The primary objective of CFMIP is to inform future assessments of cloud feedbacks through improved understanding of cloud–climate feedback mechanisms and better evaluation of cloud processes and cloud feedbacks in climate models. However, the CFMIP approach is also increasingly being used to understand other aspects of climate change, and so a second objective has now been introduced, to improve understanding of circulation, regional-scale precipitation, and non-linear changes. CFMIP is supporting ongoing model inter-comparison activities by coordinating a hierarchy of targeted experiments for CMIP6, along with a set of cloud-related output diagnostics. CFMIP contributes primarily to addressing the CMIP6 questions “How does the Earth system respond to forcing?” and “What are the origins and consequences of systematic model biases?” and supports the activities of the WCRP Grand Challenge on Clouds, Circulation and Climate Sensitivity.
A compact set of Tier 1 experiments is proposed for CMIP6 to address this
question: (1) what are the physical mechanisms underlying the range of cloud
feedbacks and cloud adjustments predicted by climate models, and which models
have the most credible cloud feedbacks? Additional Tier 2 experiments are
proposed to address the following questions. (2) Are cloud feedbacks
consistent for climate cooling and warming, and if not, why? (3) How do
cloud-radiative effects impact the structure, the strength and the
variability of the general atmospheric circulation in present and future
climates? (4) How do responses in the climate system due to changes in solar
forcing differ from changes due to CO
CFMIP also proposes a number of additional model outputs in the CMIP DECK,
CMIP6 Historical and CMIP6 CFMIP experiments, including COSP simulator
outputs and process diagnostics to address the following questions.
How
well do clouds and other relevant variables simulated by models agree with
observations? What physical processes and mechanisms are important for a
credible simulation of clouds, cloud feedbacks and cloud adjustments in
climate models? Which models have the most credible representations of
processes relevant to the simulation of clouds? How do clouds and their
changes interact with other elements of the climate system?
Inter-model differences in cloud feedbacks continue to be the largest source of uncertainty in predictions of equilibrium climate sensitivity (Boucher et al., 2013). Although the ranges of cloud feedbacks and climate sensitivity from comprehensive climate models have not reduced in recent years, considerable progress has been made in understanding (a) which types of clouds contribute most to this spread (e.g. Bony and Dufresne, 2005; Webb et al., 2006; Zelinka et al., 2013), (b) the role of cloud adjustments in climate sensitivity (e.g. Gregory and Webb, 2008; Andrews and Forster, 2008; Kamae and Watanabe, 2012; Zelinka et al., 2013), (c) the processes and mechanisms which are (and are not) implicated in cloud feedbacks, both in fine-resolution models (e.g. Rieck et al., 2012; Bretherton et al., 2015) and in comprehensive climate models (e.g. Brient and Bony, 2012; Sherwood et al., 2014; Zhao, 2014; Webb et al., 2015b), (d) the inconstancy of cloud feedbacks and effective climate sensitivity (e.g. Senior and Mitchell, 2000; Williams et al., 2008; Andrews et al., 2012; Geoffroy et al., 2013; Armour et al., 2013; Gregory and Andrews, 2016) and (e) the extent to which models with stronger or weaker cloud feedbacks or climate sensitivities agree with observations (e.g. Fasullo and Trenberth, 2012; Su et al., 2014; Qu et al., 2014; Sherwood et al., 2014; Myers and Norris, 2016). Additionally, our ability to evaluate model clouds using satellite data has benefited from the increasing use of satellite simulators. This approach, first introduced by Yu et al. (1996) for use with data from the International Satellite Cloud Climatology Project (ISCCP), attempts to reproduce what a satellite would observe given the model state. Such approaches enable more quantitative comparisons to the satellite record (e.g. Yu et al., 1996; Klein and Jakob, 1999; Webb et al., 2001; Bodas-Salcedo et al., 2008; Cesana and Chepfer, 2013). Much of our improved understanding in these areas would have been impossible without the continuing investment of the scientific community in successive phases of the Coupled Model Intercomparison Project (CMIP) and its co-evolution in more recent years with the Cloud Feedback Model Intercomparison Project (CFMIP).
CFMIP started in 2003 and its first phase (CFMIP-1) organized an
intercomparison based on perpetual July SST forced Cess style
The subsequent objective of CFMIP-2 was to inform improved assessments of
climate change cloud feedbacks by providing better tools to support
evaluation of clouds simulated by climate models and understanding of
cloud–climate feedback processes. CFMIP-2 organized further experiments as
part of CMIP5 (Bony et al., 2011; Taylor et al., 2012), introducing
seasonally varying SST perturbation experiments for the first time, as well
as fixed SST CO
Studies arising from CFMIP-2 include numerous single- and multi-model
evaluation studies which use COSP to make quantitative and fair comparisons
with a range of satellite products (e.g. Kay et al., 2012; Franklin et al.,
2013; Klein et al., 2013; Lin et al., 2014; Chepfer et al., 2014). COSP has
also enabled studies attributing cloud feedbacks and cloud adjustments to
different cloud types (e.g. Zelinka et al., 2013, 2014; Tsushima et al.,
2016). CFMIP-2 additionally enabled the finding that idealized “aquaplanet”
experiments without land, seasonal cycles or Walker circulations are able to
reproduce the essential differences between models' global cloud feedbacks
and cloud adjustments in a substantial ensemble of models (Ringer et al.,
2014; Medeiros et al., 2015). Process outputs from CFMIP have also been used
to develop and test physical mechanisms proposed to explain and constrain
inter-model spread in cloud feedbacks in the CMIP5 models (e.g. Sherwood et
al., 2014; Brient et al., 2015; Webb et al., 2015a; Nuijens et al., 2015a, b;
Dal Gesso at al., 2015). CGILS has demonstrated a consensus in the responses
of LES models to climate forcings and identified shortcomings in the physical
representations of cloud feedbacks in climate models (e.g. Blossey et al.,
2013; Zhang et al., 2013; Dal Gesso at al., 2015). The CFMIP experiments have
additionally formed the basis for coordinated experiments to explore the
impact of cloud-radiative effects on the circulation (Stevens et al., 2012;
Fermepin and Bony, 2014; Crueger and Stevens, 2015; Li et al., 2015; Harrop
and Hartmann, 2016), the impact of parametrized convection on cloud feedback
(Webb et al., 2015b) and the mechanisms of negative shortwave cloud feedback
in mid to high latitudes (Ceppi et al., 2015). Additionally, the CFMIP
experiments have, due to their idealized nature, proven useful in a number of
studies not directly related to clouds, instead analysing the responses of
regional precipitation and circulation patterns to CO
The primary objective of CFMIP is to inform future assessments of cloud
feedbacks through improved understanding of cloud–climate feedback
mechanisms and better evaluation of cloud processes and cloud feedbacks in
climate models. However, the CFMIP approach is also increasingly being used
to understand other aspects of climate change, and so a second objective has
been introduced, to improve understanding of circulation, regional-scale
precipitation, and non-linear changes. This involves bringing climate
modelling, observational and process modelling communities closer together
and providing better tools and community support for evaluation of clouds and
cloud feedbacks simulated by climate models and for understanding of the
mechanisms underlying them. This is achieved by
coordinating model inter-comparison activities which include experimental
design as well as specification of model output diagnostics to support
quantitative evaluation of modelled clouds with observations (e.g. COSP) and
in situ measurements (e.g. cfSites) as well as
process-based investigation of cloud maintenance and feedback mechanisms
(e.g. cfSites, temperature and humidity tendency terms); developing and improving support infrastructure, including COSP, CFMIP-OBS
and the CFMIP Diagnostic Codes Catalogue; and fostering collaboration with the observational and cloud process modelling
communities via annual CFMIP meetings and internationally funded projects.
CFMIP is now entering its third phase, CFMIP-3, which will run in parallel
with the current phase of the Coupled Model Intercomparison Project (CMIP6,
Eyring et al., 2016). This paper documents the CFMIP-3/CMIP6 experiments and
diagnostic outputs which constitute the CFMIP-3 contribution to CMIP6. It is
anticipated that CFMIP-3 will be broader than what is described here, for
instance including studies with process models and informal CFMIP-3
experiments which are organized
independently of CMIP6. Please refer to the CFMIP website for announcements
of these other initiatives and CFMIP annual meetings.
CFMIP-3 touches, to differing degrees, on each of the three questions around
which CMIP6 is organized. With its focus on cloud feedback, CFMIP-3 is
central to CMIP6's attempt to answer the question “How does the Earth system
respond to forcing?”, but as illustrated in the remainder of this document,
CFMIP-3 also offers the opportunity to contribute to the other two guiding
questions of CMIP6. Through its strong model evaluation component, it stands
to help to answer the question “What are the origins and consequences of
systematic model biases?”. CFMIP-3 will also help answer the question “How
can we assess future climate changes given climate variability, climate
predictability, and uncertainties in scenarios?”. For example, the
The CFMIP-3/CMIP6 experiments are outlined below in Sect. 2. Section 3 describes the diagnostics outputs proposed by CFMIP for the CFMIP-3/CMIP6 experiments and other experiments within CMIP. We provide a summary of the CFMIP-3 contribution to CMIP6 in Sect. 4.
Summary of CFMIP-3/CMIP6 Tier 1 experiments.
Summary of CFMIP-3/CMIP6 Tier 2 experiments.
Summary of CFMIP-3/CMIP6 experiments and DECK
The CFMIP-3/CMIP6 experiments are summarized in Fig. 1 and Tables 1 and 2,
and are described in detail below. Most of the CFMIP-3/CMIP6 experiments are
based on CO
Lead coordinator: Mark Webb
Science question: what are the physical mechanisms underlying the range of cloud feedbacks and cloud adjustments predicted by climate models, and which of the cloud responses are the most credible?
Equilibrium climate sensitivity (ECS) can be estimated using an idealized
AOGCM experiment such as the
A more idealized set of fixed SST experiments proposed by CFMIP-2 for CMIP5
(
The CFMIP-2/CMIP5 experiments and diagnostic outputs have thus enabled
considerable progress on a number of questions. However, participation by a
larger fraction of modelling groups is desired in CFMIP-3/CMIP6 to enable a
more comprehensive assessment of the uncertainties across the full
multi-model ensemble. Our proposal is therefore to retain the CFMIP-2/CMIP5
experiments (known in CMIP5 as
The configurations of the
We also propose using the Tier 1 experiments as the foundation for further experiments planned in the context of the Grand Challenge on Clouds, Circulation and Climate Sensitivity (Bony et al., 2015). These will include for example sensitivity experiments to assess the impacts of different physical processes on cloud feedbacks and regional circulation/precipitation responses and also to test specifically proposed cloud feedback mechanisms (e.g. Webb et al., 2015b; Ceppi et al., 2015). Additional experiments further idealizing the aquaplanet framework to a non-rotating rotationally symmetric case are also under development (e.g. Popke et al., 2013). These will be proposed as additional Tier 2 experiments at a future time or coordinated informally outside of CMIP6.
Lead coordinators: Mark Webb and Bjorn Stevens
Science question: are cloud feedbacks consistent for climate cooling and warming, and, if not, why?
There is some evidence to suggest that cloud feedbacks might operate
differently in response to cooling rather than warming. For example,
Yoshimori et al. (2009) found a positive shortwave cloud feedback in a
CO
The configuration of the
Lead coordinators: Sandrine Bony and Bjorn Stevens
Science question: how do cloud-radiative effects impact the structure, the strength and the variability of the general atmospheric circulation in present and future climates?
It is increasingly recognized that clouds, and atmospheric cloud-radiative effects in particular, play a critical role in the general circulation of the atmosphere and its response to global warming or other perturbations: they have been found to modulate the structure, the position and shifts of the ITCZ (e.g. Slingo and Slingo, 1988; Randall et al., 1989; Sherwood et al., 1994; Bergman and Hendon, 2000; Hwang and Frierson, 2013; Fermepin and Bony 2014; Voigt et al., 2014; Loeb et al., 2015), the organization of convection in tropical waves, Madden–Julian oscillations and other forms of convective aggregation (e.g. Lee et al., 2001; Lin et al., 2004; Bony and Emanuel, 2005; Zurovac-Jevtic et al., 2006; Crueger and Stevens, 2015; Muller and Bony, 2015), the extra-tropical circulation and the position of eddy-driven jets (e.g. Ceppi et al., 2012, 2014; Grise and Polvani, 2014; Li et al., 2015; Voigt and Shaw, 2015), and modes of inter-annual to decadal climate variability (e.g. Bellomo et al., 2015; Rädel et al., 2016; Yuan et al., 2016). A better assessment of this role would greatly help to interpret model biases (how much do biases in cloud-radiative properties contribute to biases in the structure of the ITCZ, in the position and strength of the storm tracks, in the lack of intra-seasonal variability, etc.) and to inter-model differences in simulations of the current climate and in climate change projections (especially changes in regional precipitation and extreme events). More generally, a better understanding of how clouds couple to the circulation is expected to improve our ability to answer the four science questions raised by the WCRP Grand Challenge on Clouds, Circulation and Climate Sensitivity (Bony et al., 2015).
These questions provided the scientific motivation for the Clouds On/Off Klima Intercomparison Experiment (COOKIE) project proposed by European consortium EUCLIPSE and CFMIP (Stevens et al., 2012). The COOKIE experiments, which have been run by four to eight climate models (depending on the experiment), switched off the cloud-radiative effects (clouds seen by the radiation code – and the radiation code only – were artificially made transparent) in an atmospheric model forced by prescribed SSTs. By doing so, the atmospheric circulation could feel the lack of cloud-radiative heating within the atmosphere, but the land surface could also feel the lack of cloud shading, which led to changes in land surface temperatures and land–sea contrasts. The change in circulation between On and Off experiments resulted from both effects, obscuring to some degree the mechanisms through which the atmospheric cloud-radiative effects interact with the circulation for given surface boundary conditions. As the longwave cloud-radiative effects are felt mostly within the troposphere (representing most of the net atmospheric cloud-radiative heating), while the shortwave effects are felt mostly at the surface (e.g. L'Ecuyer and McGarragh, 2010; Haynes et al., 2013), we could better isolate the role of tropospheric cloud-radiative effects on the circulation by running atmosphere-only experiments in which clouds are made transparent to radiation only in the longwave. In this configuration, the models will have a shortwave cloud feedback but no longwave cloud feedback. We note that the presence of clouds does affect the shortwave radiative heating of the atmosphere, although this is a much smaller effect than its longwave equivalent (e.g. Pendergrass and Hartmann, 2014).
Therefore we propose in Tier 2 a set of simple experiments similar to the
Lead coordinators: Chris Bretherton, Roger Marchand, and Bjorn Stevens
Science question: how do responses in the climate system due
to changes in solar forcing differ from changes due to CO
While rapid adjustments in clouds and precipitation can easily be separated from conventional feedbacks in SST forced experiments, such a separation in coupled models is complicated by various issues, including the response of the ocean on decadal timescales. A number of studies have examined cloud feedbacks in coupled models subject to a solar forcing, which is generally associated with much smaller global cloud and precipitation adjustment, due to a smaller atmospheric absorption for a given top of atmosphere forcing (e.g. Lambert and Faull, 2007; Andrews et al., 2010), but the regional cloud and precipitation changes have yet to be rigorously investigated across models. Solar forcing also differs from greenhouse forcing through its different fingerprint on the vertical structure of warming (Santer et al., 2013) and small changes in the radiative heating near the tropopause may project measurably on tropospheric climate (e.g. Butler et al., 2010), for instance by influencing the baroclinicity in the upper troposphere and thus the storm tracks (Bony et al., 2015).
A
Lead coordinator: Peter Good
Science question: to what extent is regional-scale climate
change per CO
Recent studies with individual, or a small number of climate models, have found substantial non-linearities in regional-scale precipitation change (Good et al., 2012; Chadwick and Good, 2013) associated with robust physical mechanisms (Chadwick and Good, 2013). Significant non-linearity has also been found in global- and regional-scale warming (e.g. Colman and McAvaney, 2009; Jonko et al., 2013; Good et al., 2015; Meraner et al., 2013) and ocean heat uptake (Bouttes et al., 2015).
To address this science question, we propose two new experiments for Tier 2,
Lead coordinator: Timothy Andrews
Science question: are climate feedbacks during the 20th century different to those acting on long-term climate change?
Recent studies have shown significant time variation in climate feedbacks in
response to CO
The previous CFMIP-2/CMIP5 design was unable to diagnose the time variation
of feedbacks of explicit relevance to the historical period, because this
requires the removal of the time-varying forcing. To address this we propose
an additional experiment called
We also consider the time variation of feedbacks in
Lead coordinators: Robin Chadwick, Hervé Douville and Christopher Skinner
Science questions:
How do regional climate responses (e.g. of precipitation) in a coupled
model
arise from the combination of responses to different aspects of CO Which aspects of forcing/warming are most important for causing inter-model
uncertainty in regional climate projections? Can inter-model differences in regional projections be related to underlying
structural or resolution differences between models through improved process
understanding, and could this help us to constrain the range of regional
projections? What impact do coupled model SST biases have on regional climate
projections?
The CFMIP-2/CMIP5 set of idealized
We propose a new set of seven 30-year atmosphere-only time slice experiments,
and one 36-year amip-style experiment, to decompose the regional responses of
each model's
The experiments are
We also propose an additional amip-based experiment,
The CFMIP-3/CMIP6 specific diagnostic request is designed to address the
following questions.
How well do clouds and other relevant variables
simulated by models agree with observations? What physical processes and
mechanisms are important for a credible simulation of clouds, cloud
feedbacks and cloud adjustments in climate models? Which models have the
most credible representations of processes relevant to the simulation of
clouds? How do clouds and their changes interact with other elements of
the climate system?
The set of diagnostic outputs recommended for CFMIP-3/CMIP6 is based on that
from CFMIP-2/CMIP5, with some modifications. The request outlined below is in
three parts. The first part describes an updated set of CFMIP process
diagnostics (based on those in CFMIP-2/CMIP5 which are documented at
CMIP mandates that for participation in CFMIP-3/CMIP6, modelling groups must commit to performing all of the Tier 1 experiments. In recognition that sufficient resources are not available for all groups to prepare all of the CFMIP-3/CMIP6-specific diagnostics, these diagnostics are considered to be Tier 2, i.e. not compulsory for participation in CFMIP-3/CMIP6. Nonetheless, these diagnostics are extremely valuable and all groups with the capacity to do so are very strongly encouraged to provide the additionally requested CFMIP-3/CMIP6-specific diagnostics.
In the case where CFMIP-3/CMIP6-specific outputs are requested in the DECK and CMIP6 Historical experiments, and modelling groups run more than one ensemble member of an experiment, we request that each set of CFMIP-3/CMIP6-specific outputs be submitted for one ensemble member only. Having different CFMIP variables in different ensemble members is acceptable, but submitting them all in the same ensemble member is preferable. We request that the modelling groups provide information on which CFMIP diagnostic sets are submitted in which ensemble members so that this information can be made available to those who may be analysing the output. Our analysis plans for the CFMIP diagnostic outputs in the CMIP DECK, CMIP6 Historical and CFMIP-3/CMIP6 experiments, including details of the CFMIP Diagnostics Code Catalogue, are summarized in Appendix A.
In CFMIP-2/CMIP5, instantaneous high-frequency “cfSites” outputs were
requested for 120 locations in the
CFMIP-3/CMIP6 cfSites locations. The contours give an indication of inter-model spread in cloud feedback from the CFMIP-2/CMIP5 amip/amip4K experiments (please refer to Webb et al., 2015a, for details).
For CFMIP-3 cfSites outputs are now requested for one ensemble member of the
AMIP DECK experiment, and the
The cfSites outputs from CFMIP-3/CMIP6 provide instantaneous outputs of a
range of quantities (including temperature and humidity tendency terms) in
experiments which can be used to evaluate the present-day relationships of
clouds to cloud controlling factors using in situ measurements, and at the
same time explore how these relationships affect cloud feedbacks and cloud
adjustments. An increasing wealth of observational data with which to
evaluate the models using these outputs is available or in the planning
stage, for example from the Barbados Cloud Observatory (Stevens et al.,
2015), the ARM Program (e.g. Wood et al., 2015; Marchand et al., 2015) or
within the German national project on high-definition clouds and
precipitation for climate-prediction, HD(CP)
CFMIP-2 also requested cloud, temperature and humidity tendency terms from
convection, radiation, dynamics, etc. in the
In CFMIP-3/CMIP6 we have improved the definitions of the temperature and humidity tendency terms, and added some additional terms such as clear-sky radiative heating rates to more precisely quantify the contributions of different processes to the temperature and humidity budget changes underlying cloud feedbacks and adjustments. We have dispensed with the cloud water tendency terms because these have been less widely used than the temperature and humidity tendencies.
A shortcoming of the CMIP5 protocol was that we were unable to interpret the
physical feedback mechanisms in coupled model experiments due to a lack of
process diagnostics. For this reason in CMIP6 we are requesting these budget
terms in the DECK
Clustering approaches (e.g. Jakob and Tselioudis, 2003) are now commonly used for assessing the contributions of different cloud regimes (e.g. stratocumulus, trade cumulus, frontal clouds) to present-day biases in cloud simulations and to inter-model differences in cloud feedbacks (e.g. Williams and Webb, 2009; Tsushima et al., 2013, 2016). We have also added some additional daily 2-D fields to the standard package of CFMIP daily outputs to allow further investigation of feedbacks between clouds and aerosols associated with the changing hydrological cycle (aerosol loadings and cloud top effective radii/number concentrations) and a clearer diagnosis of the roles of convective and stratiform clouds (convective vs. stratiform ice and condensed water paths and cloud top effective radii/number concentrations).
This section motivates and summarizes the COSP outputs requested from the DECK, CMIP6 Historical and CFMIP-3/CMIP6 experiments, as well as a corresponding set of observations.
There is no unique definition of clouds or cloud types, neither in models nor
in observations. Therefore, to compare models with observations, and even to
compare models with each other, it is necessary to use a consistent
definition of clouds between the model and the satellite product in question
(i.e. be “definition-aware”). Further complicating matters: climate model
grid boxes (typically 1
COSP is increasingly being used not only for model intercomparison activities, but also as part of the model development and evaluation process by modelling groups (e.g. Marchand et al., 2009; Zhang et al., 2010; Kay et al., 2012; Franklin et al., 2013; Lacagnina and Selten, 2014; Nam et al., 2014; Williams et al., 2015; Konsta et al., 2015). Many of the standard monthly and daily COSP outputs have been shown to be valuable in the CMIP5 experiments, not only for cloud evaluation, allowing a detailed evaluation of clouds and precipitation, and their interaction with radiation (e.g. Nam et al., 2012; Cesana and Chepfer, 2012; Kay et al., 2012; Klein et al., 2013; Tsushima et al., 2013; Gordon and Klein, 2014; Lin et al., 2014; Bodas-Salcedo et al., 2014; Bellomo and Clement, 2015), but also in quantifying the contributions of different cloud types to cloud feedbacks and forcing adjustments in climate change experiments (e.g. Zelinka et al., 2013, 2014; Chepfer et al., 2014; Tsushima et al., 2016). For a full list of studies that use COSP diagnostics for model evaluation and feedback analysis, please refer to the “CFMIP publications” section of the CFMIP website.
Here we will give only a brief overview of the COSP request; readers
interested in the complete details of the data request are referred to the
Earth System CoG website
(
Within CFMIP-3/CMIP6, COSP output is requested from six simulators as
follows.
ISCCP: pseudo-retrievals of cloud top pressure (CTP) and cloud
optical thickness (tau) (Klein and Jakob, 1999; Webb et al., 2001). CloudSat: a forward model for radar reflectivity as a function
of height (Haynes et al., 2007). CALIPSO (Chepfer et al., 2008; Cesana and Chepfer, 2013):
forward model for the lidar scattering ratio as a function of height and
cloud-phase retrieval. MODIS: pseudo-retrievals of CTP, effective particle size and tau
as a function of phase (Pincus et al., 2012). MISR: pseudo-retrievals of cloud top height (CTH) and tau
(Marchand and Ackerman, 2010). PARASOL: simple forward model of mono-directional reflectance
(Konsta et al., 2015).
The main difference to CFMIP-2 is that output is requested from a greater
number of simulators and longer periods of simulated time. MISR provides more
accurate retrievals of cloud-top height for low-level and mid-level clouds,
and more reliable discrimination of mid-level clouds from other clouds, while
MODIS provides better retrievals of high-level clouds. ISCCP and MISR
histograms can be combined to separate optically thin high-level clouds into
multi-layer and single-layer categories (Marchand et al., 2010). Aerosol
schemes are becoming more complex, with more elaborate representations of
cloud–aerosol interactions. This makes the evaluation of the phase
partitioning an important aspect of model evaluation, and height-resolved
partitioning estimates from the CALIPSO simulator are included in the COSP
request. Cloud phase and particle size estimates from the MODIS simulator
were not available in CFMIP-2, but may prove a useful complement to
investigate cloud–aerosol interactions by virtue of greater geographic
sampling and longer time records. Many of the COSP diagnostics are now
requested for the entire lengths of the DECK, CMIP6 Historical and
CFMIP-3/CMIP6 experiments to support the quantification and interpretation of
cloud feedbacks and cloud adjustments in a broader context. The new inclusion
in this COSP request of a long time series of 3-D cloud fractions will
facilitate the comparison of cloud trends with the observational record
(Chepfer et al., 2014). More details of all the changes with respect to
CFMIP-2/CMIP5 can be found in the proposal of the CMIP6-Endorsed MIPs,
available from the CMIP6 website
(
The COSP output is in six variable groups.
cfMon_sim: monthly means of ISCCP 2-D diagnostics (cloud
fraction, cloud albedo, and cloud top pressure), ISCCP CTP-tau histogram,
and CALIPSO 2-D and 3-D cloud fractions. cfDay_2d: daily means of ISCCP and CALIPSO 2-D diagnostics,
and PARASOL reflectances. cfDay_3d: daily means of ISCCP and CALIPSO 3-D diagnostics. cfMonExtra: monthly means of CloudSat reflectitivity and CALIPSO scattering
ratio histograms as a function of height, CALIPSO 3-D cloud fractions by
phase, MODIS 2-D cloud fractions, MODIS CTP-tau histogram and size-tau
histograms by phase, MISR CTH-tau histograms, and PARASOL reflectances. cfDayExtra: daily means of CALIPSO total cloud fraction, MODIS CTP-tau
histogram and size-tau histograms by phase, and PARASOL reflectances. cf3hrSim: 3-hourly instantaneous diagnostics of ISCCP CTP-tau histograms,
MISR CTH-tau histograms, MODIS CTP-tau histograms and size-tau histograms by
phase, CALIPSO 2-D and 3-D cloud fractions, CloudSat reflectitivity and
CALIPSO scattering ratio histograms as a function of height, and PARASOL
reflectances.
The variable groups cfMon_sim and cfDay_2d are requested for all years in
the
COSP is available via the CFMIP website
(
Developed over the last few years, COSP 2 substantially revises the infrastructure for integrating satellite simulators in climate models. COSP 2 makes many fewer inherent assumptions about the model representation of clouds than do previous versions but contains an optional interface allowing it to be used as a drop-in replacement for COSP 1.4 or COSP 1.4.1. At the time of this writing COSP 2 is undergoing final testing in two climate models. Availability of the final version will be announced on the CFMIP website and modelling groups are free to adopt it for use in CFMIP at that time.
The CFMIP community has developed a set of observational datasets available
via the CFMIP-OBS website
(
Summary of CFMIP-OBS observational datasets available for comparison with COSP diagnostics.
Climate models have difficulties representing the diurnal cycle of convective
clouds over land (Yang and Slingo, 2001; Stratton and Stirling, 2011), but
its evaluation is not possible with sun-synchronous satellites. Geostationary
satellites provide high-frequency sampling that can be used to evaluate model
biases in the diurnal cycle of clouds and radiation (albeit over a limited
area). The Geostationary Earth Radiation Budget instrument (GERB; Harries et
al., 2005) measures the top of atmosphere (TOA) radiation budget from a
geostationary orbit at 0
The primary goal of CFMIP is to inform improved assessments of cloud feedbacks on climate change. This involves bringing climate modelling, observational and process modelling communities closer together and providing better tools and community support for understanding and evaluation of clouds and cloud feedbacks simulated by climate models. CFMIP supports ongoing coordinated model inter-comparison activities by recommending experiments and model output diagnostics for CMIP, designed to support the understanding and evaluation of cloud processes and cloud feedbacks in models. The CFMIP approach is also increasingly being used to understand other aspects of climate change, and so a second objective has now been introduced, to improve understanding of circulation, regional-scale precipitation, and non-linear changes. CFMIP proposes a number of CFMIP-3/CMIP6 experiments and model outputs for CMIP6, building on and extending those which were part of CFMIP-2/CMIP5.
A compact set of CFMIP-3/CMIP6 Tier 1 experiments are proposed to address the
question (1) What are the physical mechanisms underlying the range of cloud
feedbacks and cloud adjustments predicted by climate models, and which models
have the most credible cloud feedbacks? The Tier 1 experiments
(
The CFMIP-3/CMIP6 experiments will continue to include outputs from the CFMIP
Observational Simulator Package (COSP) to support robust scale-aware and
definition-aware evaluation of modelled clouds with observations and to
relate cloud feedbacks to observed quantities. COSP outputs are also proposed
for inclusion in the DECK and CMIP6 Historical experiments. Process
diagnostics, including “cfSites” high-frequency outputs at selected
locations and temperature and humidity budget terms from radiation,
convection, dynamics, etc., are also retained from CFMIP-2/CMIP5. These will
help to address the following questions.
How well do clouds and other relevant
variables simulated by models agree with observations? What physical
processes and mechanisms are important for a credible simulation of clouds,
cloud feedbacks and cloud adjustments in climate models? Which models
have the most credible representations of processes relevant to the
simulation of clouds? How do clouds and their changes interact with other
elements of the climate system?
By continuing the CFMIP-2/CMIP5 experiments and diagnostic outputs within
CFMIP-3/CMIP6 we hope to apply the well-established aspects of the CFMIP
approach to a larger number of climate models. Additionally we have proposed
new CFMIP-3/CMIP6 experiments to investigate a broader range of questions
relating to the Grand Challenge on Clouds, Circulation and Climate
Sensitivity. We hope that the modelling community will participate fully in
CFMIP-3 via CMIP6 so as to maximize the relevance of our findings to future
assessments of climate change.
COSP is published under an open-source license via GitHub (please see the
CFMIP website for details). The model output from the DECK, CMIP6 Historical
and CFMIP-3/CMIP6 simulations described in this paper will be distributed
through the Earth System Grid Federation (ESGF) with digital object
identifiers (DOIs) assigned. As in CMIP5, the model output will be freely
accessible through data portals after registration. In order to document
CMIP6's scientific impact and enable ongoing support of CMIP, users are
obligated to acknowledge CMIP6, the participating modelling groups, and the
ESGF centres (see details on the CMIP Panel website at
CFMIP-2 analysis activities are ongoing and the CFMIP community is ready to analyse CFMIP-3/CMIP6 data at any time. We would like modelling groups to perform the proposed CFMIP-3/CMIP6 experiments at the same time or shortly after their DECK and CMIP6 Historical experiments. Subsequent informally organized CFMIP-3 experiments which are not included in CMIP6 will build on the proposed DECK and CFMIP-3/CMIP6 experiments and some will start as soon as CMIP6 DECK experiments start to become available. We envisage a succession of CFMIP-related intercomparisons addressing different questions arising from the Grand Challenge spanning the duration of CMIP6.
We plan to scientifically analyse, evaluate and exploit the proposed experiments and diagnostic outputs, and have identified lead coordinators within CFMIP for different aspects of this activity. The lead coordinators are responsible for encouraging analysis of the relevant experiments as broadly as possible across the scientific community. While they may lead some analysis themselves, they do not have any first claim on analysing or publishing the results. All interested investigators are encouraged to exploit the data from these experiments. While investigators may wish to liaise with the lead coordinators to avoid duplicating work that others are doing, this is not a requirement. An overview of the proposed evaluation/analysis of the CMIP DECK, CMIP6 Historical and CFMIP-3/CMIP6 experiments follows.
CFMIP will continue to exploit the CMIP DECK and CMIP6 experiments to
understand and evaluate cloud processes and cloud feedbacks in climate
models. The wide range of analysis activities described above in the context
of CFMIP-2 will be continued in CFMIP-3 using the CMIP DECK and CFMIP-3/CMIP6
experiments, allowing the techniques developed in CFMIP-2 to be applied to an
expanding number of models, including the new generation of models currently
under development. These activities will include evaluation of clouds using
additional simulators, investigation of cloud processes and cloud
feedback/adjustment mechanisms using process outputs (cfSites, tendency
terms, etc.). The inclusion of COSP and budget tendency terms in additional
DECK experiments (e.g.
Analysis of the
Analysis of non-linear climate processes is discussed in detail by Good et
al. (2016). This includes a method for validating traceability of abrupt
CO
Analysis of
An overview analysis of regional responses and model uncertainty in the piSST set of experiments will be carried out by the coordinators, in collaboration with members of contributing modelling groups. We anticipate that further detailed analysis on the processes at work in different regions will be carried out by a variety of research groups with interest and expertise in a particular region: for example, a set of similar experiments has previously been used to examine the climate response of the West African monsoon in CCSM3 (Skinner et al., 2012). The piSST set of experiments has already been successfully run using the Met Office, NCAR and CNRM CMIP5 models. Lead coordinators: Robin Chadwick, Hervé Douville and Christopher Skinner.
The analysis of the COOKIE experiments will be reviewed by the coordinators in collaboration with members of the contributing modelling groups. The role of longwave atmospheric cloud-radiative effects in large-scale circulations, regional precipitation patterns and the organization of tropical convection will be investigated in the current climate and in climate change, with the aim of highlighting both robust effects and sources of uncertainties in the model responses. Lead coordinators: Sandrine Bony and Bjorn Stevens.
When analysed together with the
The COSP data request for the AMIP DECK experiment will allow a comprehensive
multi-model evaluation of clouds and radiation, following on from CMIP5
studies (e.g. Klein et al., 2013; Bodas-Salcedo et al., 2014). The COSP data
request for the other experiments (e.g.
Analysis of output from the CFMIP-3/CMIP6 and CMIP DECK experiments will also
be facilitated by sharing of diagnostic codes via the CFMIP Diagnostics Code
Catalogue (accessible via the CFMIP website:
Aquaplanets are Earth-like planets with completely water-covered surfaces. They are often used as idealized configurations of atmospheric GCMs, and in this context the usual convention is that land masses and topography are removed. Although many flavours of aquaplanet configurations exist, another convention is to retain as much of the atmospheric model's formulation as possible. That is, the numerical grid, dynamical core, and parameterized physics are all used just as in realistic climate simulations.
The Tier 1 aquaplanet experiments follow the same experimental design as
CFMIP-2/CMIP5 (Medeiros et al., 2015). Those, in turn, were closely related
to previous aquaplanet descriptions. In particular, the control
configuration closely follows the AquaPlanet Experiment protocol (Blackburn
and Hoskins, 2013) using a prescribed SST pattern described by Neale and
Hoskins (2000). Two additional runs paralleled the CFMIP-2/CMIP5
Here we provide the detailed experimental protocol for the three aquaplanet simulations that are part of Tier 1. We note again that these follow the APE protocol and CFMIP-2/CMIP5, and therefore largely mirror previous descriptions in Blackburn and Hoskins (2013), Williamson et al. (2012), and Medeiros et al. (2015).
Orbital parameters are set to perpetual equinox conditions. This is usually
achieved by setting eccentricity and obliquity to zero to define a circular
orbit and insolation independent of calendar. The diurnal cycle is retained.
Insolation is based on a non-varying solar constant of 1365 W m
The SST is non-varying and zonally uniform. The longitudinal variation is
specified using the “Qobs” SST pattern from Neale and Hoskins (2000), given
by
Because results are sensitive to the specification of the SSTs, groups that
use a prognostic equation for the surface skin temperature are asked to set
this skin temperature to the specified SST. No sea ice is prescribed, so the
surface temperature is spatially uniform at 0
Radiatively active trace gases are well mixed, with mixing ratios following
the AMIP II recommendations: CO
Aerosols are removed to the extent possible to remove aerosol–radiation
interaction (aka direct effects) and aerosol–cloud interaction (aka indirect
effects). No external surface emissions are to be prescribed. Models
requiring aerosol for cloud condensation should use a constant oceanic
climatology that is symmetric about the Equator and zonally. Alternatively,
models with the capability should set the cloud droplet and crystal numbers
to
As in APE, it is recommended that the atmospheric dry mass be adjusted to yield a global mean of 101 080 Pa. It is also recommended to adopt the APE recommended values for geophysical constants, as listed in Table 2 of Williamson et al. (2012).
The
The
Model runs should be 10 years. We recommend discarding the initial spin-up period of a few months.
The
We are grateful to Florent Brient, Hideo Shiogama, Aiko Voigt, Mark Ringer
and two anonymous referees for helpful comments on the manuscript. We thank
the modelling groups and the wider CFMIP community for reviewing and
supporting the CFMIP contribution to CMIP6, the CMIP Panel for their
coordination of CMIP6, the WGCM Infrastructure Panel (WIP) overseeing the
CMIP6 infrastructure, and Martin Juckes for taking the lead in preparing the
CMIP6 data request. We are also grateful to Robert Pincus and Yuying Zhang
for their contributions to COSP and to CFMIP-OBS, to Dustin Swales for his
development work for COSP-2, and to Gregory Cesana and Mathieu Reverdy for
their contributions to CFMIP-OBS. We are grateful to Brian Soden for
producing the CMIP3 composite pattern dataset used for the CMIP5