CLIMATE IN A BOX

 

Narrated by Michael Starobin

 

Climate and weather models are notoriously complex beasts. It takes a small army of scientists and computer programmers a year or more to build a model. Then they need a supercomputer fast enough to run it.

Like those models, the machines themselves can be unfriendly. They're hot; they're expensive; they're big. But when you're trying to model major parts of the planet's climate, you'll put up with a prima donna to get punctilious power.

            But always in the back of your mind, a question: what if climate research had some more accessible tools?

            Like one of these...a desktop sized supercomputer, with climate models loaded and ready to run whenever and however a researcher needs it.

 

Tsendgar Lee -- High End Computing Manager, NASA HQ

           

            When I develop something I can run it, I can test it immediately. I can see the results. It increases the productivity tremendously.

 

            NASA's vision is to deliver a set of quality control procedures so researchers can perfect their own research algorithms with a standardized baseline. The plan would essentially declare a minimum standard for climate and weather research, facilitating a rapid process of sharing and evolutionary development of new ideas, and a rapid transition from research to operation.

Where big iron costs millions, these hot rods cost many orders of magnitude less. Where the giant systems require special rooms and architecture, desktops fit comfortably into ordinary equipment rooms or even office spaces. But perhaps most compelling is how these machines deliver turnkey systems, able to ingest data and start crunching shortly after installation. They democratize climate research, analogous to how ordinary personal computers democratized day to day computing tasks not too long ago. In other words: easy.

 

Tsendgar Lee -- High End Computing Manager, NASA HQ

           

            The reason we put in a common software framework is exactly for the different disciplines to come in and use the same interface to be able to exchange the data, exchange the model, exchange workflow. This is the box for everybody.

 

            NASA is one of the world leaders in climate model development. The Space Agency's new GEOS-5 model already produces highly detailed, tightly calibrated output. By facilitating heavyweight climate research on modest budgets, NASA hopes to open the door for wide groups of scientists to pursue dramatic climate research that might otherwise have been inaccessible.  The Agency's plan essentially gives away computational models to run on these systems, thus seeding the research ground with new opportunities for discovery.

            NASA's software currently runs on LINUX, but as these new desktop supercomputers are also capable of running Windows HPC, the Space Agency expects to deliver code compiled for each operating system. The modeling tools themselves are written primarily in Fortran 90, so the research community can expect to find themselves in familiar territory soon after power cords plug into walls.

 

Mike Seablom – Project Manager, Climate in a Box

 

            This capability that's provided by such a system allows us to run models at a reasonable resolution for development. Then, once we're satisfied with the results we're getting on the desktop sized system we can then do production on our very big supercomputers which are room sized systems.

 

            Climate in a Box: NASA's big idea for modeling big science. You too might have access to the whole planet...right on your desk.