ArticlePDF Available

Abstract and Figures

The Australian Community Climate and Earth System Simulator coupled model (ACCESS-CM) has been developed at the Centre for Australian Weather and Climate Research (CAWCR), a partnership between CSIRO1 and the Bureau of Meteorology. It is built by coupling the UK Met Office atmospheric unified model (UM), and other sub-models as required, to the ACCESS ocean model, which consists of the NOAA/GFDL2 ocean model MOM4p1 and the LANL3 sea-ice model CICE4.1, under the CERFACS4 OASIS3.2–5 coupling framework. The primary goal of the ACCESS-CM development is to provide the Australian climate community with a new generation fully coupled climate model for climate research, and to participate in phase five of the Coupled Model Inter-comparison Project (CMIP5). This paper describes the ACCESS-CM framework and components, and presents the control climates from two versions of the ACCESS-CM, ACCESS1.0 and ACCESS1.3, together with some fields from the 20th century historical experiments, as part of model evaluation. While sharing the same ocean sea-ice model (except different setups for a few parameters), ACCESS1.0 and ACCESS1.3 differ from each other in their atmospheric and land surface components: the former is configured with the UK Met Office HadGEM2 (r1.1) atmospheric physics and the Met Office Surface Exchange Scheme land surface model version 2, and the latter with atmospheric physics similar to the UK Met Office Global Atmosphere 1.0 including modifications performed at CAWCR and the CSIRO Community Atmosphere Biosphere Land Exchange land surface model version 1.8. The global average annual mean surface air temperature across the 500-year preindustrial control integrations show a warming drift of 0.35 °C in ACCESS1.0 and 0.04 °C in ACCESS1.3. The overall skills of ACCESS-CM in simulating a set of key climatic fields both globally and over Australia significantly surpass those from the preceding CSIRO Mk3.5 model delivered to the previous coupled model inter-comparison. However, ACCESS-CM, like other CMIP5 models, has deficiencies in various aspects, and these are also discussed.
Content may be subject to copyright.
Australian Meteorological and Oceanographic Journal 63 (2013) 41–64
The ACCESS coupled model: description,
control climate and evaluation
Daohua Bi1, Martin Dix1, Simon J. Marsland1, Siobhan O’Farrell1, Harun A. Rashid1,
Petteri Uotila1, Anthony C. Hirst1, Eva Kowalczyk1, Maciej Golebiewski5,
Arnold Sullivan1, Hailin Yan1, Nicholas Hannah1, Charmaine Franklin1, Zhian Sun2,
Peter Vohralik3, Ian Watterson1, Xiaobing Zhou2, Russell Fiedler4, Mark Collier1,
Yimin Ma2, Julie Noonan1, Lauren Stevens1, Peter Uhe1, Hongyan Zhu2,
Stephen M. Griffies6,Richard Hill7, Chris Harris7, and Kamal Puri2
1CAWCR/CSIRO Marine and Atmospheric Research, Aspendale, Australia
2CAWCR/Bureau of Meteorology, Melbourne, Australia
3CSIRO Materials Science and Engineering, Lindfield, Australia
4CAWCR/CSIRO Marine and Atmospheric Research, Hobart, Australia
5CSIRO High Performance Computing and Communications Centre, Melbourne, Australia
6NOAA/Geophysical Fluid Dynamics Laboratory, Princeton, NJ, USA
7Met Office Hadley Centre, Exeter, UK
(Manuscript received July 2012; revised December 2012)
The Australian Community Climate and Earth System Simulator coupled model
(ACCESS-CM) has been developed at the Centre for Australian Weather and Cli-
mate Research (CAWCR), a partnership between CSIRO1 and the Bureau of Mete-
orology. It is built by coupling the UK Met Office atmospheric unified model (UM),
and other sub-models as required, to the ACCESS ocean model, which consists of
the NOAA/GFDL2 ocean model MOM4p1 and the LANL3 sea-ice model CICE4.1,
under the CERFACS4 OASIS3.2–5 coupling framework. The primary goal of the
ACCESS-CM development is to provide the Australian climate community with
a new generation fully coupled climate model for climate research, and to par-
ticipate in phase five of the Coupled Model Inter-comparison Project (CMIP5).
This paper describes the ACCESS-CM framework and components, and presents
the control climates from two versions of the ACCESS-CM, ACCESS1.0 and AC-
CESS1.3, together with some fields from the 20th century historical experiments,
as part of model evaluation. While sharing the same ocean sea-ice model (except
different setups for a few parameters), ACCESS1.0 and ACCESS1.3 differ from
each other in their atmospheric and land surface components: the former is con-
figured with the UK Met Office HadGEM2 (r1.1) atmospheric physics and the Met
Office Surface Exchange Scheme land surface model version 2, and the latter with
atmospheric physics similar to the UK Met Office Global Atmosphere 1.0 includ-
ing modifications performed at CAWCR and the CSIRO Community Atmosphere
Biosphere Land Exchange land surface model version 1.8. The global average
annual mean surface air temperature across the 500-year preindustrial control
integrations show a warming drift of 0.35 °C in ACCESS1.0 and 0.04 °C in AC-
CESS1.3. The overall skills of ACCESS-CM in simulating a set of key climatic fields
both globally and over Australia significantly surpass those from the preceding
CSIRO Mk3.5 model delivered to the previous coupled model inter-comparison.
However, ACCESS-CM, like other CMIP5 models, has deficiencies in various as-
pects, and these are also discussed.
1Commonwealth Scientific and Industrial Research Organisation.
2National Oceanic and Atmospheric Administration/Geophysical Fluid
Dynamics Laboratory, Princeton, NJ, USA.
3Los Alamos National Laboratory, Los Alamos, NM, USA.
4Centre Européen de Recherche et de Formation Avancée en Calcul
Scientifique, Toulouse, France.
Corresponding author address: Daohua Bi. Email:
42 Australian Meteorological and Oceanographic Journal 63:1 March 2013
There are many outstanding limitations on our scientific
understanding of key aspects of the climate system response
to anthropogenic climate forcing. For example, there are
uncertainties concerning our understanding of the relative
strength of the climate feedbacks and the climate sensitivity
(e.g. Meehl et al. 2007), and of the response in the strength of
the hydrological cycle (e.g. Durack et al. 2012).With regard
to the Australian region, there is uncertainty concerning the
response of key features of the climate state such as storm
tracks (e.g. Trenberth and Fasullo 2010), and of the modes
of variability affecting Australian climate (e.g. Cai et al. 2009,
Zheng et al. 2010, Collins et al. 2010). All these uncertainties
impact our ability to project climate change at the regional
level and evaluate the potential impact on the Australian
community, and to respond to the risks in an informed
Puri et al. (2013) have outlined the development of a new
Australian capacity in weather and climate simulation, the
Australian Community Climate and Earth System Simulator
(ACCESS). A major goal of ACCESS is to provide state-of-
the-art climate modelling capacity to support Australian
research aimed at addressing uncertainties such as those
above. The priority initial aim in this area is to develop
the coupled climate model for climate change simulation
with the timeline driven by research community uptake
relating to: (1) utilisation of the model results in analysis
studies supporting the Fifth Assessment Report of the
Intergovernmental Panel on Climate Change (IPCC AR5)
and (2) utilisation of the model system by new Australian
initiatives such as the Australian Research Council (ARC)
Centre of Excellence for Climate System Science.
The set of papers in this special issue of the Australian
Meteorological and Oceanographic Journal (AMOJ)
describes the newly developed ACCESS climate model
and evaluates its simulations, which are participating in the
Coupled Model Intercomparison Project phase 5 (CMIP5)
(Taylor et al. 2012). The CMIP5 model output data will form
the basis of the model analyses to be used in the IPCC AR5.
Participation in CMIP5 also facilitates the benchmarking of
the ACCESS model against other models, and the ready
dissemination of the output fields to users nationally and
internationally via the earth system grid (ESG) (Williams
et al. 2009). In this paper, the ‘Model description’ section
describes all the components of the ACCESS coupled
model, including the sub-models, coupler, coupling
framework and coupling methodology. ‘Experimental
design’ documents the experimental designs, including sub-
model initialisations, atmospheric forcing setups for both
the preindustrial control runs and the 20th century historical
simulations, and spin-up processes. ‘Model results’ presents
results from the 500-year pre-industrial control and the
historical simulations, to document and evaluate key aspects
of the model performance. The last section, ‘Summary and
conclusion’, summarises the model skills and shortcomings
and gives conclusions and perspectives. More detailed
analyses of the model results from the ACCESS CMIP5
experiments in terms of performance of the individual sub-
systems (i.e. atmosphere, land, sea-ice, and ocean), and
investigations of some specific scientific issues and topics
such as climate variability will be presented in papers (Dix et
al. 2013, Kowalczyk et al. 2013, Uotila et al 2013, Marsland et
al. 2013, Rashid et al. 2013a, Rashid et al. 2013b, Bi et al. 2013,
Sun et al. 2013) that also appear in this issue.
Model description
This section describes the two versions of the ACCESS
coupled model (hereafter ‘ACCESS-CM’) contributing
to CMIP5, namely ‘ACCESS1.0’ and ‘ACCESS1.3’. The
components of ACCESS-CM are illustrated in Fig. 1. The
atmospheric component is the UK Meteorological Office
(hereafter ‘Met Office’) unified model (hereafter UM)
(Davies et al. 2005; Martin et al. 2010, 2011). The ocean is
the NOAA/GFDL MOM4p1 (Griffies 2009), and the sea-ice
is the LANL CICE4.1 (Hunke and Lipscomb 2010) model.
The atmosphere, together with the land surface model Met
Office Surface Exchange Scheme version 2 (MOSES2, for
ACCESS1.0) or CSIRO Community Atmosphere Biosphere
Land Exchange version 1.8 (CABLE1.8, for ACCESS1.3),
is coupled to ACCESS ocean model (ACCESS-OM; Bi et
al. 2013), the ocean sea-ice core of ACCESS-CM, via the
OASIS3.25 coupler (Valcke 2006).
The principal differences between the two versions of
the ACCESS-CM are in their atmospheric and land surface
components. ACCESS1.0, which may be considered
our ‘basic’ version, includes the Met Office’s well tested
HadGEM2(r1.1) atmospheric physics and MOSES land
surface model (Martin et al. 2011). ACCESS1.3 may be
considered our ‘aspirational’ version. It includes significant
new atmospheric physics similar to that of the Met Office
Global Atmosphere (GA) 1.0 configuration (Hewitt et al.
2011)5, and in particular the PC2 cloud scheme (Wilson et al.
2008). It also includes the Australian-developed CABLE1.8
land surface model (Kowalczyk et al. 2006, 2013). ACCESS1.3
is significantly more experimental than ACCESS1.0 as the
new atmospheric physics has not been tested in century-
scale climate change simulations previously6. As can be seen
via comparison with Fig. 1 of Puri et al. (2013), the ACCESS1.3
configuration is a key step on the way to attaining a full
earth system model as laid out in the original ACCESS
project plan.
5GA1.0 comprises a suite of atmospheric model implementations sharing
a common physics, and, of relevance here, includes the atmospheric com-
ponent of the UK Met Office’s HadGEM3(r1.1) coupled model (Arribas et
al.2011, Hewitt et al. 2011)
6A version of the model with HadGEM2(r1.1) atmospheric physics and
CABLE, denoted ‘ACCESS1.1’, experienced technical difficulties such
that any results would have been too late for IPCC AR5 timelines. A ver-
sion of the model the same as ACCESS1.3 but including a version of
the MOSES land surface model was also developed, and denoted ‘AC-
CESS1.2’. This version ultimately offered insufficient strategic or scientific
advantage over ACCESS1.3, and has been discontinued.
Bi et al.: The ACCESS coupled model: description, control climate and evaluation 43
Though the component codes may be different, the
ACCESS modelling program builds on a long history of
weather and climate modelling in Australia. The history of
Australian climate modelling is detailed in Smith (2007). The
ultimate climate model of the preceding series, the CSIRO
Mk3.6 (Rotstayn et al. 2012), is also featured extensively in
this issue of AMOJ. Expertise from the earlier modelling
has been beneficial in developing the weather and climate
applications of ACCESS in a timely manner.
The remainder of this section provides more detail on
each of the model components and on the coupling strategy.
Readers are directed to the associated literature including
user guides and manuals for full scientific and technical
information about each component.
Atmospheric component
The ACCESS-CM uses the Met Office unified model (UM) as
its atmospheric component. The atmospheric configuration
in ACCESS1.0 is designed to be the same as that of HadGEM2
version r1.1 (Martin et al. 2011)7, which is essentially the same
as that in the HadGEM2 model versions used for CMIP5. The
atmospheric configuration in ACCESS1.3 is as in ACCESS1.0
except in the physical parameterisation, which is similar
to that of the Met Office’s Global Atmosphere (GA) 1.0 (as
in Arribas et al. 2011, and Hewitt et al. 2011) and includes
modifications made at CAWCR.
Resolution, dynamics and orography
The resolution is the same as is standard in the Met Office’s
HadGEM2 family of models (Martin et al. 2011) and also in
early configurations of the HadGEM3 model (Arribas et al.
2011, Hewitt et al. 2011), namely a horizontal resolution of
1.25° latitude by 1.875° longitude (referred to as ‘N96’), and
38 levels in the vertical.
The dynamics is as described in Davies et al. (2005) and
is non-hydrostatic, fully compressible and uses a semi-
Lagrangian advection scheme. The Arakawa ‘C’ grid
(Arakawa and Lamb 1977) is used in the horizontal. The
vertical coordinate is height-based and terrain-following.
The orography is derived from the 30 GLOBE data
set (GLOBE Task Team and others 1999), and the basic
orography is the same for both ACCESS1.0 and ACCESS1.3.
However, the GLOBE data set is found to have substantial
deficiencies over Australia, and so instead the Geoscience
Australia high-quality data set (Hilton et al. 2003) is used
over Australia for all applications of ACCESS. This change
makes some difference even at the N96 resolution (e.g. 50 m
elevation change over the Great Dividing Range).
Physical parameterisations
The atmospheric physical parameterisations in ACCESS1.0
are as in HadGEM2 version r1.1 (Martin et al. 2011, see also
Martin et al. 2006; Collins et al. 2008; Martin et al. 2010).
A brief summary follows.
The radiation scheme is that of Edwards and Slingo
(1996). The radiative effects of the absorbing gases H2O,
CO2, O3, N2O, CH4, CFC11, CFC12 and O2 are included.
Parameterisations for 20 aerosol species are implemented in
the scheme; these include major species such as sulphate,
organic carbon, dust and sea salt. The radiation time step is
set to give eight radiation calculations per day.
The turbulent fluxes of heat, moisture and horizontal
momentum in the boundary layer are represented by the
first order K profile closure as described by Lock et al. (2000).
The scheme has a non-local mixing component for unstable
boundary layers, and uses the ‘SHARPEST’ stability function
(King et al. 2001; Edwards et al. 2006) for stable boundary
The gravity wave drag scheme includes the orographic
gravity wave component of Webster et al. (2003), which
allows for blocking by sub-gridscale orography as well
as gravity wave drag and has been shown to improve the
general circulation (e.g. Webster et al 2003).
The convection scheme is a modified mass flux scheme
based on Gregory and Rowntree (1990). Initiation of
convection is based on evaluation of undiluted parcel
ascent from the near surface, which is used to determine
whether convection is possible from the boundary layer.
Categorisation of convection as deep or shallow depends
on the level of the cloud top. Representations of convective
momentum transport (CMT) are included for both deep and
shallow convection.
Fig. 1. ACCESS-CM components and coupling framework.
The system is built under framework of the Ocean,
Atmosphere, Sea-ice, Soil (OASIS, version 3.25) cou-
pler, which is developed at the Centre Européen de
Recherche et de Formation Avancée en Calcul Scienti-
fique (CERFACS), Toulouse, France.
Land surface
UM 7.3
OASIS 3.25
ocean sea-ice core
7The base code used for the ACCESS1.0 atmospheric component is the
UM code version 7.3 (UM7.3) external release, as this contains code re-
quired for coupling via OASIS. The default atmospheric physics in this
code version is GA1.0, so the HadGEM2 (r1.1) physics codes needed to
be re-introduced.
44 Australian Meteorological and Oceanographic Journal 63:1 March 2013
The precipitation microphysics is determined by the
Wilson and Ballard (1999) single moment bulk scheme, which
features explicit calculations of transfer between vapour,
liquid and ice phases. The prognostic ice variable is split by
a diagnostic relationship into ice crystals and aggregates,
which are treated separately in the microphysical transfer
terms before being recombined after the calculations. The
condensation and evaporation of cloud water is calculated
within the diagnostic cloud scheme.
ACCESS1.0 uses the Smith (1990) diagnostic cloud
scheme that is based on a sub-grid probability distribution of
a temperature and a moisture variable, with the liquid cloud
amount then being derived using a critical relative humidity.
The scheme has been modified such that the prognostic
ice variable is used to diagnostically calculate the ice cloud
fraction (Wilson et al. 2004).
The physical parameterisations in ACCESS1.0 are mostly
common to the ACCESS NWP model described in Puri
et al. (2013). There are two exceptions involving: (1) the
specification in the stability function in the stable boundary
layer, and (2) the usage of an additional, non-orographic,
gravity wave drag term in the NWP model (Warner et al.
2005), where more detailed discussion of the above physical
parameterisations may be found.
The atmospheric physical parameterisations in ACCESS1.3
are similar to GA1.0 (Arribas et al. 2011, Hewitt et al. 2011).
Parameterisation differences between ACCESS1.0 and
ACCESS1.3 are described below.
The radiation scheme in ACCESS1.3 is modified to
include the ‘Tripleclouds’ scheme developed by Shonk and
Hogan (2008) to represent horizontal cloud inhomogeneity.
A detailed description of this scheme and its evaluation
within the ACCESS model are provided by Sun et al. (2013).
This implementation differs from the simpler scaling scheme
implemented in GA1.0 to account for cloud inhomogeneity
(Hewitt et al. 2011). Both approaches increase radiative
transmissivity in cloudy grid boxes. The radiation scheme in
ACCESS1.3 follows GA1.0 in including improved pressure
and temperature scaling (Hewitt et al. 2011). This change
mainly improves simulation of long-wave fluxes in the
middle atmosphere.
The boundary layer physics in ACCESS1.3 is mostly as in
ACCESS1.0. However, algorithms for momentum, sensible
and latent heat flux at the air-sea interface have been modified
from those in both ACCESS1.0 and in GA1.0 based on results
from field programs (Fairall et al. 2003). The modifications
involve the empirical expressions for momentum and scalar
atmospheric surface roughness lengths for 10 m neutral wind
speeds. The modified algorithms are found to ease certain
biases in the pattern of global sea surface temperature (SST)
distribution, including a mild reduction in the equatorial
Pacific cold tongue bias.
ACCESS1.3 follows GA1.0 in using a ‘buddy’ scheme
for coastal grid points to split the near-surface winds into
separate components over the ocean and the land portions
(Hewitt et al. 2011). This scheme was found in GA1.0 to
significantly increase (i.e. improve) precipitation over the
maritime continent. ACCESS1.3 does not follow GA1.0 in
using the boundary layer solver of Wood et al. (2007) due
to technical issues associated with the use of the CABLE
land surface model, and instead retains the ACCESS1.0
The convection physics in ACCESS1.3 is mostly similar
to that in ACCESS1.0. ACCESS1.3 follows GA1.0 in using
revised parcel perturbations for shallow convection that
help make vertical fluxes more consistent between the
boundary layer and convection schemes (Hewitt et al.
2011). ACCESS1.3 retains the CAPE closure scheme based
on relative humidity of ACCESS1.0, instead of the GA1.0
closure scheme based on vertical wind speed.8 9
ACCESS1.3 uses the prognostic cloud prognostic
condensate (PC2) scheme (Wilson et al. 2008), as does GA1.0.
PC2 includes prognostic variables for cloud liquid water
content, ice water content, liquid cloud fraction, ice cloud
fraction and total cloud fraction. Each process in the model
acts as a source/sink of these prognostic variables, including
the convection scheme, so that PC2 represents both
convective and large-scale cloud in the model. The GA1.0
implementation of the PC2 scheme is modified through use
of the parameterisation of Franklin et al. (2012) to modify the
PC2 ice cloud fraction and the cloud area scheme of Boutle
and Morcrette (2010) to account for the effects of coarse
vertical resolution on low-level cloud cover.
Aerosol parameterisation
The aerosol scheme used for both ACCESS1.0 and 1.3 is
the Coupled Large-scale Aerosol Simulator for Studies
in Climate (CLASSIC) (Bellouin et al. 2011). The scheme is
used to simulate seven aerosol species, some with multiple
components: mineral dust, sea salt, fossil fuel black
carbon (FFBC), fossil fuel organic carbon (FFOC), biomass
burning aerosols (with assumptions about the proportions
of organic carbon and black carbon), secondary organic
aerosols (from forest turpene), and sulphate aerosols (from
dimethylsulphide (DMS) and SO2 emissions and the sulphur
cycle). All aerosol species exert a direct radiative effect. All
species except mineral dust and FFBC aerosols (which are
considered hydrophobic) exert first and second indirect
effects. These are also referred to as the cloud albedo and
8The base code used for the ACCESS1.3 atmospheric component was a
pre-release version of the UM7.3. This version included most of the GA1.0
physics, but did not include the change to vertical wind speed-based
CAPE closure, and this was not upgraded separately for ACCESS1.3. The
base physics for this pre-release version, on which CAWCR built, is as
for the version denoted in the UM User Interface [UMUI] under job name
9A code error was inadvertently introduced during testing whereby
the convective momentum transport (CMT) for deep convection in AC-
CESS1.3 was effectively turned off. Thus the current version of AC-
CESS1.3 does have CMT for shallow convection but not for deep con-
vection. Deep convection CMT is known to have beneficial effects on the
tropical solution (e.g. Wu et al. 2007, Kim et al. 2008), and will be rectified
in a subsequent version.
Bi et al.: The ACCESS coupled model: description, control climate and evaluation 45
cloud lifetime effects. All aerosol species are prognostic
except for sea salt, which is diagnosed each time step
based on the near-surface wind speed, and secondary
organic aerosols (SAO), which are prescribed by a monthly
varying climatology. Dust concentrations in ACCESS1.3 are
essentially zero, due to further work being required on the
CABLE/dust module interface. Aerosol treatments are the
same in ACCESS1.0 and 1.3 in all other respects, and are
described in more detail in Dix et al. (2013).
Land surface process
Climate models and numerical weather prediction require
a description of the land surface and surface exchange
processes. Land surface models (LSM) provide this
information by calculating the turbulent transport of
momentum, heat and water between the land surface, canopy
and atmosphere. The thermal and hydrological processes in
the soil and snow are also simulated. The complexity and
accuracy of land surface models has increased over the last
decade. They include improved representations of canopy
processes, especially plant physiology to allow for fully
interactive terrestrial carbon cycles.
To explore the range of interactions between vegetation
behaviour and the atmosphere we are using two land
surface models coupled to the atmospheric model
in ACCESS (Kowalczyk et al. 2013). The ACCESS1.0
simulations use the same setup of the Met Office’s Surface
Exchange Scheme (MOSES) version 2.2 (Cox et al. 1999,
Essery et al. 2003) as in HadGEM2(r1.1) (Martin et al. 2011).
MOSES includes mechanistic formulations of the physical,
biophysical and biogeochemical processes that control the
exchange of momentum, radiation, heat, water and carbon
fluxes between the land surface and the atmosphere. The
land surface heterogeneity is described by having multiple
surface types in each grid cell. For MOSES there are nine
possible tiles for each grid cell and a separate energy balance
is calculated for each tile. Area-weighted grid mean fluxes
and temperatures are then calculated from the individual tile
energy balances. The surface temperatures are computed
from the same surface energy balance equation for each
vegetated and non-vegetated tile. A homogeneous soil
moisture and temperature exists for each tile within the grid.
For ACCESS1.3 the Community Atmosphere Biosphere
Land Exchange (CABLE version 1.8) has been coupled to
the UM. CABLE consists of a comprehensive description of
the surface processes that calculate momentum, heat, water
and carbon fluxes (Kowalczyk et al. 2006, Wang et al. 2010).
CABLE has 13 surface tile types (ten vegetated tile and three
non-vegetated tile types). The underlying soil is also tiled,
allowing for sub-surface soil temperature and moisture
tiling. CABLE was formulated on the basis of a multi-layer
model (Leuning 1995) and represents the canopy as a one
layer, two-leaf canopy as described in Wang and Leuning
(1998). CABLE has been extensively evaluated (e.g. Wang
et al. 2011), and a similar version has been coupled with a
low resolution global circulation model and run for multi-
hundred years to explore the impact of land use induced
land cover change on climate extremes (Avila et al. 2012).
There are a number of differences in the representation
of the canopy between CABLE and MOSES. Firstly, MOSES
places the canopy alongside a bare ground tile (a horizontal
tile approach), whereas CABLE conceptually places a
canopy above the ground allowing for aerodynamic and
radiative interaction between the canopy and the ground.
Secondly, CABLE differentiates between sunlit and shaded
leaves (two-leaf model) for the calculation of photosynthesis,
stomatal conductance and leaf temperature. Finally CABLE
includes a plant turbulence model to calculate the air
temperature and humidity within the canopy.
MOSES uses prescribed surface albedo, including
the soil albedo and canopy albedo. CABLE uses the same
prescribed, spatially varying soil albedo but resolves the
canopy albedo diurnally as a function of beam fraction, the
sun angle, canopy leaf area index, leaf angle distribution and
the transmittance and reflectance of the leaves. This results
in generally smaller surface albedo than that used in MOSES
(Kowalczyk et al. 2013).
Note that while MOSES and CABLE are both able to
calculate carbon fluxes, these have not been assessed or
submitted for the ACCESS1.0 and ACCESS1.3 simulations.
We define the UM-MOSES/CABLE grid land-sea mask,
especially its fractional land-points at coastlines, by using
the underlying ACCESS-OM land-sea mask to obtain the
best consistency of land-sea mask between the atmosphere-
land and ice-ocean subsystems. This is critical for ensuring
conservation of river runoff when passed from the UM into
CICE by the OASIS3 remapping. The resultant land fraction
is used to scale up river runoff water volume before it is sent
to the coupler. This is done to compensate for what would
be remapped by OASIS3 onto land-points of the target grid
and therefore lost in the masking. Doing so guarantees
that the volume of water actually going into the ice-ocean
system matches the real runoff amount diagnosed in the
atmosphere-land component.
The ocean component of ACCESS-CM is an implementation
of the NOAA/GFDL MOM4p1 numerical code (Griffies 2009)
which has previously been used as the ocean component for
NOAA/GFDL contributions to CMIP3 (Griffies et al. 2005,
Gnanadesikan et al. 2006) and CMIP5 (Griffies et al. 2011,
Dunne et al. 2012). The ACCESS-CM implementation uses
a Boussinesq (volume conserving) approximation for the
ocean interior and real (mass) fluxes of freshwater at the
upper surface for exchanges of precipitation, evaporation,
runoff from land, and both the melting and freezing of sea-
ice. This formulation permits a time varying ocean volume
(and sea surface height) according to the conservation (or
otherwise) of the hydrological cycling both within and
between the various components of the ACCESS-CM. More
details on the non-conservation of ocean mass in ACCESS1.0
and ACCESS1.3 are given in Marsland et al. (2013).
46 Australian Meteorological and Oceanographic Journal 63:1 March 2013
The ACCESS ocean model has 360 longitude by 300
latitude points on a logically rectangular matrix with
50 vertical levels. The horizontal discretisation is on an
orthogonal curvilinear grid nominally one degree for both
longitude and latitude. It has the following refinements: (1) a
tripolar grid (Murray 1996) is used north of 65°N to preclude
a singularity at the geographical north pole; (2) a cosine
dependent (Mercator) grid is used south of 30°S to avoid
large grid cell aspect ratios and also to better resolve zonal
currents in the Southern Ocean and at the Antarctic margin;
and, (3) a refinement of latitudinal spacing to 1/3° is applied
between 10°S and 10°N to better resolve predominantly
zonal equatorial ocean currents. The vertical discretisation
employs the z* coordinate of Adcroft and Campin (2004),
and allows for partial grid cells at the base of the water
column. The z* formulation allows for the pressure loading
of sea-ice to be accounted for when determining sea level
evolution, and avoids the possibility of a disappearing upper
ocean level in the case where sea-ice thickness exceeds the
ocean’s upper level thickness (10 m), while the atmospheric
pressure is treated as a constant and thus ignored. There are
50 vertical levels spanning the 0–6000 m depth range, with
20 levels each of nominal 10 m thickness in the upper ocean.
Below 200 m the vertical resolution smoothly decreases, with
the deepest level having a thickness of approximately 333 m.
The prognostic variables are conservative temperature
(McDougall 2003), salinity, ‘zonal’ and ‘meridional’ velocities
locally aligned to the horizontal discretisation, and sea
level displacement from an idealised sphere representing
the ocean at rest. Details of the choices of ocean physics
and sub-gridscale parameterisation settings used in the
ACCESS1.0 and ACCESS1.3 simulations are documented
in an accompanying ACCESS Ocean Model (ACCESS-
OM) benchmarking paper (Bi et al. 2013), which compares
ACCESS-OM performance in a Coordinated Ocean-ice
Reference Experiment against simulations from other
climate modelling centres (Griffies et al. 2009). The ocean
component uses the K-profile parameterisation (KPP)
scheme (Large et al. 1994); a modified skew diffusive flux
form (Griffies 1998) of the Gent and McWilliams neutral sub-
gridscale eddy advection parameterisation (Gent et al. 1995);
and mixed layer restratification by submesoscale eddies
following the scheme of Fox-Kemper et al. (2008, 2011).
There are two notable differences between the
ACCESS-OM configurations used in Bi et al. (2013) and the
ACCESS1.0 and ACCESS1.3 experiments as submitted to
CMIP5. Firstly, the ACCESS-OM used an explicit convection
scheme following Rahmstorf (1993), but the ACCESS1.0
and ACCESS1.3 models use an implicit convection scheme
where instabilities are only partially removed via a large
vertical diffusivity (0.1 m2 s–1) which introduces a timescale
for convective events rather than the instantaneous
convection of the explicit scheme. Secondly, we note that
both ACCESS-OM and the experiments considered here use
a reduced background vertical diffusivity near the equator,
with this diffusivity reducing to a minimum at the equator.
This implementation is based both on the motivations and
the scheme of Jochum (2009). Details of the implementation
in ACCESS are given in Bi et al. (2013). However, outside
of the equatorial band the ACCESS-OM experiment used a
background vertical diffusivity of 1.0 × 10–5 m2 s–1 while the
ACCESS1.0 and ACCESS1.3 simulations used half that value
(0.5 × 10–5 m2 s–1).
There are also two differences in the formulation of the
ocean model between the ACCESS1.0 and ACCESS1.3
contributions to CMIP5. Firstly, as discussed by Rashid et
al. (2013b) a key aspect of coupled model simulations is the
representation of ENSO. The critical Richardson number in
the KPP mixed layer scheme (Large et al. 1994) was halved
from 0.3 in ACCESS1.3 to 0.15 in ACCESS1.0. This change
was motivated by improved (increased) amplitude for peaks
in the power spectrum of the Nino3 SST index of interannual
variability in ACCESS1.0. Secondly, the time steps (3600 s)
of the ocean model, the sea-ice model, and the coupling
frequency between ocean and sea-ice models, were
concurrently halved from 3600 to 1800 seconds as necessary
to overcome infrequent and intermittent numerical
instabilities in the ACCESS1.3 simulations. Further details of
the ACCESS-OM component can be found in Bi et al. (2013),
while a selection of supplementary ocean results from the
ACCESS1.0 and ACCESS1.3 contributions to CMIP5 can be
found in Marsland et al. (2013).
The LANL CICE4.1 model represents the state-of-the-art
in sea-ice modelling (e.g. Flocco et al. 2012, Holland et al.
2012) and has been designed to couple with the ocean and
atmosphere components of climate models. CICE4.1 uses
an elastic-viscous-plastic dynamics scheme (Hunke and
Dukowicz 1997) for the internal ice stress, an incremental
linear remapping for the ice advection term, and computes
the ice thickness redistribution through ridging and
rafting schemes by assuming an exponential redistribution
function. The model is divided into five thickness categories
of ice and open water. CICE in the ACCESS coupled system
runs on the ocean grid which gives enhanced resolution in
the Arctic due to the orthogonal curvilinear tripolar grid and
in the Antarctic due to the Mercator grid with meridians
converging at the South Pole (Bi and Marsland 2010, Uotila
et al. 2012).
For coupling CICE to the atmospheric model (UM), a
special thermodynamic configuration is enforced because
of the implicit boundary layer scheme used in the UM
over sea-ice. The UM atmospheric boundary layer scheme
involves the direct calculation of sea-ice or snow surface
temperature, and therefore for compatibility, CICE is not
allowed to use its layered thermodynamics scheme and
compute surface temperatures. Therefore, as in the Met
Office HadGEM3(r1.1) model (Hewitt et al. 2011), CICE in the
ACCESS model uses a so called zero-layer thermodynamic
model (Semtner 1976), which is a simplified form of the
layered sea-ice thermodynamic model with no heat storage
Bi et al.: The ACCESS coupled model: description, control climate and evaluation 47
at all, except the latent heat associated with ice formation.
In addition, because of coupling to the UM, CICE employs
a simplified sea-ice albedo parameterisation, where fixed
albedo values are set in the UM radiation code for dry snow,
wet melting snow and bare ice, with a reduction dependent
on temperature near the melt point to simulate melt ponds.
Whilst ACCESS1.0 uses the ‘default’ settings of albedos
from the Met Office HADGEM2 (r1.1) model, ACCESS1.3
uses slightly larger values based on observations of Pirazzini
(2008). This is for the purpose of enhancing the simulated ice
thickness in the central Arctic and in the Southern Ocean,
which was found to be excessively thin in ACCESS1.0.
However this causes excessively thick ice next to the
Antarctic continent that does not melt out each summer
(see Fig. 12). Values of the thermodynamic and dynamic
parameters used in the ACCESS-CM CMIP5 experiments
are listed in Table 1.
Coupler and coupling strategy
The ACCESS-CM model uses the CERFACS OASIS3.2-5
coupler (hereafter OASIS3) (Valcke 2006). The implementation
in the ACCESS-CM follows closely that in the ACCESS-OM
described in Bi and Marsland (2010) and Bi et al. (2013).
At run time, OASIS3 in the coupled model remains as a
separate mono-process executable, receiving, interpolating,
and sending coupling fields between the sub-models which
also run as separate executables at the same time. As the
ACCESS-CM configures CICE4.1 and MOM4p1 on the
same global tripolar grid in the horizontal, and as both these
components use the Arakawa B-grid, coupling between
CICE4.1 and MOM4p1 is relatively straightforward.
In ACCESS-CM, as in the ACCESS-OM, the sea-ice model
is literally placed between the atmospheric model and ocean
model, working as a ‘coupling medium’ and technically being
the only one that needs to communicate with the other two
sub-models at the same time. Namely, all coupling fields from
the source model (UM/MOM4p1) are gathered, processed
jointly with the associated coupling fields from sea-ice
itself where present, and then delivered to the target model
(MOM4p1/UM). The above design has several technical
advantages, including easy control over the coupling
frequencies. The ACCESS-CM uses different frequencies
for coupling atmosphere to sea-ice (every three hours, i.e.
six atmospheric time steps) and sea-ice to ocean (every time
step of the ice and ocean models, typically one hour). This
coupling strategy is illustrated by Fig. 2 of Bi et al. (2013).
Use of OASIS3 requires the development of a coupling
interface for each sub-model. The interface for the UM in
both versions of the ACCESS-CM is based on that present
in the UM7.3 code and used in HadGEM3 (r1.1) (Hewitt et
al. 2011). In ACCESS-CM, this interface is slightly modified
for connecting atmosphere to sea-ice and handling 42 two-
dimensional coupling fields (24 from atmosphere to ice and
18 from ice to atmosphere) via OASIS3. Similarly, an interface
is implemented in MOM4p1 for connecting ocean to sea-ice
and handling 20 coupling fields (seven from ocean to ice and
13 from ice to ocean). In the ‘coupling medium’ CICE4.1, the
interface is designed to connect ice to both atmosphere and
ocean, handling (receiving, processing, and delivering) all
the 62 coupling fields.
While the exchange of coupling data between ocean
and sea-ice needs no transformation via the OASIS3 main
process because of grid compatibility, the coupling fields
between atmosphere and sea-ice are remapped by OASIS3
to the target grid using a first order conservative remapping
algorithm of the Spherical Coordinate Remapping and
Interpolation Package (SCRIP) (Jones 1997).
Experimental design
The experimental framework for the CMIP5 simulations
for both ACCESS1.0 and ACCESS1.3 involves initialisation
using near-present day conditions, a multi-century spin-up
run using the CMIP5 preindustrial conditions, followed by
commencement of the actual CMIP5 preindustrial control
(hereafter piControl) and historical simulations at the same
point. The spin-up runs for both versions were initialised
using an atmospheric and land surface state obtained for
1 January 1979 from an atmospheric/land surface model
simulation started using fields obtained from the Met Office
for 30 September 1978, and ocean climatological temperature
and salinity fields for January from the World Ocean Atlas
2005 (WOA2005; Locarnini et al. 2006, Antonov et al. 2006).
The sea-ice model is initialised using the WOA2005 January
sea surface temperature (SST) and salinity (SSS). Any grid
point that has SST no higher than the SSS-dependent
freezing point is set to have five thickness category ice
areas which jointly fully cover the cell. While the 3 m thick
category ice has the largest area within the cell, the average
ice thickness for the cell is less than 3 m due to the nonlinear
distribution of ice thickness.
Details of the forcing data for the CMIP5 simulations are
given in Dix et al. (2013). For the preindustrial spin-up and
CMIP5 piControl simulations, standard CMIP5 preindustrial
(circa 1850) prescriptions are used for atmospheric
concentrations of CO2, CH4, N2O and O3, the solar constant,
and aerosol emissions. All these factors, together with
atmospheric halocarbon concentrations and volcanic
stratospheric aerosol load are set to vary according to the
standard CMIP5 prescription in the historical simulation.
The option to treat the preindustrial stratosphere as clear
of volcanic aerosols (Taylor et al. 2009) is taken. (Note,
however, that this will result overall in a slight cold bias in
the historical simulation relative to the piControl, see Dix et
al. (2013).) Finally, the seasonally-varying biogenic aerosol
concentration and a background volcanic SO2 out-gassing
flux (into the lower to mid troposphere) are maintained
throughout all simulations.
The preindustrial spin-up simulations for ACCESS1.0
and ACCESS1.3 are continued for 300 years and 250 years,
respectively. Ideally, the spin-up simulation should continue
until equilibrium is achieved prior to commencement of the
48 Australian Meteorological and Oceanographic Journal 63:1 March 2013
historical simulation. However, the deep ocean is known
to require thousands of years to reach equilibrium (e.g.
Stouffer 2004) which is not possible given current resource
constraints. Therefore, like other CMIP5 models (e.g. Griffies
et al. 2011, Voldoire et al. 2012), the spin-up simulations for
the ACCESS-CM are performed for an affordable length that
yield adequate quasi-equilibria of the surface fields before
the start of the CMIP5 historical and piControl simulations
It should be noted that in the ACCESS1.3 spin-up
phase, sea-ice albedos used for the first 149 years are also
the ‘default’ setting for UM, i.e. the same as that used in
ACCESS1.0. It was found that CABLE results in warmer
surface air temperature which allows larger sea-ice albedos
(closer to the Pirazzini 2008 observations) to be used for
better model climate in terms of both ice and surface air
temperature (SAT) distributions. Therefore, from year 150 of
the spin-up onwards, ACCESS1.3 uses larger sea-ice albedos
(Table 1). The effect of the albedo change on the surface
energy budget contributes to the differences in residual
drifts seen in the ACCESS1.0 and ACCESS1.3 solutions (see
section ‘Model results’).
The 500 years of the CMIP5 piControl simulation for
ACCESS1.0 covers years 300 through 799, and ACCESS1.3
years 250 through 749, from time of original initialisation.
The historical simulations also commence at the beginning
of year 300 and year 250 for ACCESS1.0 and 1.3, respectively.
Hereafter, we will redesignate the first year of the CMIP5
piControl simulation in each case as ‘1850’, for ease of
comparison to the respective historical simulation. In the
following, we will examine the extent of the residual model
adjustments during the course of the piControl simulations.
We will see that ACCESS1.0 displays a slight residual drift in
surface properties, more than for ACCESS1.3, and this was
the motivation for continuing the ACCESS1.0 pre-industrial
spin-up for the slightly longer period. We will then provide an
initial evaluation of certain aspects of the historical simulations
during the instrumental period. Subsequent manuscripts
in this issue of AMOJ will provide much more detailed
documentation and evaluation of the model behaviour.
Model results
In this section we evaluate ACCESS-CM by presenting a
selection of key fields from the piControl and historical
simulations performed with the two versions, and comparing
them against observations or their reanalysis estimations
where appropriate. We define the ‘control climatology’
as the time average over the last 100 years of the 500-year
piControl integrations and the ‘present climate’ as the 30-
year average over 1976–2005 from the historical simulations.
Control run global mean thermal equilibrium and drift at
the surface
It is desirable that a coupled model used for climate change
simulation and climate sensitivity research be able to
hold a stable, realistic control climate, especially a thermal
equilibrium at the surface. Fig. 2(a) shows the time series
of the globally averaged annual mean SAT for the 500-year
period of the ACCESS1.0 and ACCESS1.3 piControl runs. In
the ACCESS1.0 case, we see a persistent increase of SAT in
the course of the integration, and the final warming reaches
about 0.35 °C, with the drift rate being 0.07 °C/century. This
drift rate is modest in comparison to that found in the majority
of CMIP3 models, where the median drift is approximately
Table 1. Values of selected important dynamic and thermodynamic sea-ice model parameters used in the ACCESS CMIP5
Short name Value Full name
cosw, sinw 16° Ocean-ice turning angle
mu_rdg 3 m1/2 Ridging parameter value
ALPHAC 0.78 Cold deep snow albedo in ACCESS 1.0
ALPHAC 0.84 Cold deep snow albedo in ACCESS 1.3
ALPHAB 0.61 Bare ice albedo in ACCESS 1.0
ALPHAB 0.68 Bare ice albedo in ACCESS 1.3
ALPHAM 0.65 Melting deep snow albedo in ACCESS1.0
ALPHAM 0.72 Melting deep snow albedo in ACCESS1.3
DTICE 0.5 Temperature range to determine snow melting in albedo calculation
DT_BARE 0.25 Temperature range to determine bare ice melting in albedo calculation
DALB_BARE_WET –0.075 Albedo change to determine bare ice melting in albedo calculation
Dragio 0.00536 Ice-ocean drag
ustar_min 0.0005 m/s Minimum ice-ocean friction velocity
Conduct Bubbly Ice conductivity option
Iceruf 0.0005 m Surface roughness of ice
Chio 0.004 Ice-ocean heat exchange coefficient
Bi et al.: The ACCESS coupled model: description, control climate and evaluation 49
0.12 °C/century (Sen Gupta et al. 2012). For ACCESS1.3, the
whole 500-year integration shows a minimal increase of SAT.
The final warming is less than 0.04 °C and the trend is 0.008
Such a difference between the two versions indicates that
ACCESS1.0 has a bigger deficiency in energy balance at the
top of the atmosphere (TOA). This is confirmed by the time
series of energy budget at the TOA shown in Fig. 2(b) where
we see a considerable imbalance of heat fluxes throughout
the ACCESS1.0 control run. It starts from a net energy gain
of above 0.5 W m–2 and ends with a smaller value of around
0.4 W m–2 which is still quite far from the desirable ±0.1 W
m–2 criteria for long control runs (e.g. Gent et al. 2011). In
contrast the ACCESS1.3 shows a much smaller imbalance at
the TOA through the 500-year control period. It starts from
just above 0.1 W m–2 and ends with a very small positive
imbalance of about 0.03 W m–2.
Surface air temperature
Figure 3 presents the bias maps of the modelled near surface
air temperature against the ERA-Interim reanalysis data (Dee
et al. 2011) for the period of 1979–2008. Table 2 gives some
details of the area averaged biases and root mean square
errors (RMSEs) over land and oceans, in selected zonal
bands for both ACCESS1.0 and ACCESS1.3, including the
difference between the two models. The two model versions
show generally very similar bias patterns. Particularly they
both bear large cold biases over the polar regions, especially
on the sea-ice adjacent to the Antarctic continent, the
coasts of Greenland and the Canadian archipelagos. The
ACCESS1.0 historical run simulates a lower present-day
global mean SAT than ACCESS1.3 because it starts from a
considerably cooler initial condition and never catches up
with ACCESS1.3 during the course of the 156-year simulation,
similar to that shown in Fig. 2(a) for the piControl runs. This
difference in part reflects the different effects of the land
surface models (i.e. MOSES and CABLE) in determining the
model surface thermal states. Table 2 reveals that, except
for Antarctica, all land is simulated significantly colder
in ACCESS1.0 than in ACCESS1.3. ACCESS1.0 has a cold
bias as large as 0.70 °C over the northern hemisphere land,
in contrast to the small cold bias of 0.02 °C in ACCESS1.3.
The southern hemisphere land excluding Antarctica sees a
larger warm bias in ACCESS1.3 (0.92 °C) than in ACCESS1.0
(0.26 °C). Such a contrast in the surface air temperature over
land is mainly attributed to CABLE in ACCESS1.3 yielding
significantly lower land surface albedo than MOSES in
ACCESS1.0 (Kowalczyk et al. 2013).
Over the oceans, ACCESS1.3 is warmer than ACCESS1.0
in the northern hemisphere and the tropics but colder over
the Southern Ocean. Particularly, ACCESS1.3 is 1.66 °C
colder than ACCESS1.0 in high latitudes of the Southern
Fig. 2. Evolution of the piControl annual mean global average: (a) SAT (°C) and (b) TOA energy budget (W m–2). Thick lines are the
linear regressions.
Fig. 3. Historical run present-day SAT biases (relative to ERA-Interim reanalysis 1979–2008 data): (a) ACCESS1.0 and (b)
ACCESS1.3. Units are °C.
50 Australian Meteorological and Oceanographic Journal 63:1 March 2013
Ocean (>60°S). This is the effect of the larger sea-ice albedos
used in ACCESS1.3, which overcompensate the cloud
radiative forcing warming error shown in Fig. 6. It benefits
ACCESS 1.3 by producing a greater sea-ice extent at the
summer minimum, though leaves thick ice adjacent to the
coast (see the ‘Sea-ice’ subsection and also Uotila et. al (2013)
for more details).
Figure 4 shows the global maps of the present climate of
precipitation from the ACCESS1.0 historical run, together
with the observed climatology of precipitation and the
model biases of the present climate from the historical runs.
The two models simulate very similar global precipitation
distribution patterns and intensities, both fairly close to the
GPCP observation data (Huffman et al. 2009). The common
model problem of a double intertropical convergence zone
(ITCZ), which is seen in most of the IPCC AR4 models (e.g.
Lin 2007) with excess rainfall simulated over the south
side of the equator and reduction of rainfall over the other
side, is still there weakly in the ACCESS models. Globally,
ACCESS1.0 has a somewhat closer match to the observation
than ACCESS1.3 in both the mean bias (0.39 vs. 0.42 mm/day)
and RMSE (1.41 vs. 1.45 mm/day). Away from the tropical
region, the models simulate excessive precipitation nearly
everywhere; especially over the mid to high latitude oceans
because of the dominant warm biases of SST there (see Fig.
14). In fact, the rainfall biases over the oceans are primarily
determined by the SST biases, as evidenced by comparing
the rainfall bias pattern shown here and the SST bias pattern
shown in Fig. 14(c) and 14(d). For example, except for the
tropical oceans, the regions with major warm biases (such
as the oceanic frontal zones, the upwelling regions off the
west coasts of South America and North America, but with
the exception of South Africa) also have noticeable positive
rainfall errors. Over the land, notable errors of excess
precipitation are found over the Himalayas, southeast Africa,
and the Andes in particular.
Total cloud amount and cloud radiative forcing (CRF)
One of the substantial differences between the ACCESS1.0
and the ACCESS1.3 atmospheric configurations is the cloud
schemes. It is well recognised that clouds play a crucial role
in modulating the climate with changes in the location or
frequency of cloud distributions impacting both regional
and global climate. This is because clouds constitute one of
the major factors in determining the earth radiation budget,
largely controlling the atmospheric circulation, hydrological
cycle, and the surface energy budget (e.g. Stephens et al.
1990, Yao and Del Genio 1999, Williams et al. 2006, IPCC
2007). The uncertainty about the magnitude and sign of the
cloud feedback on climate is considered one of the major
obstacles in improving climate change prediction, and
therefore improvements in the representation of clouds are
an extremely challenging but crucial goal for climate and
climate change modelling (e.g. IPCC, 2007). The availability
of satellite observations for the past few decades provides
a critical measurement of model performance in simulating
cloud cover and cloud radiative forcing (CRF).
Figure 5 compares the modelled present day annual
mean total cloud amount against the D2 ISCCP observations
(Rossow and Schiffer 1999) for the period of 1983–2005.
Both models simulate the global distribution patterns
quite well. The highest cloud cover over the Tropical Warm
Pool, Southern Ocean and North Atlantic Ocean are well
captured, as are the minima over the subtropical desert
regions. However, model deficiencies are evident and are
different for ACCESS1.0 and ACCESS1.3 over various
regions, as detailed by the two error maps and Table 3 which
presents the modelled cloud amount biases and RMSEs over
Table 2. The historical run present-day surface air temperature biases and RMSEs in ACCESS1.0 and ACCESS1.3 against the ERA-
Interim reanalysis (Dee et al. 2011) data for 1979–2008. Values are presented for the groups of land and oceans, and for
different zonal bands. (Units: °C).
Bias RMSE bias RMSE Bias RMSE
All land –0.38 1.96 0.14 2.21 –0.53 –0.25
0–90°N –0.70 1.88 –0.02 2.12 –0.68 –0.24
60°S–0° 0.26 1.44 0.92 1.93 –0.66 –0.49
90°S–60°S 0.36 3.20 –0.53 3.25 0.89 –0.06
All oceans –0.15 1.26 –0.06 1.40 –0.09 –0.13
60°N–90°N –1.76 2.74 –1.16 2.73 –0.59 0.01
20°N–60°N –0.51 1.17 –0.10 1.12 –0.41 0.05
20°S-20°N –0.13 0.86 0.17 0.95 –0.30 –0.09
60°S–20°S 0.45 0.98 0.34 0.85 0.11 0.12
90°S–60°S –0.99 2.52 –2.65 3.63 1.66 –1.11
Global -0.21 1.48 –0.01 1.65 –0.20 –0.17
Bi et al.: The ACCESS coupled model: description, control climate and evaluation 51
different zonal bands. For ACCESS1.0, the globally averaged
cloud fraction is 53.8 per cent, considerably lower than the
ACCESS1.3 result which is 65.8 per cent, very close to the
observed value of 66.5 per cent. The error in ACCESS1.0
results from the systematic underestimation of cloud cover
between 60°N and 60°S, particularly the subtropical regions
around 20°N and 20°S in the Indo-Pacific oceans where the
largest underestimate is 30 per cent. For ACCESS1.3 the cloud
cover error is noticeably more complicated. While the model
generally underestimates cloud cover over the majority of
the oceans between 60°S–60°N, it also overestimates cloud
cover in parts of the tropics, with an average bias of 2.6 per
cent within 10°S–10°N where ACCESS1.0 in contrast has an
average cloud cover bias of –13.9 per cent. In polar regions,
Fig. 5 shows that ACCESS1.3 overestimates cloud cover.
However, ISCCP has significant uncertainties in these regions
and Franklin et al. (2013) has shown that ACCESS1.3 cloud
cover is in good agreement with CALIPSO observations in
the high latitudes. Globally, ACCESS1.3 has a somewhat
smaller RMSE than ACCCESS1.0, and the difference would
be considerably enlarged when high latitudes are excluded
from the calculation, as shown in Table 3. This indicates a
significantly better simulation of cloudiness in ACCESS1.3
(with the PC2 cloud scheme) than in ACCESS1.0 (with the
Smith cloud scheme).
CRF is a measure of how clouds affect the radiation
budget at the top of the atmosphere and is calculated as the
difference between the clear-sky and all-sky radiation. CRF
influences surface heating gradients, and consequently can
impact large-scale circulations and ocean heat transports. It
is not only the total cloud cover that determines the CRF but
also the vertical distribution of cloud, the condensate amount
and the microphysical properties such as the effective radius.
Figure 6 shows the short-wave, long-wave and net CRF errors
Fig. 4. Precipitation climatology and biases: (a) GPCP precipitation (Huffman et al. 2009) 1979–2005 mean, (b) ACCESS1.0 present
climate, (c) ACCESS1.0 biases, and (d) ACCESS1.3 biases. Units are mm/day.
Table 3. Historical run present-day annual mean cloud amount biases and RMSEs in ACCESS1.0 and ACCESS1.3 against the D2 IS-
CCP observations for 1983–2005. Values are presented for different zonal bands and units are in per cent. Note the 60°–90°
bands are excluded from the comparison because of the poor quality of observations in high latitudes.
0–60°N –13.8 15.6 –2.3 9.3 –11.5 6.3
60°S–0 –15.0 16.6 –4.8 10.6 –10.2 6.0
10°S–10°N –13.9 15.7 2.6 10.8 –16.5 4.9
60°S–60°N –14.4 16.1 –3.6 10.0 –10.9 6.1
52 Australian Meteorological and Oceanographic Journal 63:1 March 2013
for ACCESS1.0 and ACCESS1.3 compared to the ISCCP
observations. ACCESS1.3 simulates a global average annual
mean total CRF of –17.9 W m–2, weaker than the ACCESS1.0
result of –19.7 W m–2 which is closer to the observed value
of –24.3 W m–2. This is despite ACCESS1.0 simulating much
less total cloud cover globally than the observations (Fig. 5)
and demonstrates the importance of the cloud properties
and vertical distribution of clouds for determining the
CRF. The long-wave (LW) CRF is better simulated than
short-wave (SW) CRF in both models in terms of the global
mean error and spatial patterns. ACCESS1.3 has a smaller
LW-CRF error than ACCESS1.0 in the tropics, suggesting
that ACCESS1.3 has a better representation of the high
clouds associated with deep convection. ACCESS1.3 shows
larger SW-CRF errors than ACCESS1.0 over the Southern
Ocean, particularly south of 60°S where the average SW
CRF biases (over ocean) are 13.1 W m–2 and 5.5 W m–2 for
ACCESS1.3 and ACCESS1.0, respectively. Given that Fig. 5
shows that ACCESS1.3 typically has higher cloud cover over
the Southern Ocean, the weaker SW-CRF in ACCESS1.3
suggests that the clouds in this region are optically too thin.
On the whole the total CRF errors are generally lower for
ACCESS1.0 compared to ACCESS1.3 and this is partly due
to better cancellation of errors between the SW-CRF and
LW-CRF in ACCESS1.0, particularly in the tropics.
Model skill scores in simulating a selection of key fields
For a graphical depiction of skill in simulating the present
climate, we adapt the histogram of skill scores of Gordon et al.
(2010), following Watterson (1996). For each of 13 quantities
and each of the four seasons, the global field from a model
is compared to the best available observational field using
the non-dimensional statistic M, as defined in the caption.
The average of the four seasonal scores for each quantity is
shown in Fig. 7, for each of three models. A value of one for
M indicates perfect agreement, while zero indicates no skill.
As can be seen each model produces considerable skill in
each quantity. In most variables, in particular u500 and psl,
Fig. 5. Annual mean total cloud amount: (a) ISCCP climatology for 1983–2005, (b) ACCESS1.0 present-day climate, (c) ACCESS1.3
present-day climate, (d) ACCESS1.0 biases, and (e) ACCESS1.3 biases. Units are in per cent. Note that the ISCCP data has a
notable discontinuity in the Indian Ocean because of a gap in coverage by geostationary satellites between Meteosat and
GMS (Rossow and Garner 1993).
Bi et al.: The ACCESS coupled model: description, control climate and evaluation 53
the ACCESS models provide a step up in skill compared to
Mk3.5. Consistent with the RMSE values noted previously,
ACCESS1.0 performed a little better than ACCESS1.3 for
tas, pr and crf, but less well for clt. Averaging over the 13
variables gives M values of 0.77 for ACCESS1.0, 0.76 for
ACCESS1.3 and 0.72 for Mk3.5.
The ACCESS model data are 30-year averages from
the historical run, as above. The Mk3.5 fields are as in
Gordon et al. (2010, from CMIP3 20C3M, 1961–90). For
observations, we use here the ERA-interim data set (Dee et
al. 2011), with ISCCP cloud (Rossow and Schiffer 1999) data
as above. For the TOA quantities (rlut, rsut, crf), the fields
are based on CERES satellite data over 2000–2005 provided
by Pincus et al. (2008). As shown by Gordon et al. (2010),
there is considerable uncertainty in the true climatologies,
particularly for precipitation, cloud forcing and total cloud,
which impacts on the scores. For further assessment of
atmospheric circulations, see Rashid et al. (2013a).
Time series of sea-ice extents and volume in the
piControl runs
The sea-ice extent and volume time series from the ACCESS-
CM piControl runs are shown in Fig. 8. The Arctic time
series of ACCESS1.3 and ACCESS1.0 are relatively close, but
in the Antarctic, the ACCESS1.3 sea-ice extent and volume
are clearly higher than their ACCESS1.0 counterparts due to
the effect of higher sea-ice albedo more than compensating
for the larger CRF warming bias in ACCESS1.3 (Fig. 6(f)).
The time series show large interannual and multi-decadal
variability. ACCESS1.0 has a statistically significant 500-
year sea-ice volume trend of –0.36 (–0.42) per cent/decade
in the northern (southern) hemisphere. The corresponding
ACCESS1.3 trends are –0.07 (+0.13) per cent/decade, which
are also statistically significant, but clearly smaller than the
ACCESS1.0 sea-ice volume drift. ACCESS1.0 is losing ice
in both hemispheres, while ACCESS1.3 has hemispheric
trends with opposite signs.
Fig. 6. Simulated annual mean cloud radiative forcing biases, relative to the ISCCP observations 1979–2008 mean, for short-wave
(top panels), long-wave (middle panels) and total CRF (bottom panels). Left and right panels are for ACCESS1.0 and
ACCESS1.3, respectively. Units are W m–2.
54 Australian Meteorological and Oceanographic Journal 63:1 March 2013
Ice coverage
The annual cycle of the sea-ice extent for the ACCESS1.0
and ACCESS 1.3 piControl runs is shown in Fig. 9.
Modelled extents are rather close to the observed sea-ice
extent climatology based on 1979–2005 monthly means.
This indicates that the modelled sea-ice extents might be
too small, especially in summer in the Arctic (Fig. 9(a)),
because the climate during the historical period, used in the
observations, is likely to be warmer due to the influence of
greenhouse gas emissions than the piControl climate. On the
other hand, the pre-industrial runs show a significant natural
interannual variability having periods with temperatures
similar to the historical era (Uotila et al. 2013). The impact of
higher sea-ice albedo in the ACCESS1.3 simulation results
in a systematic difference between the ACCESS1.3 and
ACCESS1.0 sea-ice extent climatologies in the Antarctic
(Fig. 9(b)), but not in the Arctic (Fig. 9(a)), where the location
of land masses controls the sea-ice extent. A relatively low
summer Arctic sea-ice extent could be due to the warm
atmosphere and/or the ocean opening leads and polynyas
in a thin ice field, which then enables the effective ice-albedo
feedback process during the melting season and dominates
Fig. 7. The skill of models ACCESS1.0, ACCESS1.3 and Mk3.5
in reproducing observational climatological fields of
various quantities. The bars give M scores for the
global domain, averaged over four seasons, where M
= (2/π) arcsin [1 – mse / (VX+VY + (GX – GY)2)], with mse
the mean square error between the model field X and
observed field Y, and V and G are spatial variance and
mean of the fields (as subscripted). The 13 fields pre-
sented here are: tas—surface air temperature; t500,
u500, and r500—temperature, zonal wind and relative
humidity, respectively, at 500 hPa; psl—sea level
pressure; rsds—short-wave radiation down at sur-
face; rlut, rsut, crf—long-wave radiation up, short-
wave radiation up and cloud radiative forcing, re-
spectively, at the TOA; prw—atmosphere water
vapour content; pr—precipitation; evap—evapora-
tion; clt—total cloud amount.
Fig. 8. The annual mean time series of ACCESS1.0 (red line)
and ACCESS1.3 (blue line) piControl simulations for
(a) Arctic and (b) Antarctic sea-ice extent, and for (c)
Arctic and (d) Antarctic sea-ice volume. Sea-ice extent
is defined as the area of grid cells comprising at least
15% of ice.
Fig. 9. Annual cycle of sea-ice extents for: (a) Arctic and (b)
Antarctic. Black lines are the NSIDC observations for
the 1979–2005 period, and red and blue lines are the
piControl climatology (last 100-year mean) of AC-
CESS1.0 and ACCESS1.3, respectively. Units: 106 km2.
Fig. 10. Annual average sea-ice concentration of the last 100-
year piControl simulations: (a) ACCESS1.0 Arctic, (b)
ACCESS1.0 Antarctic, (c) ACCESS1.3 Arctic, and (d)
ACCESS1.3 Antarctic. The 15% isoline of the observed
SSMI sea-ice concentration for 1979–2000 is marked
as a black continuous line.
Bi et al.: The ACCESS coupled model: description, control climate and evaluation 55
the annual cycle of the sea-ice thickness (Laxon et al. 2003),
but it is not clear if this is the case in the model results.
Annual geographical mean sea-ice concentrations look
reasonable and generally follow closely the observed 15 per
cent sea-ice concentration isolines (Fig. 10). Additionally, the
differences between the ACCESS1.0 and the ACCESS1.3
sea-ice concentration fields look small. In the Barents Sea,
the ice edge in both ACCESS runs is too far north indicating
that the ice is melting too much in summer. The ACCESS1.3
run seems to have a slightly larger area of very high sea-
ice concentration of 99 per cent in the Central Arctic than
the ACCESS1.0 run, probably due to the higher albedo. In
the Antarctic, the simulated sea-ice edge is relatively close
to the observed, which might indicate too small ice coverage
for pre-industrial runs. The ACCESS1.3 ice concentration is
clearly higher than the ACCESS1.0 ice concentration, and,
as in the Arctic, this signifies the impact of the higher sea-ice
albedo in the ACCESS1.3 simulation.
Ice thickness
The sea-ice volume between the ACCESS simulations
(Fig. 11) is systematically higher in ACCESS1.3 than in
ACCESS1.0 every month, which is not the case for the sea-
ice extent (Fig. 10). This indicates the well known fact that
the albedo impacts more on the sea-ice thickness than the
sea-ice concentration (see for example Uotila et al. 2012),
because the thinning of ice always reduces the ice volume,
while the ice concentration decreases only after the ice has
melted completely. Accordingly, the difference between
the ACCESS1.0 sea-ice volume and the ACCESS1.3 sea-ice
volume is larger during summer than in winter in both the
Antarctic and the Arctic. As with the Antarctic sea-ice extent,
ACCESS1.3 has a systematically higher sea-ice volume than
ACCESS1.0, which indicates significantly thicker ice.
Figure 12 shows the annual mean sea-ice thickness of the
last 100 years of the piControl simulations by ACCESS1.0
and ACCESS1.3. The thickest Arctic sea-ice in the ACCESS
model appears in the narrow channels of the Canadian
Arctic Archipelago, where air temperatures are cold and the
ice is almost stationary. ACCESS1.0 has relatively thin ice,
2–2.5 m, in the Arctic Ocean north of Greenland, where ice
thicknesses of 4–5 m have been observed (Fig. 12(a); Kwok
and Cunningham 2008).The ACCESS1.3 ice is thicker north
of Greenland, 3–4 m, but the relatively thick ice is transported
by the strong anticyclonic Beaufort Gyre resulting in too
thick ice in the central Arctic, while the ice is rather thin
north of the Barents Sea close to the sea-ice edge.
In the Southern Ocean, the sea-ice is too thin especially
in ACCESS1.0 (Fig. 12(b) and 12(d)). Close to the Antarctic
continent, however, the sea-ice is thicker. Away from the
Antarctic coast in the Southern Ocean, the ACCESS ice
thickness is generally 0.1–0.5 m, which is less than observed
(Worby et al. 2008), although the ACCESS1.3 sea-ice
thickness looks more realistic than ACCESS1.0. The thin
Antarctic ice suggests that the models do not simulate the
deformation of ice due to rafting adequately in the Southern
Ocean and/or the ice is melted by the oceanic heat (Uotila et
al. 2012, Marsland et al. 2013).
The oceans play a fundamental role in adjusting and
stabilising the global climate. Ocean properties such as
the sea surface temperature (SST) and salinity (SSS) are
commonly used for model evaluation. The ocean surface
thermohaline condition is the joint product of ocean
thermodynamic and dynamic processes, along with ocean
coupling to the overlying atmosphere and sea-ice (via heat,
mass and momentum exchange) and land (river runoff
input). The modelled evolutions of global mean SST (which
is strongly coupled to the SAT over the open ocean) and
Fig. 11. The annual cycle of sea-ice volume averaged over
last 100-year control simulations of ACCESS1.0 (red
line) and ACCESS1.3 (blue line).
Fig. 12. Annual mean sea-ice thickness of the last 100-year pi-
Control simulations: (a) ACCESS1.0 Arctic, (b) AC-
CESS1.0 Antarctic, (c) ACCESS1.3 Arctic, and (d) AC-
CESS1.3 Antarctic. The 15% isoline of the observed
SSMI sea-ice concentration for 1979–2000 is marked
as a black continuous line. Units: m.
56 Australian Meteorological and Oceanographic Journal 63:1 March 2013
SSS demonstrate the climate stability, sensitivity, and, to a
great degree, the reliability of a coupled model. The spatial
distribution of the climatology and biases of SST and SSS
provide a primary metric useful for evaluating model
performance, including buoyancy and momentum fluxes,
and surface currents.
Sea surface temperature
Figure 13(a) presents the annual mean global average
SST evolution in the ACCESS-CM piControl simulations.
As expected, SST evolves in a manner very similar to that
of the SAT shown in Fig. 2(a). ACCESS1.0 has a small but
noticeable trend in the control simulation with a final
warming of about 0.42 °C and an average drift rate of 0.08
°C/century, and ACCESS1.3 holds a very stable sea surface
equilibrium, with a nearly unnoticeable SST trend (<0.01 °C/
century) throughout the control integration.
Figure 13(b) shows evolution of the ocean surface heat
budget which consists of all the heat items (radiative and
turbulent heat fluxes, ice formation-melt heat flux and
precipitation associated heat flux), for the control runs of
ACCESS1.0 and ACCESS1.3. Both models have a small
but persistent imbalance of energy across the control
integrations. The 500-year mean imbalances are only +0.15
and –0.15 W m–2 for ACCESS1.0 and ACCESS1.3, respectively,
both being well below a traditional level of acceptance of
±1.0 W m–2 (e.g. Voldoire et al. 2012, Lucarini and Ragone
2011). These imbalances reflect the thermal adjustments
occurring at the sea surface and in the ocean interior. For
ACCESS1.0, this positive energy imbalance and the large
warm biases found in the subsurface layers (see Marsland
et al. 2013) jointly explain the lasting SST warm drift shown
in Fig. 13(a). For ACCESS1.3, the continuous heat loss at the
surface does not lead to a decrease in the global SST because,
as in ACCESS1.0, the subsurface layers are overheated and
the warm impact on the surface layer cancels and slightly
outweighs the heat loss at the surface. It is noted that the
trend of the ocean surface energy budget for the last 100-year
has changed direction in the ACCESS1.3 case, indicating an
improving heat balance for the ACCESS1.3 ocean.
Figures 13(c) and 13(d) show the SST drift maps for the
two piControl runs using their last century average minus
first century average. For ACCESS1.0 (Fig. 13(c)), the global
mean SST change is just below 0.2 °C, but large regional
drifts are evident in mid to high latitudes, and a maximum
warming of about 1.4 °C is found in the northern hemisphere
high latitude oceans. The Southern Ocean sees a maximum
warming drift of nearly 1.0 °C south of Australia. The drifts in
low latitudes are generally small. For ACCESS1.3 (Fig. 13(d)),
the global mean warming drift is less than 0.04 °C, and there
are no evident regional drift patterns in the world ocean. It is
Fig. 13. (a) piControl time series of SST (in °C) , (b) piControl time series of sea surface energy budget (in W m–2), (c) map of SST
drift in ACCESS1.0, and (d) map of SST drift in ACCESS1.3. The time series is for global (sea surface) average annual mean,
with thick lines being the linear regressions of the curves over the entire 500-year integration and the last 100-year period
(green for ACCESS1.0 and cyan for ACCESS1.3). The maps of drift (°C) are for years 401–500 mean – years 1–100 mean.
Bi et al.: The ACCESS coupled model: description, control climate and evaluation 57
interesting to note that, while the low latitudes see nearly no
SST trend at all in the 500-year run, the northern hemisphere
high latitudes generally warm up, but the majority of regions
in the Southern Ocean slightly cool down.
Figure 14 shows the SST biases of the ACCESS-CM
piControl climate and historical present-day climates. We
take the SST reconstruction for 1870–1899 from the HadISST
dataset (Rayner et al. 2003) as the observation for the control
runs to compare against. It is already noticed from Fig. 13(a)
that the ACCESS1.0 piControl starts from a colder initial
SST condition than ACCESS1.3 (by about 0.2 °C), warming
up slowly in the course of the run, and catching up with
ACCESS1.3 in the last 100 years of integration. Hence the
two runs show similar SST piControl climatology, with global
mean biases being about 0.25 °C and 0.26 °C, respectively.
The RMS errors, 1.12 °C for ACCESS1.0 and 1.02 °C for
ACCESS1.3, are moderate in both runs, but the difference
indicates that ACCESS1.3 is in slightly better agreement
with the observation, as evidenced by the details of the
biases shown in Fig. 14(a) and 14(b). While the two runs
have similar bias patterns nearly everywhere in the global
ocean, ACCESS1.3 has larger areas than ACCESS1.0, where
the biases are less than 1 °C, especially in the Pacific Ocean.
Over the whole Southern Ocean, particularly the Indian
Ocean section, the warming biases are significantly smaller
in ACCESS1.3 than ACCESS1.0. As discussed above (i.e.
the SAT subsection), this warming bias reduction is mainly
attributed to the larger ice albedos used in ACCESS1.3,
which result in much better simulation of the Antarctic sea-
ice (as shown in the ‘Sea-ice’ subsection) and thus produce a
more realistic SST near the Antarctic. However, ACCESS1.0
shows somewhat smaller cold biases in the tropical Pacific,
which has important implications for the simulated ENSO
behaviour (see Rashid et al. 2013b).
Apart from the above features, the two runs have
comparable large biases in various regions. In the North
Atlantic, for example, the cold error southeast of the
Labrador Sea and the adjoining warm errors to the south are
associated with a poor representation of the Gulf Stream path
(too far south) and the model’s inability to explicitly resolve
eddy transports. As pointed out by Griffies et al. (2011), this
region, as well as some other major frontal zones such as the
North Pacific western boundary current (Kuroshio Current),
has strong SST gradients, and small errors in the simulated
flow location and intensity can lead to large SST errors. In
addition, broad cooling biases are evident at both sides of the
equator in the Atlantic Ocean, mainly attributed to the model
producing too strong SW cloud radiative forcing there, as
shown in Fig. 6. This is the case for both ACCESS1.0 and
Fig. 14. SST biases of: (a) ACCESS1.0 piControl climatology, (b) ACCESS1.3 piControl climatology, (c) ACCESS1.0 historical present
climate, and (d) ACCESS1.3 historical present climate. For the piControl and historical biases, the reference data is the
HadISST (Rayner et al. 2003) 1870–1899 SST reconstruction and the 30-year observation over 1976–2005, respectively.
Units: °C.
58 Australian Meteorological and Oceanographic Journal 63:1 March 2013
Another apparent deficiency seen from the bias maps
is the warming bias off the west coast of America and
Africa. As shown by Griffies et al. (2009), this is an intrinsic
deficiency shared by nearly all state-of-the-art ocean models,
most likely attributable to the model underestimating the
upwelling and associated westward mass transport due to
the coarse resolution of the model grids in both the ocean
and atmosphere.
All the major features of the piControl SST biases
discussed above also appear in the historical present-day
climate, as shown by Fig. 14(c), 14(d). Again, ACCESS1.3
has significantly less warm biases in the Southern Ocean
but broader cold biases across the equatorial Pacific Ocean
than ACCESS1.0. While the other major similarities and
differences between the pair of control runs are well reflected
in the historical runs, the global mean biases (–0.072 °C and
–0.016 °C) and the RMS errors (1.04 °C for ACCESS1.0 and
0.96 °C for ACCESS1.3) are mostly reduced, partly attributed
to the noticeable reduction of warming over the North
Pacific region. This may be associated with the impact of
aerosol forcing which cools the surface temperature and
thus enhances the cold biases, particularly in the northern
hemisphere, as evidenced in these SST bias maps.
Sea surface salinity
Figure 15(a) shows the evolution of annual mean globally
averaged SSS through the piControl runs. ACCESS1.3 has
an apparent increase of SSS, with interannual and decadal
variations, over the course of the piControl simulation. This
drift suggests the ocean recovering from a ‘coupling shock’
(not shown) which brings the global mean SSS down to a
minimum of 34.57 psu from the initial value of around 34.72
psu within the first 50 years of the spin-up integration. The
global SSS then starts rising and takes 200 years to reach
a level of about 34.63 psu when the piControl simulation
starts. The 250-year spin-up for the model is far from enough
to stabilise the global mean SSS. It still has the appearance
of ongoing ‘recovery’ through at least the first 400 years of
the piControl integration. Note one may not expect prefect
equilibrium for the global mean SSS because of the imbalance
of fresh water budget at the ocean surface, as shown in Fig.
15(b). ACCESS1.0 experiences a similar sudden freshening
in the early stage of its 300-year spin-up, but the global
mean SSS recovers much faster than that in ACCESS1.3,
and appears to reach a quasi-equilibrium within 100 years
of the start of the piControl simulation. Fig. 15(b) presents
the freshwater budget (precipitation – evaporation + runoff)
at the global ocean surface. The overall very minor negative
imbalance of water flux (~0.1 mm/year) has little association
with the SSS evolution, but it has direct impact on the global
ocean sea level and the volume-average salinity, as discussed
below. Note that, like the energy budget trend shown in Fig.
13(b), the water imbalance is evidently improving during the
last 100-year simulation, as indicated by the change in trend
for this period.
Figure 16 shows the spatial distributions of the historical
run present day SSS biases relative to the WOA2009 data
(Antonov et al. 2010). The two models have a similar global
bias pattern. The pattern is very similar to what was found in
the GFDL CM2.1 and CM3 models (Griffies et al. 2011). Large
fresh biases are located in the eastern Mediterranean Sea,
east Asian marginal seas, North Sea, Baltic Sea and Gulf of
Bothnia, and particularly Hudson Bay. These extreme fresh
biases are possibly associated with river routing errors in
the land surface model and also due to the difficulty in a
coarse resolution model of representing the circulation in an
enclosed basin with limited connection to the open ocean. In
the North Atlantic, a fresh bias is found off Newfoundland.
This freshening bias, along with the corresponding cooling
bias discussed earlier, is associated with the model’s
deficiency in representing the Gulf Stream path which in
reality turns east at higher latitudes so that the warm, salty
waters from low latitudes are transported into the northern
North Atlantic. Over other regions, large positive salinity
Fig. 15. Time series of annual mean global average: (a) SSS (in psu), and (b) P–E+R (in mm/year) at the ocean surface, from the
ACCESS1.0 (red) and ACCESS1.3 (blue) piControl runs. Also presented in (b) are the linear regressions (thick lines) of the
water budgets over the entire 500-year integration and the last 100-year period (red/green for ACCESS1.0 and blue/cyan for
Bi et al.: The ACCESS coupled model: description, control climate and evaluation 59
biases are found in several places. A large salinity error is
located in the Arctic Ocean (north of Siberia), likely resulting
from poor representation of river runoff in the atmosphere-
land component and a summer bias in the observational
data in ice covered waters. Also notable is the large area
of high salinity errors in the Arabian Sea and the Bay of
Bengal, which is associated with the negative rainfall bias
there, as shown in Fig. 4(c) and 4(d). In the tropical oceans
the SSS biases are also largely determined by the rainfall
errors. The small area of strong bias (up to 20 psu) located off
northeast South America is attributed to the weak outflow of
the Amazon River.
Table 4 summarises the simulated present-day SSS biases
and RMSEs in ACCESS1.0 and ACCESS1.3 against the
WOA2009 data for different zonal bands and global mean.
The models have the largest SSS biases and RMSEs in the
Arctic Ocean and the smallest SSS errors in the southern
tropical oceans. The globally averaged SSS biases are
0.020 psu for ACCESS1.0 and 0.015 psu for ACCESS1.3,
respectively, indicating a reasonable match of the simulated
SSS to the observations. However, the RMS errors, over
1.29 psu for both models show large spatial deviation of
the modelled distributions from the observations. Such
large RMS errors are due to the extreme, localised biases
mentioned above. In fact, excluding extreme freshening of
more than 8 psu (i.e. in Hudson Bay and Gulf of Bothnia)
alone would have the SSS RMS errors significantly reduced
(see the last two columns of Table 4). For the three zonal
bands highlighted in Table 4, the two ACCESS models show
RMS errors comparable to the GFDL CM3 results presented
by Griffies et al. (2011), except the northern band where
ACCESS-CM has larger salinity errors in the Arctic Ocean.
Ocean interior properties
The ocean is spun up from present-day temperature and
salinity conditions with preindustrial atmospheric forcing,
and such an approach requires thousands of years of
integration for the ocean interior, particularly the deep
ocean to reach its equilibrium (e.g. Stouffer 2004). Because of
our very short spin-up periods (i.e. 300 years for ACCESS1.0
and 250 years for ACCESS1.3), it is unsurprising that during
the 500-year control simulation the models show trends in
the evolution of the ocean interior. Fig. 17 displays the time
series of global ocean volume-averaged temperature and
salinity in the control runs.
Both model versions undergo slow and steady changes of
the water properties, but the drifts are in opposite directions.
While ACCESS1.0 is persistently warming up with an
average rate of 0.03 °C/century, ACCESS1.3 is cooling down
at a similar rate, resulting in a thermal deviation of 0.45 °C
Fig. 16. Historical run present-day annual mean SSS biases (model – WOA2009 data): (a) ACCESS1.0, and (b) ACCESS1.3.
Units: psu.
Table 4. Historical run present-day SSS biases and RMSEs in ACCESS1.0 and ACCESS1.3 against theWOA2009 data. Values are pre-
sented for selected zonal bands and global mean. (Units: psu). Note the underlined numbers are for comparison with the
GFDL CM3 model results (Griffies et al. 2011).
ACCESS1.0 ACCESS1.3 RMSE excluding Hudson Bay
and Gulf of Bothnia
70°N–90°N 1.324 3.066 1.532 3.001
30°N–70°N –0.436 2.678 –0.461 2.689 1.199 1.211
30°N–90°N –0.127 2.750 –0.114 2.746 1.690 1.677
0-30°N 0.423 0.773 0.403 0.823
30°S–30°N 0.216 0.686 0.163 0.687
30°S–0 0.021 0.593 –0.062 0.529
90°S–30°S –0.236 0.441 –0.172 0.414
Global 0.020 1.295 0.015 1.291 0.902 0.894
60 Australian Meteorological and Oceanographic Journal 63:1 March 2013
between the two model oceans by the end of the piControl
runs (including the initial difference of 0.13 °C at the end of
the spin-up phase). Such contrasting thermal evolutions are
attributed to the different surface energy budgets achieved
for the two models. As shown in Fig. 13(b), whereas the
ACCESS1.0 ocean sees a net energy gain and therefore
warms up, the ACCESS1.3 ocean is losing heat and thus
cools down. In fact, at the early stage of spin-up (not shown),
both models are evidently warming up due to a large energy
imbalance at the TOA which results in a net energy gain into
the ocean. As described in the ‘Experimental design’ section,
since year 150 in the spin-up phase (when the global ocean is
warmed to about 4 °C from an initial temperature of 3.67 °C),
ACCESS1.3 uses larger sea-ice albedos (Table 1) to produce a
more realistic climate. This results in a much smaller energy
imbalance at the TOA, and the heat exchange at the ocean
surface reaches a temporary balance which lasts for a few
decades. Then the surface heat budget changes sign (from
net gain to net loss), and the ocean starts cooling down.
This cooling continues in the piControl run, as seen in Fig.
17(a). One may expect that this ocean cooling will gradually
slow down following the evolution of the ocean surface heat
budget as shown in Fig. 13(b). In the ACCESS1.0 case, with
a large positive heat budget at the TOA, the ocean surface
never achieves an energy balance as seen in ACCESS1.3, and
the ocean warming remains through the spin-up integration
and then the whole control run.
The global ocean warming or cooling signal found in
the two models is not uniform in distribution. These trends
are the net effect of complicated temperature changes
within ocean basins and on isopycnal surfaces. Taking the
zonal mean, depth dependent features (not shown, see
Marsland et al. 2013) as an example, both models have slow
but prolonged cooling throughout the water column in the
Southern Ocean (south to 60°S), and this cooling spreads
northwards into the global deep ocean below 2000 m via
a strong Southern Ocean abyssal cell. Meanwhile, the top
layers (above 2000 m, north to 60°S), particularly the 200–
1500 m layer, see considerable warming. For ACCESS1.0, the
deep ocean cooling is outweighed by the top layer warming,
resulting in an overall warming drift shown in Fig. 17(a).
The opposite situation is seen in ACCESS1.3 which shows
an overall cooling drift in the ocean. This trend in the ocean
interior is also found in the ACCESS-OM benchmarking
experiment (Bi et al. 2013).
For the piControl run global ocean volume-averaged
salinity, ACCESS1.0 shows a weak decrease whilst
ACCESS1.3 undergoes a considerable increase. Examination
of the ocean surface fresh water budget (Fig. 15(b)) reveals
that both models are losing water. With a loss rate of 0.121
mm/year, ACCESS1.3 has a sea level decrease of 60.5 cm
by the end of the 500-year piControl, which largely, but
not completely, explains the salinity increase of about 0.007
psu seen in Fig. 17(b). In fact, with the model’s volume
conserving configuration (excluding change associated
with the surface mass fluxes of freshwater), an increase of
0.00756 psu (from 34.7265 psu to 34.7341 psu) in the global
ocean salinity indicates a water volume decrease of 2.896
× 1014 m3, equivalent to a sea level drop of 80.1 cm. In the
ACCESS1.0 case, however, the global ocean is freshening
throughout the piControl run, despite the ocean surface
budget showing a net loss of water (39.2 cm/500 years) from
the ocean. These mismatches indicate some unknown leak
of water in ACCESS1.3 but a spurious source of water in
ACCESS1.0, requiring further investigation. The differences
in the non-conservation of hydrology between ACCESS1.0
and ACCESS1.3 impede the diagnosis of the models sea
level as discussed in Marsland et al. (2013) and require
further investigation.
Summary and conclusion
This paper documents the ACCESS-CM, a new coupled
climate model developed at the Centre for Australian
Weather and Climate Research, a partnership between
CSIRO and the Bureau of Meteorology. Two versions of
this model, ACCESS1.0 and ACCESS1.3, have been used in
Fig. 17. piControl run annual mean global ocean volume-averaged evolution of (a) temperature (°C), and (b) salinity (psu). Red for
ACCESS1.0 and blue for ACCESS1.3.
Bi et al.: The ACCESS coupled model: description, control climate and evaluation 61
parallel to conduct a set of CMIP5 experiments, and basic
evaluation has been performed in this study based on the
500-year preindustrial control integrations and 156-year
historical simulations (from 1850 to 2005) in comparison with
observations or reanalysis estimations whenever applicable.
Results of all the ACCESS-CM CMIP5 experiments have been
made available to the international community10 for model
intercomparison studies within the CMIP5 framework.
The basic evaluation of model performance presented
in this paper shows that, despite some deviations, the two
versions of ACCESS-CM generally simulate similar global
average annual mean climate under both preindustrial and
the reconstructed historical atmospheric forcing conditions,
and the results are either close to the available observations/
reanalysis estimations or comparable to the results of
other CMIP5 models (see e.g. Watterson et al. 2013). For
the preindustrial control simulations, ACCESS1.3 shows
nearly no drift in the global average annual mean surface
air temperature while ACCESS1.0 has a weak warming
trend of 0.07 °C/century. The drift found in the ACCESS1.0
piControl has very limited effect on the climate change
signals simulated in the historical and RCP forcing scenario
runs (Dix et al. 2013). The deep ocean shows significant long
term drifts in both temperature and salinity, which persist
through the course of the 500-year piControl simulation,
and which will need to be taken into account when using
the model output in analysis such as for sea level change.
As shown by Dix et al. (2013), the historical simulations also
successfully capture the major features of the observed
SAT evolution, especially the rapid warming epoch of the
past few decades. In addition, the ACCESS-CM model
skill scores in simulating the present climate of a selection
of key atmospheric quantities, globally and over Australia
(Watterson et al. 2013), significantly surpass that of the
preceding CSIRO Mk3.5 model delivered to IPCC AR4.
Because of the substantial difference in their atmosphere-
land surface configurations and a few different parameters
used for the ice and ocean, evident distinctions between
the two models are found in some of the simulated fields,
in terms of both global mean and geographical distribution.
For example, ACCESS1.3 produces a higher global mean
SAT than ACCESS1.0, which is largely attributed to the
lower land surface albedo from the CABLE model over
majority of the continents (see Kowalczyk et al. 2013) that
results in a considerably warmer land surface condition.
Another example is that the PC2 scheme (ACCESS1.3)
shows significant improvement over the Smith scheme
(ACCESS1.0) in simulating the global cloud coverage,
particularly in the tropics (Fig. 5, also see Franklin 2013).
However, this does not necessarily lead to a corresponding
improvement in the cloud radiative forcing. In fact, the solar
CRF and therefore the total CRF simulated by ACCESS1.3
are overall poorer than that by ACCESS1.0, particularly over
the Southern Ocean.
Both ACCESS1.0 and ACCESS1.3 yield solutions which
overall show significant improvement over those for the
Australian CMIP3 contributions. Nevertheless, the models
and their solutions possess certain imperfections, as
discussed previously. Since the CMIP5 simulations were
performed, a program of model testing and development has
continued to address these imperfections. The primary focus
is on improvements to ACCESS1.3 which, with its inclusion
of CABLE and the first of the GA series of atmospheric
components, is the preferred model for a range of activities
undertaken by the Australian climate research community,
and a starting point for development of the next generation
of model for application beyond CMIP5.
This work has been undertaken as part of the Australian
Climate Change Science Program, funded jointly by the
Department of Climate Change and Energy Efficiency, the
Bureau of Meteorology and CSIRO. The computation for
this work was performed at the NCI National Facility at the
Australian National University, and we particularly thank
Dr Ben Evans and the other NCI staff members for their
support. We further thank Cath Senior, Chris Gordon and
Gill Martin of the Met Office for helpful discussions during
the course of this work, and Gill Martin and Charline Marzin
for HadGEM and proto-HadGEM3 model output fields useful
for comparison and benchmarking. We thank Leon Rotstayn
for his comments on a pre-submission draft. We finally thank
other members of CAWCR and the Australian Research
Council Centre of Excellence for Climate System Science
who have contributed their views and encouragement
during the course of this work.
Adcroft, A. and Campin, J.-M. 2004. Rescaled height coordinates for ac-
curate representation of free-surface flows in ocean circulation mod-
els. Ocean Modell., 7, 269–284.
Adcroft, A., Hill, C., Marshall, J. 1997. Representation of topography by
shaved cells in a height coordinate ocean model. Mon. Weather Rev.,
125, 2293–2315.
Antoniou, J., Locarnini, R., Boyer, T., Mishonov, A., Garcia, H. 2006. World
Ocean Atlas 2005, Volume 2: Salinity. U.S. Government Printing Office
62, NOAA Atlas NESDIS, Washington, DC, 182 pp.
Antonov, J.I., Seidov, D., Boyer, T.P., Locarnini, R.A., Mishonov, A.V.,
Garcia, H.E., Baranova, O.K., Zweng, M.M., and Johnson, D.R. 2010.
World Ocean Atlas 2009, Volume 2: Salinity. S. Levitus, Ed. NOAA At-
las NESDIS 69, U.S. Government Printing Office, Washington, D.C.,
184 pp.
Arakawa, A. and Lamb, V.R. 1977. Computational design and the basic
dynamical processes of the UCLA general circulation Model. Methods
in Computational Physics, 17, p173.
Arribas, A., Glover, M., Maidens, A., Peterson, K., Gordon, M., MacLach-
10The ACCESS raw model outputs have been written out to CMIP5 speci-
fications that include 1) providing many of the requested parameters, 2)
following the Data Reference Syntax and Controlled Vocabularies, i.e. file
and directory naming conventions, and 3) ensuring metadata contained
in the requested netCDF files following specified conventions such as
CMIP5 experiment instructions (Taylor et al. 2011). Datasets that pass
the CMIP5 quality control standards have been published to the Earth
System Grid (ESG) and available to the international community via the
CMIP5 data portal.
62 Australian Meteorological and Oceanographic Journal 63:1 March 2013
lan, C., Cusack, S., and Scaife, A. 2011. The GloSea4 ensemble predic-
tion system for seasonal forecasting. Mon. Weather Rev., 139, 1891–
1910. doi: 10.1175/2010MWR3615.1
Avila, F.B., Pitman, A.J., Donat, M.G., Alexander L.V., and Abramow-
itz, G. 2012. Climate model simulated changes in temperature ex-
tremes due to land cover change, J. Geophys. Res., 117, D04108,
Bellouin, N., Rae, J., Jones, A., Johnson, C., Haywood, J., and Boucher, O.
2011. Aerosol forcing in the Climate Model Intercomparison Project
(CMIP5) simulations by HadGEM2-ES and the role of ammonium ni-
trate. J. Geophys. Res., 116, D20206, doi:10.1029/2011JD016074.
Bi, D. and Marsland, S.J. 2010. Australian Climate Ocean Model (Aus-
COM) Users Guide. CAWCR Technical Report No. 027, The Centre for
Australian Weather and Climate Research, a partnership between
CSIRO and the Bureau of Meteorology. 72 pp. ISBN: 978-1-921605-
92-5 (PDF).
Bi, D., Marsland, S.J., Uotila, P., O’Farrell, S., Fielder, R., Sullivan, A.,
Griffies, S.M., Zhou, X., and Hirst, A.C. 2013. ACCESS OM, the Ocean-
Sea Ice Core of the ACCESS Coupled Model. Aust. Met. Oceanogr. J.,
63, 213–232.
Boutle, I.A. and Morcrette, C.J. 2010. Parametrization of area cloud frac-
tion. Atmos. Sci. Let., 11, 283–289.
Cai, W., Sullivan, A., and Cowan, T. 2009. Climate change contributes
to more frequent consecutive positive Indian Ocean Dipole events.
Geophys. Res. Lett., 36, L19783, doi:10.1029/2009GL040163.
Collins, M., An, S-I., Cai, W., Ganachaud, A., Guilyardi, E., Jin, F-F., Jo-
chum, M., Lengaigne, M., Power, S., Timmermann, A., Vecchi, G. and
Wittenberg, A. 2010. The Impact of Global Warming on the Tropical
Pacific and El Niño. Nature Geoscience, 3, 391–397.
Collins, W.J., Bellouin, N., Doutriaux-Boucher, M., Gedney, N., Hinton,
T., Jones, C.D., Liddicoat, S., Martin, G., O’Connor, F., Rae, J., Senior,
C., Totterdell, I., Woodward, S., Reichler, T., Kim, J., Halloran, P. 2008.
Evaluation of the HadGEM2 model. Hadley Centre Technical Note
HCTN 74, Met Office Hadley Centre, Exeter, U.K.,http://www.metof- publications/science/climate-science.
Cox, P., Betts, R., Bunton, C., Essery, P., Rowntree, R. and Smith, J. 1999.
The impact of new land surface physics on the GCM simulation of
climate and climate sensitivity. Clim. Dyn., 15, 183–203.
Davies, T., Cullen, M.J.P., Malcolm, A.J., Mawson, M.H., Staniforth, A.,
White, A.A., and Wood, N. 2005. A new dynamical core for the Met
Office’s global and regional modelling of the atmosphere. Q. J. R. Me-
teorol. Soc., 131, 1759–1782.
Dee, D.P., Uppala, S.M., Simmons, A.J., Berrisford, P., Poli, P., and others.
2011. The ERA-Interim reanalysis: configuration and performance of
the data assimilation system. Q. J. R. Meteorol. Soc., 137, 553–597.
Dix, M., Vohralik, P., Bi, D., Rashid, H., Marsland, S.J., O’Farrell, S., Uotila,
P., Hirst, A.C., Kowalczyk, E., Sullivan, A., Yan, H., Franklin, C., Sun,
Z., Watterson, I., Collier, M., Noonan, J. Rotstayn, J. Stevens, L., Uhe
P., and Puri, K. 2013. The ACCESS Coupled Model: Documentation
of core CMIP5 simulations and initial results. Aust. Met. Oceanogr. J.,
63, 83–99.
Dunne, J.P., John, J.G., Adcroft, A.J., Griffies, S.M., Hallberg, R.W., Shev-
liakova, E.N., Stouffer, R.J., Cooke, W., Dunne, K.A., Harrison, M.J.,
Krasting, J.P., Levy, H., Malyshev, S.L., Milly, P.C.D., Phillipps, P.J.,
Sentman, L.A., Samuels, B.L., Spelman, M.J., Winton, M., Wittenberg,
A.T., and Zadeh, N. 2012 GFDL’s ESM2 global coupled climate-car-
bon Earth System Models Part I: Physical formulation and baseline
simulation characteristics. J. Climate, 25, 6646–6665. doi:10.1175/JCLI-
Durack, P.J., Wijffels S.E., and Matear, R.J. 2012. Ocean Salinities Re-
veal Strong Global Water Cycle Intensification During 1950 to 2000.
Science, 336, 455–458. DOI: 10.1126/science.1212222.
Edwards, J.M., Beare, R.J., and Lapworth, A.J. 2006. Simulations of the
observed evening transition and nocturnal boundary layers: Single-
column modelling. Q. J. R. Meteorol. Soc., 132, 61–80.
Edwards, J.M. and Slingo, A. 1996. Studies with a new flexible radiation
code. I: Choosing a configuration for a large-scale model. Q. J. R. Me-
teorol. Soc, 122, 689–720.
Elliott, S.M., Deal, C., Humphries, G., Hunke, E.C., Jeffery, N., Jin, M.,
Levasseur, M., and Stefels, J. 2012. Pan-Arctic simulation of coupled
nutrient-sulfur cycling due to sea ice biology: Preliminary results,
J. Geophys. Res., 117, 1–16, doi:10.1029/2011JG001649.
Essery, R.L.H., Best, M.J., Betts, R.A., Cox, P.M., and Taylor, C.M. 2003.
Explicit Representation of Subgrid Heterogeneity in a GCM Land
Surface Scheme. J. Hydrometeor., 4, 530–543, doi:10.1175/1525-
Fairall, C.W., Bradley, E.F., Hare, J.E., Grachev, A.A., and Edson, J.B. 2003.
Bulk Parameterization of Air- Sea Fluxes: Updates and Verification for
the COARE Algorithm. J. Clim., 16, 571–591.
Flocco, D., Schroeder, D., Feltham, D.L., and Hunke, E.C. 2012. Impact of
melt ponds on Arctic sea ice simulations from 1990 to 2007. J. Geo-
phys. Res., 117, 1–17, doi:10.1029/2012JC008195.
Fox-Kemper, B., Ferrari, R., and Hallberg, R. 2008. Parameterization of
mixed layer ddies. Part I: Theory and diagnosis. J. Phys. Oceanogr.,
38, 1145–1165.
Fox-Kemper, B., Danabasoglu, G., Ferrari, R., Griffies, S.M., Hallberg,
R.W., Holland, M.M., Maltrud, M.E., Peacock, S.L., and Samuels, B.L.
2011. Parameterization of mixed layer eddies. III: Implementation
and impact in global ocean climate simulations. Ocean Modelling, 39,
Franklin, C.N., Jakob, C., Dix, M., Protat, A., and Roff, G. 2012. Assessing
the performance of a prognostic and a diagnostic cloud scheme using
single column model simulations of TWP-ICE. Q. J. R. Meteorol. Soc,
138, 734–754, doi: 10.1002/qj.954
Franklin, C.N., Sun, Z., Bi, D., Dix, M., Yan, H., and Bodas-Salcedo, A.
2013. Evaluation of clouds in ACCESS using the satellite simulator
package COSP. Global, seasonal and regional cloud properties. J. Geo-
phys. Res., 118, 732–748. doi: 10.1029/2012JD018469.
Gent, P.R., and McWilliams, J.C. 1990. Isopycnal mixing in ocean circula-
tion models. J. Phys. Oceanogr., 20, 150–155.
Gent, P.R. Willebrand, J., McDougall, T.J., and McWilliams, J.C. 1995.
Parameterizing eddy-induced tracer transports in ocean circulation
models. J. Phys. Oceanogr., 25, 463–474.
Gent, P.R., Danabasoglu, G., Donner, L.J., Holland, M.M., Hunke, E.C.,
Jayne, S.R., Lawrence, D.M., Neale, R.B., Rasch, P.J., Vertenstein, M.,
Worley, P.H., Yang, Z.L., and Zhang, M. 2011. The Community Climate
System Model Version 4. J. Clim., 24, 4973–4991.
GLOBE Task Team and others (D.A. Hastings, P.K. Dunbar, G.M. Elphin-
gstone, M. Bootz, H. Murakami, H. Maruyama, H. Masaharu, P. Hol-
land, J. Payne, N.A. Bryant, T.L. Logan, J.-P. Muller, G. Schreier, and
J.S. MacDonald), eds., 1999. The Global Land One-kilometer Base Ele-
vation (GLOBE) Digital Elevation Model, Version 1.0. National Oceanic
and Atmospheric Administration, National Geophysical Data Center,
325 Broadway, Boulder, Colorado 80303, U.S.A. Digital data base on
the World Wide Web (URL:
globe.html) and CD-ROMs.
Gnanadesikan, A., Dixon, K.W., Griffies, S.M., Balaji, V., Barreiro, M.,
Beesley, J.A., Cooke, W.F., Delworth, T.L., Gerdes, R., Harrison, M.J.,
Held, I.M., Hurlin, W.J., Lee, H.C., Liang, Z., Nong, G., Pacanowski,
R.C., Rosati, A., Russell, J.L., Samuels, B.L., Song, Q., Spelman, M.J.,
Stouffer, R.J., Sweeney, C., Vecchi, G.A., Winton, M., Wittenberg, A.T.,
Zeng, F., Zhang, R., and Dunne, J.P. 2006. GFDL’s CM2 Global Coupled
Climate Models. Part II: The baseline ocean simulation. J. Clim., 19,
Gordon, H., O’Farrell, S., Collier, M., Dix, M., Rotstayn, L., Kowalczyk,
E., Hirst, T., and Watterson, I. 2010. The CSIRO Mk3.5 Climate Model
[Electronic resource]. Centre for Australian Weather and Climate Re-
search. GPO Box 1289, Melbourne, Victoria 3001, Australia. CAWCR
Technical Report No. 021. 61 pp. ISBN: 978-1-921605-666.
Gregory, D. and Rowntree, P.R. 1990. A mass-flux convection scheme with
representation of cloud ensemble characteristics and stability depen-
dent closure. Mon. Weather Rev., 118, 1483–1506.
Griffies, S.M, 1998. The Gent-McWilliams skew-flux. J. Phys. Oceanogr.,
28, 831–841.
Griffies, S.M., Gnanadesikan, S. Dixon, K.W., Dunne, J.P., Gerdes, R., Har-
rison, M.J., Rosati, A., Russell, J., Samuels, B.L., Spelman, M.J., Win-
ton, M., Zhang, R. 2005. Formulation of an ocean model for global
climate simulations. Ocean Science, 1, 45–79.
Griffies, S.M. 2009. Elements of MOM4p1: GFDL Ocean Group Tech. Rep.
No. 6, NOAA/Geophysical Fluid Dynamics Laboratory, 444 pp.
Griffies, S.M., Biastoch, A., Boening, C.W., Bryan, F., Chassignet, E.,
England, M., Gerdes, R., Haak, H., Hallberg, R.W., Hazeleger, W.,
Jungclaus, J., Large, W.G., Madec, G., Samuels, B.L., Scheinert, M.,
Gupta, A.S., Severijns, C.A., Simmons, H.L., Treguier, A.M., Winton,
Bi et al.: The ACCESS coupled model: description, control climate and evaluation 63
M., Yeager, S., and Yin, J. 2009. Coordinated ocean-ice reference
experiments (COREs). Ocean Modelling, 26, 1–46, DOI:10.1016/j.oc-
Griffies, S.M., Winton, M., Donner, L.J. Horowitz, L.W., Downes, S.M.,
Farneti, R., Gnanadesikan, A., Hurlin, W.J., Lee, H.C., Liang, Z. Palter,
J.B., Samuels, B.L., Wittenberg, A.T., Wyman, B.L., Yin, J., and Zadeh,
N. 2011. The GFDL CM3 Coupled Climate Model: Characteristics
of the Ocean and Sea Ice Simulations. J. Clim., 24, 3520–3544, DOI:
Hewitt, H.T., Copsey, D., Culverwell, I.D., Harris, C.M., Hill, R.S.R., Keen,
A.B., McLaren, A.J., and Hunke, E.C. 2011. Design and implementa-
tion of the infrastructure of HadGEM3: the next-generation Met Office
climate modelling system. Geosci. Model Dev., 4, 223–253, doi:10.5194/
Hilton, R.D., Featherstone, W.E., Berry, P.A.M., Johnson, C.P.D., and Kir-
by, J.F. 2003. Comparison of digital elevation models over Australia
and external Validation using ERS-1 satellite radar altimetry. Aust. J.
Earth Sci., 50, 157–168, doi:10.1046/j.1440-0952.2003.00982.x.
Holland, M.M., Bailey, D.A., Briegleb, B.P., Light, B., and Hunke, E. 2012.
Improved Sea Ice Shortwave Radiation Physics in CCSM4: The Impact
of Melt Ponds and Aerosols on Arctic Sea Ice*. J. Clim., 25, 1413–1430.
Huffman, G.J, Adler, R.F., Bolvin, D.T., and Gu, G. 2009. Improving the
Global Precipitation Record: GPCP Version 2.1. Geophys. Res. Lett., 36,
L17808, doi:10.1029/2009GL040000.
Hunke, E.C., and Dukowicz, J.K. 1997. An Elastic–Viscous–Plastic
Model for Sea ice Dynamics. J. Phys. Oceanogr., 27, 1849–1867,
doi:10.1175/1520-0485(1997)027 <1849:AEVPMF>2.0.CO;2.
Hunke, E.C. and Lipscomb, W.H., 2010. CICE: The Los Alamos sea ice
model documentation and software user’s manual, Version 4.1. LA-
CC-06-012, Los Alamos National Laboratory, N.M., 76 pp.
Intergovernmental Panel on Climate Change. 2007. IPCC Fourth Assess-
ment Report, Climate Change 2007, UNEP/WMO.
Jochum, M. 2009. Impact of latitudinal variations in vertical dif-
fusivity on climate simulations. J. Geophys. Res., 114, C01010,
Jones, P.W. 1997. A User’s Guide for SCRIP: A spherical coordinate re-
mapping and interpolation package. LA-CC- 98-45, Los Alamos Na-
tional Laboratory, N.M., 27pp.
Kim, D., Kug, J.-S., Kang, I.-S., Jin F.-F., and Wittenberg, A.T. 2008. Tropi-
cal Pacific impacts of convective momentum transport in the SNU
coupled GCM. Clim. Dyn., 31, 213–226. DOI 10.1007/s00382-007-0348-
King, J.C., Connolley, W.M. and Derbyshire, S.H. 2001. Sensitivity of
modelled Antarctic climate to surface and boundary-layer parametri-
zations. Q. J. R. Meteorol. Soc., 127, 779–794.
Kowalczyk, E.A., Wang, Y.P., Law, R.M., Davies, H.L., McGregor, J.L.,
and Abramowitz, G. 2006. The CSIRO Atmosphere Biosphere Land
Exchange (CABLE) model for use in climate models and as an offline
model. CSIRO Marine and Atmospheric Research paper 013, 43pp.
Kowalczyk, E.A., Stevens, L., Law, R.M., Dix, M., Wang, Y.P., Harman, I.N.,
Hayens, K., Srbinovsky, J., Pak, B., and Zhien, T. 2013. The land surface
model component of ACCESS: description and impact on the simu-
lated surface climatology. Aust. Met. Oceanogr. J., 63, 65–82..
Kwok, R., and Cunningham, G.F. 2008. ICESat over Arctic sea-ice: Estima-
tion of snow depth and ice thickness. J. Geophys. Res., 113, C08010,
Large, W.G., McWilliams, J.C., and Doney, S.C. 1994. Oceanic vertical
mixing: A review and a model with a nonlocal boundary layer param-
eterization. Rev. Geophys., 32, 363–403, doi:10.1029/94RG01872.
Laxon, S., Peacock, N. and Smith, D. 2003. High interannual variability of
sea ice thickness in the Arctic region. Nature, 425, 947–950.
Lock, A.P, Brown, A.R., Bush, M.R., Martin G.M. and Smith, R.N.B. 2000.
A new boundary layer mixing scheme. Part I: Scheme description and
single-column model tests. Mon. Weather Rev., 128, 3187–3199
Leuning, R. 1995. A critical appraisal of a combined stomatal pho-
tosynthesis model for C3 plants. Plant Cell Environ., 18, 339–355,
Lin, J. L. 2007. The double-ITCZ problem in IPCC AR4 coupled GCMs:
ocean-atmosphere feedback analysis. J. Clim., 20, 4497–4525.
Locarnini, R., Mishonov, A., Antonov, J., Boyer, T., Garcia, H. 2006. World
Ocean Atlas 2005, volume 1: Temperature. U.S. Government Printing
Office 61, NOAA Atlas NESDIS, Washington, DC, 182 pp.
Lock, A.P., Brown, A.R, Bush, M.R., Martin, G.M., and Smith, R.N.B. 2000.
A new boundary layer mixing scheme. Part I: Scheme description and
SCM tests. Mon. Weather Rev., 128, 3187–3199.
Lucarini, V., and Ragone, F. 2011. Energetics of PCMDI/CMIP3 climate
models: energy budget and meridional enthalpy transport. Rev Geo-
phys., 49, RG1001. DOI:10.1029/2009RG000323.
McDougall, T. J., 2003: Potential enthalpy: a conservative oceanic vari-
able for evaluating heat content and heat fluxes. J. Phys. Oceanogr.,
33, 945–963.
Marsland, S.J., Bi, D., Uotila, P., Fiedler, R., Griffies, S.M., Lorbacher, K.,
O’Farrell, S., Sullivan, A., Uhe, P., Zhou, X., and Hirst, A.C. 2013. Eval-
uation of ACCESS Climate Model ocean diagnostics in CMIP5 simula-
tions. Aust. Met. Oceanogr. J., 63, 101–119.
Martin, G.M., Milton, S.F., Senior, C.A., Brooks, M.E., Ineson, S., Reichler,
T., and Kim, J. 2010. Analysis and Reduction of Systematic Errors
through a Seamless Approach to Modelling Weather and Climate. J.
Clim., 23, 5933–5957, doi:10.1175/2010JCLI3541.1.
Martin, G.M., Ringer, M.A., Pope, V.D., Jones, A., Dearden, C., and Hin-
ton, T.J. 2006. The physical properties of the atmosphere in the new
Hadley Centre Global Environment Model, HadGEM1. Part I: Model
description and global climatology. J. Clim., 19, 1274–1301.
Martin, G.M., Bellouin, N., Collins, W.J., Culverwell, I.D., Halloran, P.,
Hardiman, S., Hinton, T.J., Jones, C.D., McLaren, A., O’Connor, F.,
Rodriguez, J., Woodward, S., et al. 2011. The HadGEM2 family of Met
Office Unified Model climate configurations. Geosci. Model Dev. Dis-
cuss., 4, 723–757, doi:10.5194/gmd-4-723-2011.
Meehl, G.A., Stocker, T.F., Collins, W.D., Friedlingstein, P., Gaye, A.T.,
Gregory, J.M., Kitoh, A., Knutti, R., Murphy, J.M., Noda, A., Raper,
S.C.B., Watterson, I.G., Weaver A.J., and Zhao. Z.-C. 2007. Global Cli-
mate Projections. In: Climate Change 2007: The Physical Science Basis.
Contribution of Working Group I to the Fourth Assessment Report of
the Intergovernmental Panel on Climate Change [Solomon, S., D. Qin,
M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M. Tignor and H.L.
Miller (eds.)]. Cambridge University Press, Cambridge, United King-
dom and New York, NY, USA.
Murray, R.J. 1996. Explicit generation of orthogonal grids for ocean mod-
els. J. Comput. Phys., 126, 251–273.
Pacanowski, R.C. and Gnanadesikan, A. 1998. Transient response in a z-
level ocean model that resolves topography with partial-cells. Mon.
Weather Rev., 126, 3248–3270.
Pincus R., Batstone, C.P., Patrick-Hofmann, R.J., Taylor, K.E., and Gleckler,
P.E. 2008. Evaluating the present-day simulation of clouds, precipita-
tion and radiation in climate models. J. Geophys. Res., 113, D14209,
Pirazzini, R. 2008. Factors controlling the surface energy budget over
snow and ice. Finnish Meteorolog. Inst. Contributions, 75, 1–55.
Puri, K., G. Dietachmayer, P. Steinle, M. Dix, L. Rikus, L. Logan, M.
Naughton, C. Tingwell, Y Xiao, V. Barras, I. Bermous, R. Bowen, L.
Deschamps, C. Franklin, J. Fraser, T. Glowacki, B. Harris, J. Lee, T. Le,
G. Roff, A. Sulaiman, H. Sims, X. Sun, Z. Sun, H. Zhu, M. Chattopad-
hyay, C. Engel. 2013. Implementation of the initial ACCESS Numerical
Weather Prediction system. Aust. Met. Oceanogr. J., in press.
Rashid, H.A., Hirst, A.C. and Dix, M. 2013a. Atmospheric circulation fea-
tures in the ACCESS coupled model simulations for CMIP5: Historical
simulation and future projections. Aust. Met. Oceanogr. J., 63, 145–160.
Rashid, H.A., Sullivan, A., Hirst, A.C., Bi, D., and Marsland, S.J. 2013b.
Evaluation of El Nino-Southern Oscillation in the ACCESS coupled
model simulations for CMIP5. Aust. Met. Oceanogr. J., 63, 161–180.
Rahmstorf, S. 1993. A fast and complete convection scheme for ocean
models. Ocean Modelling, 101, 9–11.
Rayner, N.A., Parker, D.E., Horton, E.B., Folland, C.K., Alexander, L.V.,
Rowell, D.P., Kent, E.C. and Kaplan, A. 2003. Global analyses of sea
surface temperature, sea ice, and night marine air temperature since
the late nineteenth century. J. Geophys. Res., 108, No. D14, 4407,
Rossow, W.B. and Sciffer, R.A. 1999. Advances in understanding clouds
from ISCCP. Bull. Am. Meteorol. Soc., 80, 2261–2287
Rossow, W.B. and Garner, L.C. 1993. Validation of ISCCP Cloud Dections.
J. Clim., 6, 2370–2393.
64 Australian Meteorological and Oceanographic Journal 63:1 March 2013
Rotstayn, L.D., Jeffrey, S.J., Collier, M.A., Dravitzki, S.M., Hirst, A.C., Syk-
tus, J.I., and Wong, K.K. 2012. Aerosol and greenhouse gas-induced
changes in summer rainfall and circulation in the Australasian region:
a study using single-forcing climate simulations. Atmos. Chem. Phys.,
12, 6377–6404, doi:10.5194/acp-12-6377-2012.
Scaife, A.A., Butchart, N., Warner, C.D. and Swinbank, R. 2002. Impact of
a spectral gravity wave parameterization on the stratosphere in the
Met. Office Unified Model. Q. J. R. Meteorol. Soc., 59, 1473–1489.
Semtner, A.J. 1976. A model for the thermodynamic growth of sea ice
in numerical investigations of climate. J. Phys. Oceanogr., 6, 379–389.
Sen Gupta, A., Muir, L.C., Brown, J.N., Phipps, S.J., Durack, P.J., Mon-
selesen, D., and Wijffels, S.E. 2012. Climate drift in CMIP3 models. J.
Clim., 25, 4621–4640.
Shonk, J.K.P. and Hogan, R.J. 2008. Tripleclouds: an efficient method for
representing cloud inhomogeneity in 1D radiation schemes by using
three regions at each height. J. Clim., 21, 2352–2370.
Smith, R.N.B. 1990. A scheme for predicting layer clouds and their water
contents in a general circulation model. Q. J. R. Meteorol. Soc., 116,
Smith, I. 2007. Global climate modelling within CSIRO: 1981 to 2006. Aust.
Meteorol. Mag., 56, 153–166.
Stephens, G.L., Tsay, S.C., Stackhouse Jr, P.W., and Flatau, P. 1990. The rel-
evance of the microphysical and radiative properties of cirrus clouds
to climate and climate feedback. J. Atmos. Sci., 47, 1742–1753.
Stouffer, R.J. 2004. Time Scales of Climate Responses. J. Clim., 17, 209–217.
Sun, Z.C. Franklin, X. Zhou, Y. Ma, P. Okely, D. Bi, M. Dix, A.C. Hirst, J.
K.P. Shonk, and K. Puri, 2013. Modifications in atmospheric physical
parameterization for improving SST simulation in the ACCESS cou-
pled model. Aust. Met. Oceanogr. J., 63, 233–247.
Taylor, K.E., Stouffer, R.J. and Meehl, G.A., 2009. A summary of the CMIP5
experiment design. PCDMI Rep., 33 pp. [
Taylor, K.E., Stouffer, R.J., Meehl, G.A. 2012. An Overview of CMIP5
and the experiment design. Bull. Amer. Meteor. Soc., 93, 485–498,
Trenberth, K.E., and Fasullo, J.T. 2010. Simulation of present day and 21st
century energy budgets of the southern oceans. J. Clim., 23, 440–454,
Uotila, P., O’Farrell, S., Marsland, S.J., and Bi, D. 2012. A sea ice sensitivity
study with a global ocean-ice model. Ocean Modelling, 18, doi:http://
Uotila, P., O’Farrell, S., Marsland, S.J., and Bi, D. 2013. The sea ice perfor-
mance of the Australian models participating in the CMIP5. Aust. Met.
Oceanogr. J., 63, 121–143.
Valcke, S. 2006. OASIS3 User Guide (prism 2-5). PRISM Support Initiative
No 3, 68 pp.
Voldoire, A., Sanchez-Gomez, E., Salas y Me´lia, D., Decharme, B., Cassou,
C., Se´ne´si, S., Valcke, S., Beau, I., Alias, A., Chevallier, M., De´que´,
M., Deshayes, J., Douville, H., Fernandez, E., Madec, G., Maisonnave,
E., Moine, M.P., Planton, S., Saint-Martin, D., Szopa, S., Tyteca, S., Al-
kama, R., Belamari, S., Braun, A., Coquart, L., and Chauvin, F. 2012.
The CNRM-CM5.1 global climate model: description and basic evalu-
ation. Clim. Dyn., DOI 10.1007/s00382-011-1259-y.
Wang, Y.P. and Leuning, R. 1998. A two-leaf model for canopy conduc-
tance, photosynthesis and partitioning of available energy I. Model
description and comparison with a multi-layered model. Agric. Forest
Meteor., 91, 89–111.
Wang, Y.P., Kowalczyk, E.A., Leuning, R., Abramowitz, G., Raupach, M.R.,
Pak, B., van Gorsel, E. and Luhar, A. 2011. Diagnosing errors in a land
surface model (CABLE) in the time and frequency domains. J. Geophys.
Research-Biogeosciences, 116. G01034, doi:10.1029/2010JG001385.
Warner, C.D. and McIntyre, M.E. 2001. An Ultrasimple Spectral Param-
eterization for Nonorographic Gravity Waves. J. Atmos. Sci., 58, 1837–
Watterson, I.G. 1996. Non-dimensional measures of climate model perfor-
mance. Int. J. Climatol., 16, 379–391.
Watterson, I.W., Hirst, A.C. and Rotstayn, L.D. 2013. A skill-score based
evaluation of simulated Australian climate. Aust. Met. Oceanogr. J., 63,
Webster, S., Brown, A.R., Cameron, D.R. and Jones, C.P. 2003. Improve-
ments to the representation of orography in the Met. Office Unified
Model. Q. J. R. Meteorol. Soc., 129, 1989–2010.
Williams, D.N., Ananthakrishnan, R., Bernholdt, D.E., Bharathi, S.,
Brown, D., Chen, M., Chervenak, A.L., Cinquini, L., Drach, R., Foster,
I.T., Fox, P., Fraser, D., Garcia, J., Hankin, S., Jones, P., Middleton, D.E.,
Schwidder, J., Schweitzer, R., Schuler, R., Shoshani, A., Siebenlist, F.,
Sim, A., Strand, W.G., Su, M., and Wilhelmi, N. 2009. The Earth System
Grid - Enabling Access to Multimodel Climate Simulation Data.Bul.
Am. Met. Soc., 90, 195–205.
Williams, K.D., Ringer, M.A., Senior, C.A., Webb, M.J., McAvaney, B.J.,
Andronova, N., Bony, S., Dufresne, J.L., Emori, S., and Gudgel, R. 2006.
Evaluation of a component of the cloud response to climate change in
an intercomparison of climate models. Clim. Dyn., 26, 145–165.
Wilson, D.R., Smith, R.N.B., Gregory, D., Wilson, C., Bushell, A.C. and
Cusack, S. 2004. The Large-Scale Cloud Scheme and Saturated Spe-
cific Humidity. Unified Model Documentation Paper 26. Available from
Met Office.
Wilson, D.R. and Ballard, S.P. 1999. A microphysically based precipita-
tion scheme for the UK Meteorological Office Unified Model. Q. J. R.
Meteorol. Soc, 125, 1607–1636
Wilson, D.R., Bushell, A.C., Kerr-Munslow, A.M., Price, J.D., and Mor-
crette, C.J. 2008. PC2: A prognostic cloud fraction and condensation
scheme. I: Scheme description, Q. J. R. Meteorol. Soc., 134, 2093–2107.
Wood, N., Diamantakis, M., Staniforth, A. 2007. A monotonically damp-
ing, second-order accurate, unconditionally stable, numerical scheme
for diffusion An improved implicit predictor-corrector scheme for
boundary layer vertical diffusion. Q. J. R. Meteorol. Soc. 133, 1559–
Worby, A., Geiger, C.A., Paget, M.J., Van Woert, M.L., Ackley, S.F., and
DeLiberty, T.L. 2008. Thickness distribution of Antarctic sea-ice. J.
Geophys. Res., 617, 1-14, doi:10.1029/2007JC004254.
Wu, X.Q., Deng, L., Song, X., Zhang, G.J. 2007. Coupling of convective
momentum transport with convective heating in global climate simu-
lations. J. Atmos. Sci., 64, 1334–1349.
Yao, M.S., and Del Genio, A.D. 1999. Effects of cloud parameterization
on the simulation of climate changes in the GISS GCM. J. Clim., 12,
Yao, M.S. and Del Genio, A.D. 1999. Effects of cloud parameterization
on the simulation of climate changes in the GISS GCM. J. Clim., 12,
Zheng, X.T., Xie, S.P., Vecchi, G.A., Liu, Q. and Hafner, J. 2010. Indian
Ocean dipoleresponse to global warming: Analysis of ocean-atmo-
spheric feedbacks in a coupled model. J. Clim., 23, 1240–1253.
... Different models share historical predecessors (Masson and Knutti, 2011;Knutti et al., 2013), conceptual frameworks, and, in some cases, source code (Boé, 2018;Brands, 2022b;Brands et al., 2023). An active field of research has developed to identify and manage these "hidden dependencies" through weighting or subselection of the broader CMIP archives (e.g., Bishop and Abramowitz, 2013;Sanderson et al., 2015;Knutti et al., 2017;Brunner and Sippel, 2023), but open questions remain, particularly with regards to how best to determine dependence within multi-model ensembles Annan and Hargreaves, 2017). ...
... In prior studies, it has been shown that a climate model's origins and evolution can be traced via statistical properties of its outputs (e.g., Masson and Knutti, 2011;Bishop and Abramowitz, 2013;Knutti et al., 2013). Output-based model identification can uncover hidden dependencies within the ensemble, e.g., models that are similar because they share components or lineages but not names. ...
... ACCESS1-0 and HadGEM2-ES also share HadGAM2 atmospheres, while ACCESS1-3 features a modified version of the UK Met Office Global Atmosphere 1.0 AGCM (UM7.3/GA1; Bi et al., 2012;Brands, 2022a). ACCESS1-3 is closer to ACCESS1-0 and HadGEM2-ES than to other CMIP5 models and thus joins the family group despite the differing atmospheric component, demonstrating that a family designation is more complex than just a single shared model component. ...
Full-text available
As the number of models in Coupled Model Intercomparison Project (CMIP) archives increase from generation to generation, there is a pressing need for guidance on how to interpret and best use the abundance of newly available climate information. Users of the latest CMIP6 seeking to draw conclusions about model agreement must contend with an “ensemble of opportunity” containing similar models that appear under different names. Those who used the previous CMIP5 as a basis for downstream applications must filter through hundreds of new CMIP6 simulations to find several best suited to their region, season, and climate horizon of interest. Here we present methods to address both issues, model dependence and model subselection, to help users previously anchored in CMIP5 to navigate CMIP6 and multi-model ensembles in general. In Part I, we refine a definition of model dependence based on climate output, initially employed in Climate model Weighting by Independence and Performance (ClimWIP), to designate discrete model families within CMIP5 and CMIP6. We show that the increased presence of model families in CMIP6 bolsters the upper mode of the ensemble's bimodal effective equilibrium climate sensitivity (ECS) distribution. Accounting for the mismatch in representation between model families and individual model runs shifts the CMIP6 ECS median and 75th percentile down by 0.43 ∘C, achieving better alignment with CMIP5's ECS distribution. In Part II, we present a new approach to model subselection based on cost function minimization, Climate model Selection by Independence, Performance, and Spread (ClimSIPS). ClimSIPS selects sets of CMIP models based on the relative importance a user ascribes to model independence (as defined in Part I), model performance, and ensemble spread in projected climate outcome. We demonstrate ClimSIPS by selecting sets of three to five models from CMIP6 for European applications, evaluating the performance from the agreement with the observed mean climate and the spread in outcome from the projected mid-century change in surface air temperature and precipitation. To accommodate different use cases, we explore two ways to represent models with multiple members in ClimSIPS, first, by ensemble mean and, second, by an individual ensemble member that maximizes mid-century change diversity within the CMIP overall. Because different combinations of models are selected by the cost function for different balances of independence, performance, and spread priority, we present all selected subsets in ternary contour “subselection triangles” and guide users with recommendations based on further qualitative selection standards. ClimSIPS represents a novel framework to select models in an informed, efficient, and transparent manner and addresses the growing need for guidance and simple tools, so those seeking climate services can navigate the increasingly complex CMIP landscape.
... Furthermore, ACCESS-M is simply built from time-averaged model volume fluxes, while OCIM2 has a steady transport that is optimized to yield propagated tracer concentrations that are as close as possible to observations. ACCESS-M therefore inherits documented circulation and thermodynamic biases from the parent ocean model (Marsland et al., 2013;Bi et al., 2013bBi et al., , 2020. ...
... The unrealistically deep mixed layer in the Weddell Sea was reported for the 500 years AC-CESS1.3 benchmark run of Bi et al. (2013b), although that run did not have the deep mixed layer in the Ross Sea that was present in the ACCESS1.3 runs on which ACCESS-M is based. Appendix D: Biogeochemical diagnostic computations D1 Export production Export production via a given carbon species (POC f , POC s , DOC, or PIC), referred to here as a specific export pathway, is calculated using a Green function approach. ...
Full-text available
Accurate predictive modeling of the ocean's global carbon and oxygen cycles is challenging because of uncertainties in both biogeochemistry and ocean circulation. Advances over the last decade have made parameter optimization feasible, allowing models to better match observed biogeochemical fields. However, does fitting a biogeochemical model to observed tracers using a circulation with known biases robustly capture the inner workings of the biological pump? Here we embed a mechanistic model of the ocean's coupled nutrient, carbon, and oxygen cycles into two circulations for the current climate. To assess the effects of biases, one circulation (ACCESS-M) is derived from a climate model and the other from data assimilation of observations (OCIM2). We find that parameter optimization compensates for circulation biases at the expense of altering how the biological pump operates. Tracer observations constrain pump strength and regenerated inventories for both circulations, but ACCESS-M export production optimizes to twice that of OCIM2 to compensate for ACCESS-M having lower sequestration efficiencies driven by less efficient particle transfer and shorter residence times. Idealized simulations forcing complete Southern Ocean nutrient utilization show that the response of the optimized system is sensitive to the embedding circulation. In ACCESS-M, Southern Ocean nutrient and dissolved inorganic carbon (DIC) trapping is partially short circuited by unrealistically deep mixed layers. For both circulations, intense Southern Ocean production deoxygenates Southern-Ocean-sourced deep waters, muting the imprint of circulation biases on oxygen. Our findings highlight that the biological pump's plumbing needs careful assessment to predict the biogeochemical response to ecological changes, even when optimally matching observations.
Climate projections simulated by the Brazilian Earth System Model (BESM2.5) under RCP4.5 and RCP8.5 climate scenarios are analyzed based on future changes of surface air temperature (SAT) and precipitation with respect to the historical reference period 1971–2000. Since BESM2.5 is the only climate model developed in a South American country, this study gives a particular emphasis to South American future climate projections. Regarding the surface air temperature, BESM2.5 projects a steady warming throughout the 21st century, with the highest warming over eastern Amazonia, northern Chile and central South America for both scenarios. The SAT changes range between 2 °C and 3–4 °C for RCP4.5 and RCP8.5, respectively. On the other hand, projected precipitation varies over different regions of South America, with decreasing and increasing trends over the Amazon and southern South America, respectively. Interestingly, this study shows contrasting results with respect to extreme precipitation indicators, projecting enhanced extreme events with higher numbers of both consecutive dry days and days in which the precipitation exceeds 20 mm over the southeastern region. The model projects a meridional dipole pattern in the precipitation, with decreasing precipitation and longer dry spells over Northeast Brazil and the East Amazon region and increasing precipitation and shorter dry spells over Northwest South America and West Amazon, that is driven by future changes in the SLP that imposes a meridional gradient over these regions, causing the increase of westerlies that are likely to increase the moisture transport from the Pacific Ocean into western South America and the weakening of the easterlies that transport moisture over eastern South America and East Amazon.
Full-text available
Harnessing energy from the sun is crucial for locations battling with energy poverty and generation, especially in Africa, where equity in energy distribution and generation is a daily challenge. However, the evaluation and analysis of solar radiation has been limited by the paucity of atmospheric data in the African region. This study used monthly downward surface solar radiation (SSRD) from ERA5 as reference data to evaluate simulations of solar radiation from CORDEX, CMIP5 and CMIP6 models spanning the period 1990−2020 (present-day), mid-future (2020−2050), and far-future (2070−2100) across 4 climatic zones (Coastal, Forest, Guinea and Sahel) in Nigeria. Solar radiation were found to be overestimated in the Guinea and Sahel zones of the country, but fairly good performance were made in the Coastal and Forest zones. CMIP5, CMIP6 and CORDEX individual models all exhibit strong agreement in the projection of solar dimming across the four climatic zones in the mid- and far-future under both RCP4.5/SSP5\(-\)4.5 and RCP8.5/SSP5\(-\)8.5 scenarios. However, under the RCP8.5/SSP5\(-\)8.5 the greatest magnitude of dimming (\(-\,35 W/m^2\)) was found in CMIP6 models in the far-future and (\(-12 W/m^2\)) in the mid-future. The projected solar dimming was also predominant in all climatic regions under SSP5\(-\)4.5 for CORDEX, CMIP5, and CMIP6 models but at a much lower magnitude.
Full-text available
The objectives of the work were to understand the potential future climate changes in the Mediterranean region, assess the drought tolerance of the black calla lily (Arum palaestinum Boiss.), and investigate the mechanisms associated with its ability to withstand drought conditions. The Shared Socioeconomic Pathways (SSPs) of the Intergovernmental Panel on Climate Change (IPCC) were used to predict future temperature and precipitation changes. Both the SSP2-4.5 and SSP5-8.5 scenarios predicted a general increase in minimum and maximum temperatures and a decrease in precipitation. The projected increase in minimum temperature ranged from 2.95 °C under SSP2-4.5 to 5.67 °C under SSP5-8.5. The projected increase in maximum temperature ranged from 0.69 °C under SSP2-4.5 to 3.34 °C under SSP5-8.5. The projected decrease in precipitation ranged from −1.04 mm/day under SSP2-4.5 to −1.11 mm/day under SSP5-8.5. Results indicated that drought significantly impacted the physiological responses of the black calla lily. As drought increased, the black calla lily showed a reduction in leaf characteristics and non-structural carbohydrates, while proline content and reducing sugar content were increased, enhancing drought tolerance through osmoregulation. The black calla lily tolerates drought at a total ET of up to 50%. It has the potential to adapt to expected climate change through osmoregulation or by building a carbon and nitrogen sink for stress recovery.
Full-text available
Climate scenarios are important information for water planning, but, in some cases, they disagree with recent climate alterations, which affects their robustness and reliability. Robustness evaluation can help identifying areas that should be prioritized by in water sector adaptation to climate change. Although crucial, this kind of analysis has been overlooked in most climate change assessments, for instance in South America. This study assesses the robustness and reliability of river discharge scenarios by comparing them with observed and modelled data. Areas where current changes and scenarios agree are more likely to experience changes and, therefore, water planners should pay special attention to them. Tocantins-Araguaia, São Francisco, Western Northeast Atlantic and upper La Plata basins agreed with a discharge decrease, indicating that climate change should be prioritized in planning. Orinoco and upper-western Amazon basins showed strong disagreement between recent and projected discharge alterations, with positive change in last decades, showing that scenarios in these regions should be carefully interpreted. With this, water planners could interpret Northeastern and upper-central South America as presenting more likely scenarios in comparison to Amazon and Orinoco basins. Keywords: Climate change impacts; South America; Discharge alteration; Robustness
Motivated by the limited understanding of future changes in Mesoscale Convective Systems (MCSs), we investigated characteristics of warm‐season (June‐August) MCSs in the central United States based on high‐resolution convection‐permitting Weather Research and Forecasting (WRF) simulations. We examined two 15‐year simulations, which include current simulations (2004‐2018) forced by ECMWF reanalysis version 5 (ERA5) and future simulations (2086‐2100) forced by perturbed ERA5 (i.e., ERA5 + climate change signal derived from 28 Coupled Intercomparison Projected Phase 6 models under the Shared Socioeconomic Pathway‐Representative Concentration Pathway 8.5 emission scenario). The initiations and longevities of MCSs were determined using the object tracking algorithm MODE‐Time Domain (MTD) from observation (OBS), current simulations (ERA), and future simulations (PGW). MTD‐identified objects were divided into short‐/long‐lived (based on 75 th percentiles of longevity) and daytime (initiated during 00‐11 UTC)/nighttime (initiated during 12‐23 UTC). We found that ERA and OBS have comparable occurrences of MCSs. MCSs in PGW are associated with intensified rain rates (RR) in New Mexico, Colorado, and Kansas and lower RR in Texas, Louisiana, and Arkansas than in ERA. Moreover, the statistical analysis based on 15 parameters before MCSs initiation indicates that short‐lived MCSs in PGW are characterized by prominent changes in precipitable water (PW) and the most unstable convective available potential energy (MUCAPE). We also found that long‐lived MCSs in PGW are associated with prominent changes in PW, MUCAPE, and isentropic potential vorticity at 345K. According to the statistical results, PW is the most important variable in determining the longevity of MCSs and in understanding future changes. This article is protected by copyright. All rights reserved.
Full-text available
Contemporary general circulation models (GCMs) and Earth system models (ESMs) are developed by a large number of modeling groups globally. They use a wide range of representations of physical processes, allowing for structural (code) uncertainty to be partially quantified with multi‐model ensembles (MMEs). Many models in the MMEs of the Coupled Model Intercomparison Project (CMIP) have a common development history due to sharing of code and schemes. This makes their projections statistically dependent and introduces biases in MME statistics. Previous research has focused on model output and code dependence, and model code genealogy of CMIP models has not been fully analyzed. We present a full reconstruction of CMIP3, CMIP5, and CMIP6 code genealogy of 167 atmospheric models, GCMs, and ESMs (of which 114 participated in CMIP) based on the available literature, with a focus on the atmospheric component and atmospheric physics. We identify 12 main model families. We propose family and ancestry weighting methods designed to reduce the effect of model structural dependence in MMEs. We analyze weighted effective climate sensitivity (ECS), climate feedbacks, forcing, and global mean near‐surface air temperature, and how they differ by model family. Models in the same family often have similar climate properties. We show that weighting can partially reconcile differences in ECS and cloud feedbacks between CMIP5 and CMIP6. The results can help in understanding structural dependence between CMIP models, and the proposed ancestry and family weighting methods can be used in MME assessments to ameliorate model structural sampling biases.
Full-text available
Deep learning (DL) methods have recently garnered attention from the climate change community, as an innovative approach for downscaling climate variables from Earth System and Global Climate Models (ESGCMs) with horizontal resolutions still too coarse to represent regional-to-local-scale phenomena. In the context of the Coupled Model Intercomparison Project phase 6 (CMIP6), ESGCMs simulations were conducted for the Sixth Assessment Report (AR6) of the Intergovernmental Panel on Climate Change (IPCC), at resolutions ranging from 0.70º to 3.75º. Here, four Convolutional Neural Network (CNN) architectures were evaluated for their ability to downscale, to a resolution of 0.1º, seven CMIP6 ESGCMs over the Iberian Peninsula - a known climate change hotspot, due to its increased vulnerability to projected future warming and drying conditions. The study is divided into three stages: (1) evaluating the performance of the four CNN architectures in predicting mean, minimum, and maximum temperatures, as well as daily precipitation, trained using ERA5 data, and compared with the Iberia01 observational dataset; (2) downscaling the CMIP6 ESGCMs using the trained CNN architectures and further evaluating the ensemble against Iberia01; and (3) constructing a multi-model ensemble of CNN-based downscaled projections for temperature and precipitation over the Iberian Peninsula at 0.1º resolution throughout the 21st century, under four Shared Socioeconomic Pathway (SSP) scenarios. Upon validation and satisfactory performance evaluation, the DL downscaled projections demonstrate overall agreement with the CMIP6 ESGCM ensemble in terms of temperature and precipitation projections. Moreover, the advantages of using a high-resolution DL downscaled ensemble of ESGCM climate projections are evident, offering substantial added value in representing regional climate change over Iberia. Notably, a clear warming trend is observed, consistent with previous studies in this area, with projected temperature increases ranging from 2 ºC to 6 ºC depending on the climate scenario. Regarding precipitation, robust projected decreases are observed in western and southwestern Iberia, particularly after 2040. These results may offer a new tool for providing regional climate change information for adaptation strategies based on CMIP6 ESGCMs prior to the next phase of the European branch from the Coordinated Regional Climate Downscaling Experiment (EURO-CORDEX) experiments.
Full-text available
Coordinated Ocean-ice Reference Experiments (COREs) are presented as a tool to explore the behaviour of global ocean-ice models under forcing from a common atmospheric dataset. We highlight issues arising when designing coupled global ocean and sea ice experiments, such as difficulties formulating a consistent forcing methodology and experimental protocol. Particular focus is given to the hydrological forcing, the details of which are key to realizing simulations with stable meridional overturning circulations. The atmospheric forcing from [Large, W., Yeager, S., 2004. Diurnal to decadal global forcing for ocean and sea-ice models: the data sets and flux climatologies. NCAR Technical Note: NCAR/TN-460+STR. CGD Division of the National Center for Atmospheric Research] was developed for coupled-ocean and sea ice models. We found it to be suitable for our purposes, even though its evaluation originally focussed more on the ocean than on the sea-ice. Simulations with this atmospheric forcing are presented from seven global ocean-ice models using the CORE-I design (repeating annual cycle of atmospheric forcing for 500 years). These simulations test the hypothesis that global ocean-ice models run under the same atmospheric state produce qualitatively similar simulations. The validity of this hypothesis is shown to depend on the chosen diagnostic. The CORE simulations provide feedback to the fidelity of the atmospheric forcing and model configuration, with identification of biases promoting avenues for forcing dataset and/or model development.
Full-text available
This paper describes the development of a technically robust climate modelling system, HadGEM3, which couples the Met Office Unified Model atmosphere component, the NEMO ocean model and the Los Alamos sea ice model (CICE) using the OASIS coupler. Details of the coupling and technical solutions are documented in the paper in addition to a description of the configurations of the individual submodels. The paper demonstrates that the implementation of the model has resulted in accurate conservation of heat and freshwater across the model components. The model performance in early versions of this climate model is briefly described to demonstrate that the results are scientifically credible. HadGEM3 is the basis for a number of modelling efforts outside of the Met Office, both within the UK and internationally. This documentation of the HadGEM3 system provides a detailed reference for developers of HadGEM3-based climate configurations.
Full-text available
A subgrid-scale form for mesoscale eddy mixing on isopycnal surfaces is proposed for use in non-eddy-resolving ocean circulation models. The mixing is applied in isopycnal coordinates to isopycnal layer thickness, or inverse density gradient, as well as to passive scalars, temperature and salinity. The transformation of these mixing forms to physical coordinates is also presented.
Full-text available
The Australian Community Climate and Earth-System Simulator (ACCESS) is a coupled climate and earth system simulator being developed as a joint initiative of the Bureau of Meteorology and CSIRO in cooperation with the university community in Australia. The main aim of ACCESS is to develop a national approach to climate and weather prediction model development. Planning for ACCESS development commenced in 2005 and significant progress has been made subsequently. ACCESS-based numerical weather prediction (NWP) systems were implemented operationally by the Bureau in September 2009 and were marked by significantly increased forecast skill of close to one day for three-day forecasts over the previously operational systems. The fully-coupled ACCESS earth system model has been assembled and tested, and core runs have been completed and submitted for the Intergovernmental Panel for Climate Change (IPCC) Fifth Assessment Report. Significant progress has been made with ACCESS infrastructure including successful porting to both Solar and Vayu (National Computational Infrastructure (NCI)) machines and development of infrastructure to allow usage by university researchers. This paper provides a description of the NWP component of ACCESS and presents results from detailed verification of the system.
Full-text available
We evaluate the performance of two versions of the ACCESS model (1.0 and 1.3) in simulating both the historical (1979-2008) and projected (2071-2100) atmospheric circulations during, principally, the austral winter under two CMIP5 emission scenarios (RCP4.5 and RCP8.5). The model biases are estimated relative to two recent reanalysis datasets, while the projected circulation changes are assessed against the simulated historical circulations. Overall, both ACCESS models display model biases comparable in magnitude to other CMIP5 model biases. The most significant biases include the upper-tropospheric cold and polar warm biases, a westerly wind bias in the tropical upper troposphere and easterly wind biases in the southern and northern mid-latitudes, a narrower than observed Hadley circulation cell, a stronger Walker circulation cell, and drying (moistening) near the outer edges of the ascending (descending) branch of the Hadley cell. The projected circulation changes for the late 21st century in ACCESS simulations are largely similar to those found in the previous generation climate models including upper-tropospheric and polar warmings, a stronger subtropical jet, a poleward shifted mid-latitude jet, a deeper Hadley cell with its descending branch expanding poleward, and a weakened Walker circulation. However, our analysis also re-veals a moderate intensification of the projected Hadley cell in the RCP4.5, but not the RCP8.5, simulations. Most of the projected changes are similar in ACCESS1.0 and ACCESS1.3, except that the Walker circulation change in ACCESS1.3 is es-sentially an eastward shift of its ascending branch to the east of the dateline, while that in ACCESS1.0 is an in-place weakening of the circulation.
Full-text available
This atlas consists of a description of data analysis procedures and horizontal maps of annual, seasonal, and monthly climatological distribution fields of salinity at selected standard depth levels of the world ocean on a one-degree latitude-longitude grid. The aim of the maps is to illustrate large-scale characteristics of the distribution of ocean salinity. The fields used to generate these climatological maps were computed by objective analysis of all scientifically quality-controlled historical salinity data in the World Ocean Database 2009. Maps are presented for climatological composite periods (annual, seasonal, monthly, seasonal and monthly difference fields from the annual mean field, and the number of observations) at selected standard depths.
Many numerical weather prediction models are moving towards more prognostic schemes for the prediction of ice and liquid water contents within clouds. This paper describes a large-scale cloud and precipitation scheme developed for the UK Meteorological Office's Unified Model. It uses physically based transfer equations to predict ice as a prognostic variable. We review similar schemes and then describe our new scheme, giving examples of its performance in mesoscale forecasts compared with the current operational scheme and with observations. The microphysical processes occurring in a frontal cloud are well modelled. The prediction of supercooled stratocumulus cloud is much improved.
The Hadley Centre climate model version HadAM2 is used to study the sensitivity of modelled Antarctic climate to the parametrization of surface and boundary-layer heat fluxes under stable conditions. Specifically, the impact of changing the dependence of surface exchange coefficients and eddy diffusivities on the Richardson number is investigated. Three alternative parametrizations are implemented; in all of these the exchange coefficients decrease more rapidly with increasing stability than they do in the standard parametrization used in the model. When only the surface flux scheme is replaced by one of these alternatives, cooling is largely restricted to the surface, with some compensating warming occurring at the lowest atmospheric levels, and little change is seen in the low-level wind field over Antarctica. if alternative schemes are implemented both at the surface and in the boundary layer, widespread cooling occurs at the surface and at the lowest one or two atmospheric levels. The increased negative buoyancy thus generated causes significant increases in the speed of katabatic winds blowing down the coastal slopes of Antarctica. Colder and stronger offshore winds lead to increased cooling of the Antarctic coastal waters. In a coupled model, this could impact on the production of sea ice and ocean-bottom water. The modelled temperature changes appear to show both a direct response to changed boundary-layer heat-flux divergence and an indirect response as a result of the consequent changes to the low-level circulation.
The three Australian submissions to CMIP5, based on two versions of the ACCESS coupled climate model and the upgraded CSIRO model Mk3.6, are evaluated using skill scores for both global fields and for features of Australian climate. Global means of surface air temperature and precipitation are similar to those from observational datasets, except for a cool bias in Mk3.6. The agreement of the climatological seasonal mean fields for these and eleven other quantities, as measured by the M score, is comparable to the Australian CMIP3 models, in the case of Mk3.6, but mostly improved on by both ACCESS1.0 and ACCESS1.3. The ACCESS mean sea-level pressure and winds are notably better. An overall skill score, calculated for both global and Australian land, shows that the ACCESS models are among the best performing of 25 CMIP5 models. ACCESS1.0 improves slightly on the U.K. Met Office submissions. A suite of tests developed for the CAPTIVATE project is applied, and again ACCESS1.0 matches the U.K. Met Office reference model. ACCESS1.3 has improved cloud cover and a better link between equatorial sea surface temperatures (SSTs), using the Pacific-Indian Dipole index, and Australian rainfall. Mk3.6 has excessive variability in the SST index. In all, both versions of ACCESS are highly successful, while the ensembles of simulations of Mk3.6 make it also a very worthwhile submission to CMIP5.