Cpdnboinc dev

CPDN dev
Project
Status	Active
Category	Development
Compute	CPU
Requires	None
Development
Developer	University of Oxford — Oxford e-Research Centre
Initial release	December 9, 2003  (23 years ago)
Completed	No
Software
Operating system	Windows, Linux, macOS
Metadata
Website	https://dev.cpdn.org/
License	Proprietary

cpdnboinc dev is the development instance of climateprediction.net (CPDN), a volunteer distributed computing project dedicated to investigating and reducing uncertainties in climate modelling. The development server hosts test work-units and experimental configurations before they are deployed on the main production BOINC server at main.cpdn.org. The production project is one of the largest and longest-running climate-science experiments ever conducted, harnessing the idle processing power of hundreds of thousands of personal computers around the world to run climate simulations that would otherwise be impossible on any single supercomputer.

Why cpdnboinc dev?

Climate modelling is one of science's grand computational challenges. The global climate system involves the atmosphere, oceans, land surface, ice, and the biosphere — all interacting across vastly different scales of time and space. Running a single high-resolution simulation of even one century of Earth's climate can take weeks on a supercomputer.

The central problem is uncertainty: climate models contain dozens of physical parameters — such as how clouds form, how aerosols scatter sunlight, or how ocean eddies mix heat — that cannot be measured perfectly. To understand how sensitive the climate is to these unknowns, scientists must run the same model thousands of times, each time with slightly different parameter settings. This technique is called a perturbed-physics ensemble (PPE), and the sheer volume of computing required puts it far beyond the reach of any single institution's hardware.

Volunteer computing solves this problem elegantly: by distributing individual model runs to tens of thousands of home computers simultaneously, CPDN generates ensemble sizes that dwarf anything previously possible. The IPCC itself identified "long-term ensemble simulations using complex models" as a high-priority research need as far back as 2001.

Goal

The overarching goal of climateprediction.net — and by extension its development server cpdnboinc dev — is to quantify and reduce the uncertainties in 21st-century climate projections, giving policymakers and scientists a firmer scientific foundation for understanding climate change.

More specifically, the project aims to:

• Investigate parameterisation uncertainties in state-of-the-art climate models by running the model thousands of times with slight, physically plausible perturbations to atmospheric, oceanic, and sulphur-cycle physics.
• Determine the range of possible climate sensitivities — how much global temperature rises when atmospheric CO₂ doubles — and assess which parameter combinations best reproduce observed past climate (known as hindcasting). Models that faithfully reproduce the historical climate record are weighted more heavily in probabilistic projections.
• Extend to regional scales through the weather@home sub-project (launched 2010), which nests a high-resolution regional climate model inside the global driver model to examine changes in extreme weather events at the local level.
• Support event-attribution science: assessing how much human-caused greenhouse-gas emissions have changed the probability of specific extreme weather events (e.g., floods, droughts, heatwaves).
• Expand into new application domains, including energy-system planning, infrastructure resilience, and health impacts of climate change, through dedicated study types offered to external researchers.

In summary, CPDN has produced over 100 million model-years of simulation data — more than any other climate modelling project in history — and has registered users in over 220 countries.

Methods

Global carbon monoxide measurements from NASA's Atmospheric Infrared Sounder

Why BOINC?

The BOINC logo. BOINC was adopted by CPDN in August 2004.

The project originally ran on bespoke "classic" software at its September 2003 launch. On 26 August 2004, CPDN migrated to BOINC (Berkeley Open Infrastructure for Network Computing), the open-source volunteer-computing middleware developed at the University of California, Berkeley by David Anderson and the SETI@home team.

BOINC offered several decisive advantages:

• Cross-platform support — the BOINC client runs on Windows, Linux, and macOS, immediately broadening the pool of potential volunteers beyond the Windows-only classic client.
• Multi-project flexibility — volunteers can run CPDN alongside other BOINC projects (e.g., Einstein@home, Rosetta@home) without having to choose between them.
• Screensaver visualisation — the BOINC client can display an animated visualisation of the running climate model as a screensaver, making participation visible and engaging.
• Robust security and credit accounting — BOINC provides redundant result validation, ensuring that volunteer-computed results can be trusted before they enter the scientific database.

Oxford's contribution to BOINC development during the CPDN project was itself significant enough that the University of Oxford's Department of Computer Science cites it as a major research-impact case study. BOINC is now recognised as the key open-source resource for volunteer computing globally.

How the models work

Each volunteer's computer receives a self-contained climate model — currently a version of the Met Office Hadley Centre climate model (HadSM3 / HadCM3) — pre-configured with a specific set of perturbed physics parameters. The model runs in the background using only idle CPU time, and periodically "trickles" progress reports back to the server. When complete, the full result is uploaded automatically.

The suite of model types that CPDN has distributed over its history (in chronological order) includes:

Model	Notes
Classic Slab Model	The original 2003 experiment; equilibrium response to doubled CO₂, atmosphere and static "slab" ocean. Windows only.
BOINC Slab Model	The same experiment re-issued under BOINC from August 2004.
ThermoHaline Circulation (THC) Model	Launched May 2004; investigates climate response to a 50% slowdown of the North Atlantic thermohaline circulation. Now closed to new participants.
Sulfur Cycle Model	Launched August 2005; models the effect of sulfate aerosols and dimethyl sulfide on climate. A five-phase, 75 model-year simulation.
Coupled Spin-Up Model	Internal preparation run (not publicly released); added dynamic ocean physics needed for the coupled model.
Transient Coupled Model (BBC Climate Change Experiment)	Launched February 2006; 80-year hindcast (1920–2000) plus 80-year forecast (2000–2080) with dynamic ocean. The most realistic model released to that date.
Seasonal Attribution Project	High-resolution single-year runs using HadAM3-N144; at least 1.5 GB RAM required. Studies extreme precipitation events.
weather@home	Launched November 2010; pairs a global driver model (HadAM3P at N96 resolution) with the Met Office regional model (HadRM3P at ~50 km resolution) to study regional weather extremes.

The development server cpdnboinc dev hosts experimental versions of these model configurations and new application types before they are promoted to the main project. It allows the team to perform integration testing and validation with a smaller pool of technically experienced volunteers.

Ensemble size and scale

By November 2005 alone, the project had completed over 135,000 individual model runs across all experiment types, representing more than 6 million model-years processed. As of June 2016, the project reported approximately 55 teraflops of processing power from more than 12,000 active participants across 223 countries. The total registered user base surpassed 700,000 by 2013.

Key scientific milestone: first results in Nature (2005)

In January 2005, the first scientific results from the CPDN experiment were published in the journal Nature. The landmark paper — Stainforth et al. (2005), "Uncertainty in predictions of the climate response to rising levels of greenhouse gases" (Nature 433, 403–406, doi:10.1038/nature03301) — analysed over 2,000 model runs and showed that, with only physically plausible changes to parameters, climate sensitivity ranged from less than 2 °C to more than 11 °C in response to a doubling of CO₂. This was the first time that a full general circulation model (GCM) had produced such extreme sensitivities, and the result attracted worldwide media attention.

History

Climate change deniers cherry picking time periods

Origins (1997–2002)

The seeds of climateprediction.net were planted in 1997, when Myles Allen — then at the University of Oxford — recognised the scientific need for very large climate-model ensembles. He was introduced to the success of SETI@home in 1999 and was inspired by the potential of volunteer computing. His first funding proposal, submitted in April 1999, was rejected as "utterly unrealistic."

Following a presentation at the World Climate Conference in Hamburg and the publication of his landmark commentary article Do-it-yourself climate prediction in Nature (October 1999), thousands of would-be volunteers signed up — though the software was not yet ready. The dot-com bubble bursting forced the team to develop most of the infrastructure themselves rather than outsourcing it.

The project was initially named Casino-21 — a reference both to Monte Carlo simulations and to 21st-century climate. It was renamed climateprediction.com and then refined to climateprediction.net to make clear it was not a commercial enterprise. By 2002, funding from the Natural Environment Research Council (NERC) and the UK Department of Trade and Industry enabled the team to grow significantly, drawing in expertise from the Open University, the Knowledge Media Institute (KMi), and Oxford's Computing Laboratory.

Launch and rapid growth (2003–2006)

The full public launch occurred on 12 September 2003. Within 24 hours the project attracted 25,000 registered users, and by the following day it had exceeded the processing capacity of Japan's Earth Simulator — then the world's most powerful supercomputer — to become the world's largest climate modelling facility.

In May 2004, a thermohaline circulation slowdown experiment was launched to coincide with the release of the film The Day After Tomorrow. Three months later, on 26 August 2004, CPDN migrated to BOINC, opening participation to Mac and Linux users for the first time.

In February 2006, the BBC Climate Change Experiment (also known as the transient coupled model) was launched in conjunction with the BBC's climate change season. It attracted approximately 23,000 participants on the first day and eventually recruited around 300,000 new volunteers. It was officially declared complete on 8 March 2009.

weather@home and regional modelling (2010–present)

In November 2010, a new experiment called weather@home was launched in collaboration with the Guardian newspaper and with support from Met Office colleagues Dr. Richard Jones and Dr. Simon Wilson. For the first time, CPDN ran a nested regional climate model (the Met Office's PRECIS/HadRM3P regional model) driven by a global driver model (HadAM3P), allowing scientists to study changes in extreme weather at much finer spatial resolution (~50 km).

The first weather@home results, focusing on the European and western US regions, were published in a special issue of the Bulletin of the American Meteorological Society. The framework has since been extended to cover most of the globe, with regional configurations for Africa, North America, South Asia, Australia–New Zealand, Mexico, and more.

Project team / Sponsors

John Speed's map of Oxford. University of Oxford, the primary institution hosting CPDN.

climateprediction.net / CPDN is operated primarily by the University of Oxford, hosted within:

• Oxford e-Research Centre (Department of Engineering Science)
• Atmospheric, Oceanic and Planetary Physics (AOPP)
• Environmental Change Institute (ECI)

A team of approximately 13 climate scientists, computing experts, and graduate students runs the day-to-day operations, alongside international partners and collaborators.

Key personnel

• Prof. Myles Allen — Founder of climateprediction.net. He wrote the seminal 1999 Nature commentary Do-it-yourself climate prediction that sparked the project's creation and has led the team's scientific direction ever since.
• Dr. David Stainforth — Led the 2005 Nature publication and the early perturbed-physics ensemble design.
• Dr. Simon Sparrow — Current senior scientific computing specialist involved in ongoing ensemble design and publications.
• Dr. Dave Wallom — Oxford e-Research Centre; central to the computing infrastructure.

Funders and partners

Historic and ongoing funding and partnerships have included:

• Natural Environment Research Council (NERC) — primary UK research funder since the early 2000s
• UK Department of Trade and Industry (now the Department for Business, Energy and Industrial Strategy)
• BBC — co-launched the 2006 BBC Climate Change Experiment
• Met Office Hadley Centre — provided the base climate models (HadSM3, HadCM3, HadAM3P, HadRM3P)
• Open University / Knowledge Media Institute (KMi) — contributed to early project development
• Oxford University Computing Laboratory (ComLab)
• Carnegie Mellon University — collaboration for the 2008 geoengineering experiment (Kate Ricke and Granger Morgan)
• The Guardian — media partner for weather@home launch (2010)
• British Council — sponsor of the weather@home Mexico / RECLIM-UK project

Scientific publications

CPDN has produced a substantial body of peer-reviewed literature. A full list is maintained at climateprediction.net/publications and results are also indexed on the BOINC publications list. Selected landmark papers include:

Foundational papers

• Allen, M.R..(1999}).Do-it-yourself climate prediction. Nature. pp. 627. DOI: 10.1038/44176. — The commentary article that launched the project concept.

• Stainforth, D.A..(2005}).Uncertainty in predictions of the climate response to rising levels of greenhouse gases. Nature. pp. 403–406. DOI: 10.1038/nature03301. — First major results from over 2,000 CPDN model runs, showing climate sensitivity ranging 2–11 °C.

• Murphy, J.M..(2004}).Quantification of modelling uncertainties in a large ensemble of climate change simulations. Nature. pp. 768–772. DOI: 10.1038/nature02771.

Event attribution and regional studies

• Pall, P..(2011}).Anthropogenic greenhouse gas contribution to flood risk in England and Wales in autumn 2000. Nature. pp. 382–385. DOI: 10.1038/nature09762.

• Massey, N..(2015}).weather@home — development and validation of a very large ensemble modelling system for probabilistic event attribution. Quarterly Journal of the Royal Meteorological Society. pp. 1528–1545. DOI: 10.1002/qj.2455.

Recent publications (selected)

• Calafat, F..(2022}).Trends in Europe storm surge extremes match the rate of sea-level rise. Nature. pp. 841–845. DOI: 10.1038/s41586-022-04426-5.

• Miranda, N..(2023}).Change in cooling degree days with global mean temperature rise increasing from 1.5 °C to 2.0 °C. Nature Sustainability. DOI: 10.1038/s41893-023-01155-z.

• Ye, K..(2024}).Response of winter climate and extreme weather to projected Arctic sea-ice loss in very large-ensemble climate model simulations. npj Climate and Atmospheric Science. pp. 20.

Participating

To join the main (production) project:

1. Download and install the BOINC client from boinc.berkeley.edu
2. Add the project URL: https://climateprediction.net/
3. Create or log in with a CPDN account

To join the development / test server (cpdnboinc dev):

1. Use the project URL: https://dev.cpdn.org/
2. Note: the dev server issues experimental work-units and is intended primarily for experienced BOINC volunteers willing to assist with testing.

See also

• BOINC
• Volunteer computing
• Climate ensemble
• Global climate model
• Myles Allen
• BBC Climate Change Experiment
• SETI@home

External links

• cpdnboinc dev — Development server
• climateprediction.net — Main BOINC server
• CPDN project website (University of Oxford)
• CPDN project history
• CPDN scientific publications
• Wikipedia: climateprediction.net
• BOINC publications list
• Archived CPDN pages — Wayback Machine