GMDGeoscientific Model DevelopmentGMDGeosci. Model Dev.1991-9603Copernicus PublicationsGöttingen, Germany10.5194/gmd-10-3715-2017The BRIDGE HadCM3 family of climate models: HadCM3@Bristol v1.0ValdesPaul J.p.j.valdes@bristol.ac.ukhttps://orcid.org/0000-0002-1902-3283ArmstrongEdwardhttps://orcid.org/0000-0001-9561-0159BadgerMarcus P. S.https://orcid.org/0000-0001-8195-5244BradshawCatherine D.BraggFranhttps://orcid.org/0000-0002-8179-4214CrucifixMichelhttps://orcid.org/0000-0002-3437-4911Davies-BarnardTarakahttps://orcid.org/0000-0001-7389-6571DayJonathan J.https://orcid.org/0000-0002-3750-649XFarnsworthAlexhttps://orcid.org/0000-0001-5585-5338GordonChrisHopcroftPeter O.https://orcid.org/0000-0003-3694-9181KennedyAlan T.https://orcid.org/0000-0001-5143-8932LordNatalie S.https://orcid.org/0000-0002-5312-6593LuntDan J.https://orcid.org/0000-0003-3585-6928MarzocchiAlicehttps://orcid.org/0000-0002-3430-3574ParryLouise M.PopeVickyRobertsWilliam H. G.StoneEmma J.TourteGregory J. L.https://orcid.org/0000-0002-2819-392XWilliamsJonny H. T.https://orcid.org/0000-0002-0680-0098School of Geographical Sciences, University of Bristol, Bristol, UKSchool of Environment, Earth and Ecosystem Sciences, The Open University, Milton Keynes, UKApplied Science group, Met Office Hadley Centre, Exeter, UKEarth and Life Institute, Georges Lemaître Centre for Earth and Climate Research, Université catholique de Louvain, Louvain-la-Neuve, BelgiumCollege of Engineering, Mathematics, and Physical Sciences, University of Exeter, Laver Building, North Park Road Exeter, EX4 4QE, UKDepartment of Meteorology, University of Reading, Reading, UKCentre for Climate Research Singapore, Meteorological Service Singapore, SingaporeDepartment of the Geophysical Sciences, The University of Chicago, Chicago, IL, USAScottish Environment Protection Agency (SEPA), Perth, UKNational Institute of Water and Atmospheric Research (NIWA), Wellington, New ZealandMet Office, Hadley Centre, Fitzroy Road, Exeter, UKPaul J. Valdes (p.j.valdes@bristol.ac.uk)12October201710103715374320January20178February201728June20177July2017This work is licensed under the Creative Commons Attribution 3.0 Unported License. To view a copy of this licence, visit https://creativecommons.org/licenses/by/3.0/This article is available from https://gmd.copernicus.org/articles/10/3715/2017/gmd-10-3715-2017.htmlThe full text article is available as a PDF file from https://gmd.copernicus.org/articles/10/3715/2017/gmd-10-3715-2017.pdf
Understanding natural and anthropogenic climate change processes involves
using computational models that represent the main components of the Earth
system: the atmosphere, ocean, sea ice, and land surface. These models have
become increasingly computationally expensive as resolution is increased and
more complex process representations are included. However, to gain robust
insight into how climate may respond to a given forcing, and to meaningfully
quantify the associated uncertainty, it is often required to use either or
both ensemble approaches and very long integrations. For this reason, more
computationally efficient models can be very valuable tools. Here we provide
a comprehensive overview of the suite of climate models based around the
HadCM3 coupled general circulation model. This model was developed at the UK
Met Office and has been heavily used during the last 15 years for a range of
future (and past) climate change studies, but has now been largely superseded
for many scientific studies by more recently developed models. However, it
continues to be extensively used by various institutions, including the
BRIDGE (Bristol Research Initiative for the Dynamic Global Environment)
research group at the University of Bristol, who have made modest adaptations
to the base HadCM3 model over time. These adaptations mean that the original
documentation is not entirely representative, and several other relatively
undocumented configurations are in use. We therefore describe the key
features of a number of configurations of the HadCM3 climate model family,
which together make up HadCM3@Bristol version 1.0. In order to differentiate
variants that have undergone development at BRIDGE, we have introduced the
letter B into the model nomenclature. We include descriptions of the
atmosphere-only model (HadAM3B), the coupled model with a low-resolution
ocean (HadCM3BL), the high-resolution atmosphere-only model (HadAM3BH), and
the regional model (HadRM3B). These also include three versions of the land
surface scheme. By comparing with observational datasets, we show that these
models produce a good representation of many aspects of the climate system,
including the land and sea surface temperatures, precipitation, ocean
circulation, and vegetation. This evaluation, combined with the relatively
fast computational speed (up to 1000 times faster than some CMIP6 models),
motivates continued development and scientific use of the HadCM3B family of
coupled climate models, predominantly for quantifying uncertainty and for
long multi-millennial-scale simulations.
Introduction
This paper describes the variants of the HadCM3 family of climate models (all
of which can be classed as general circulation models, GCMs), produced by the
UK Hadley Centre/Met Office, and which remain in regular use by a number of
research groups, including the Bristol Research Initiative for the Dynamic
Global Environment group (BRIDGE,
http://www.bristol.ac.uk/geography/research/bridge). HadCM3 originated
in the late 1990s with developments to the atmosphere model, HadAM3
. Together with improvements to the ocean GCM, this enabled
the development of HadCM3 , which was one of the first
coupled atmosphere–ocean GCMs that did not require flux correction to
maintain a reasonable present-day climate, i.e. the artificial adjustments of
water, heat, and momentum in order to maintain a stable climate. It has been
extensively used for scientific studies of future climate change
e.g. and
is heavily cited, including in the 2007 IPCC report , and was
still included in the 2013 report. The family of models has the advantage of
now being very well known in terms of their strengths and weaknesses, as
numerous studies have shown and classified model biases and forecast skill at
representing the mean climate state as well as variability
e.g.. The model family has now been
superseded by the HadGEM2 and HadGEM3
families of models.
Compared to more recent models, HadCM3 is relatively low resolution but
continues to perform reasonably well, at least with respect to its mean
climate . It also has the great benefit of
computational speed, being more than 1000 times faster than some of the most
recent and complex versions of the UK Met Office Unified Model (UM). This
computational speed is particularly valuable for long-term simulations
(necessary for many palaeoclimate simulations, studies which investigate the
carbon cycle and the evolution of ice sheets) and for large ensembles
(necessary for investigating the model's sensitivity to multiple parameters
and quantifying the uncertainty in the model's response to forcing). Long
model runs are also crucial for understanding unforced variability in the
climate system e.g..
Palaeoclimate simulations typically need many hundreds of model years to
reach near equilibrium in the surface and intermediate ocean and many
thousands of years to reach equilibrium in the deep ocean. Moreover, there
recently has been an increasing need to be able to consider the transient
behaviour of past climate change. This has previously been tackled using the
HadCM3 family of models by using either multiple “snapshot” simulations
e.g. or by
performing fully transient simulations for multi-centennial or -millennial
timescales e.g..
Faster models are also invaluable for investigating the sensitivity and
robustness of results to changes in the initial conditions of the model and
changes in boundary conditions such as topography, as numerous simulations
can be performed . Additionally, they are
ideal for investigating anthropogenic changes on long timescales
and for performing perturbed parameter
ensembles to rigorously calculate the probability density functions of either
the mean or extreme climates e.g. using HadCM3
variants,. Computational speed also aids more speculative
studies. For instance, many early geoengineering simulations were run using
variants of the HadCM3 family of models such
as.
In response to the need for fast models, Earth system models of intermediate
complexity (EMICs) have been developed. These models
frequently achieve their speed by heavily parameterising the atmospheric
response, even though atmospheric processes transport two-thirds of the total
heat from Equator to pole and play a vital role in the hydrological cycle. It
is, therefore, also important to have a class of fast models that is
equivalent to full atmosphere–ocean general circulation models (GCMs). Some
EMICS do represent the dynamics of the atmosphere; for instance, LOVECLIM
uses a three-level quasi-geostrophic atmosphere.
Similarly, FAMOUS (part of the HadCM3 family) includes a full primitive
equation atmosphere but at low resolution
. Hence the division between EMICs
and full complexity models is becoming increasingly blurred and we consider
that the HadCM3 family provides a further bridge in the spectrum of models
between intermediate complexity models and full complexity, state-of-the art
models.
Since its introduction, HadCM3 (and related models) has undergone a number of
changes, bug fixes, and adaptations.
The original model described in and , i.e.
HadCM3 with MOSES1, is still used, but now many other versions exist. Some
groups have largely stuck to the standard release of the model
e.g.. Other groups have incorporated a
variety of bug fixes and scientific changes; in particular, many papers have
used a revised land surface scheme, MOSES2.1 e.g., but
this is relatively poorly documented.
Therefore, in this paper, we aim to rectify this for the wide range of HadCM3
variants currently in use within the BRIDGE modelling group. Our
implementations of the models have diverged from other versions, and so here
we aim to provide clear documentation of our version of each model. In order
to do this more clearly, we use the nomenclature HadCM3B, in order to
differentiate model variants that have undergone development at Bristol from
those originally developed at the Met Office. We have followed a specific
modelling philosophy in which we attempt to minimise the differences between
model configurations, particularly when changing resolution. For instance,
previously published descriptions of HadRM3 and HadAM3H
use slightly improved physics to
HadCM3, but we choose to keep the same physics (except for specific changes
related to resolution).
We include detailed descriptions of each module of the models, differences
between variants, and comparison with observations across a range of metrics.
This will increase transparency, traceability, and scientific openness. By
detailing the changes and variations of these models, and providing an
extensive comparison to observational data, we hope to show that these models
remain useful tools for climate simulation and are suitable for further
scientific use. Furthermore, we shall show that despite their relative
simplicity, the models simulate the modern climate with comparable accuracy
to many of the latest CMIP5 models.
To this end, we first describe the “base” model, which we term HadCM3B-M1.
This is essentially almost identical to that of , but with
some minor modifications made by BRIDGE detailed in Sect. , to
which all the other models will be compared (Sect. ). As such,
because it is largely simply bug fixes, it could be argued that HadCM3B-M1 is
not a different model to the original, but we include it for completeness. We
then subsequently discuss different land surface schemes (Sect. ),
followed by model variants with different ocean or atmospheric resolutions
(Sect. ). Finally, we evaluate the models' performance when
compared to observations and CMIP5 models, to show that they recreate many
key aspects of the climate system, and show which models are more suitable
for certain applications (Sect. ).
Overview of HadCM3@Bristol
The family of models has at its
core HadCM3. From this core, variants are derived according to resolution,
land surface scheme, and components. In order to distinguish variants that have
undergone further development at Bristol to those originally developed at the
Met Office, we include the letter B for Bristol in the model
acronym. As discussed in the text, the changes between the Bristol and Met
Office variants are small in some cases; however, we believe they warrant
documentation to remove ambiguity. The model family is then split into
groups: HadCM3B, HadCM3BL (HadCM3B but lower ocean resolution), HadAM3B
(HadCM3B but atmosphere-only), HadAM3BH (HadAM3B but higher resolution),
HadRM3B (HadAM3B but regional), and FAMOUS (HadCM3L but lower atmosphere
resolution).
FAMOUS is a low-resolution model derived from HadCM3, sharing much of the
same physics, but with some numerical modifications suitable for the low
resolution and which give quicker run times. It is well documented elsewhere
and will not be described again in detail
here, although some comparisons with FAMOUS are included for completeness.
Run times for M1 model versions are compared in Table
for typical configurations. This demonstrates the efficiency of FAMOUS-M1 at
around several modelled centuries per day on just eight cores, and the
relatively high computational cost of the two high-resolution model versions
(HadAM3BH and HadRM3B). This compares with 1.87 model years per day on 1152
cores for the higher-resolution version of HadGEM3-GC2 .
As it is normal for these types of numerical models, it is to be noted that
the relationship between the number of cores used and the model efficiency is
not linear, i.e. using twice as many cores does not make the model twice as
fast, and eventually adding more cores has no positive effect on the model
efficiency.
The nomenclature adopted for the HadCM3@Bristol model variants is
Had〈Com〉M3B〈Res〉-〈Land〉〈Veg〉,
where〈Com〉
(components) is one of
A – atmosphere-only model;
C – coupled model;
R – regional model.
〈Res〉
(resolution) is one of
L – lower than standard-resolution ocean;
H – higher than standard-resolution atmosphere;
blank – standard resolution.
〈Land〉
(land surface scheme) is one of
M1 – MOSES1 land surface exchange scheme;
M2.1a – MOSES2.1a land surface exchange scheme;
M2.2 – MOSES2.2 land surface exchange scheme.
〈Veg〉
(vegetation) is one of
blank or N – no change to vegetation (i.e. static vegetation distribution);
E – vegetation predicted using TRIFFID, but in “equilibrium”
mode;
D – same as E above, but fully dynamic model.
As such, the original “base” model described in which has
undergone some minor modifications (see Sect. ) is named HadCM3B-M1.
HadCM3B-M1
This section describes the “core” model, HadCM3B-M1, to which all other
variants will be compared in this paper. This variant of the family was
originally the most commonly used, and is still used for studies where the
vegetation is known, and as such can be prescribed where relatively short
simulations are sufficient for the science questions being addressed and
where the ocean plays a critical role and as such high resolution is
desirable e.g..
This model is a three-dimensional, fully dynamic, coupled atmosphere–ocean
global climate model without flux adjustment. Our version of the model is
very similar to that described by . Our aim is to provide a
full description of how our version differs from that in
followed by a brief description of the core model. A full description of the
version can be found in the UK Met Office technical notes
http://cms.ncas.ac.uk/wiki/Docs/MetOfficeDocs; the base model code is
currently available to view at
http://cms.ncas.ac.uk/code_browsers/UM4.5/UMbrowser/index.html, but it
should be noted that additional modifications are required for the full
scientific definition of the model as described here (see Sect. “Code
availability”).
Modifications for the Bristol versions of HadCM3B-M1 and HadAM3B
We have benchmarked the standard version of HadCM3-M1 supplied by the UM Met
Office against existing model results from the published Hadley Centre
version of and confirmed that we could reproduce the
results within the normal statistical variability of the model. Subsequently
a few relatively minor changes have been made. These include the following.
HadCM3B-M1
Correction of a small bug in the Visbeck horizontal eddy mixing scheme
which was originally included in the standard
configuration of the model to ensure compatibility with previous
versions.
Use of versions of the radiation and primary field advection schemes
that are scientifically identical to the standard version and which make
the model faster but which are not bit reproducible.
HadAM3B
Fixes to a few array bounds errors (which may or may not have an
impact on the scientific results).
Multiple other bug fixes which did not change the science but
corrected problems with some aspects of the code and diagnostic outputs.
There were two small bugs in the conservation of atmospheric mass, and
the computation of vertical velocity, fixes to which are not included in
the standard release version of HadAM3 but are included in HadCM3B. We
include these bug fixes in all versions of the code so that our
atmosphere model (HadAM3B) is 100 % identical to the
atmosphere component of our version of HadCM3B.
There is also another important code fix (Steenman-Clark, personal communication)
which is vital to include. If this is not included, then some compilers
will lead to a large (e.g. 0.75 ∘C bias in global
mean surface air temperature) error in mean climate.
These modifications are included in the supplementary information. The overall
impact of these changes on the climate simulation is very small.
Atmosphere (HadAM3B)
The atmosphere component of HadCM3B is almost identical to the atmosphere
component of HadAM3B, which is the atmosphere-only variant with fixed sea
surface temperatures (SSTs). The modifications made at Bristol are
highlighted in Sect. , beyond which the model is the same as that
described by . HadAM3B has a Cartesian grid with a horizontal
resolution of 96×73 grid points (3.75∘
longitude × 2.5∘ latitude) with 19 hybrid levels (sigma
levels near the surface, changing smoothly to pressure levels near the top of
the atmosphere) in the vertical and uses a
30 min time step. HadAM3B solves the primitive equation set of
which includes certain terms necessary to conserve both
energy and angular momentum. Equations are solved through the use of a
grid-point scheme, specifically the Arakawa staggered B-grid
, on a regular latitude–longitude grid in the
horizontal. At high latitudes, Fourier filtering of higher wave-number
dynamics is used to remove sub-grid-scale variability. A split-explicit time
scheme conserves mass, mass-weighted potential temperature, moisture, and
angular momentum, and ensures the reliability for solving equations on long
timescales, which is particularly important for climate modelling
.
As with any climate model, a number of parameterisation schemes are needed
within HadAM3B to represent certain physical processes which occur on
sub-grid scales.
Precipitation is dealt with in two schemes: (i) the large-scale
precipitation scheme, and (ii) the convection scheme. The large-scale
precipitation scheme removes cloud water resolved on the grid scale, i.e.
frontal precipitation. This is done via a simple bulk parameterisation scheme
converting water content into precipitation
. The convection scheme uses a
mass-flux scheme with the addition of convective downdrafts.
A first-order scheme for turbulent vertical mixing of momentum and
thermodynamic quantities is used within the boundary layer, which can
occupy up to the first five layers of the model. Sub-grid-scale gravity wave
and orographic drag parameterisations include the impact of
orographic variance anisotropy . The scheme comprises
four elements: (i) “triggering” which determines whether the physical
conditions within the grid box constitute convection taking place;
(ii) “cloudbase closure” controlling the intensity of convection which is
determined by the mass transported through the cloudbase; (iii) a
transport model where temperature, moisture, wind fields, and thus
precipitation are determined, and (iv) a “convective cloud scheme” where
cloud fractions derived from convection are calculated which
will be used by the radiation scheme .
In the real world, clouds are formed on scales far below that of the
coarse grid used in HadAM3B; therefore, there is the need for a statistical
parameterisation of this variable. Probability density functions are used on
the total water content over the grid-box mean to parameterise cloud
amount/distribution and longevity . Clouds are modelled as
either water, ice, or mixed phase when the temperature in the model level is
between 0 and -9 ∘C. Clouds form when the
mean plus the standard deviation of the grid-cell moisture content
exceeds a threshold of relative humidity (see RHCrit in
Table for numerical values). This cloud water content
can then be used to produce a cloud fraction for each grid box
. The threshold of total water content for
precipitation to occur varies between land and ocean cells to account for
the different levels of available cloud condensation nuclei. The scheme
uses temperature through the vertical levels to determine the ice and
water phases to determine cloud water content.
Radiation is represented using the radiation scheme of
. This scheme has six short-wave and eight
long-wave bands and represents the effects of water vapour, carbon dioxide,
ozone, and minor trace gases. A background aerosol climatology following
increases the atmospheric absorption of short-wave
radiation relative to previous versions, representing a significant
improvement. The long-wave and short-wave spectral scheme used
“3A” of is an improvement over the previous
versions as it allows the freedom of choices of cloud parameterisation,
gases, and aerosols to be included through spectral input files
.
Horizontal diffusion takes the form k∇N where both k
and N can vary with vertical levels and with variable. The standard
resolution of the model uses a formulation k1∇6 where k1=5.47×108m6s-1 corresponding to a e-folding
timescale for the two-grid wave of approximately 12 h. The topmost
level in the model uses a stronger diffusion of the form k2∇2
where k2=4.0×106m2s-1. Moisture also has
stronger diffusion in the five levels below the top (approximately from
150 hPa) corresponding to km∇4 where km=1.5×108m4s-1. The functional form and strength of
diffusion for other resolutions are summarised in
Table .
Summary of the key differences between model variants. For further
details of these differences and description of the features common to all
variants, see the relevant sections of the text. Note that HadAM3B is
identical to the atmosphere of HadCM3B.
ItemHadCM3BHadCM3BLFAMOUSHadAMB3HadAM3BHHadRM3BAtmosphere Horizontal resolution (n)96 × 7396 × 7348 × 3796 × 73288 × 217Varies with selected regionHorizontal resolution (deg)3.75∘× 2.5∘3.75∘× 2.5∘7.5∘× 5∘3.75∘× 2.5∘1.25∘× 0.83∘0.4425∘× 0.4425∘ or 0.22∘× 0.22∘Vertical resolution19 levels19 levels11 levels19 levels30 levels19 levelsTime step (mins)30306030105 or 2Dynamics sweeps/physics time step111 or 2121Max wind test for half time step dynamics (ms-1)––––240–Convective precipitation grid-box fraction (conv_eps)0.30.30.30.30.30.65 or 1.0Large-scale precipitation grid-box fraction (ls_eps)1.01.01.01.01.00.75 or 1.0Boundary layer top and number of levels (eta/level)0.835/50.835/50.9/30.835/50.8/60.835/5Cloud levels (eta/level)0.02/180.02/180.125/100.02/180.02/290.02/18Pure pressure level start (eta/level)0.04/170.04/170.06/110.04/170.04/280.04/17Gravity wave drag start (eta/level)0.956/30.956/30.9/30.956/30.956/30.956/3Surface gravity wave constant (m)2.0 × 1042.0 × 1042.0 × 1042.0 × 1041.6 × 1042.0 × 104Trapped lee wave constant (m-3/2)3.0 × 1053.0 × 1053.0 × 1053.0 × 1052.4 × 1053.0 × 105Filtering safety multiplying factor0.010.010.0110.010.1–Filtering wave numbers checked every1 time step1 time step1 time step1 time step6 h–Steep slope horizontal diffusion off until pressure level (kPa)202020202050Diffusion coefficient (m6s-1)*5.47 × 1085.47 × 1084.19 × 1095.47 × 1084.0 × 1071.7 × 107Diffusion power (dimensionless)*668644Humidity diffusion coefficient (m6s-1/m4s-1)*5.47 × 1081.5 × 1085.47 × 1081.5 × 1082.4 × 1085.47 × 1081.5 × 1082.0 × 1074.0 × 1071.7 × 107Humidity diffusion power (dimensionless)*646446444RHcrit*0.95 0.70.95 0.70.91 0.6870.95 0.70.95 0.80.91 0.84 0.95Ocean Horizontal resolution (n)288 × 14496 × 7396 × 73–––Horizontal resolution (deg)1.25∘× 1.25∘3.75∘× 2.5∘3.75∘× 2.5∘Vertical resolution20 levels to 5500 m20 levels to 5500 m20 levels to 5500 m–––North Atlantic bathymetryStandard Met OfficeNo IcelandNo Iceland–––Vertical tracer diffusivityRichardson number dependenceConstant background valueConstant background value–––Coefficient for solar penetration (ratio)03.8 × 10-13.8 × 10-1–––Horizontal momentum diffusion coefficient (m2s-1)3 × 1031.5 × 1051.5 × 105–––Isopycnal diffusion coefficients (m2s-1) latitudinally varying schemeConstant valuesConstant values–––Sea ice diffusion (m2s-1)6.7 × 1022.0 × 1032.0 × 103–––
* Level-dependent parameters (where
multiple values are given, this indicates the range from surface to
top-of-atmosphere (TOA)).
Boundary conditions for the model include the land–sea mask, orography, and
its sub-grid-scale variability (originally derived from the US Navy updates
10' dataset), and a range of soil and vegetation parameters (originally
derived from data in . The model also needs to be
initialised with soil moisture and snow cover based
on, and deep soil temperatures (empirically derived using
). When the model is run in atmosphere-only mode, i.e.
HadAM3B, sea surface temperature and sea ice (concentration and depth) are
required to be prescribed. These can be derived from observational data or
from coupled model simulations.
Ocean
The ocean component has a horizontal resolution of 288×144 grid
points (1.25∘× 1.25∘) and, as
with the atmosphere, also uses Fourier filtering at high latitudes. The
higher resolution means that six ocean grid cells correspond to each
atmosphere grid cell. In order to simplify the coupling of the atmosphere and
ocean models, the land–sea mask is defined at the atmosphere resolution;
therefore, the ocean model's coastlines appear relatively coarse. In the
vertical there are 20 depth levels with finer definition at the ocean
surface, with the topmost model layer being 10 m thick and the
bottommost 616 m thick. The ocean time step is 1 h. The ocean and
atmosphere modules are coupled once a day with no flux adjustment necessary.
The ocean model is based on the model of and is a full
primitive equation, three-dimensional model of the ocean. A second-order
numerical scheme is used along with centred advection to remove non-linear
instabilities. The Arakawa B-grid is used for staggering of tracer and
velocity variables, allowing for more accurate numerical calculations of
geostrophically balanced motion. It uses a rigid lid which eliminates fast
external mode gravity waves found in the real ocean, thus allowing for longer
time steps, and with the result that there is no variation in the volume of
the ocean. The barotropic solver requires the pre-specification of
“islands” around which the barotropic circulation may occur (see
Sect. ).
Availability of alternative land surface schemes.
ItemHadCM3BHadCM3BLFAMOUSHadAM3BHadAM3BHHadRM3BMOSES1HadCM3B-M1HadCM3BL-M1FAMOUS-M1HadAM3B-M1HadAM3BH-M1HadRM3B-M1MOSES2.1HadCM3B-M2.1HadCM3BL-M2.1–HadAM3B-M2.1HadAM3BH-M2.1*MOSES2.1 TRIFFID (D and E)HadCM3B-M2.1HadCM3BL-M2.1–HadAM3B-M2.1HadAM3BH-M2.1*MOSES2.2HadCM3B-M2.2HadCM3BL-M2.2FAMOUS-M2.2***MOSES2.2 TRIFFID (D and E)HadCM3B-M2.2HadCM3BL-M2.2FAMOUS-M2.2***
* Variant currently does not exist, but
there is no barrier to creation.
As with the atmosphere, the ocean model also requires a number of
parameterisations.
The ocean mixed layer is represented by the
model which assigns 15 % of gravitational potential energy
and 70 % of wind-stress energy to turbulent kinetic energy,
which is mixed out exponentially with depth. At all depths, five
iterations of convective mixing are carried out at each time step. Tracer and
momentum mixing is modelled using the K-theory scheme. Within the
mixed layer a simplified version of the scheme is
employed: below this the K-theory
parameterisation is used.
Momentum mixing is approximated using diffusion that is governed by a
coefficient that consists of two terms: a constant background value and a
term dependent on the local Richardson number. For tracers, diffusion
increases with depth as detailed in Table A of .
Horizontal eddy mixing of tracers is carried out using the isopycnal
parameterisation of , with thickness diffusion
coefficients modified following the method of .
Isopycnal mixing uses the implementation of the
scheme. The along-isopycnal diffusion coefficient is
1000 m2s-1. Horizontal mixing of momentum is performed
using a latitudinally varying formulation which, coupled with the finer
resolution of the ocean grid, enables western boundary currents to be
resolved.
There is no dynamic connection between the Mediterranean Sea and
Atlantic Ocean, so it is modelled as a “diffusive pipe” by completely
mixing the easternmost point of the Atlantic with the westernmost point
of the Mediterranean. Mixing occurs over the top 13 layers, to a depth of
1200 m, on the assumption that Mediterranean water will sink
to at least this depth. A similar parameterisation is applied in the outflow
of Hudson Bay.
Ice sheets are not modelled dynamically; therefore, the snow
accumulation on surface land ice points and over isolated water bodies
must be balanced by loss through a notional iceberg calving that is
represented as a time-invariant freshwater flux (which, because of the
rigid lid, is converted to a virtual salinity flux). This is distributed
around the edge of the ice sheets and polar oceans. The virtual salinity
flux is calculated using a globally constant reference salinity, which
can distort the local response to the surface water forcing. River runoff
is instantaneously transferred to the ocean using a prescribed runoff
map.
The modern bathymetry for the model is derived from the ETOPO5 reconstruction
using a simple smoothing algorithm. The geometry of some
significant channels is modified from the resulting coarse interpolation to
ensure a more realistic model performance . For example,
the Greenland–Scotland Ridge and Denmark Strait have significant
sub-grid-scale channels which are lost in the smoothing and so have been
re-created by deepening channels (single cell width) in three locations along
the ridge to reproduce the mean outflow to match observations, and the
bathymetry around Indonesia is modified to ensure that flow occurs between
Indonesia and Papua New Guinea but not between Indonesia and the mainland of
Asia.
Sea ice
Sea ice is calculated as a zero layer model on top of the ocean grid. Partial
cell coverage of sea ice is possible up to 0.995 in the Arctic and 0.98 in
the Antarctic. This is based on the parameterisation of sea ice concentration
from . Ice forms primarily by freezing in leads, although
ice can also form from snow falling on existing ice. It is assumed to freeze
at the base of the sea ice at the freezing point of -1.8 ∘C. A
constant salinity is assumed for ice, with the excess salt on
melting/formation added as a flux into the ocean. Sea ice dynamics are simply
parameterised: the surface wind stress over sea ice is applied to the ocean
beneath the ice, and the ice thickness, concentration, and accumulated snow
then drift following the ocean currents in the top model layer
. The maximum depth that sea ice can reach due to
convergence from drift is limited to 4 m in depth, although it may
subsequently thicken further due to freezing. The albedo of sea ice is set at
0.8 for temperatures below -10 ∘C and 0.5 for temperatures above
0 ∘C, with a linear variation between these values.
Land surface scheme: MOSES1
The MOSES (Met Office Surface Exchange Scheme) land surface
scheme is built upon the previous
Met Office land surface scheme (UKMO) . In the
version of HadCM3, MOSES version 1, MOSES1, is used. A
technical overview of MOSES1, a comparison to its predecessor (UKMO) and its
climatological impact are provided by .
In addition to calculating the fluxes of water and energy, MOSES1
incorporates the physiological impact of atmospheric carbon dioxide, water
vapour, and temperature on photosynthesis and stomatal conductance. It
accounts for the effects of freezing and melting of soil moisture in four
soil layers, the proportion of frozen soil moisture being a function of the
soil heat capacity and conductivity of the grid cell. Both vegetated and
non-vegetated land surface types are characterised by a set of surface
properties that are not updated during the model run. The canopy scheme is
based on that used in .
MOSES1 has two sets of prescribed land surface property attributes, which are
input into the model via two external files. The soil attributes are
volumetric soil moisture concentration at the wilting point, critical point,
field capacity, and saturation, the saturated hydrological soil conductivity,
the Clapp–Hornberger B exponent, the thermal capacity of soil, thermal
conductivity of soil, and the saturated soil water suction. (The
Clapp–Hornberger exponent is a measure of the pore volume distribution and
the formulation was originally devised in .) The
vegetation attributes are root depth, snow-free albedo, stomatal resistance
to evaporation, surface roughness, canopy water capacity, infiltration
enhancement rate, deep snow albedo, leaf area index, and canopy height of
vegetation fraction. All of these attributes are derived from the
dataset.
Alternative land surface schemes
Section describes the MOSES1 land
surface
scheme which is used in the standard version of HadCM3. Here we describe two
other versions, MOSES2.1 and MOSES2.2, as well as the vegetation component
TRIFFID.
MOSES2
MOSES1 requires maps of vegetation properties, such as root depth and leaf
area index, to be prescribed (normally in a set of external files). As such,
it is not very suitable for an interactive vegetation model. As part of the
process of developing a dynamic vegetation module for HadCM3, an upgraded
land surface scheme, MOSES2, was also developed. The first version of this
scheme, MOSES2.1, is the original scheme used in early work with dynamic
vegetation . This version was originally coupled to HadCM3LC
, which is a flux-corrected low-resolution version of HadCM3
which includes a carbon cycle. MOSES2.1 was further developed for use in
HadCM3 as part of the Paleoclimate Modelling Intercomparison Project Phase II
(PMIP2) . Subsequently, a second version of MOSES2 was
developed, MOSES2.2 , which was similar
scientifically to MOSES2.1 but had improved code structure. This has become
the initial core of the JULES land surface model .
At the University of Bristol, we have mainly used MOSES2.1, with MOSES2.2
only being used in a few specific contexts such as for investigating changes
in atmospheric chemistry because it can
include additional parameterisations of isoprene emissions. MOSES2.2 can also
be used in FAMOUS , though the majority of FAMOUS
publications have used MOSES1.
A detailed discussion of the upgrades between MOSES1 and MOSES2.2 is provided
in and a full and complete technical overview of MOSES2.2 in
. But so far there have been no clear comparisons as to how
MOSES2.2 differs scientifically or technically from MOSES2.1, despite MOSES2.1
being the core version used at Bristol. The following sections aim to rectify
this and clarify the differences between MOSES2.2 and MOSES2.1. First we
outline how MOSES2.2 differs from MOSES1.
Differences between MOSES2.2 and MOSES1
Compared to MOSES1, MOSES2.2 has major upgrades to all aspects of the land
surface exchange and the surface radiation scheme . The
surface radiation scheme has an updated coupling between the land surface and
atmosphere, including the calculation of surface net radiation and surface
heat and moisture fluxes. MOSES2.2 allows fractional coverage of different
surface types on a sub-grid scale. There are nine land surface types
explicitly modelled at a sub-grid scale, each with a set of characteristic
parameters. MOSES2.2 can be fully coupled to the TRIFFID dynamic vegetation
model (see Sect. ) via the five plant functional types
(PFTs): broadleaf trees, needleleaf trees, shrubs, C3 (temperate) grasses,
and C4 (tropical) grasses. The remaining four are non-vegetated surface
types: urban, inland water, bare soil, and ice. Excluding ice type, each land
surface grid box can be made up of any mixture of the other eight surface
types. Land ice must have a fractional cover of 0 or 1 only. The fractional
coverage for each surface type is specified for each grid point from an
external file. In addition, another file is supplied specifying the necessary
parameters for the five vegetation types at each grid point: leaf area index
(LAI), canopy height, and canopy conductance (not PFT dependent). The
vegetation fractions and parameters will be updated by TRIFFID if it is being
used. Other PFT-dependent parameters, including root depth and values of
albedo under a variety of conditions, are hard-wired into the code.
In MOSES1, the surface energy and moisture fluxes are calculated based on
grid-box average values of parameters (such as roughness and length). In
MOSES2.2, the surface energy balance is explicitly solved for each surface
type and then weighted by the fractional area of the surface types within the
grid box. This produces the grid-box average surface temperature and soil
moisture and fluxes of long-wave, short-wave, sensible, latent, and ground
heat. Above the surface, air temperature, humidity, and wind speed on
atmospheric levels are treated as homogeneous across the grid box. Similarly,
soil temperatures and moisture contents are also treated as homogeneous. The
aerodynamic surface roughness lengths are calculated explicitly according to
the canopy height and the rate of change of roughness length with canopy
height for each tile. This roughness length is used to calculate
surface–atmosphere fluxes of heat, water, momentum, and CO2. The
surface albedo determines the amount of downward short-wave heat flux that is
reflected at the surface. The surface albedo for fractional covered vegetated
surface types (unweighted) is described by the snow-free and cold deep snow
albedos. The soil albedo is defined according to colour and moisture content.
LAI is also used in determining the surface albedo for surfaces covered by
vegetation.
The hydrological cycle in MOSES2.2 is similar to MOSES1, with small changes
for the interactions with vegetation. However, it continues to treat each
tile separately, so extraction of water from the soil is calculated for each
tile and then weighted summed to give the grid-box average. Precipitation is
partitioned into interception (via the canopy), throughfall, runoff, and
infiltration into the ground. Different parameters apply to each vegetation
type. Canopy water refers to the precipitation intercepted by plant leaves
available for free evaporation. MOSES2 uses the same four soil layers as
MOSES1, with thicknesses from the surface downwards set to 0.1, 0.25, 0.65,
and 2 m. The moisture content of the upper soil layer (0.1 m) is
increased via snowmelt and throughfall and decreased according to evaporation
from the soil layer, flow of water into lower layers, and draw-up of water
via plant roots. The extraction of water from any particular soil layer is
proportional to the water lost by evapotranspiration, reflecting the vertical
distribution of roots. The five PFTs have different root depths, such that
trees are able to access moisture from soil layers at deeper depths compared
with grasses and shrubs. The soil moisture content and soil water phase
changes and the associated latent heat describe the thermal characteristics
of soil that determine, via a discretised form of the heat diffusion
equation, the subsurface temperatures. Subsurface soil temperatures are
determined by the diffusive heat fluxes into and out of a soil layer and the
heat flux advected from the layer by the moisture flux.
MOSES2 requires similar soil parameter inputs to MOSES1, although it
additionally requires bare soil albedo and soil carbon content of the soils.
However, the vegetation properties are very different. MOSES1 required inputs
of grid-box average LAI, root depth, etc., whereas MOSES2 requires prescribed
inputs of the fractional types of each surface type, the LAI and canopy
height of each vegetated PFT, and the overall canopy conductance. It also
includes a disturbance fraction that represents agriculture. When using
dynamic vegetation (TRIFFID), these fields (except for disturbance) are only
used for initialisation and the model will dynamically update them.
Differences between MOSES2.2 and MOSES2.1
There are a number of key differences between MOSES2.1 and MOSES2.2, and a
number of smaller modifications between the versions. These major changes
include the following.
MOSES2.2 uses a spectral albedo scheme to calculate separately
the diffuse and direct-beam surface albedos. This scheme is not
used in MOSES2.1, although modifications can be added to
include it.
MOSES2.2 uses a spectral snow albedo model that includes a
prognostic grain size that characterises the ageing of snow and
its impact on snow albedo. This is not present in MOSES2.1.
MOSES2.2 also introduces a new calculation of evapotranspiration
from soil moisture stores, as well as a different
parameterisation of bare soil evaporation.
Supersaturation in the soil layer is treated differently in the
two versions of MOSES2. In MOSES2.2, supersaturation results in
an increase in surface runoff. In contrast, supersaturation in
MOSES1 and MOSES2.1 is managed via an increase in downward flow
into the deeper soil layers and so is removed via subsurface
runoff.
Tests carried out in which MOSES2.1 is gradually changed to MOSES2.2 show
that the first two changes affect surface temperature, whereas the third
difference substantially alters soil moisture. Supersaturation changes impact
the partitioning of runoff between surface and sub-surface and also influence
the soil moisture, and to a lesser extent the evapotranspiration changes.
There are also a number of smaller changes (such as using an implicit soil
moisture scheme in MOSES2.2 compared to an explicit scheme in MOSES2.1 and
MOSES1), but these do not result in a major change to the climate. MOSES2.2
also had some major restructuring of the Fortran code.
Additionally, in the default version of MOSES2.1 (used until recently), the
rate of respiration increases almost exponentially with temperature (Julia C. Tindall,
personal communication, 2015). As a result, in some conditions such as during the Amazon
dry season, respiration excessively increases, and this decreases soil
moisture, which consequently inhibits tree growth. In MOSES2.2, the impact of
temperature on respiration rate declines at high temperatures. This revised
respiration rate reduces drying and dieback of trees. This has a relatively
limited impact on the simulation of vegetation for the pre-industrial (the
broadleaf tree fraction is slightly increased in the Amazon), but does have a
bigger effect on very warm climates such as the early Eocene. This has now
become the default for the Bristol variant and will be referred to as
HadCM3B-M2.1a.
TRIFFID
MOSES2.1 and MOSES2.2 both have the capacity to be run in coupled mode with a
dynamic vegetation and terrestrial carbon cycle scheme, TRIFFID (Top-down
Representation of Interactive Foliage and Flora Including Dynamics)
. TRIFFID predicts the distribution and properties of
global vegetation based on plant functional types using a competitive,
hierarchical formulation. The performance and sensitivity of TRIFFID have
been compared with a variety of other dynamic vegetation models
and an updated version of TRIFFID is used in both the
latest Coupled Model Intercomparison Project (CMIP5) model HadGEM2-ES
and in JULES .
In the model configurations presented here, TRIFFID is normally only used
with MOSES2.1 because of a dry bias in MOSES2.2 which is manifested by an
overly dry surface climate over the Eurasian continent in summer. This
results in loss of vegetation if used with dynamic vegetation. The cause of
this drying is unclear, but is partially linked to the changes in evaporation
and evapotranspiration parameterisations discussed above.
TRIFFID updates the five vegetation PFTs and the bare soil fraction, all of
which can change dynamically. TRIFFID can be run in two different modes.
Equilibrium mode, where TRIFFID runs for 50 years of TRIFFID for
each 5 years of the climate model run. The fluxes between the land and
the atmosphere are calculated and averaged over 5 years. This is
particularly valuable for quick spin-up of the vegetation and soil
carbon.
Dynamic mode, where TRIFFID is run every 10 days. Fluxes are averaged
over 10 days; as such high-frequency variability is accounted for. This
mode is the standard for full runs of the coupled model.
MOSES2 passes the averaged fluxes of carbon to TRIFFID, which calculates the
growth and expansion of the existing vegetation, and updates the land surface
parameters based on the new vegetation distribution and structure. TRIFFID
calculates areal coverage, leaf area index (LAI), and canopy height for five
defined plant functional types (PFTs): broadleaf tree, needleleaf tree, C3
grass, C4 grass, and shrub. These PFTs respond differently to climate and
CO2 forcing (e.g. C3 and C4 grasses use different photosynthetic
pathways) and also impact differently on the physical properties of the land
surface, i.e. possessing different aerodynamic roughness lengths and albedo
properties. Broadleaf and needleleaf trees and C3 and C4 grasses react
independently within the model due to their unique parameter sets. C4 plants
use water more efficiently than C3 plants, requiring less water to produce
the same amount of biomass. Overall, C4 plants have the highest critical
humidity deficit and temperature range, meaning that in high-temperature,
low-moisture environments they will do better than other PFTs, even though
the competition model would normally favour trees.
All PFTs can co-exist within the same grid box, each possessing a fractional
coverage that is equivalent to the population size. The fractional coverage
co-existence approach allows smooth transitions to occur when the vegetation
distribution changes rather than the sudden discontinuities that would occur
in a “dominant” PFT-only approach . However, the
Lotka–Volterra equations used in TRIFFID mean that each grid cell in the
model tends to converge on one dominant plant functional type
. Competition is essentially based on a height hierarchy of
trees > shrubs > grasses. Each terrestrial grid square has a small
minimum content of each plant functional type, regardless of location and
competition, as a “seeding” fraction . This ensures that no
PFT can become extinct and can regenerate when conditions become appropriate.
TRIFFID can specify areas of agricultural crops as C3 and C4 grasses, without
competing land types .
The terrestrial net primary productivity (NPP) is calculated by a coupled
photosynthesis–stomatal conductance model . Factors affecting
the rate of photosynthesis are the humidity deficit, the photochemically
active radiation, soil moisture, and LAI. The maximum rate of photosynthesis
is directly related to the leaf temperature and the upper and lower
temperatures for photosynthesis (defined individually for each PFT). Carbon
is stored in the vegetation and soil stores.
The predicted vegetation in each grid box feeds back into the climate system
in a number of ways, principally through evapotranspiration from the canopy,
alteration of surface albedo, and alteration of mixing at the boundary layer
between the surface and the atmosphere (due to changes in roughness length).
Variants with differing resolutionsHadCM3BL
HadCM3BL comprises the same model components as HadCM3B, but with a
lower-resolution ocean which matches the standard atmosphere resolution of
96 × 73 grid points (3.75∘× 2.5∘)
. Note that the Bristol version, HadCM3BL, is very different
from the Met Office version. The Met Office version was mainly used for the
early carbon cycle work , but required significant flux
corrections to ensure that the Atlantic surface climate was reasonable. Our
version does not require flux correction because of changes in bathymetry
described below. It can be run with all versions of MOSES, with or without
TRIFFID, in the same manner as HadCM3B. We tend to use HadCM3BL when long
simulations are required. For instance, when the land–sea mask and/or
bathymetry are substantially changed from those of modern ones, it can take many thousands of
years of integration to get the deep ocean into equilibrium. As such,
HadCM3BL has been used extensively for our pre-Quaternary climate modelling
work e.g..
The implementation of the atmosphere and land surface schemes is identical to
HadCM3B. There are some differences in the ocean due to its lower resolution,
some of which are substantive differences required either to maintain
stability or to reproduce the present-day climate without the need for the
flux corrections used in earlier versions of the model, some of which are
simple scalings of parameters to give the same scientific behaviour as
HadCM3B at the lower resolution. These differences between HadCM3B and
HadCM3BL, which are described below, are generally consistent with work done
to optimise the FAMOUS model , which has the same ocean
resolution as HadCM3BL.
North Atlantic bathymetry: “No Iceland”
As described in Sect. , care was taken when developing HadCM3
to define the bathymetry of the North Atlantic in order to ensure that the
appropriate flow through the Denmark Strait was captured. This flow is lost
when the ocean resolution is reduced in HadCM3BL as the channel between
Iceland and Greenland becomes less than a single grid cell wide (on the
velocity grid) and thus no flow is permitted. investigated
potential modifications to allow increased heat transport through this
region, thus alleviating the unrealistic build-up of sea ice in the Nordic
Sea, and concluded that the removal of Iceland was the preferred solution.
With this modification, the improved meridional overturning circulation leads
to more realistic heat transports in the coupled system and alleviates the
need for flux correction.
This change also has a knock-on effect on the land surface (and ultimately
the atmosphere) in that the two cells that define Iceland have been removed.
Ocean vertical diffusion
In HadCM3BL, the Richardson number dependence of the vertical tracer
diffusivity is replaced with a constant background rate, as it is in FAMOUS.
describes problems encountered with FAMOUS in the
interaction between the mixed layer and deep vertical diffusion schemes, but
this was found to have little impact on the solution because of the
relatively low resolution.
For the calculation of vertical diffusion, HadCM3BL uses a different
calculation for the density of seawater from HadCM3B. HadCM3B calculates all
densities relative to a reference level at the surface using the updated
equation of state for seawater of . This can result in
negative density gradients in the deep ocean and hence a negative Richardson
number, which in turn can produce very high diffusivities at depth which
was never intended to handle
. HadCM3BL instead derives
third-order polynomials for each 250 m depth span of the ocean
to fit the Knudsen–Ekman equation for the density of
seawater and does not produce negative density gradients ,
but the range of salinities covered may be insufficient for some
applications. This choice of diffusion scheme is consistent with that used in
FAMOUS.
Ocean isopycnal diffusion
HadCM3BL uses different coefficients for a number of aspects of the diffusion
formulation, as described in Table . All of these values
are consistent with those used in FAMOUS. In addition, the
scheme for the calculation of isopycnal thickness
diffusion coefficients, introduced in HadCM3B to improve resolution of
currents such as western boundary currents on the 1.25∘ grid, is not
used in HadCM3BL. Instead, fixed values of the coefficients for surface ocean
diffusion, deep ocean diffusion, and scale depth are specified, as in FAMOUS.
Solar penetrative radiation
In HadCM3 the penetration of solar radiation is represented by a double
exponential decay with depth, with coefficients determined from observations.
The ratio between the shallower decay and deeper decay exponential is
controlled by a parameter called RSOL. This is set to 0.0 in HadCM3B and is
set to 3.8 × 10-1 in HadCM3BL, as it is in FAMOUS.
Islands
HadCM3B defines six islands in the barotropic solution, around which non-zero
depth-integrated flow is permitted: Antarctica, Australia, New Zealand, the
Caribbean, Madagascar and Iceland. In HadCM3BL, there is no island for
Iceland as this is entirely absent from HadCM3BL and Madagascar is also not
defined as an island due to its proximity to Africa.
HadAM3BH
HadAM3BH is a higher-resolution version of the atmosphere-only variant,
HadAM3B. This model is different to that used by the Met Office
e.g., which keeps to 19 levels in
the vertical but has some changes to the parameterisations, particularly in
the boundary layer. It is used for studies in which the atmospheric
circulation is critical, and as such is best represented at high resolution.
Its horizontal resolution is 3 times greater than HadAM3B both latitudinally
and longitudinally, i.e. 288×217 grid points
(1.25∘× 0.83∘). The number of vertical levels is
increased from 19 to 30, with the extra levels being concentrated close to
the Earth's surface and the upper levels remaining similar to HadAM3B. The
higher spatial resolution requires a smaller time step of 10 min. It may be
used with either the MOSES1 or MOSES2.1 land surface schemes, and can be used
with TRIFFID, though this has rarely been done. The time stepping algorithm
is slightly different, in that the dynamics can be updated multiple times
between the full physics time steps. In HadAM3BH, we use two dynamic per
physics time steps to allow for improved numerical stability of the model.
Various diffusion coefficients, critical relative humidity, and parameters
for the gravity wave drag scheme have been re-tuned to account for the change
in resolution, as documented in Table . Otherwise the model
has identical physics to HadAM3B and has had no further changes.
HadRM3B
HadRM3B is the regional climate model (RCM) version of HadAM3B which has been
used when representation of high-resolution atmospheric processes is
important, such as around orography or studying extreme events. It can be
configured for any domain size and location has commonly been used for
studies over Europe , the Arctic and
Svalbard , and the East Asian Monsoon region
. It has also been used to model deep time
.
The BRIDGE version is based on the same fundamental physics and model
structure as the Met Office HadAM3, and currently is only available with the
MOSES1 land surface scheme. We again do not make any substantial changes to
the physical parameterisations, so the model is largely identical to HadAM3
except for parameters sensitive to resolution.
Regional climate models require either fixed or time evolving data on the
large scale and global atmospheric and ocean response to climate forcings to
be provided to them at their lateral (atmospheric) and sea surface
boundaries, such as potential temperature and specific humidity. The common
experiment set-up, used here, is a one-way nested approach, where no
information is fed back into the GCM simulation, but the large-scale
atmospheric circulation patterns, such as the location of the jet streams,
are fed in through the lateral boundary conditions (LBCs). For a RCM to have
a “parent” GCM is rare, offering a unique opportunity to investigate the
effects of dynamical downscaling without modification (or contradiction) of
the physics between the driving GCM and the RCM at the lateral boundaries.
LBCs are updated every 6 h and linearly interpolated for time steps in
between. A four-grid smoothing is applied to global model data entering the
regional model domain. Therefore, typically, HadRM3B has been run here using
HadAM3B or HadCM3B to produce the LBCs, sea surface temperature, and sea ice
concentration data, although there have been experiments using SSTs from
HadISST and HadGEM as well as other models in the CMIP5 experiment to analyse
the sensitivity of the model to its boundary conditions.
HadRM3B is run on a standard lat–lon grid with the pole rotated so that the
centre of the domain of interest lies across the Equator within the RCM's
grid of reference (see Fig. ) to reduce variation in the
areas of the grid cells. The time step of the model is 5 min to maintain
numerical stability with the increase in spatial resolution which is commonly
0.44∘× 0.44∘
(∼ 50 km × 50 km) but has also been run at
0.22∘× 0.22∘
(∼ 25 km × 25 km). Lateral boundary conditions are
typically provided every 6 h and linearly interpolated to each time step.
The main difference between HadRM3B and HadCM3B/HadAM3B in terms of
atmospheric dynamics is in the sub-grid-scale diffusion applied to the
horizontal wind component to prevent the accumulation of energy at the
smallest scales and noise (see Table ). In addition, the
parameters which control the proportion of a grid box over which convective
precipitation and large-scale precipitation are assumed to fall, as well as
diffusion parameters, vary compared to HadAM3B (see Table ,
variables conv_eps and ls_eps).
Land–sea mask and orography (sea coloured grey, land height in
metres) for four configurations of
HadRM3B. (a) shows the standard
European domain at 0.44∘ resolution, (b) shows the
equivalent domain for East Asia, (c) shows a configuration for the
Arctic
and Svalbard at 0.22∘as used in,
and (d) a North America/European configuration for the early
Cretaceous at 0.44∘ resolution as used
in.
Simulations using the regional climate models have enabled improved spatial
representation of temperature and precipitation patterns and response to
climate forcings, particularly around mountains and coastlines. The increase in
resolution also improves the simulated temporal variability, including
simulation of extremes .
Comparison with data
The aim of this section is to qualitatively and quantitatively evaluate the
suite of HadCM3@Bristol models in terms of their ability to recreate key
aspects of the climate system relative to observations, and other models
within the CMIP5 family. In the following subsections, a selection of
observational datasets is compared to multiple modelled climatic variables.
Details on the datasets used for each variable are briefly outlined in each
subsection. This is not intended to be a complete model evaluation; however,
it will highlight that some variants do a more realistic job than others at
representing various environmental processes. Where appropriate, stronger or
weaker models will be highlighted, and some other CMIP5 models will be shown
for comparison. Because much of our work at Bristol involves carrying out
palaeoclimate or idealised simulations, our standard control simulations are
static pre-industrial simulations, similar to the CMIP5 DECK pre-industrial
simulation . However, most observational datasets are from
the instrumental record, typically the last few decades. This is to be
considered when interpreting our evaluation, although it is likely that
differences between the pre-industrial and the instrumental period are
generally small relative to the model biases.
A quantitative evaluation (global root-mean square difference (RMSE)
analysis) of the four base state BRIDGE models, namely HadCM3B with the
MOSES1 land surface scheme, HadCM3B with MOSES2.1a, HadCM3BL with MOSES2.1a,
and FAMOUS with MOSES1, is performed against reanalysis and/or observational
data and shown alongside new and predecessor models from the CMIP5 database
(Fig. ; BRIDGE models highlighted in red). Here we
make use of the ESMValTool(v1.0), a community diagnostic and performance tool
to assess and compare the magnitude of known systematic
biases inherent in all climate models. Better understanding of these biases
is instrumental in diagnosing their origin and a model's ability to reproduce
observed spatial and temporal variability and trends in various atmospheric
(e.g. large-scale circulation) and oceanic phenomena (e.g. ENSO). CMIP5 model
data are provided from http://www.ceda.ac.uk, while observational
obs4MIPs; and re-analysis
ana4MIPs; data are provided from
https://www.earthsystemcog.org, all conforming to the CMIP5 format.
Here the BRIDGE models have also been standardised to the CMIP5 format.
Further, models and observations are re-gridded to the coarsest resolution
within the ESMValTool framework for evaluation. Table
details the different metrics used for the evaluation of the historical model
simulations in Fig. . The BRIDGE models are only
pre-industrial climatologies (30 years) without any year-on-year historical
forcing; however, this is not expected to be detrimental for the evaluation.
The results in Fig. demonstrate that the BRIDGE suite
of models, with the exception of FAMOUS-M1H, accurately reproduces observed
global spatio-temporal patterns. Indeed, HadCM3B-M1, HadCM3B-M2.1a, and in
most respects HadCM3BL-M2.1a, outperform many of the higher-fidelity CMIP5
models with lower RMSE when compared to the observations, particularly with
respect to global air temperature (at 850 and 200 hPa), U-wind (at 850 and
200 hPa), and 1.5 m surface temperature. It is likely that the course
resolution of FAMOUS has a detrimental impact on its performance. The
following sections provide a more detailed evaluation of various atmosphere,
ocean, and land surface variables in the BRIDGE model suite.
Observational and reanalysis datasets used for the evaluation in Fig. .
Relative error measure of the CMIP5 models (21 in total; in black)
and the BRIDGE models (4 in total; in red) performance. Error measure is
calculated from a time–space root-mean square error (RMSE) of
contemporary and predecessor CMIP5 model historical climatological
(1980–2005) seasonal cycle simulations and BRIDGE pre-industrial
seasonal cycle climatologies against observations (1980–2005) for a set
of nine different atmospheric variables. Error for each individual
variable is characterised as a relative error by normalising the result of
the median error of all model results ; the BRIDGE
models are not included in the mean/median error. For instance, a
value of 0.20 indicates that a model's RMSE is 20 %
larger than the median CMIP5 error for that variable, whereas a value of
-0.20 means the error is 20 % smaller than the median
error. The diagonal split grid square shows the relative error for the
reference observed/reanalysis dataset (lower right triangle) and
the alternative dataset (top left triangle). White triangles/boxes
indicate where no data were available. Evaluated global atmospheric variables
are TOA outgoing all-sky short-wave radiation (rsut_Glob), TOA outgoing
all-sky outgoing long-wave radiation (rlut_Glob), precipitation
(pr_Glob), near-surface temperature (tas_Glob), specific humidity at
400 hPa (hus_Glob-400), geopotential at 500 hPa
(zg_Glob-500), V-wind at 200 hPa (va_Glob-200), V-wind at
850 hPa (va_Glob-850), U-wind at 200 hPa
(ua_Glob-200), U-wind at 850 hPa (ua_Glob-850), temperature
at 200 hPa (ta_Glob-200), and temperature at 850 hPa
(ta_Glob-850).
AtmosphereSurface temperature patterns
We compare the modelled temperature and precipitation to observational data
provided by the University of East Anglia high-resolution climatology for
1960–1990 (CRU CL v2.0) . This record is based on a range of
weather stations totalling more than 10 000 stations for temperature and
more than 25 000 stations for precipitation, with the best spatial coverage
over North America, Europe, and India and the sparsest spatial coverage over
the interiors of South America and Africa and Antarctica. Modelled SAT fields
were masked to model land points only and differences to observations were
done at the same resolution as the relevant model, as shown in
Fig. .
(a) The difference between the annual mean surface air
temperature (in ∘C) of HadCM3B-M1
and the CRU CL v2.0 for the period 1960–1990 regridded onto the
HadCM3B-M1 grid, (b) As (a) for the HadCM3B-M2.1a version, (c) As
(a) but for HadCM3B-M2.2, (d) As (a) but for the HadCM3BL-M2.1aN
version, (e) As (a) but for HadAM3B-M2.1a, (f) As (a) but for
HadAM3BH-M2.1a, (g) As (a) but for FAMOUS-M1, and (h) As (a) but
for FAMOUS-M2.2. Panels (i, j, k, l) show comparable results
for four CMIP5 models, ACCESS1-0, CCSM4, GISS-E2-H and
IPSL-CM5A-LR respectively. These were chosen to represent two
models which were above the CMIP5 average in terms of their
RMSE with respect to surface air temperature, and two models
which were below average. All differences are calculated by
regridding the CRU data onto the corresponding model grid, using simple
bi-linear interpolation.
It should be noted that the comparison between the versions of the HadCM3B
family and the observed CRU CL v2.0 data is not a “clean” comparison. The
observed data are for 1960–1990, whereas all model simulations are for the
pre-industrial period. In the case of HadAMB3 simulations, the SSTs used are
the 1870–1900 means of HadISST. To evaluate the impact of this effect, we
examined the CMIP5 historical experiment of HadCM3-M1 done at the Hadley
Centre . The differences between the 1960–1990
climate means compared to the 1860–1890 climate means were generally small
compared to the model biases, with the overall mean warming between the two
periods being 0.6 ∘C. Similarly, the four CMIP5 simulations are
averages from 1860 to 1890 of the historical runs (using one ensemble member
only, r1i1p1), and so the comparisons to the HadCM3B family are not perfectly
clean.
HadCM3B-M1 (Fig. a) generally has a small cold bias compared
to the data, with most regions experiencing colder temperatures by 2 to
3 ∘C. The area-weighted RMSE is 2.8 ∘C, but with smaller
errors in the tropics and a small warm bias in South America. There is also a
small warm bias over Greenland, but this should be treated with some caution
since there are issues about elevation effects and the data
are relatively sparse in this region. The results for
Fig. a are largely identical to those calculated using the
CMIP5 HadCM3-M1 archived data (run by the UK Hadley Centre) for the
historical run averaged between 1860 and 1899 inclusive (not shown). The
differences are mostly less than 0.5 ∘C and never exceed
1 ∘C, with a RMSE of 0.5 ∘C. Differences between the
1860–1889 average and the 1960–1989 average for the CMIP5 historical run
are small, verifying that the model biases greatly exceed any differences
between pre-industrial and modern temperatures. However, the small warming
that does occur between 1860–1889 and 1960–1989 does reduce the cold bias
marginally (RMSE decreased by 0.1 ∘C).
Using MOSES2, HadCM3B-M2.1a (Fig. b) shows a significant
reduction in the cold bias, resulting in a RMSE of 2.1 ∘C. The cold
bias has reduced but still remains over northern Russia and Scandinavia,
while over South America (Amazon) and Greenland the warm anomalies have
intensified. Over the Amazon this is likely due to the difficulties in the
vegetation model (see Sect. ), while difficulties with
Greenland were mentioned above. Elsewhere, the general cool bias seen in
Fig. a has gone, replaced by anomalies of ±2 to
5 ∘C, with few widespread regional anomalies. Similarly,
HadCM3B-M2.2N (Fig. c) also shows a reduced cold bias, with
a RMSE of 2.1 ∘C. This model variant shows a slight reduction in the
warm anomaly observed over the Amazon compared to Fig. b,
but has a more extensive warm bias of 1 to 2 ∘C at higher northern
latitudes, e.g. over North America.
HadCM3BL-M2.1a (Fig. d) has a RMSE of 2.6 ∘C and a
comparable cold bias to HadCM3B-M1. As with the HadCM3B model variants, using
MOSES2 with HadCM3BL reduces the cold bias and RMSE compared to using MOSES1,
with HadCM3BL-M1 having a much higher RMSE (not shown). Once again, the high
northern latitudes (particularly over Russia and Scandinavia) are too cold,
which is the result of an exaggerated seasonal cycle due to an overly cold
winter. This is also the case for other HadCM3B model variants, but it is
most pronounced for the HadCM3BL variants. Similarly to the other simulations
using MOSES 2, the Amazon remains slightly warmer than the observations with
slightly reduced broadleaf forest cover (see Sect. ).
The atmosphere-only models vary significantly depending on their resolution.
At standard resolution, HadAM3B-M2.1a (Fig. e) shows similar
spatial anomalies and RMSE to Fig. a–d, but greater warm
biases over North America and Greenland of up to 5 ∘C and cool
biases over Africa and southern Asia of 2 to 5 ∘C. However, it has
the smallest anomaly over the Amazon compared to the other standard
resolution model variants, and a comparable RMSE (2.3 ∘C).
HadAM3BH-M2.1a (Fig. f) on the other hand shows a markedly
different spatial pattern in its temperature biases to the model versions
already described. It is the only simulation not to show a global cold bias.
This is due to warmer than observed temperatures of 2 to 5 ∘C over
the majority of land surfaces north of 30∘ N (with the exception of
the southern tip of Greenland and mountainous regions). It has a slight cold
bias of 1 to 2 ∘C over areas south of 30∘ N (with the
exception of some regions in South America). Although these biases are
extensive spatially, they are not of greater magnitude than the regional
biases found in other model variants or CMIP5 models and the RMSE of the
HadAM3BH-M2.1a simulation is 2.2 ∘C.
The FAMOUS model variants (Fig. g and h) have larger RMSE
values than the higher-resolution model variants and the other CMIP5 models.
FAMOUS-M2.2 (Fig. h) is the worse of the two, with a RMSE of
4.1 ∘C and extreme cold biases over Northern Hemisphere continents,
which exceed 10 ∘C around Scandinavia. The cold bias in
Fig. g is less extreme, but instead has a warm bias in South
America of up to 5 to 10 ∘C and up to 2 to 5 ∘C over India
and Australia. There is some improvement in the RMSE for FAMOUS-M1, but it is
still much higher (3.3 ∘C) than the higher-resolution model
variants.
Some of the differences between the mean annual temperature biases in the
models are due to changes in the models themselves. For instance, the
improvements generally seen between models with MOSES1 and MOSES2.1 are
primarily due to the better representation of the land surface, particularly
the snow cover, as discussed above. It is also notable that the
lower-resolution ocean models tend to be cooler at the higher latitudes, as
the lower-resolution ocean makes it more difficult to move heat away from the
Equator.
For comparison, we show the SAT fields from four CMIP5 models
(Fig. i–l), selected based on the results of the IPCC AR5
WG1 model evaluation (Sect. 9). We selected two models which were above
average for their simulation of SAT (ACCESS1-0 and CCSM4) and two models
which were below average (GISS-E2-H and IPSL-CM5A-LR). In all cases these
models are not the best or worst extremes, but represent the typical range of
model skill. Again, the observations have been interpolated onto the
appropriate resolution of the model from which the RMSE was calculated. As
can be seen, the general picture that emerges is that most of the varieties
of HadCM3B (except perhaps for FAMOUS) are well within the skill of the CMIP5
ensemble. The CMIP5 models all show large regional biases of up to
±5 ∘C (with little consistency on the sign of the anomaly
between them) and the RMSE scores range from 2.3 to 3.3 ∘C, which
are similar to the varieties of the HadCM3B model. Indeed HadCM3B-M2.1aN and
HadCM3B-M2.2 have the smallest RMSE values of the models sampled.
Precipitation
As for Fig. but showing the difference in mean
annual precipitation, expressed as a % difference to the CRU CL v2.0
observations.
Figure shows that the BRIDGE models with the
exception of FAMOUS produce annual precipitation amounts comparable to other
CMIP5 models, suggesting that our models are capturing the general
synoptic-scale features (frontal, convective and mesoscale).
While global annual RMSEs for the BRIDGE models compare favourably, it is
also key to investigate the mean spatial patterns of precipitation to
ascertain whether the models are reproducing these patterns in accordance
with the observations. We assess annual climatological precipitation for the
BRIDGE model suite against CRU CL v2.0 , a high-resolution
(0.5∘× 0.5∘) global land surface product
(excluding Antarctica). The resolution is transformed (bi-linear
interpolation) to the appropriate grid in the model. We are again comparing
our pre-industrial simulations with 1960–1990 observations, but the model
biases are generally much larger than any trends.
Figure shows the regional biases in mean annual
precipitation, expressed as a % error compared to the CRU CL data. For
consistency with the previous figure, we also include the same four CMIP5
models. Regionally, spatial patterns in precipitation bias are generally
consistent between the different BRIDGE models and broadly comparable to
their CMIP5 models.
The BRIDGE simulations have a similar problem to many CMIP5 models in that
they overestimate precipitation in regions of topography. This is
particularly noticeable around the Himalayas and Tibet, but is also visible
on the upstream side of the Rockies and Andes. This may be due to poor
representation of moisture gradients and regional dynamics. However, the
apparent discrepancy with observations can be amplified by a known negative
bias in gauge stations . Gauge stations also underestimate
precipitation leeward of mountain ranges (e.g. Himalayas and Andes), as well
as over arid regions, which can contribute to model–data discrepancy in
these regions also.
The annual mean northward heat transport in total, the atmosphere,
and the ocean. The black line shows the observational estimate, blue
HadCM3B-M2.1a, green HadCM3BL-M2.1a, and red FAMOUS-M2.2; grey lines
show transports for a selection of CMIP5 models. These transports
are calculated as the implied heat transports from the TOA and
surface energy fluxes (see , for details).
Observational estimates for the total transport are derived from the CERES
data.
Monsoonal regions of South-east Asia, Australasia, southern South America,
and western and central Africa overestimate precipitation by 0.5 to
2 mm day-1 but underestimate precipitation in the Indian and northern
arm of the South American region by ∼ 1 to 4 mm day-1. There are
however some exceptions with HadAM3B-M2.1a (Fig. ), an
atmosphere-only GCM using observed SST (HadISST 1870–1900) producing a more
reasonable precipitation signal compared to the observations, suggesting the
importance in the accuracy of SST/local ocean circulation dependency (this is
also seen in Australia). There is still a problem with the ITCZ location over
South America being too far south, giving this north–south dipole in
negative/positive anomalies.
It is also noted that an increase in resolution does not produce a noticeable
improvement in spatial annual precipitation bias in certain monsoon regions
in the BRIDGE models (Fig. f compared to
Fig. j), again suggesting the importance of accuracy in SST
and ocean circulation, with the exception of South America, where there is
improvement. Spatially, increased resolution does not affect the sign of
anomaly or the spatial patterns of precipitation regionally (with the
exception of South America) throughout the BRIDGE suite of models; however,
the magnitude of the precipitation bias does progressively decrease.
Horizontal heat transports
There is broad agreement between the observed and simulated total northward
heat transport. Similarly, the partitioning between the ocean and the
atmosphere is qualitatively similar to that estimated by .
We find that all versions of the model simulate heat transport and are
consistent with CMIP5 models (see the grey lines in
Fig. ). However, in common with almost all other
climate models, we find that on the Equator, although the total heat
transport is northward, in agreement with the observations, the atmospheric
heat transport is also northward, contrary to the observed southward
transport . The cause of this in any of the models in which
it is a feature is unclear. The three versions of HadCM3B show remarkably
similar amounts of total heat transport; the major difference is FAMOUS,
which underestimates the southward heat transport in the Southern Hemisphere
subtropics rather more than HadCM3B and HadCM3BL. This is due to the smaller
amount of ocean heat transport in this region in FAMOUS. This discrepancy is
not due to the coarse resolution of the FAMOUS ocean because, interestingly,
in this region the ocean heat transport in HadCM3BL is very similar to
HadCM3B, whose ocean resolution is quite different. Therefore it is more
likely that the difference arises from the atmospheric forcing of the surface
ocean. In the Northern Hemisphere the HadCM3BL ocean heat transport is more
similar to FAMOUS, suggesting that the ocean resolution is more important
here. This is likely due to the processes that determine the ocean's
overturning circulation being simulated rather differently in the higher- and
lower-resolution models.
OceanSea surface temperature
The BRIDGE suite of models is capable of reproducing the broad global
latitudinal patterns and gradients in SST (Fig. ).
Nonetheless, some cold and warm biases of over 8 ∘C are present,
especially where sharp fronts and boundary currents are not resolved. Other
biases of similar magnitude also appear in the upwelling regions (e.g. west
of Africa and of South America), and again these are likely associated with
processes that are not fully resolved by the model. Colder SSTs in the
sub-polar North Atlantic for all models are not uncommon and likely due to
the coarse resolution e.g.. This can be seen by
comparing HadCM3B and HadCM3BL, which are models that differ most in their
ocean resolution. Cold biases in the Northern Hemisphere are more extensive
in HadCM3BL than in HadCM3B. Warmer SSTs of up to 8 ∘C are present
in the Southern Hemisphere, especially in the Southern Ocean, in both HadCM3B
and HadCM3BL. FAMOUS is characterised by colder than observed SSTs in the
Northern Hemisphere, in common with HadCM3B and HadCM3BL, and warmer SSTs by
up to 8 ∘C almost everywhere in the Southern Hemisphere despite the
bias in ocean heat transport. HadCM3B and HadCM3BL do not show any notably
larger biases when compared to typical CMIP5 models. All of the HadCM3B
models, including FAMOUS, show smaller temperature biases in the Southern
Ocean than GISS-E2-H, and the biases in the North Pacific are of a similar
magnitude to those in IPSL-CM5A-LR.
Annual mean sea surface temperature differences (in ∘C) for
a range of coupled model simulations, and also for the same four
CMIP5 models used in Figs. and . The
observational dataset is the Levitus World Ocean Atlas (2009)
. The figure shows the difference in SST between model
and observations for (a) HadCM3B-M1, (b) HadCM3B-M2.1a, (c) HadCM3B-M2.2,
(d) HadCM3BL-M2.1a, (e) FAMOUS-M1, (f) FAMOUS-M2.2, (g) ACCESS1.0, (h) CCSM4,
(i) GISS-E2-H, and (j) IPSL-CM5A-LR. Model output is regridded to the
same resolution of the observations.
As Fig. but showing the differences in sea
surface salinity (in gkg-1) between models and observations.
(a) HadCM3B-M1, (b) HadCM3B-M2.1a, (c) HadCM3B-M2.2, (d) HadCM3BL-M2.1a,
(e) FAMOUS-M1, (f) FAMOUS-M2.2, (g) GISS-E2-H, and (h) IPSL-CM5A-LR. The
observational dataset is the Levitus World Ocean Atlas (2009)
. Model output is regridded to the same resolution of
the observations.
Sea surface salinity
The broad global latitudinal patterns of sea surface salinity are
realistically reproduced by the suite of BRIDGE simulations
(Fig. ). However, on the global average, the models
show a fresh bias of about 0.5 g kg-1; as we shall show in the
following section, this is likely related to the rather different vertical
structure of the ocean in the model than in the observations. In all models,
substantial differences from the observations are found in the Arctic Ocean,
exhibiting higher salinities (up to 10 g kg-1) in the Kara Sea and
generally north of Russia. Generally lower salinities (of up to
5 gkg-1) are found in the Chukchi and Beaufort seas. The largest
differences are found in enclosed or semi-enclosed basins, such as the
Mediterranean Sea, where it is more saline, or the Black Sea, Caspian Sea,
and Hudson Bay, where it is markedly fresher. In all versions of the model
the subtropical North Atlantic tends to be more saline than the observations.
Substantial differences from the observations can also be found in CMIP5
models (Fig. g–h), with magnitudes comparable to the
BRIDGE models. We note that some of the differences at high latitudes could
be due to biases in the simulation of sea ice concentration and distribution.
The Atlantic Meridional Overturning Circulation
Annual depth profiles of the Atlantic Meridional Overturning
Circulation (AMOC) at 26.25∘ N showing the range of values for
variants of HadCM3B, HadCM3BL, and FAMOUS. Annual data from the RAPID array
at 26.5∘ N are highlighted in grey. The depth at
which the AMOC reaches its maximum is indicated with a point.
Figure shows the mean strength of the Atlantic
Meridional Overturning Circulation (AMOC) for the three main model families
(HadCM3B, HadCM3BL, and FAMOUS). Values are shown as zonally integrated depth
profiles measured in terms of the northward flow of water at
26.25∘ N. The modelled AMOC is compared to observations from the
Rapid Climate Change-Meridional Overturning Circulation and Heatflux Array
(RAPID-MOCHA) at 26.5∘ N , which have been
calculated from daily data spanning 2 April 2004 to 30 March 2015.
The strength of the AMOC varies on an annual basis, so a range of values is
shown for both the models and observations, with the depth at which the AMOC
peaks highlighted with a point. The peak flow of the North Atlantic Deep
Water (NADW) cell identified by the RAPID-MOCHA array lies at around
1000 m and varies from year to year between 14 and 19 Sv. All
three models do a reasonable job of modelling the NADW cell in terms of the
magnitude of maximum flow. However, maximum overturning is too shallow for
all model variants, peaking at approximately 800 m. HadCM3BL shows
larger year to year variability than the observations: approximately twice as
large as that in the observations. This results in years with a lower minimum
volume transport than are seen in the observations. FAMOUS model variants
tend to underestimate the year to year variation by approximately 50 %,
although this is in contrast to the study of , who showed
that FAMOUS exhibited greater short-term variability than the RAPID-MOCHA
array. HadCM3B variants have a realistic year to year variability, at least
in the upper 1500 m of the ocean.
All of the models do a poor job at representing the flow of the NADW cell
below 2000 m depth. showed that at this
latitude, approximately 60 % of the southward return flow is comprised of
upper NADW (between 1100 and 3000 m) and 40 % of the lower NADW
(between 3000 and 5000 m). The modelled stream functions show that
the return flow is shifted to shallower depths, indicating a shallower
overturning in all of the model variants.
The CMIP5 models exhibit a wide spread in the mean strength of the AMOC,
ranging from 13 to 31 Sv and peaking at latitudes between 20 and
60∘ N e.g.. It was not possible to include
the CMIP5 models in Fig. ; however, the studies of
and produced similar plots of AMOC
zonally integrated depth profiles for a range of models compared to
observations (their Figs. 1 and 3 respectively). The HadCM3B and FAMOUS
variants are shown to have very similar streamfunction profiles to GFDL
Climate Model 2.1, NCAR CCSM4 models, and the MPI models, and more accurately
simulate the maximum overturning than the NorESM1 model variants. A similar
pattern of biases is apparent in the vertical structure for these models,
i.e. a too shallow overturning cell; however, the point of maximum
overturning is shallower in the HadCM3B and FAMOUS variants.
This bias in the vertical structure has been attributed in some studies to
inaccurate transport in the Nordic Sea overflows, which in the case of
HadCM3B includes a greater than observed overflow across the Denmark Strait,
in addition to sub-grid-scale processes see.
An additional cause of the shallow overturning may be the excessive surface
salinity in the North Atlantic in all model versions, particularly around the
subtropics as shown in Fig. . The study of
investigated the freshwater budget in HadCM3B,
concluding that in the North Atlantic saline conditions are primarily a
result of excessive evaporation. Other components, such as insufficient
subtropical runoff from the western coast, may also have an influence. This
results in the Atlantic being too stratified and consequently too stable,
which may reduce the depth of overturning.
Maps of the dominant plant functional type for
observations (a) and model simulations of the pre-industrial. The
models shown are (b) HadCM3B-M2.1aD (c) HadCM3BL-M2.1aD and
(d) HadAM3B-M2.1aD. The observed dataset for comparison is
.
A further consequence of this is a net northward transport of freshwater into
the Atlantic , which may result in a monostable stability
regime of the AMOC in HadCM3B instead of a bistable regime . In contrast, have demonstrated a bistable
regime in FAMOUS. Approximately 60 % of the CMIP5 models have been shown
to exhibit monostability . However, this is contrary to
what is indicated in the palaeorecord and inferred from the measurements of
diagnostic indicators in the present-day ocean – that there is a net export
of freshwater from the Atlantic and consequently that the AMOC may be in a
bistable regime. This indicates that the AMOC may be artificially stable in
the HadCM3B and FAMOUS model variants in addition to a range of other CMIP5
models. There remains uncertainty about this hypothesis however, with
concluding that freshwater export may not be a reliable
indicator of AMOC stability.
LandVegetation distribution
These models have a simple representation of terrestrial vegetation, with
five plant functional types that each covers a large climatic range.
Comparing the dominant PFT in the model to a reconstruction of pre-industrial
vegetation , we can see the model captures the overall
correct pattern (Fig. ), with slight errors of extent
and/or exact location. Previous studies which compared
TRIFFID PFT distributions to the IGBP-DIS land cover dataset (which
represents the modern distribution of vegetation as derived from satellite
image interpretation, found much of the same patterns.
The broadleaf trees in the tropics tend to extend too far, especially in the
Southern Hemisphere, as can be seen in Fig. b–d. The
southern mid-latitudes are difficult to capture accurately, for a variety of
reasons, including the challenge of precipitation patterns in this region.
The HadCM3BL model is significantly worse than either HadCM3B or HadAM3B in
this regard. This is because of its decreased ocean resolution, which affects
the sea surface temperature and therefore the water transport to the Amazon
region.
A feature which appears in the HadCM3 and HadCM3 models is a tendency for the
Amazon broadleaf forest extent to be underestimated at the mouth of the River
Amazon, even at relatively low carbon dioxide concentrations
(Fig. b–d compared to Fig. a). At higher
carbon dioxide levels, this is a known feature of the model caused by ocean circulation resulting in insufficient
precipitation to sustain the forest. The tendency of the coupled models to
underestimate precipitation in this area is apparent in
Fig. and is particularly notable in HadCM3. This leads to
TRIFFID modelling the presence of C4 grasses instead of broadleaf trees.
Grasses tend to be globally slightly underestimated with the position of
vegetation in the Sahara and other arid regions well reproduced, but the
density is modelled to be too sparse, particularly in south-western Africa,
central and south-western Asia, south-western North America, and Australia.
The shrub PFT is overestimated at high latitudes, perhaps as a result of the
high-latitude cold bias in the model. We can see this in these simulations
(Fig. b–d compared to observations in Fig. a). The models simulate fewer needleleaf
trees than observations for , instead simulating shrubs. The
models also underestimate the amount of grasses and bare soil.
The observational dataset is re-gridded from the original to the nine surface
types in our models, which introduces more uncertainty. In particular, the
dominant PFT obviously is a difficult metric to consider precisely, as it does
not represent mixed vegetation systems such as Savannah, well. Some
difficulties mainly originate in how areas such as tundra are allocated – to
bare soil or to C3 grasses. Because of the limited number of PFTs in the model,
C3 grass represents a large range of low-lying vegetation types, arguably also
encompassing mosses and lichen and very sparse tundra vegetation.
There are also some uncertainties associated with the
dataset, which is a reconstruction of pre-industrial vegetation. Other
model–observation discrepancies have been suggested to be a combination of
orographic representation leading to underestimation of precipitation and the
inadequate treatment of natural disturbance mechanisms such as fire
.
Though not shown, the equilibrium (run for 50 years every 5 years)
simulations are very similar to the dynamic (run ever 10 days) ones,
especially in the tree PFTs. That the equilibrium and dynamic simulations
from the same model are very closely related suggests that although the
inter-annual variability does have some influence on the vegetation, in
general the mean climate is more important.
Net primary productivity
The NPP of the models, compared to MODIS 2001 NPP observations, is good at
capturing the global latitudinal patterns, with higher NPP in the tropics and
lower in other regions (see Fig. ). One notable exception
is the failure of the model to capture sub-tropical spikes in productivity,
especially at around 20∘ N, which is also underestimated in the
CMIP5 models analysed here (shown by grey lines). The HadCM3B productivity
peak over the Amazon tropical area is lower in the model than observations.
Overall, the NPP performance of our models compares favourably with that of
CMIP5 models. The large range of NPP values of these CMIP5 models encompasses
our models at nearly all latitudes.
The latitudinal average NPP in gCm-2yr-1. CMIP5
models without dynamic vegetation plotted here are CCSM4 and IPSL-CM5A-LR.
CMIP5 models with dynamic vegetation plotted here are MIROC-ESM and
MPI-ESM.
The Amazon forest extends a little too far south in all the models, but this
is a key area of difference as well, with HadAM3B models better capturing the
observed distribution, and the lower-resolution ocean of HadCM3BL suffering
the most from excess tropical forest. However, HadCM3BL models do better in
the Southern Hemisphere, and better than HadCM3B in other parts of the
tropics. As in the case of the PFT distribution (upon which the NPP is
based), there is a close relationship between the equilibrium (not shown) and
dynamic simulations of NPP.
Summary and future directions
This paper provides an overview of a variety of versions of the HadCM3 family
of coupled climate models used in BRIDGE at the University of Bristol. In this
study we have termed the BRIDGE variants HadCM3B, in order to distinguish our
versions from those originally developed at the Met Office. We provide updated
documentation of these variants, including atmosphere-only, low-resolution
ocean, and high-resolution atmosphere-only models, and including three
alternative versions of the MOSES land surface scheme. Using an up-to-date set
of observational benchmarks we show through detailed comparisons, that the
models provide a good representation of large-scale features of the climate
system, both over land and for the ocean. We additionally show that they remain
comparable to most CMIP5 models.
The speed and relative complexity of HadCM3B and its variants creates
opportunities for tackling a range of problems. Large ensembles are possible
because of the relatively small number of processors required. Ensembles can
explore probabilistic approaches to climate change quantification, model
parametric uncertainty, or boundary condition uncertainty. Long integrations
of many millennia are also possible, so that longer-term climate changes, for
example covering the last deglaciation, can be investigated.
Several versions of the model are under continued development and
improvement. For example, FAMOUS has been coupled to an interactive ice-sheet
model to allow predictions of sea level and land ice on
longer timescales. Further developments in this approach will allow more
detailed investigation of climate–sea-level interactions for a variety of
times in the past e.g.. FAMOUS now also includes a
marine carbon cycle (HadOCC) and an oceanic oxygen cycle
, allowing direct comparisons to biogeochemical cycles.
Currently a very high-resolution version of HadAM3BH is finalising
development in Bristol. This uses a resolution of
0.625∘× 0.4166∘ (576 × 433 grid
points, N288) as this has been suggested as a minimum resolution for
realistic simulation of the hydrological cycle . The model
appears to be significantly computationally more efficient (approximately
10 × faster) than a similar-resolution version of the more recent
UK Met Office HadGEM3 model , because of the lower model
top, simplified aerosol physics, and major differences in the underlying
atmospheric dynamical core.
This paper motivates the continued development and scientific application of
the HadCM3B family of coupled climate models. Future updates will cover new
developments to the presented model version, bug corrections and enhancements.
The UK Met Office made available the source code of HadCM3
via the Ported Unified Model release
(http://www.metoffice.gov.uk/research/collaboration/um-partnership).
Enquiries regarding the use of HadCM3 should be directed in the first
instance to the UM Partnership team, who can be contacted at
um_collaboration@metoffice.gov.uk.
The main repository for the Met Office Unified Model (UM) version corresponding
to the model presented here can be viewed at
http://cms.ncas.ac.uk/code_browsers/UM4.5/UMbrowser/index.html.
The code detailing the advances described in this paper is completely
contained within the files available as a Supplement to this paper. These
files are known as code modification files or “mod” files and should be
applied to the original code of the model. This is protected under UK Crown
Copyright, as is the base code linked above.
The UM basis files for the simulations described in this paper can be found
on the puma.nerc.ac.uk facility
(please contact Andy Heaps for access: andy.heaps@ncas.ac.uk). The simulation
names are the following.
tcsyf: HadCM3B-M1
tcywd: HadCM3B-M2.1aN
tcyxc: HadCM3B-M2.2N
tdbad: HadCM3BL-M2.1aN
tdekd: HadAM3B-M2.1aN
tdewb: HadAM3BH-M2.1aN
tdexb: FAMOUS-M1
tdeyb: FAMOUS-M2.2N
tdkym: HadCM3B-M2.1aD
tdkyn: HadCM3BL-M2.1aD
tdkyo: HadAM3B-M2.1aD
The CERES data were obtained from the NASA Langley Research Center
CERES ordering tool at http://ceres.larc.nasa.gov/.
We acknowledge the World Climate Research Programme's Working Group
on Coupled Modelling, which is responsible for CMIP, and we thank the
climate modelling groups for producing and making available their model
output. For CMIP the U.S. Department of Energy's Program for Climate
Model Diagnosis and Intercomparison provides coordinating support and
led development of software infrastructure in partnership with the
Global Organization for Earth System Science Portals.
We acknowledge the MODIS/TERRA project for the NPP data, acquired
from
http://neo.sci.gsfc.nasa.gov/view.php?datasetId=MOD17A2_M_PSN&year=2001.
Data from the RAPID-WATCH MOC monitoring project are funded by the
Natural Environment Research Council and are freely available from
http://www.rapid.ac.uk/rapidmoc.
The simulations' output for the experiments used in this paper can be
accessed from the BRIDGE website at
http://www.paleo.bristol.ac.uk/ummodel/scripts/papers/Valdes_et_al_2017.html.
The Supplement related to this article is available online at https://doi.org/10.5194/gmd-10-3715-2017-supplement.
The authors declare that they have no conflict of
interest.
Acknowledgements
The efforts of all of the developers of HadCM3 at the Hadley Centre
and UK Met Office is gratefully acknowledged. Without their hard
and successful work we would not be able to tackle some important
scientific problems.
Climate simulations were carried out using the computational
facilities of the Advanced Computing Research Centre, University of
Bristol (http://www.bris.ac.uk/acrc) (Bluecrystal).
Edward Armstrong was funded by NERC (NE/L501554/1).
Marcus P. S. Badger was funded by NERC (NE/J008591/1).
Taraka Davies-Barnard was funded by the European Commission's Seventh
Framework Program grant agreement 282672 (EMBRACE) and EU grant
ERC-2013-CoG-617313 (PaleoGenie).
Alex Farnsworth was funded by NERC (NE/K014757/1; NE/I005722/1; NE/I005714/1)
Peter O. Hopcroft was funded by NERC (NE/I010912/1).
Alan T. Kennedy was funded by NERC (NE/L002434/1).
Natalie S. Lord was funded by RWM Limited via a framework contract with
Amec Foster Wheeler, who were supported by Quintessa.
Alice Marzocchi was funded by National Science Foundation (NSF)
award no. 1536454.
William H. G. Roberts was funded by the Leverhulme Trust.
Gregory J. L. Tourte was funded through “The
Greenhouse Earth System” advanced ERC grant (T-GRES, project reference
340923), awarded
to Richard D. Pancost.
Edited by: Paul
HalloranReviewed by: Christopher Brierley and
two anonymous referees
ReferencesAdam, J. C., Clark, E. A., Lettenmaier, D. P., and Wood, E. F.: Correction of
Global Precipitation Products for Orographic Effects, J. Climate, 19,
15–38, 10.1175/JCLI3604.1, 2006.
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.: Volume 2:
Salinity, World Ocean Atlas 2009, NOAA Atlas NESDIS, U.S. Government Printing Office, Washington, D.C., 2010.Arakawa, A. and Lamb, V. R.: Computational Design of the Basic Dynamical
Processes of the UCLA General Circulation Model, in: General Circulation
Models of the Atmosphere, edited by: Chang, J., Vol. 17, Methods in
Computational Physics: Advances in Research and Applications,
173–265, Elsevier, 10.1016/B978-0-12-460817-7.50009-4, 1977.Armstrong, E., Valdes, P., House, J., and Singarayer, J.: The Role of CO2
and
Dynamic Vegetation on the Impact of Temperate Land-Use Change in the HadCM3
Coupled Climate Model, Earth Interact., 20, 1–20,
10.1175/EI-D-15-0036.1, 2016.Arnell, N. W., Hudson, D. A., and Jones, R. G.: Climate change scenarios from
a
regional climate model: Estimating change in runoff in southern Africa,
J. Geophys. Res.-Atmos., 108, 4519,
10.1029/2002JD002782, 2003.Beerling, D. J., Fox, A., Stevenson, D. S., and Valdes, P. J.: Enhanced
chemistry-climate feedbacks in past greenhouse worlds, P. Natl. Acad. Sci.
USA, 108, 9770–9774, 10.1073/pnas.1102409108, 2011.Best, M. J., Pryor, M., Clark, D. B., Rooney, G. G., Essery, R. L. H.,
Ménard, C. B., Edwards, J. M., Hendry, M. A., Porson, A., Gedney, N.,
Mercado, L. M., Sitch, S., Blyth, E., Boucher, O., Cox, P. M., Grimmond, C.
S. B., and Harding, R. J.: The Joint UK Land Environment Simulator (JULES),
model description – Part 1: Energy and water fluxes, Geosci. Model Dev., 4,
677–699, 10.5194/gmd-4-677-2011, 2011.
Betts, R. A., Cox, P. M., Collins, M., Harris, P. P., Huntingford, C., and
Jones, C. D.: The role of ecosystem-atmosphere interactions in simulated
Amazonian precipitation decrease and forest dieback under global climate
warming, Theor. Appl. Climatol., 78, 157–175, 2004.Bhaskaran, B., Jones, R. G., Murphy, J. M., and Noguer, M.: Simulations of
the
Indian summer monsoon using a nested regional climate model: domain size
experiments, Clim. Dynam., 12, 573–587, 10.1007/BF00216267, 1996.Booth, B. B. B., Jones, C. D., Collins, M., Totterdell, I. J., Cox, P. M.,
Sitch, S., Huntingford, C., Betts, R. A., Harris, G. R., and Lloyd, J.: High
sensitivity of future global warming to land carbon cycle processes,
Environ. Res. Lett., 7, 024002,
10.1088/1748-9326/7/2/024002, 2012.Braconnot, P., Otto-Bliesner, B., Harrison, S., Joussaume, S., Peterchmitt,
J.-Y., Abe-Ouchi, A., Crucifix, M., Driesschaert, E., Fichefet, Th., Hewitt,
C. D., Kageyama, M., Kitoh, A., Laîné, A., Loutre, M.-F., Marti, O.,
Merkel, U., Ramstein, G., Valdes, P., Weber, S. L., Yu, Y., and Zhao, Y.:
Results of PMIP2 coupled simulations of the Mid-Holocene and Last Glacial
Maximum – Part 1: experiments and large-scale features, Clim. Past, 3,
261–277, 10.5194/cp-3-261-2007, 2007.
Bradshaw, C. D., Lunt, D. J., Flecker, R., and Davies-Barnard, T.:
Disentangling the roles of late Miocene palaeogeography and vegetation –
Implications for climate sensitivity, Palaeogeogr. Palaeocl., 417, 17–34, 2015.Bragg, F. J., Lunt, D. J., and Haywood, A. M.: Mid-Pliocene climate modelled
using the UK Hadley Centre Model: PlioMIP Experiments 1 and 2, Geosci. Model
Dev., 5, 1109–1125, 10.5194/gmd-5-1109-2012, 2012.
Brooks, R. H. and Corey, A. T.: Hydraulic properties of porous media,
Hydrology Papers, Colorado State University, Fort Collins, Co, 3, 1964.Bryan, K. and Cox, M. D.: An Approximate Equation of State for Numerical
Models
of Ocean Circulation, J. Phys. Oceanogr., 2, 510–514,
10.1175/1520-0485(1972)002<0510:AAEOSF>2.0.CO;2, 1972.Bushell, A.: Unified Model User Guide, chap. Clouds, p. 27, in:
Matthews (1998), available at:
http://www.ukscience.org/_Media/UM_User_Guide.pdf (last access: 8 September 2017),
1998.Clark, D. B., Mercado, L. M., Sitch, S., Jones, C. D., Gedney, N., Best, M.
J., Pryor, M., Rooney, G. G., Essery, R. L. H., Blyth, E., Boucher, O.,
Harding, R. J., Huntingford, C., and Cox, P. M.: The Joint UK Land
Environment Simulator (JULES), model description – Part 2: Carbon fluxes and
vegetation dynamics, Geosci. Model Dev., 4, 701–722,
10.5194/gmd-4-701-2011, 2011.Claussen, M., Mysak, L., Weaver, A., Crucifix, M., Fichefet, T., Loutre,
M.-F.,
Weber, S., Alcamo, J., Alexeev, V., Berger, A., Calov, R., Ganopolski, A.,
Goosse, H., Lohmann, G., Lunkeit, F., Mokhov, I., Petoukhov, V., Stone, P.,
and Wang, Z.: Earth system models of intermediate complexity: closing the gap
in the spectrum of climate system models, Clim. Dynam., 18, 579–586,
10.1007/s00382-001-0200-1, 2002.Collins, M., Tett, S. F. B., and Cooper, C.: The internal climate variability
of HadCM3, a version of the Hadley Centre coupled model without flux
adjustments, Clim. Dynam., 17, 61–81, 10.1007/s003820000094, 2001.Collins, W. J., Bellouin, N., Doutriaux-Boucher, M., Gedney, N., Halloran,
P., Hinton, T., Hughes, J., Jones, C. D., Joshi, M., Liddicoat, S., Martin,
G., O'Connor, F., Rae, J., Senior, C., Sitch, S., Totterdell, I., Wiltshire,
A., and Woodward, S.: Development and evaluation of an Earth-System model –
HadGEM2, Geosci. Model Dev., 4, 1051–1075,
10.5194/gmd-4-1051-2011, 2011.
Cox, P. M.: A primitive equation, 3-dimensional model of the ocean, GFDL
Ocean
Group Technical Report No. 1, Geophysical Fluid Dynamics
Laboratory, Princeton, New Jersey, 1984.
Cox, P. M.: Description of the TRIFFID dynamic global vegetation model, Tech.
rep., Met Office Hadley Centre, Exeter, UK,
2001.Cox, P. M., Huntingford, C., and Harding, R. J.: A canopy conductance and
photosynthesis model for use in a GCM land surface scheme, J.
Hydrol., 212–213, 79–94, 10.1016/S0022-1694(98)00203-0, 1998.Cox, P. M., Betts, R. A., Bunton, C. B., Essery, R. L. H., Rowntree, P. R.,
and
Smith, J.: The impact of new land surface physics on the GCM simulation of
climate and climate sensitivity, Clim. Dynam., 15, 183–203,
10.1007/s003820050276, 1999.Cox, P. M., Betts, R. A., Jones, C. D., Spall, S. A., and Totterdell, I. J.:
Acceleration of global warming due to carbon-cycle feedbacks in a coupled
climate model, Nature, 408, 184–187, 10.1038/35041539, 2000.Cusack, S., Slingo, A., Edwards, J. M., and Wild, M.: The radiative impact of
a
simple aerosol climatology on the Hadley Centre atmospheric GCM, Q.
J. Roy. Meteor. Soc., 124, 2517–2526,
10.1002/qj.49712455117, 1998.Davies-Barnard, T., Valdes, P. J., Jones, C. D., and Singarayer, J. S.:
Sensitivity of a coupled climate model to canopy interception capacity,
Clim. Dynam., 42, 1715–1732, 10.1007/s00382-014-2100-1, 2014.Day, J. J., Bamber, J. L., Valdes, P. J., and Kohler, J.: The impact of a
seasonally ice free Arctic Ocean on the temperature, precipitation and
surface mass balance of Svalbard, The Cryosphere, 6, 35–50,
10.5194/tc-6-35-2012, 2012.Day, J. J., Bamber, J. L., and Valdes, P. J.: The Greenland Ice Sheet's
surface
mass balance in a seasonally sea ice-free Arctic, J. Geophys.
Res.-Earth, 118, 1533–1544, 10.1002/jgrf.20112, 2013.Demory, M.-E., Vidale, P. L., Roberts, M. J., Berrisford, P., Strachan, J., Schiemann, R., and Mizielinski, M. S.: The role of horizontal
resolution in simulating drivers of the global hydrological cycle, Clim.
Dynam., 42, 2201–2225, 10.1007/s00382-013-1924-4, 2014.Dolan, A. M., Haywood, A. M., Hunter, S. J., Tindall, J. C., Dowsett, H. J.,
Hill, D. J., and Pickering, S. J.: Modelling the enigmatic Late Pliocene
Glacial Event – Marine Isotope Stage M2, Global Planet. Change, 128,
47–60, 10.1016/j.gloplacha.2015.02.001, 2015.Durman, C. F., Gregory, J. M., Hassell, D. C., Jones, R. G., and Murphy,
J. M.:
A comparison of extreme European daily precipitation simulated by a global
and a regional climate model for present and future climates, Q.
J. Roy. Meteor. Soc., 127, 1005–1015,
10.1002/qj.49712757316, 2001.Edwards, J.: “Radiation” Unified Model User Guide, Version 4.4., Report,
The Meteorological Office, UK, Bracknell, Berkshire, UK,
available at: http://www.ukscience.org/_Media/UM_User_Guide.pdf (last access: 8 September 2017),
1998.Edwards, J. M. and Slingo, A.: Studies with a flexible new radiation code. I:
Choosing a configuration for a large-scale model, Q. J.
Roy. Meteor. Soc., 122, 689–719, 10.1002/qj.49712253107,
1996.
Edwards, M. O.: Global gridded elevation and bathymetry on 5-minute
geographic
grid (ETOPO5), NOAA, National Geophysical Data Center, 1989.
Essery, R. L. H., Best, M. J., and Cox, P. M.: MOSES 2.2 Technical
Documentation, Tech. rep., Hadley Centre, Met Office,
2001.Essery, R. L. H., Best, M. J., Betts, R. A., Cox, P. M., and Taylor, C. M.:
Explicit Representation of Subgrid Heterogeneity in a GCM Land Surface
Scheme, J. Hydrometeorol., 4, 530–543,
10.1175/1525-7541(2003)004<0530:EROSHI>2.0.CO;2, 2003.Eyring, V., Bony, S., Meehl, G. A., Senior, C. A., Stevens, B., Stouffer, R.
J., and Taylor, K. E.: Overview of the Coupled Model Intercomparison Project
Phase 6 (CMIP6) experimental design and organization, Geosci. Model Dev., 9,
1937–1958, 10.5194/gmd-9-1937-2016, 2016a.Eyring, V., Righi, M., Lauer, A., Evaldsson, M., Wenzel, S., Jones, C., Anav,
A., Andrews, O., Cionni, I., Davin, E. L., Deser, C., Ehbrecht, C.,
Friedlingstein, P., Gleckler, P., Gottschaldt, K.-D., Hagemann, S., Juckes,
M., Kindermann, S., Krasting, J., Kunert, D., Levine, R., Loew, A.,
Mäkelä, J., Martin, G., Mason, E., Phillips, A. S., Read, S., Rio,
C., Roehrig, R., Senftleben, D., Sterl, A., van Ulft, L. H., Walton, J.,
Wang, S., and Williams, K. D.: ESMValTool (v1.0) – a community diagnostic
and performance metrics tool for routine evaluation of Earth system models in
CMIP, Geosci. Model Dev., 9, 1747–1802,
10.5194/gmd-9-1747-2016, 2016b.Ferraro, R., Waliser, D. E., Glecker, P., Taylor, K. E., and Eyring, V.:
Evolving obs4MIPS to support the Sixth Coupled Model Intercomparison Project
(CMIP6), Am. Metereol. Soc., 131–133, 10.1175/BAMS-D-14-00216.1,
2015.Flato, G., Marotzke, J., Abiodun, B., Braconnot, P., Chou, S., Collins, W.,
Cox, P. M., Driouech, F., Emori, S., Eyring, V., Forest, C., Gleckler, P.,
Guilyardi, E., Jakob, C., Kattsov, V., Reason, C., and Rummukainen, M.:
Evaluation of Climate Models, in: Climate Change 2013: The Physical Science
Basis. Contribution of Working Group I to the Fifth Assessment Report of the
Intergovernmental Panel on Climate Change, edited by: Stocker, T., Qin, D.,
Plattner, G.-K., Tignor, M., Allen, S., Boschung, J., Nauels, A., Xia, Y.,
Bex, V., and Midgley, P., chap. 9, 741–866, Cambridge University Press,
Cambridge, UK and New York, NY, USA,
10.1017/CBO9781107415324.020, 2013.Foreman, S. J.: Unified Model Documentaiton Paper Number 40, The Ocean Model,
Report, The Met. Office, available at:
http://cms.ncas.ac.uk/documents/vn4.5/p040.pdf (last access: 8 September 2017),
2005.
Gent, P. R. and McWilliams, J. C.: Isopycnal Mixing in Ocean Circulation
Models, J. Phys. Oceanogr., 20, 150–155, 1990.Gleckler, P. J., Taylor, K. E., and Doutriaux, C.: Performance metrics for
climate models, J. Geophys. Res.-Atmos., 113, d06104,
10.1029/2007JD008972, 2008.Goosse, H., Brovkin, V., Fichefet, T., Haarsma, R., Huybrechts, P., Jongma,
J., Mouchet, A., Selten, F., Barriat, P.-Y., Campin, J.-M., Deleersnijder,
E., Driesschaert, E., Goelzer, H., Janssens, I., Loutre, M.-F., Morales
Maqueda, M. A., Opsteegh, T., Mathieu, P.-P., Munhoven, G., Pettersson, E.
J., Renssen, H., Roche, D. M., Schaeffer, M., Tartinville, B., Timmermann,
A., and Weber, S. L.: Description of the Earth system model of intermediate
complexity LOVECLIM version 1.2, Geosci. Model Dev., 3, 603–633,
10.5194/gmd-3-603-2010, 2010.Gordon, C., Cooper, C., Senior, C. A., Banks, H., Gregory, J. M., Johns,
T. C.,
Mitchell, J. F. B., and Wood, R. A.: The simulation of SST, sea ice extents
and ocean heat transports in a version of the Hadley Centre coupled model
without flux adjustments, Clim. Dynam., 16, 147–168,
10.1007/s003820050010, 2000.Grant, A.: Unified Model User Guide, chap. Convection, p. 32, in:
Matthews (1998), available at:
http://www.ukscience.org/_Media/UM_User_Guide.pdf (last access: 8 September 2017),
1998.Gregory, D., Kershaw, R., and Inness, P. M.: Parametrization of momentum
transport by convection. II: Tests in single-column and general circulation
models, Q. J. Roy. Meteor. Soc., 123,
1153–1183, 10.1002/qj.49712354103, 1997.Gregory, D., Shutts, G. J., and Mitchell, J. R.: A new gravity-wave-drag
scheme
incorporating anisotropic orography and low-level wave breaking: Impact upon
the climate of the UK Meteorological Office Unified Model, Q. J.
Roy. Meteor. Soc., 124, 463–493,
10.1002/qj.49712454606, 1998.Gregory, J. M., Ingram, W. J., Palmer, M. A., Jones, G. S., Stott, P. A.,
Thorpe, R. B., Lowe, J. A., Johns, T. C., and Williams, K. D.: A new method
for diagnosing radiative forcing and climate sensitivity, Geophys.
Res. Lett., 31, L03205, 10.1029/2003gl018747, 2004.Gregory, J. M., Browne, O. J. H., Payne, A. J., Ridley, J. K., and Rutt, I.
C.: Modelling large-scale ice-sheet-climate interactions following glacial
inception, Clim. Past, 8, 1565–1580, 10.5194/cp-8-1565-2012,
2012.
Griffies, S., Gnanadesikan, A., Pacanowski, R., Larichev, V. D., Dukowicz, J. K., and Smith, R. D.: Oceanic isopycnal
mixing by coordinate rotation, J. Phys. Oceanogr., 28, 805–830, 1998.The HadGEM2 Development Team: G. M. Martin, Bellouin, N., Collins, W. J.,
Culverwell, I. D., Halloran, P. R., Hardiman, S. C., Hinton, T. J., Jones, C.
D., McDonald, R. E., McLaren, A. J., O'Connor, F. M., Roberts, M. J.,
Rodriguez, J. M., Woodward, S., Best, M. J., Brooks, M. E., Brown, A. R.,
Butchart, N., Dearden, C., Derbyshire, S. H., Dharssi, I., Doutriaux-Boucher,
M., Edwards, J. M., Falloon, P. D., Gedney, N., Gray, L. J., Hewitt, H. T.,
Hobson, M., Huddleston, M. R., Hughes, J., Ineson, S., Ingram, W. J., James,
P. M., Johns, T. C., Johnson, C. E., Jones, A., Jones, C. P., Joshi, M. M.,
Keen, A. B., Liddicoat, S., Lock, A. P., Maidens, A. V., Manners, J. C.,
Milton, S. F., Rae, J. G. L., Ridley, J. K., Sellar, A., Senior, C. A.,
Totterdell, I. J., Verhoef, A., Vidale, P. L., and Wiltshire, A.: The HadGEM2
family of Met Office Unified Model climate configurations, Geosci. Model
Dev., 4, 723–757, 10.5194/gmd-4-723-2011, 2011.Hawkins, E., Smith, R. S., Allison, L. C., Gregory, J. M., Woollings, T. J.,
Pohlmann, H., and de Cuevas, B.: Bistability of the Atlantic overturning
circulation in a global climate model and links to ocean freshwater
transport, Geophys. Res. Lett., 38, L10605, 10.1029/2011gl047208, 2011.Haywood, A. M., Valdes, P. J., and Markwick, P. J.: Cretaceous (Wealden)
climates: a modelling perspective, Cretaceous Res., 25, 303–311,
10.1016/j.cretres.2004.01.005, 2004.Hewitt, C. D., Broccoli, A. J., Mitchell, J. F. B., and Stouffer, R. J.: A
coupled model study of the Last Glacial Maximum: Was part of the North
Atlantic relatively warm?, Geophys. Res. Lett., 28, 1571–1574,
10.1029/2000GL012575, 2001.Hibler, W. D.: A Dynamic Thermodynamic Sea Ice Model, J. Phys.
Oceanogr., 9, 815–846,
10.1175/1520-0485(1979)009<0815:ADTSIM>2.0.CO;2, 1979.Hopcroft, P. O., Valdes, P. J., Wania, R., and Beerling, D. J.: Limited
response of peatland CH4 emissions to abrupt Atlantic Ocean circulation
changes in glacial climates, Clim. Past, 10, 137–154,
10.5194/cp-10-137-2014, 2014.Hudson, D. A. and Jones, R. G.: Regional Climate Models Simulations of
Present-Day and Future Climate over Southern Africa, Tech. rep., Met Office
Hadley Center, Exeter, UK, available at:
https://digital.nmla.metoffice.gov.uk/download/file/sdb:digitalFile|a55db57e-b268-49ba-9cc3-d48553b244f5/ (last access: 8 September 2017),
2002.Hughes, J. K., Valdes, P. J., and Betts, R. A.: Dynamics of a global-scale
vegetation model, Ecol. Model., 198, 452–462,
10.1016/j.ecolmodel.2006.05.020, 2006.
Ingram, W. S., Woodward, S., and Edwards, J.: Unified Model Documentation
Paper: Radiation, Report 23, The Meteorological Office, UK, 1997.
IPCC: Climate Change 2007: The Physical
Science Basis. Contribution of Working Group I to the Fourth Assessment
Report of the Intergovernmental Panel on Climate Change, edited by: Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Avery, K., Tignor,
M., and Miller, H., Cambridge University
Press, 2007.Irvine, P. J., Ridgwell, A., and Lunt, D. J.: Assessing the regional
disparities in geoengineering impacts, Geophys. Res. Lett., 37,
L18702, 10.1029/2010GL044447, 2010.Jackson, L. and Vellinga, M.: Multidecadal to Centennial Variability of the
AMOC: HadCM3 and a Perturbed Physics Ensemble, J. Climate, 26,
2390–2407, 10.1175/JCLI-D-11-00601.1, 2013.Johns, T. C., Gregory, J. M., Ingram, W. J., Johnson, C. E., Lowe, J. A.,
Mitchell, J. F. B., Roberts, D. L., Sexton, D. M. H., Stevenson, D., Tett, S.
F. B., and Woodage, M. J.: Anthropogenic climate change for 1860 to 2100
simulated with the HadCM3 model under updated emissions scenarios, Clim.
Dynam., 20, 583–612, 10.1007/s00382-002-0296-y, 2003.Jones, C.: A Fast Ocean GCM without Flux Adjustments, J. Atmos. Ocean.
Technol., 20, 1857–1868,
10.1175/1520-0426(2003)020<1857:AFOGWF>2.0.CO;2, 2003.Jones, C., Gregory, J., Thorpe, R., Cox, P., Murphy, J., Sexton, D., and
Valdes, P. J.: Systematic optimisation and climate simulation of FAMOUS, a
fast version of HadCM3, Clim. Dynam., 25, 189–204,
10.1007/s00382-005-0027-2, 2005.Jones, R. G., Murphy, J. M., and Noguer, M.: Simulation of climate change
over
europe using a nested regional-climate model. I: Assessment of control
climate, including sensitivity to location of lateral boundaries, Q.
J. Roy. Meteor. Soc., 121, 1413–1449,
10.1002/qj.49712152610, 1995.Kennedy, A. T., Farnsworth, A., Lunt, D. J., Lear, C. H., and Markwick,
P. J.:
Atmospheric and oceanic impacts of Antarctic glaciation across the
Eocene–Oligocene transition, Philos. T. R.
Soc., A, 373, 20140419, 10.1098/rsta.2014.0419, 2015.Kraus, E. B. and Turner, J. S.: A one-dimensional model of the seasonal
thermocline II. The general theory and its consequences, Tellus, 19, 98–106,
10.1111/j.2153-3490.1967.tb01462.x, 1967.Large, W. G., McWilliams, J. C., and Doney, S. C.: Oceanic vertical mixing: A
review and a model with a nonlocal boundary layer parameterization, Rev. Geophys., 32, 363–403, 10.1029/94RG01872, 1994.Legg, S., Briegleb, B., Chang, Y., Chassignet, E. P., Danabasoglu, G., Ezer,
T., Gordon, A. L., Griffies, S., Hallberg, R., Jackson, L., Large, W.,
Ozgokmen, T. M., Peters, H., Price, J., Riemenschneider, U., Wu, W. L., Xu,
X. B., and Yang, J. Y.: Improving Oceanic Overflow Representation in Climate
Models: The Gravity Current Entrainment Climate Process Team, B.
Am. Meteorol. Soc., 90, 657–670, 10.1175/2008bams2667.1,
2009.Liu, W., Liu, Z. Y., and Brady, E. C.: Why is the AMOC Monostable in Coupled
General Circulation Models?, J. Climate, 27, 2427–2443,
10.1175/Jcli-D-13-00264.1, 2014.
Locarnini, R. A., Mishonov, A. V., Antonov, J. I., Boyer, T. P., Garcia,
H. E.,
Baranova, O. K., Zweng, M. M., and Johnson, D. R.: Volume 1: Temperature,
World Ocean Atlas 2009, NOAA Atlas NESDIS, U.S. Government Printing Office, Washington, D.C.,
2010.Loeb, N., Wang, H., Cheng, A., Kato, S., Fasullo, J., Xu, K.-M., and Allan, R.:
Observational constraints on atmospheric and oceanic cross-equatorial heat
transports: revisiting the precipitation asymmetry problem in climate models,
Clim. Dynam., 46, 3239–3257, 10.1007/s00382-015-2766-z, 2016.Loptson, C. A., Lunt, D. J., and Francis, J. E.: Investigating
vegetation–climate feedbacks during the early Eocene, Clim. Past, 10,
419–436, 10.5194/cp-10-419-2014, 2014.Loveland, T. R. and Belward, A. S.: The IGBP-DIS global 1 km land cover
data
set, DISCover: First results, Int. J. Remote Sens., 18,
3289–3295, 10.1080/014311697217099, 1997.Lunt, D. J., Ridgwell, A., Valdes, P. J., and Seale, A.: “Sunshade World”:
A fully coupled GCM evaluation of the climatic impacts of geoengineering,
Geophys. Res. Lett., 35, L12710, 10.1029/2008GL033674, 2008.Lunt, D. J., Farnsworth, A., Loptson, C., Foster, G. L., Markwick, P.,
O'Brien, C. L., Pancost, R. D., Robinson, S. A., and Wrobel, N.:
Palaeogeographic controls on climate and proxy interpretation, Clim. Past,
12, 1181–1198, 10.5194/cp-12-1181-2016, 2016.Malhi, Y., Aragão, L. E. O. C., Galbraith, D., Huntingford, C., Fisher,
R.,
Zelazowski, P., Sitch, S., McSweeney, C., and Meir, P.: Exploring the
likelihood and mechanism of a climate-change-induced dieback of the Amazon
rainforest, P. Natl. Acad. Sci., 106,
20610–20615, 10.1073/pnas.0804619106, 2009.Marzocchi, A., Hirschi, J. J.-M., Holliday, N. P., Cunningham, S. A., Blaker,
A. T., and Coward, A. C.: The North Atlantic subpolar circulation in an
eddy-resolving global ocean model, J. Marine Syst., 142, 126–143,
10.1016/j.jmarsys.2014.10.007, 2015a.Marzocchi, A., Lunt, D. J., Flecker, R., Bradshaw, C. D., Farnsworth, A., and
Hilgen, F. J.: Orbital control on late Miocene climate and the North African
monsoon: insight from an ensemble of sub-precessional simulations, Clim.
Past, 11, 1271–1295, 10.5194/cp-11-1271-2015, 2015b.Matthews, D. (Ed.): Unified Model User Guide, The Meteorological Office,
Bracknell, Berkshire, UK, available at:
http://www.ukscience.org/_Media/UM_User_Guide.pdf (last access: 8 September 2017),
1998.McCarthy, G., Frajka-Williams, E., Johns, W. E., Baringer, M. O., Meinen,
C. S., Bryden, H. L., Rayner, D., Duchez, A., Roberts, C., and Cunningham,
S. A.: Observed interannual variability of the Atlantic meridional
overturning circulation at 26.5∘ N, Geophys. Res. Lett., 39,
10.1029/2012gl052933, 2012.Msadek, R., Johns, W. E., Yeager, S. G., Danabasoglu, G., Delworth, T. L.,
and
Rosati, A.: The Atlantic Meridional Heat Transport at 26.5∘ N and Its
Relationship with the MOC in the RAPID Array and the GFDL and NCAR Coupled
Models, J. Climate, 26, 4335–4356, 10.1175/Jcli-D-12-00081.1,
2013.Murphy, J. M., Booth, B. B. B., Collins, M., Harris, G. R., Sexton, D. M. H.,
and Webb, M. J.: A methodology for probabilistic predictions of regional
climate change from perturbed physics ensembles, Philos. T. R. Soc. Lond. A, 365, 1993–2028, 10.1098/rsta.2007.2077, 2007.New, M., Lister, D., Hulme, M., and Makin, I.: A high-resolution data set of
surface climate over global land areas, Clim. Res., 21, 1–25,
10.3354/cr021001, 2002.
Pacanowski, R. and Philander, S.: Parametrisation of vertical mixing in
numerical models of tropical oceans, J. Phys. Oceanogr., 11, 1443–1451, 1981.Pardaens, A. K., Banks, H. T., Gregory, J. M., and Rowntree, P. R.:
Freshwater
transports in HadCM3, Clim. Dynam., 21, 177–195,
10.1007/s00382-003-0324-6, 2003.Pope, V. D., Gallani, M. L., Rowntree, P. R., and Stratton, R. A.: The impact
of new physical parametrizations in the Hadley Centre climate model: HadAM3,
Clim. Dynam., 16, 123–146, 10.1007/s003820050009, 2000.Ramankutty, N. and Foley, J. A.: Estimating historical changes in global land
cover: Croplands from 1700 to 1992, Global Biogeochem. Cy., 13,
997–1027, 10.1029/1999GB900046, 1999.Redi, M. H.: Oceanic Isopycnal Mixing by Coordinate Rotation, J. Phys.
Oceanogr., 12, 1154–1158,
10.1175/1520-0485(1982)012<1154:OIMBCR>2.0.CO;2, 1998.Reichler, T. and Kim, J.: How well do coupled models simulate today's
climate?, B. Am. Meteorol. Soc., 89, 303–311,
10.1175/BAMS-89-3-303,
2008.Rickard, G.: Unified Model Documentation No. 59: Ocean Models and the
Implementation of Vertical Diffusion and Vertical Mixing, Report, The Met.
Office, available at: http://cms.ncas.ac.uk/documents/vn4.5/p059.pdf (last access: 8 September 2017),
1999.Ridgwell, A., Singarayer, J. S., Hetherington, A. M., and Valdes, P. J.:
Tackling Regional Climate Change By Leaf Albedo Bio-geoengineering, Curr. Biol., 19, 146–150, 10.1016/j.cub.2008.12.025, 2009.Ridley, J. K., Huybrechts, P., Gregory, J. M., and Lowe, J. A.: Elimination
of
the Greenland ice sheet in a high CO2 climate, J. Climate, 18,
3409–3427, 10.1175/Jcli3482.1, 2005.Roberts, C. D., Garry, F. K., and Jackson, L. C.: A Multimodel Study of Sea
Surface Temperature and Subsurface Density Fingerprints of the Atlantic
Meridional Overturning Circulation, J. Climate, 26, 9155–9174,
10.1175/Jcli-D-12-00762.1, 2013.Roberts, W. H. G. and Valdes, P. J.: Green Mountains and White Plains: the
effect of Northern Hemisphere ice sheets on the global energy budget, J.
Climate., 30, 3887–3905, 10.1175/JCLI-D-15-0846.1, 2017.Roberts, W. H. G., Valdes, P. J., and Payne, A. J.: Topography's crucial role
in Heinrich Events, P. Natl. Acad. Sci. USA, 111, 16688–16693,
10.1073/pnas.1414882111, 2014.Balan Sarojini, B., Gregory, J. M., Tailleux, R., Bigg, G. R., Blaker, A. T.,
Cameron, D. R., Edwards, N. R., Megann, A. P., Shaffrey, L. C., and Sinha,
B.: High frequency variability of the Atlantic meridional overturning
circulation, Ocean Sci., 7, 471–486, 10.5194/os-7-471-2011,
2011.Schaller, N., Kay, A. L., Lamb, R., Massey, N. R., van Oldenborgh, G. J.,
Otto,
F. E. L., Sparrow, S. N., Vautard, R., Yiou, P., Ashpole, I., Bowery, A.,
Crooks, S. M., Haustein, K., Huntingford, C., Ingram, W. J., Jones, R. G.,
Legg, T., Miller, J., Skeggs, J., Wallom, D., Weisheimer, A., Wilson, S.,
Stott, P. A., and Allen, M. R.: Human influence on climate in the 2014
southern England winter floods and their impacts, Nature Clim. Change, 6,
627–634, 10.1038/nclimate2927, 2016.Sijp, W. P.: Characterising meridional overturning bistability using a
minimal set of state variables, Clim. Dynam., 39, 2127–2142,
10.1007/s00382-011-1249-0, 2012.Simmons, A. J. and Strüfing, R.: Numerical forecasts of stratospheric
warming events using a model with a hybrid vertical coordinate, Q.
J. Roy. Meteor. Soc., 109, 81–111,
10.1002/qj.49710945905, 1983.Singarayer, J. S. and Valdes, P. J.: High-latitude climate sensitivity to
ice-sheet forcing over the last 120 kyr, Quaternary Sci. Rev., 29,
43–55, 10.1016/j.quascirev.2009.10.011, 2010.Singarayer, J. S., Ridgwell, A., and Irvine, P.: Assessing the benefits of
crop
albedo bio-geoengineering, Environ. Res. Lett., 4, 045110, 10.1088/1748-9326/4/4/045110, 2009.Sitch, S., Huntingford, C., Gedney, N., Levy, P. E., Lomas, M., Piao, S. L.,
Betts, R., Ciais, P., Cox, P. M., Friedlingstein, P., Jones, C. D., Prentice,
I. C., and Woodward, F. I.: Evaluation of the terrestrial carbon cycle,
future plant geography and climate-carbon cycle feedbacks using five Dynamic
Global Vegetation Models (DGVMs), Glob. Change Biol., 14, 2015–2039,
10.1111/j.1365-2486.2008.01626.x, 2008.Smeed, D., McCarthy, G., Rayner, D., Moat, B. I., Johns, W. E., Baringer, M.,
and Meinen, C. S.: Atlantic meridional overturning circulation observed by
the RAPID-MOCHA-WBTS (RAPID-Meridional Overturning Circulation and Heatflux
Array-Western Boundary Time Series) array at 26N from 2004 to 2014, British
Oceanographic Data Centre – Natural Environment Research Council, UK,
10/6qb, 2015.Smith, D. M., Cusack, S., Colman, A. W., Folland, C. K., Harris, G. R., and
Murphy, J. M.: Improved Surface Temperature Prediction for the Coming Decade
from a Global Climate Model, Science, 317, 796–799,
10.1126/science.1139540, 2007.Smith, D. M., Eade, R., Dunstone, N. J., Fereday, D., Murphy, J. M.,
Pohlmann,
H., and Scaife, A. A.: Skilful multi-year predictions of Atlantic hurricane
frequency, Nat. Geosci., 3, 846–849, 10.1038/ngeo1004, 2010.Smith, R. S.: The FAMOUS climate model (versions XFXWB and XFHCC):
description update to version XDBUA, Geosci. Model Dev., 5, 269–276,
10.5194/gmd-5-269-2012, 2012.Smith, R. S., Gregory, J. M., and Osprey, A.: A description of the FAMOUS
(version XDBUA) climate model and control run, Geosci. Model Dev., 1, 53–68,
10.5194/gmd-1-53-2008, 2008.Spencer, H., Sutton, R., and Slingo, J. M.: El Niño in a Coupled Climate
Model: Sensitivity to Changes in Mean State Induced by Heat Flux and Wind
Stress Corrections, J. Climate, 20, 2273–2298,
10.1175/JCLI4111.1, 2007.Stainforth, D. A., Aina, T., Christensen, C., Collins, M., Faull, N., Frame,
D. J., Kettleborough, J. A., Knight, S., Martin, A., Murphy, J. M., Piani,
C., Sexton, D., Smith, L. A., Spicer, R. A., Thorpe, A. J., and Allen, M. R.:
Uncertainty in predictions of the climate response to rising levels of
greenhouse gases, Nature, 433, 403–406, 10.1038/nature03301, 2005.
Stott, P. A. and Kettleborough, J. A.: Origins and estimates of uncertainty
in
predictions of twenty-first century temperature rise, Nature, 416, 723–726,
2002.
Stott, P. A., Tett, S. F. B., Jones, G. S., Allen, M. R., Mitchell, J. F. B.,
and Jenkins, G. J.: External control of 20th century temperature by natural
and anthropogenic forcings, Science, 290, 2133–2137, 2000.Svirezhev, Y. M.: Thermodynamics and ecology, Ecol. Model., 132,
11–22, 10.1016/S0304-3800(00)00301-X, 2000.Tett, S. F. B., Betts, R., Crowley, T. J., Gregory, J., Johns, T. C., Jones,
A., Osborn, T. J., Öström, E., Roberts, D. L., and Woodage, M. J.:
The impact of natural and anthropogenic forcings on climate and hydrology
since 1550, Clim. Dynam., 28, 3–34, 10.1007/s00382-006-0165-1,
2006.Toniazzo, T., Collins, M., and Brown, J.: The variation of ENSO
characteristics
associated with atmospheric parameter perturbations in a coupled model,
Clim. Dynam., 30, 643–656, 10.1007/s00382-007-0313-2, 2007.Trenberth, K. E. and Fasullo, J. T.: An apparent hiatus in global warming?,
Earth's Future, 1, 19–32, 10.1002/2013EF000165, 2013.UNESCO: Tenth report of the joint panel on oceanographic tables and
standards,
Report, available at:
http://unesdoc.unesco.org/images/0004/000461/046148eb.pdf (last access: 8 September 2017),
1981.Valdes, P. J., Beerling, D. J., and Johnson, C. E.: The ice age methane
budget,
Geophys. Res. Lett., 32, L02704, 10.1029/2004GL021004, 2005.Van der Wal, A.: Unified Model User Guide, chap. Radiation, p. 11, in:
Matthews (1998), available at:
http://www.ukscience.org/_Media/UM_User_Guide.pdf (last access: 8 September 2017),
1998.
Visbeck, M., Marshall, J., Haine, T., and Spall, M.: On the specification of
eddy transfer coefficients in coarse resolution ocean circulation models,
J.
Phys. Oceanogr., 27, 381–402, 1997.Walters, D. N., Williams, K. D., Boutle, I. A., Bushell, A. C., Edwards, J.
M., Field, P. R., Lock, A. P., Morcrette, C. J., Stratton, R. A., Wilkinson,
J. M., Willett, M. R., Bellouin, N., Bodas-Salcedo, A., Brooks, M. E.,
Copsey, D., Earnshaw, P. D., Hardiman, S. C., Harris, C. M., Levine, R. C.,
MacLachlan, C., Manners, J. C., Martin, G. M., Milton, S. F., Palmer, M. D.,
Roberts, M. J., Rodríguez, J. M., Tennant, W. J., and Vidale, P. L.: The
Met Office Unified Model Global Atmosphere 4.0 and JULES Global Land 4.0
configurations, Geosci. Model Dev., 7, 361–386,
10.5194/gmd-7-361-2014, 2014.
Warrilow, D.:
Modelling of land surface processes and
their influence on European climate, Met Office 20 (Dynamical Climatology
Branch), Meteorological Office, Exeter, UK, 1986.
Warrilow, D. A. and Buckley, E.: The Impact of Land Surface Processes on the
Moisture Budget of a Climate Model, Ann. Geophys., 7, 439–449, 1989.Weaver, A. J., Sedlacek, J., Eby, M., Alexander, K., Crespin, E., Fichefet,
T.,
Philippon-Berthier, G., Joos, F., Kawamiya, M., Matsumoto, K., Steinacher,
M., Tachiiri, K., Tokos, K., Yoshimori, M., and Zickfeld, K.: Stability of
the Atlantic meridional overturning circulation: A model intercomparison,
Geophys. Res. Lett., 39, L20709, 10.1029/2012gl053763, 2012.White, A. A. and Bromley, R. A.: Dynamically consistent, quasi-hydrostatic
equations for global models with a complete representation of the Coriolis
force, Q. J. Roy. Meteor. Soc., 121, 399–418,
10.1002/qj.49712152208, 1995.Williams, J. H. T., Smith, R. S., Valdes, P. J., Booth, B. B. B., and Osprey,
A.: Optimising the FAMOUS climate model: inclusion of global carbon cycling,
Geosci. Model Dev., 6, 141–160, 10.5194/gmd-6-141-2013,
2013.Williams, J. H. T., Totterdell, I. J., Halloran, P. R., and Valdes, P. J.:
Numerical simulations of oceanic oxygen cycling in the FAMOUS Earth-System
model: FAMOUS-ES, version 1.0, Geosci. Model Dev., 7, 1419–1431,
10.5194/gmd-7-1419-2014, 2014.Williams, K. D., Harris, C. M., Bodas-Salcedo, A., Camp, J., Comer, R. E.,
Copsey, D., Fereday, D., Graham, T., Hill, R., Hinton, T., Hyder, P., Ineson,
S., Masato, G., Milton, S. F., Roberts, M. J., Rowell, D. P., Sanchez, C.,
Shelly, A., Sinha, B., Walters, D. N., West, A., Woollings, T., and Xavier,
P. K.: The Met Office Global Coupled model 2.0 (GC2) configuration, Geosci.
Model Dev., 8, 1509–1524, 10.5194/gmd-8-1509-2015, 2015.Willmott, C. J., Rowe, C. M., and Mintz, Y.: Climatology of the terrestrial
seasonal water cycle, J. Climatol., 5, 589–606,
10.1002/joc.3370050602, 1985.Wilson, D.: Unified Model User Guide, chap. Precipitation, p. 32, in:
Matthews (1998), available at:
http://www.ukscience.org/_Media/UM_User_Guide.pdf (last access: 8 September 2017),
1998.
Wilson, M. F. and Henderson-Sellers, A.: A global archive of land cover and
soils data for use in general circulation climate models, J.
Climatol., 5, 119–143, 10.1002/joc.3370050202, 1985.
Zhang, L. P. and Wang, C. Z.: Multidecadal North Atlantic sea surface
temperature and Atlantic meridional overturning circulation variability in
CMIP5 historical simulations, J. Geophys. Res.-Oceans, 118,
5772–5791, 10.1002/jgrc.20390, 2013.