensemble, it does not represent an unbiased
sample relative to some choice of the prior distribution of uncertain model
parameters, so the plumes in figure 9 should be regarded as sample-dependent
frequency distributions rather than pdfs. However, information from this
ensemble can be used to construct an emulator (Rougier & Sexton 2007),
which is a statistical model that allows us to predict the model response at any
untried point in parameter space. By sampling this emulated equilibrium
response and time scaling, we can derive pdfs for our model’s predictions of
transient regional climate change for any underlying distribution of uncertain
model parameters. A further step in the derivation of pdfs is comparison with
observations of current climate and historical trends, in order to down-weight
parts of the model pdf that compare less well with observations than with those
parts which compare better, for example in examining the response to past
historical or palaeoclimate forcings. This application of observational constraints
will be the next stage in our work in ensemble climate prediction.
Phil. Trans. R. Soc. A (2007)
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
Hadley Centre climate modelling capability
(a)
5.0
(b)
2651
30
global
land
3.0
2.0
1.0
∆ precipitation (%)
∆T1.5 m (°C)
4.0
20
10
0
0
–1.0
1850 1900 1950 2000 2050 2100
year
–10
1850 1900 1950 2000 2050 2100
year
Figure 9. Uncertainty ‘plumes’ under historical and A1B forcing with HadCM3, showing evolution
in the median, 80, 90 and 95% confidence ranges for: (a) global and 20-year mean surface
temperature anomalies (K) and (b) global and 20-year mean percentage change in precipitation
over land. Uncertainties accounted for include model internal variability, modelling uncertainty
sampled by 245 equilibrium simulations with different combinations of model parameters and
uncertainty due to time-scaling equilibrium to transient climate change.
The ensemble simulations and pdfs shown in figures 8 and 9 only have
perturbations made to parameters relevant to atmospheric and surface processes.
While these processes do exert a leading-order control on the sensitivity of the
physical climate system, other processes in the ocean and related to chemistry
and carbon cycle feedbacks must also be quantified. Work to run ensembles in
which parameters in those model components are varied is currently underway.
In addition, the sensitivity to various methodological assumptions needs to be
tested and the output pdfs compared with those derived using alternative
approaches (e.g. Stott & Kettleborough 2002).
The science of probabilistic and ensemble climate prediction is still rather
young but is growing quickly. We expect probabilities and quantitative risk
assessment of climate change to become a key tool in the policy arena in the
near future.
The ensemble approach provides useful information about the probability of
extreme events. The rarity of such events means that some estimate of
uncertainty is required to make meaningful statements about extremes.
Figure 10a,b shows the change of hot and average days in summer and wet
and average days in winter, respectively, for two grid boxes. These show that
average days change in a different way to extreme days. The average summer’s
day representative of Washington is between 2 and 88 warmer (with a few
outliers over 108). The hottest day of summer, on the other hand, has
temperature increases with a much larger range of values between 2.5 and 148.
The average winter’s day representative of Southeast England is between 25%
drier and 50% wetter, whereas the wettest day is between 0 and 50% wetter.
Hence, changes in global mean temperature, or even changes in the average local
values of a particular quantity, cannot be used to infer the pattern of change in
extreme events.
Phil. Trans. R. Soc. A (2007)
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
2652
V. Pope et al.
relative likelihood (%)
(a) 20
change in normal day of summer
change in hottest day of summer
(b) 40
15
30
10
20
5
10
0
5
10
temperature change (°C)
15
0
–50
change in wettest day of winter
change in normal day of winter
0
rainfall change (%)
50
Figure 10. (a) Relative likelihood changes in temperature for the hottest and average summer day
due to the doubling of CO2 concentrations for the grid box containing Washington, DC.
(b) Relative likelihood changes in precipitation for the wettest and average winter day due to the
doubling of CO2 concentrations for the grid box containing the southeast of England. Wettest
(hottest) day is calculated by ranking 20 years of winter (summer) data for each ensemble member
and each CO2 concentration and selecting the 20th highest value. The average day is taken as the
median of the ranked data. Distributions are calculated from the differences between 1!CO2 and
2!CO2 pairs for each ensemble member. Sample contains 128 perturbed physics members where
all the perturbed parameters are perturbed simultaneously.
5. Summary and further work
This paper has outlined some of the current approaches to climate modelling
used by the Met Office Hadley Centre by illustrating some of our latest results.
The competing requirements for complexity, resolution and addressing
uncertainty have led us to develop a number of modelling families within
which there are a hierarchy of models whose physical mechanisms can be thought
of as being consistent or ‘traceable’ to one another.
The new HadGEM family of models includes substantial improvements in the
representation of clouds and sea ice, which are important for climate sensitivity.
The transport of water vapour and other tracers has improved, improving the
water balance and representation of chemistry and aerosols. Boundary layer and
land surface representation has improved, together with gravity wave
representation. In HadGEM1, some aspects of variability are less well simulated,
in particular tropical Pacific SSTs, ENSO variability and monsoon rainfall.
These problems were the focus of targeted model development, which has led to
the definition of HadGEM2-AO, the new atmosphere–ocean component model,
within the HadGEM2 model family. Ongoing research will define an Earth
system version of this model, HadGEM2-ES.
HadGEM2-AO is currently being documented in detail, but an outline of the
improvements is given here. In the atmosphere submodel, the representation of
aerosols is much improved relative to HadGEM1; two natural aerosol species,
mineral dust and secondary organic aerosols, have been added to the model.
Improvements to the existing sulphate and biomass burning aerosols have also
been made, based on observations from dedicated field campaigns.
Phil. Trans. R. Soc. A (2007)
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
Hadley Centre climate modelling capability
2653
Improvements to both the atmosphere and the ocean have been included,
which contribute to improvements in the wind stresses and SSTs in the tropical
Pacific and associated improvements in ENSO variability. The convection
scheme now exhibits a smoother and more realistic mass flux profile due to new
functionality that allows the scheme to detrain moisture less sharply near the
tropopause. Consequently, this change improves the vertical structure of
diabatic heating and was shown to have a positive effect in reducing the
strength of easterly wind stresses in the equatorial Pacific region. Changes to the
ocean submodel for HadGEM2-AO are designed to improve the tropical
simulation by reducing mixing. A revised vertical diffusivity profile in the top
1000 m (but still within the uncertainties of observational estimates) was used in
HadGEM2-AO. This helps to warm the tropical SSTs by reducing the mixing of
cooler subsurface water. Horizontal momentum mixing is now made a function of
latitude such that the mixing coefficient is a minimum at the equator and
increasing towards the poles, instead of the constant value used in HadGEM1.
The result is a reduction in tropical viscosity that leads to a reduction in the
westward currents on the equator, in better agreement with observations.
HadGEM2-AO uses a new method for returning soil water content in
supersaturated soil layers to the saturation value; excess soil water is now
drained out of the bottom of the layer instead of being pushed back out the top of
the layer. This helps to reduce the Northern Hemispheric warm biases over land
as seen in HadGEM1, as does a simple exponential decay lifetime applied to
convective cloud amount, which alleviates the intermittency of the convection
scheme. Early tests show that these improvements mean that carbon cycle
feedbacks can realistically be included in the Earth system model, HadGEM2ES, within the HadGEM2 model family.
Overall, the general performance of the HadGEM2 atmosphere is at least very
similar to, and in most cases better than, HadGEM1, using an r.m.s.-based index
of global model skill. Aspects of variability such as the Indian summer monsoon
and ENSO are also improved in HadGEM2-AO, although there is still scope for
further improvement and this is being addressed as part of our development
towards the next model family, HadGEM3.
Despite recent improvements in climate modelling, including those outlined
above in HadGEM2-AO, one of the most challenging issues in model
development is tracing back systematic deficiencies in observable climate
properties (e.g. ENSO, monsoon, blocking, etc.) to fundamental improvements
in model formulation. Recent research has shown that some of the key climate
model shortcomings (e.g. aspects of the monsoon and equatorial wind simulations
in HadGEM1) can be traced back to short NWP simulations. Many of the
fast physics processes can also be investigated on NWP time scales and upper
ocean–atmosphere interaction on seasonal time scales. The Met Office Unified
Model, which is the basis of the HadGEM models, is in a prime position to
exploit these various aspects of model traceability across space and time scales.
A new challenge for climate research is to adapt to the changing requirements
of policy makers towards mitigation and adaptation advice and research into
climate impacts. Mitigation advice requires quantification of the climate impact
on the global carbon cycle and other Earth system feedbacks, but the current
uncertainty is large. Understanding the cause of this uncertainty and
constraining it through improved process representation and observational
Phil. Trans. R. Soc. A (2007)
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
2654
V. Pope et al.
constraints is a key challenge of climate modelling. A traceable hierarchy of
models is required to address these issues. Complex models are required to
understand processes and ensure that they are accurately represented. Simpler
and cheaper versions of the same models are required to explore and quantify
uncertainty and to perform multiple ‘what if’ scenarios. As we move forward to
developing a full range of traceable models within our new model families such as
HadGEM3, these questions will continue to be addressed using cheaper and less
complex models (such as the standard and FAMOUS versions of HadCM3).
Adaptation advice and detailed impacts information require detailed
information on climate change at a regional scale from seasonal, through decadal
to centennial time scales. Improved representation of physical processes and
increased vertical and horizontal resolutions will be required to improve regional
climate and important aspects of variability. Developments in NWP (as outlined
above) and seasonal forecasting using the HadGEM families of models will all
contribute to such improvements. However, it is probable that substantial
increases in computer power will also be needed to fully realize the benefits of
improved models.
This work was funded by the Department for Environment, Food and Rural Affairs under the
Climate Prediction Programme, contract PECD 7/12/37 and by the Ministry of Defence under the
Government Meteorological Research Contract. q Crown Copyright 2007, the Met Office, UK.
References
Boer, G. J. & Yu, B. 2003 Climate sensitivity and response. Clim. Dynam. 20, 415–429.
Bony, S. & Dufresne, J.-L. 2005 Marine boundary layer clouds at the heart of tropical feedback
uncertainties in climate models. Geophys. Res. Lett. 32, L20806. (doi:10.1029/2005GL023851)
Buonomo, E., Jones, R., Huntingford, C. & Hannaford, J. 2007 On the robustness of changes in
extreme precipitation over Europe from two high resolution climate change simulations.
Q. J. R. Meteorol. Soc. 133, 65–81. (doi:10.1002/qj.13)
Charlson, R. J., Lovelock, J. E., Andreae, M. O. & Warren, S. G. 1987 Oceanic phytoplankton,
atmospheric sulphur, cloud albedo and climate. Nature 326, 665–661. (doi:10.1038/326655a0)
Collins, M. 2007 Ensembles and probabilities: a new era in the prediction of climate change. Phil.
Trans. R. Soc. A 365, 1957–1970. (doi:10.1098/rsta.2007.2068)
Collins, W. J., Derwent, R. G., Johnson, C. E. & Stevenson, D. S. 2002 The oxidation of organic
compounds in the troposphere and their global warming potentials. Clim. Change 52, 453–479.
(doi:10.1023/A:1014221225434)
Collins, W. J., Derwent, R. G., Garnier, B., Johnson, C. E., Sanderson, M. G. & Stevenson, D. S.
2003 The effect of stratosphere–troposphere exchange on the future tropospheric ozone trend.
J. Geophys. Res 108, 8528. (doi:10.1029/2002JD002617)
Collins, M., Booth, B. B. B., Harris, G. R., Murphy, J. M., Sexton, D. M. H. & Webb, M. J. 2006
Towards quantifying uncertainty in transient climate change. Clim. Dynam. 27, 127–147.
(doi:10.1007/s00382-006-0121-0)
Cox, P. M., Betts, R. A., Jones, C. D., Spall, S. A. & Totterdell, I. J. 2000 Acceleration of global
warming due to carbon-cycle feedbacks in a coupled climate model. Nature 408, 184–187.
(doi:10.1038/35041539)
Cox, P. M., Betts, R. A., Bunton, C. B., Essery, R. L. H., Rowntree, P. R. & Smith, J. 1999 The
impact of new land surface physics on the GCM simulation of climate and climate sensitivity.
Clim. Dyn. 15, 183–203. (doi:10.1007/s003820050276)
Cubasch, U. et al. 2001 Projections of future climate change. In Climate change 2001: the scientific
basis. Contribution of working group 1 to the third assessment of the Intergovernmental Panel on
Climate Change (eds Houghton et al.). Cambridge, UK: Cambridge University Press.
Phil. Trans. R. Soc. A (2007)
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
Hadley Centre climate modelling capability
2655
Cullen, M. 1993 The unified forecast/climate model. Meteor. Mag. 122, 81–94.
Cusack, S., Edwards, J. M. & Crowther, J. M. 1999a Investigating k-distribution methods for
paraeterizing gaseous absorption in the Hadley Centre climate model. J. Geophys. Res. 104,
2051–2057. (doi:10.1029/1998JD200063)
Cusack, S., Edwards, J. M. & Kershaw, R. 1999b Estimating the subgrid variance of saturation,
and its parametrization for use in a GCM cloud scheme. Q. J. R. Meteor. Soc. 125, 3057–3076.
Davies, T., Cullen, M. J. P., Malcolm, A. J., Mawson, M. H., Staniforth, A., White, A. A. & Wood,
N. 2005 A new dynamical core for the Met Office’s global and regional modelling of the
atmosphere. Q. J. R. Meteor. Soc. 131, 1759–1782. (doi:10.1256/qj.04.101)
Derwent, R. G., Collins, W. J., Johnson, C. E. & Stevenson, D. S. 2001 Transient behaviour of
tropospheric ozone precursors in a global 3-D CTM and their indirect greenhouse effects. Clim.
Change 49, 463–487. (doi:10.1023/A:1010648913655)
Dubal, M., Wood, N. & Staniforth, A. 2004 Analysis of parallel versus sequential splittings for
time-stepping physical parameterizations. Mon. Weather Rev. 132, 121–132. (doi:10.1175/15200493(2004)131!0121:AOPVSSO2.0.CO;2)
Edwards, J. M. & Slingo, A. 1996 Studies with a flexible new radiation code. I: choosing a configuration
for a large-scale model. Q. J. R. Meteor. Soc. 122, 689–719. (doi:10.1002/qj.49712253107)
Essery, R., Best, M. & Cox, P. 2001 MOSES 2.2 technical documentation. Hadley Centre
Technical Note 30. Available online at http://www.metoffice.gov.uk/research/handleycentre/
pubs/HCTN/index.html.
Friedlingstein, P. et al. 2006 Climate–carbon cycle feedback analysis, results from the C4MIP
model intercomparison. J. Clim. 19, 3337–3353. (doi:10.1175/JCLI3800.1)
Gauss, M. et al. 2006 Radiative forcing since preindustrial times due to ozone change in the
troposphere and the lower stratosphere. Atmos. Chem. Phys. 6, 575–599.
Gedney, N., Cox, P. M. & Huntingford, C. 2004 Climate feedback from wetland methane
emissions. Geophys. Res. Lett. 31, L20503. (doi:10.1029/2004GL020919)
Gordon, C., Cooper, C., Senior, C. A., Banks, H., Gregory, J. M., Johns, T. C., Mitchell, J. F. B. &
Wood, R. A. 2000 The simulation of SST, sea ice extents and ocean heat transports in a version
of the Hadley Centre coupled model without flux adjustments. Clim. Dynam. 16, 147–168.
(doi:10.1007/s003820050010)
Grant, A. L. M. 2001 Cloud-base fluxes in the cumulus-capped boundary layer. Q. J. R. Meteor.
Soc. 127, 407–421. (doi:10.1002/qj.49712757209)
Grant, A. & Brown, A. R. 1999 A similarity hypothesis for shallow cumulus transports. Q. J. R.
Meteor. Soc. 125, 1913–1936. (doi:10.1002/qj.49712555802)
Gregory, D. 1995 A consistent treatment of the evaporation of rain and snow for use in large-scale models.
Mon. Wea. Rev. 123, 2716–2732. (doi:10.1175/1520-0493(1995)123!2716:ACTOTEO2.0.CO;2)
Gregory, J. 1999 Representation of the radiative effects of convective anvils. Hadley Centre Tech.
Note 7, pp. 20.
Gregory, D. & Rowntree, P. R. 1990 A mass flux convection scheme with representation of cloud
ensemble characteristics and stability-dependent closure. Mon. WeatherRev. 118, 1483–1506.
(doi:10.1175/1520-0493(1990)118!1483:AMFCSWO2.0.CO;2)
Gregory, D. & Allen, S. 1991 The effect of convective scale downdrafts upon NWP and climate
simulations. 9th Conf. Numerical Weather Prediction, Denver, CO, Amer. Meteor. Soc. 9th
Conf. Numerical Weather Prediction, Denver, CO, Amer. Meteor. Soc. pp. 122–123.
Gregory, D., Kershaw, R. & Inness, P. M. 1997 Parametrization of momentum transport by
convection. II: Tests in single-column and general circulation. Q. J. R. Meteor. Soc. 123,
1153–1183. (doi:10.1002/qj.49712354103)
Gregory, D., Shutts, G. J. & Mitchell, J. R. 1998 A new gravity-wave-drag scheme incorporating
anisotropic orography and low-level wave breaking: impact upon the climate of the UK Meteorological
Office Unified Model. Q. J. R. Meteor. Soc. 124, 463–493. (doi:10.1002/qj.49712454606)
Gunson, J. R., Spall, S. A., Anderson, T. R., Jones, A., Totterdell, I. J. & Woodage, M. J. 2006
Climate sensitivity to ocean dimethylsulphideemissions. Geophys. Res. Lett. 33, L07701.
(doi:10.1029/2005GL024982)
Phil. Trans. R. Soc. A (2007)
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
2656
V. Pope et al.
Harris, G. R., Sexton, D. M. H., Booth, B. B. B., Collins, M., Murphy, J. M. & Webb, M. J. 2006
Frequency distributions of transient regional climate change from perturbed physics ensembles of
general circulation model simulations. Clim. Dynam. 27, 257–375. (doi:10.1007/s00382-006-0142-8)
Johns, T. C. et al. 2004 Evaluation of the atmospheric performance of HadGAM/GEM1. Hadley
Centre Technical Note 54. Available online at http://www.metoffice.gov.uk/research/
handleycentre/pubs/HCTN/index.html.
Johns, T. C. et al. 2006 The new Hadley Centre climate model HadGEM1: evaluation of coupled
simulations. J. Clim. 19, 1327–1353. (doi:10.1175/JCLI3712.1)
Johnson, C. E., Stevenson, D. S., Collins, W. J. & Derwent, R. G. 2001 Role of climate feedback on
methane and ozone studied with a coupled Ocean–Atmosphere–Chemistry model. Geophys.
Res. Lett. 28, 1723–1726. (10.1029/2000GL011996)
Jones, A., Roberts, D. L., Woodage, M. J. & Johnson, C. E. 2001 Indirect sulphate aerosol forcing
in a climate model with an interactive sulphur cycle. J. Geophys. Res. 106, 20 293–20 310.
(doi:10.1029/2000JD000089)
Jones, C. D., Gregory, J., Thorpe, R., Cox, P., Murphy, J., Sexton, D. & Valdes, P. 2005
Systematic optimisation and climate simulation of FAMOUS, a fast version of HadCM3. Clim.
Dynam. 25, 189–204. (doi:10.1007/s00382-005-0027-2)
Jones, C. D., Cox, P. M. & Huntingford, C. 2006 Impact of climate-carbon cycle feedbacks on
emissions scenarios to achieve stabilisation. In Avoiding dangerous climate change (eds H. J.
Schellnhuber, W. Cramer, N. Nakicenovic, T. Wigley & G. Yohe), ch. 34, pp. 323–331.
Cambridge, UK: Cambridge University Press.
Lawrence, D. M. & Slingo, J. M. 2004 An annual cycle of vegetation in a GCM. Part I: Implementation
and impact on evaporation. Climate Dyn. 22, 87–105. (doi:10.1007/s00382-003-0366-9)
Le Quere, C. et al. 2007 Saturation of the Southern Ocean CO2 sink due to recent climate change.
Science 316, 1735–1738. (doi:10.1126/science.1136188)
Lock, A. P., Brown, A. R., Bush, M. R., Martin, G. M. & Smith, R. N. B. 2000 A new boundary
layer mixing scheme. Part I: scheme description and SCM tests. Mon. Weather Rev. 128,
3187–3199. (doi:10.1175/1520-0493(2000)128!3187:ANBLMSO2.0.CO;2)
McKee, I. F. & Long, S. P. 2001 Plant growth regulators control ozone damage to wheat yield. New
Phytol. 152, 41–51. (doi:10.1046/j.0028-646x.2001.00207.x)
McLaren, A. J. et al. 2006 Evaluation of the sea ice simulation in a new coupled atmosphere–ocean
climate model (HadGEM1). J. Geophys. Res. 111, C12014. (doi:10.1029/2005JC003033)
Martin, G. M., Ringer, M. A., Pope, V. D., Jones, A., Dearden, C. & Hinton, T. J. 2006 The
physical properties of the atmosphere in the new Hadley Centre Global Environmental Model,
HadGEM1. Part I. Model description and global climatology. J. Clim. 19, 1274–1301. (doi:10.
1175/JCLI3636.1)
Morgan, P. B., Ainsworth, E. A. & Long, S. P. 2003 How does elevated ozone impact soybean?
A meta-analysis of photosynthesis, growth and yield. Plant Cell Environ. 26, 1317–1328.
(doi:10.1046/j.0016-8025.2003.01056.x)
Morgan, P. B., Bernacchi, C. J., Ort, D. R. & Long, S. P. 2004 An in vivo analysis of the effect of
season-long open-air elevation of ozone to anticipated 2050 levels on photosynthesis in soybean.
Plant Physiol. 135, 2348–2357. (doi:10.1104/pp.104.043968)
Murphy, J. M., Sexton, D. M. H., Barnett, D. N., Jones, G. S., Webb, M. J., Collins, M. &
Stainforth, D. A. 2004 Quantification of modelling uncertainties in a large ensemble of climate
change simulations. Nature 430, 768–772. (doi:10.1038/nature02771)
Murphy, J. M., Booth, B. B. B., Collins, M., Harris, G. R., Sexton, D. & Webb, M. 2007
A methodology for probabilistic predictions of regional climate change from perturbed physics
ensembles. Phil. Trans. R. Soc. A 365, 1993–2028. (doi:10.1098/rsta.2007.2077)
Oki, T. & Sud, Y. C. 1998 Design of Total Runoff Integrating Pathways (TRIP)—A global river
channel network. Earth interactions, 2. Available online at http://EarthInteractions.org.
Pope, V. D., Gallani, M. L., Rowntree, P. R. & Stratton, R. A. 2000 The impact of new physical
parametrizations in the Hadley Centre climate model—HadAM3. Clim. Dynam. 16, 123–146.
(doi:10.1007/s003820050009)
Phil. Trans. R. Soc. A (2007)
Downloaded from http://rsta.royalsocietypublishing.org/ on May 10, 2017
Hadley Centre climate modelling capability
2657
Ringer, M. A. et al. 2006 Global mean cloud feedbacks in idealized climate change experiments.
Geophys. Res. Lett. 33, L07718. (doi:10.1029/2005GL025370)
Roberts, D. L. & Jones, A. 2004 Climate sensitivity to black carbon aerosol from fossil fuel
combustion. J. Geophys. Res. 109, D16202. (doi:10.1029/2004JD004676)
Rougier, J. & Sexton, D. S. 2007 Inference in ensemble experiments. Phil. Trans. R. Soc. A 365,
2133–2143. (doi:10.1098/rsta.2007.2071)
Sanderson, M. G., Jones, C. D., Collins, W. J., Johnson, C. E. & Derwent, R. G. 2003 Effect of
climate change on isoprene emissions and surface ozone levels. Geophys. Res. Lett. 30. (doi:10.
1029/2003GRL017642)
Sanderson, M. G., Collins, W. J., Hemming, D. L. & Betts, R. A. 2007 Stomatal conductance
changes due to increasing carbon dioxide levels: projected impact on surface ozone levels. Tellus
B 59, 404–411. (doi:10.1111/j.1600-0889.2007.00277.x)
Senior, C. A. & Mitchell, J. F. B. 1993 CO2 and climate: the impact of cloud parameterization.
J. Clim. 6, 393–418. (doi:10.1175/1520-0442(1993)006!0393:CDACTIO2.0.CO;2)
Sitch, S., Cox, P., Huntingford, C., Collins, W., Hemming, D. & Gedney, N. 2005 Report on
possible impacts of tropospheric ozone changes on land carbon uptake. Contract deliverable
08.04.04, Defra Climate Prediction Programme.
Smith, R. N. B. 1990 A scheme for predicting layer clouds and their water content in a general
circulation model. Q. J. R. Meteor. Soc. 116, 435–460. (doi:10.1002/qj.49711649210)
Smith, S. 1993 Experience and developments with the layer cloud and boundary layer mixing
schemes in the UK Meteorological Office Unified Model. In Proc. ECMWF/GCSS Workshop on
Parametrization of the Cloud-Topped Boundary Layer, ECMWF, Reading, UK, pp. 319–339.
Smith, R., Gregory, D., Wilson, C. A., Bushell, A. C. & Cusack, S. 1999 Calculation of saturated specific
humidity and largescale cloud. UM Documentation Paper 29, Met Office, Exeter, UK, pp. 27.
Soden, B. J. & Held, I. M. 2006 An assessment of climate feedbacks in coupled Ocean–Atmosphere
Models. J. Clim. 19, 3354–3360. (doi:10.1175/JCLI3799.1)
Staniforth, A., White, A., Wood, N., Thuburn, J., Zerroukat, M. & Cordero, E. 2003 Joy of
UM5.4—model formulation. Unified Model Documentation Paper 15, Met Office, Exeter, UK.
Stainforth, D. A. et al. 2005 Uncertainty in predictions of the climate response to rising levels of
greenhouse gases. Nature 433, 403–406. (doi:10.1038/nature03301)
Stott, P. A. & Kettleborough, J. A. 2002 Origins and estimates of uncertainty in predictions of
twenty first century temperature rise. Nature 416, 723–726. (doi:10.1038/416723a)
Webb, M. J. et al. 2006 On the contribution of local feedback mechanisms to the range of climate
sensitivity in two GCM ensembles. Clim. Dynam. 27, 17–38. (doi:10.1007/s00382-006-0111-2)
Webster, S., Brown, A. R., Cameron, D. R. & Jones, C. P. 2003 Improvements to the
representation of orography in the Met Office Unified Model. Q. J. R. Meteor. Soc. 129,
1989–2010. (doi:10.1256/qj.02.133)
Williams, K. D., Senior, C. A. & Mitchell, J. F. B. 2001 Transient climate change in the Hadley
Centre models: the role of physical processes. J. Clim. 14, 2659–2674. (doi:10.1175/15200442(2001)014!2659:TCCITHO2.0.CO;2)
Wilson, D. R. & Ballard, S. P. 1999 A microphysically based precipitation scheme for the UK
meteorological office unified model. Q. J. R. Meteor. Soc. 125, 1607–1636. (doi:10.1002/
qj.49712555707)
Wilson, S., Hassell, D., Hein, D., Jones, R. & Taylor, R. 2006 Installing and using the Hadley
Centre regional climate modelling system, p. 140. PRECIS, Met Office Hadley Centre, Exeter,
UK. See http://precis.metoffice.com/docs/tech_man.pdf.
Woodward, S., Roberts, D. L. & Betts, R. A. 2005 A simulation of the effect of climate changeinduced desertification on mineral dust aerosol. Geophys. Res. Lett. 32, L18810. (doi:10.1029/
2005GL023482)
Woodage, M., Davison, P. & Roberts, D. L. 2003 Aerosol processes. Unified Model Documentation
Paper 20, Met Office, Exeter, UK.
Phil. Trans. R. Soc. A (2007)