Climate change impact models are driven by climate projections from Global Circulation Models (GCMs). There are twenty-two GCMs used in the 4th IPCC assessment, and their climate projections differ. What are the implications of this uncertainty for ecosystem properties such as primary production and soil carbon storage globally? Answering such questions takes huge computing resources which are not routinely available to the climate change impacts community. How do we increase accessibility of these models to this community? And, is a cloud approach a possible solution?
A storyboard was created for a typical question from a climate change impact researcher: “Where do uncertainty in climate projections translate into greatest uncertainty on the ground in terms of carbon and water fluxes?”
This question led to a set of scalable climate patterns all on a common grid representing the 22 GCMs. These were used to enhance the IMOGEN modelling systems. These link the circulation patterns to the JULES model which describes how the land influences the atmosphere. A commercial cloud-provider was used to run the combined IMOGEN framework. The skills and expertise of the bioinformatics and genomics communities were harnessed to run it as efficiently as possible.
A researcher can now run the IMOGEN modelling framework simply by entering an emission scenario of greenhouse gases, the years in which they are interested, and the type of outputs that are required. For example, soil or vegetation carbon. If such a combination has been run before, the user is sent the output from the portal’s library. If not, a new model-run is completed. Output data or maps are emailed to the user to save costs. This can be as quick as 2-3 days; a saving of 95% over a non-cloud approach.
This exemplar has led to funding being secured to undertake a full cloud-implementation of the JULES model under the UK’s Big Data initiative. This will make this large land-atmosphere model much more accessible to the climate change impacts community.
The cost-barrier of using a commercial cloud provider was highlighted by many academics during the course of this work. NERC has responded to this through the provisioning of cloud services under the Big Data initiative. It is intended that future work will focus on more web-enabling of a wide range of models to increase accessibility to the whole NERC community and its stakeholders. Exploitation of the cloud to make data and modelling resources more accessible and transferable between communities will lead to greater efficiencies and more integrated science crossing traditional boundaries.