The Climateprediction.net program harnesses over 600,000 volunteers and their computers to conduct large ensemble climate simulations through the BOINC distributed computing platform. It has run over 130 million model years across more than 25 subprojects exploring uncertainties in climate predictions and extreme weather attribution. Current work includes super-ensembles examining stratosphere-troposphere coupling and mid-latitude dynamics, as well as the HAPPI project simulating 1.5°C and 2°C warming scenarios consistent with the goals of the Paris Agreement.
4. Our laboratory: the world’s largest climate
modelling facility
13 years, >25 sub projects, >600,000 volunteers, >130M model-years
5. Unlimited ensemble size: exploring
uncertainties in climate predictions
Results of the BBC Climate Change Experiment:
Rowlands et al, Nature Geosci., 2012
6. Berkley Open Infrastructure for
Network Computing (BOINC)
• Developed at UCB Space Science Laboratory by the SETI@home group
• Public-resource rather than grid computing
• Goals of BOINC
– Reduce the barriers of entry to public-resource computing
– Share resources among autonomous projects
– Support diverse applications
– Reward participation
• Offers tools for
– Creating, starting, stopping and querying projects
– Adding new applications, new platforms, …
– Creating workunits/tasks
– Monitoring server performance
7. Project ScientistCitizen Scientist
Web server
DB server
Up/download
serverClimate models
Experiment
Results
Communication
Papers
Volunteer Distributed Computing
Very large ensembles of simulations can be generated by using this framework.
8.
9. weather@home regional climate models
High impact weather
events are typically rare
and unpredictable.
They also involve small
scales.
Resolution provided by
nested regional model.
10. Creating work for CPDN/W@H
• External collaborators fully supported by CPDN team
• Work is sent out in batches
– Allows easier management and attribution of data usage
– A batch contains an arbitrary number of workunits
– A batch is defined within an XML file containing details on all contained workunits, final data
destination etc
• Application distributed seperatly outside of workunit structure to participating
volunteer systems
• An individual work unit is;
– Namelist & ancillary files
• Results
– Results returned on a predetermined sub-schedule (e.g. Monthly within an 1 year model.
– ‘Start dump’ also returned to allow restart from this time point
• Chain together short WU to create longer running model which wouldn’t be possible to run individually.
• New: investigating publishing results using Nature Scientific Data to make
unique resources openly available
11. Models
• All models part of Hadley CM3 family (currently)
• Climateprediction.net
– HadCM3
• Weather@Home
– HadAM3P
• Global Atmosphere only model with prescribed SST
• Mainly used as driver of regional model but capable of individual
operation
– HadRM3P
• Regional Climate Model with flexible user defined region of interest
• MOSES1 in W@H1, MOSES2 + TRIFFID in W@H2
12. Super-ensemble projects linked to
atmospheric dynamics
1. Strat-trop coupling to extremes
2. Mid-latitude dynamics (DOCILE)
1. Dynamics of extremes
3. The Paris Agreement on Climate Change
(HAPPI)
13. 1. Strat-trop coupling to extremes
Caption: NAM signal in reanalysis
(Mitchell et al, Jclim. 2013).
1. Similar evolution is seen in
intermediate GCM
(O’Callaghan et al, GRL, 2015)
2. Similar evolution in full GCMs
(Seviour et al, JGR, 2015)
What sample sizes are
needed to study
extremes?
• Central theme: To understand how different
stratospheric variability leads to extreme
events at the surface.
14. 2. Mid-latitude extremes (DOCILE)
• Central theme: To increase reliability of
attribution statements by considering more
realistic dynamics.
• 2003 case study
(Mitchell et al, ERL, 2016)
15. 2. Mid-latitude
extremes
(DOCILE)
Wave resonance at
mid-latitudes?
Normal: July 1980 Extreme: May 2013
(Huntingford et al, in prep)
Data source: PIK/Stefan Rahmstorf; Illustration: Focus 2013,Nr. 28/13
16. “Holding the increase in the global average
temperature to well below 2 °C above pre-
industrial levels and to pursue efforts to limit
the temperature increase to 1.5 °C above pre-
industrial levels…”
3. The Paris Agreement on Climate
Change (HAPPI)
17. 3. The Paris Agreement on Climate
Change (HAPPI)
(Mitchell et al, NCC, 2016)
18. If you are keen to get involved;
Websites:
www.happimip.org
www.climateprediction.net
20. Current status of HAPPI data
Model name Historical
decade
(ems)
1.5 decade
(ems)
2 decade
(ems)
Validation (for bias
correction)
CAM4 testing testing testing
CAM5.1-1d testing testing testing
CAM5.1-0.25d 2 2 N/A 1995-2005 (1 em)
CanAM4
HadAM3P 1000 200 2 1985-2015 (50 ems)
HadGEM3 15 testing N/A 1960-2015 (15 ems)
MIROC 100 100 testing 1950-2015 (10 ems)
NorESM_happi 100 testing testing
ems = ensemble members
Red = still running additional ems
Blue = not expecting to run any more
Editor's Notes
Combines the resources of personal computers and game consoles belonging to the general public to perform scientific computations
started in 1999 and still runs today
Currently supporting ~40 projects in 7 areas
12 regions [3@ 25km, 9 @ 50km resolution]
Flooding – UK, central Europe
Drought California
Heatwave – Australia
Health – UK
Biosphere – Bark Beetle forest destruction – USA/Canada
Largest ever single batch had ~196k members for schaller et al
Of recent submissions ~ 43k for NERC MaRIUS project
HadCM3
Atmosphere: 2.5 x 3.75 degrees lat-lon resolution, 19 vertical levels (known as N48; comparable resolution to ~T42)., 30 mins timestep for dynamics, 3 hours for radiative transfer.
Ocean: 1.25 x 1.25 degrees lat-lon resolution, 20 vertical levels, 1 hour timestep.
Sulphur and carbon cycles, dynamic vegetation etc. are optional.
HadAM3P
N96 resolution
(1.25 x 1.875 degrees resolution, 19 levels) with 15 minutes timestep for dynamics and improved physics.
HadRM3P
Either 50km or 25km horizontal resolution