HighResMIP
•
•
•
•
Rein Haarsma KNMI (lead)
Malcolm Roberts Met Office (co-lead)
Important weather and climate processes emerge at
sub-50km resolution
They contribute significantly to both large-scale
circulation and local impacts, hence vital for
understanding and constraining regional variability
How robust are these effects?
Is there any convergence with resolution across
models?
Need coordinated, simplified experimental design
to find out
http://www.wcrp-climate.org/index.php/modelling-wgcm-mipcatalogue/429-wgcm-hiresmip
Experimental protocol:
Global models – AMIP-style and coupled
Physical climate system only
Integrations: 1950-2050
Ensemble size: >=1 (ideally 3)
Resolutions: <25km HI and ~60-100km STD
Aerosol concentrations specified
e.g. Zhao et al, 2009; Haarsma et al, 2013; Demory et al, 2013
Global drivers
Regional
variability
Feedbacks
to large
scale
Local processes
Impacts, extremes
Malcolm Roberts, Met Office (coordinator)
Pier Luigi Vidale, Univ. of Reading (scientific coordinator)
PRocess-based climate sIMulation: AdVances in high resolution
modelling and European climate Risk Assessment
Goal:
•
to deliver novel, advanced and well-evaluated high-resolution global climate
models (GCMs), capable of simulating and projecting regional climate with
unprecedented fidelity, out to 2050.
To deliver:
•
innovative climate science and a new generation of European advanced GCMs.
•
improve understanding of the drivers of variability and change in European
climate, including extremes, which continue to be characterised by high
uncertainty
•
new climate information that is tailored, actionable and strengthens societal risk
management decisions with sector-specific end-users
•
new insights into climate processes using eddy-resolving ocean and explicit
convection atmosphere models
To run for 4 years from Nov 2015 including 19 partners across Europe, funded by the
Horizon 2020 call SC5-1-2014 - Advanced Earth System Models
proj.badc.rl.ac.uk/primavera
Core integrations in PRIMAVERA will form much of the European contribution to CMIP6
HighResMIP
http://www.wcrp-climate.org/index.php/modelling-wgcm-mip-catalogue/429-wgcm-hiresmip
Modelling groups expressing interest in HighResMIP
(at least for Tier 1 simulations)
Country
Group
Model
China
BCC
BCC-CSM2-HR
Brazil
INPE
BESM
China
Chinese Academy of Meteorological Sciences
CAMS-CSM
China
Institute of Atmospheric Physics, Chinese Academy of Sciences
FGOALS
USA
NCAR
CESM
China
Center for Earth System Science/Tsinghua University
CESS/THU
Italy
Centro Euro-Mediterraneo sui Cambiamenti Climatici
CMCC
France
CNRM-CERFACS
CNRM
Europe
EC-Earth consortium (11 groups)
EC-Earth
USA
GFDL
GFDL
Russia
Institute of Numerical Mathematics
INM
Japan
AORI, University of Tokyo / JAMSTEC / National Institute for Environmental Studies
MIROC6-CGCM
Japan
AORI, University of Tokyo / JAMSTEC / National Institute for Environmental Studies
NICAM
Germany
Max Planck Institute for Meteorology (MPI-M)
MPI-ESM
Japan
Meteorological Research Institute
MRI-AGCM3.xS
UK
Met Office
UKESM
/HadGEM3
European HighResMIP resolutions
(as part of PRIMAVERA)
Institution
Model names
Atmosph.
Res., core
Oceanic
Res., core
Oceanic
Res.,
Frontiers
MO
NCAS
MetUM
NEMO
60-25km
KNMI IC3
SMHI CNR
ECEarth
NEMO
T255-799
CERFACS
MPI
AWI
CMCC
ECMWF
Arpege
NEMO
T127-359
ECHAM
MPIOM
T63-255
ECHAM
FESOM
T63-255
CCESM
NEMO
100-25km
IFS
NEMO
T319-799
¼o
¼o
¼
0.4-¼o
¼
¼
1/12˚
1/12˚
1-¼
spatially
variable
1/10˚
Spatially
variable
1/10˚
• Example resolutions which will definitely contribute to HighResMIP
• Concentrate on horizontal resolution – keep vertical resolution the same
• ECMWF is an NWP partner – present day simulations, extra ensemble
members, and including very high resolution atmosphere simulations – we
welcome other NWP centres’ submissions
Anticipated outcomes from this
meeting
• Agree to as much of the protocol as possible
• Suggested way forward for any other aspects
• Complete email contact list to participate in
GMD HighResMIP protocol manuscript (open
April – December 2015)
• PRIMAVERA starts 1 Nov – need protocol in
place (and ideally tested) by then.
HighResMIP model configurations (1)
• Parallel standard and high resolution integrations
– STD likely to be default CMIP5/CMIP6-DECK resolution
(~100km atmosphere resolution)
– Hence DECK is benchmark for STD, HI is sensitivity test
– HI being ~25km atmosphere resolution
• Strongly encourage absolutely minimal differences
between STD and HI configurations
– Vital part of HighResMIP is to look for systematic differences
with model resolution across multi-model ensemble
– If extra tuning is made between different resolutions, it will
make it extremely hard to pick apart the causes of differences
– NOT a beauty contest to have the perfect HI model, we are
most interested in the delta between resolutions
• Similarly vertical resolution should be the same in STD and
HI
HighResMIP model configurations (2)
• Diagnostics list
– Initial submission to CMIP6 for HighResMIP diagnostic list (having
attempted to consult widely on required diagnostics)
– Considerable number of extra high frequency (1h, 3h, 6h) diagnostics
compared to CMIP5 list, to look at extreme processes in particular
– Balance between scientific interest and manageable data volumes
• Integration protocol (STD + HI)
– Physical climate system only, minimal Earth System components
– Tier 1: simple 1950-2050 forced-atmosphere integrations using SST and
sea-ice datasets to be provided
– Tier 2: coupled simulations 1950-2050
– Tier 3: extension of forced integration to 2100
– Tier 4: aquaplanet
– Tier 5: switch-on 4xCO2
– Further discussions on more detailed aspects of protocol are ongoing, and
will be published in GMD Special Issue later in 2015
– Will also be published on HighResMIP wiki, together with helpful scripts
for producing forcing datasets
– Contacts: haarsma@knmi.nl, malcolm.roberts@metoffice.gov.uk
HighResMIP protocol choices (1)
• 1950-2050 time period
– Positives:
• Longer period than typical 1979-2008 AMIP period (overlap with
CMIP6 DECK period will give extra ensemble member at low
resolution)
• Sample different periods of variability from the past, influence
of phases of global modes of variability
• Enables examination of decadal variability of different signs in
tropical cyclones, Sahel rainfall etc
• Idealised climate change signal several decades into future
– Negatives
• More expensive
• Fewer ensemble members possible
• Methodology for future period will be new, and not so
interesting to some, e.g. NWP centres
• signal to noise may be small
HighResMIP protocol choices (2)
• All models use the same aerosol concentrations
timeseries (provided by e.g. RFMIP)
– Positives:
• Multi-model comparison concentrates on physical resolution
impact
• Cleaner comparison than e.g. CMIP5
• Influences on e.g. decadal variability the same
– Negatives
• Complicated with different aerosol schemes in different
models
• Providing aerosol datasets reliant on other projects, e.g.
RFMIP
• Models may need retuning from their standard aerosol
climatology (though this may have been tuned at low
resolution anyway)
HighResMIP protocol choices (3)
• No retuning of model between low and high
resolution
– Positives:
• Enables assessment of impact of horizontal resolution
alone
• Hence essential for the basic outcome of HighResMIP
• High resolution expensive to tune anyway
– Negatives
• Higher resolution may not be ideal configuration
• May be issues with TOA radiation
– Document all changes made between resolutions
– If insist, then show sensitivity studies so can
understand what is resolution and what is tuning
HighResMIP protocol choices (4)
• Diagnostics and data output at high frequency
(more at 6hr, 3hr and 1hr) as well as columnintegrated
– Positives:
• Significant new analysis becomes possible, including
extreme processes and events (polar lows, Medicanes,
intense precipitation, atmospheric rivers etc)
• Encourages central analysis platform and sharing common
analysis tools (rather than downloading huge volumes of
data to local space)
• Makes best use of HPC usage (run once, analyse many)
– Negatives
• Significant storage requirements (~400GB per model year at
25km resolution for priority 1 data)
• More difficult to share data – no longer possible to store all
diagnostics on CMIP5 data node (use of shared JASMIN
platform will help this for coordinated analysis
SST and sea-ice protocol for forced-atmosphere
(current plan)
• 1950 - present day
– Based on HadISST
– HadISST machinery can produce daily, 1/4˚ fields
– Use fixed SST and sea-ice (rather than slab-ocean where model
comparison will be harder)
• Present day – 2050 (2100)
– SST
• Find period in past with similar phase of major modes of variability
• Stitch this to the present day
• Use trend in HadISST into the future (or use CMIP5 models for future
trend)
– Sea-ice
• use Mizuta et al methodology to scale present-day sea-ice shape into
future based on global SST change
• KNMI methodology
Coupled model protocol
(current plan)
• 1950 ocean initial condition
– EN4 from observed profiles and analyses in 1950
– Present day ocean (more observations but longer spinup)
• Coupled spin-up strategy
– Start from 1950 state, run with 1950’s atmosphere
forcing until TOA at acceptable level (both high and low
resolution ocean) - ~decades
– Start from present day, run with 1950’s atmosphere
until TOA acceptable (both high and low resolution
ocean) - ~decades
– Use lower resolution model for spin-up and interpolate
final ocean state to higher resolution – either ~decades
or use CMIP6 DECK (will lock-in biases from low
resolution)
Q&A
Joint Weather and Climate
Research Programme
A partnership in climate research