A brand new science for the planet

“When the problem gets complicated, networking is the solution”. This is nothing new in principle but what Bob Bishop has in mind is one of those novelties that have the potential to change the course of history. He proposes to start networking sciences to create new knowledge. All this for the benefit – and the survival – of the planet.

At the end of the fifteenth and the start of the sixteenth centuries, Leonardo da Vinci was not only an engineer but also a painter, a mathematician and an architect. But in more recent years the sciences have evolved more towards specialization. “We have been treating the sciences as separate stovepipes and silos for over 200 years”, says Bob Bishop, former CEO at Silicon Graphics and a physicist with more than 40 years’ experience in scientific, technical and engineering computing.

On 29 January, Bob Bishop visited CERN and gave a seminar on the role of computing in climate science. He is the President of the newly formed International Centre for Earth Simulation (ICES) Foundation, whose ultimate goal is to build a supercomputer capable of modeling the whole Earth and thus simulating its behaviour. This would allow us to predict the occurrence of natural disasters such as tsunamis or hurricanes. “In current science one has theory and experimentation. Modeling and simulation is the third branch of knowledge generation. In order to build a model with some prediction ability, one has to understand all the various processes involved and then bring in the mathematics that best represents those processes”, he says.

The goal of setting up a worldwide organization for climate science is to allow the construction of a new large-scale, high performance supercomputer capable of computing all the different parameters, linkages and input data and producing sensible predictions from them. “In order for a model to be validated, one has to compare the output production with reality”, explains Bob Bishop. “If you cannot get the validation, you have to come back and go through the whole circle again. You can do that rather fast with a powerful supercomputer”.

Bishop’s supercomputer will have to process about one million billion floating point operations (i.e. one petaflop), and deal with exabytes (billion billions of bytes) of data. At full operation intensity, the LHC will produce roughly 15 petabytes of data annually, which will ultimately also accumulate to the exabyte level. The strategy adopted by CERN to deal with such a large quantity of data has been the Grid, the network of thousands of “ordinary” computers distributed across the world and hierarchically arranged in tiers. “In the case of Earth sciences the option of building a single supercomputer has to be preferred because a speed of about 1000 times real time is needed to get ahead of the problem. If you break and scatter the problem across the Grid, the latency of the connection between each machine will be enough to slow down the speed of the calculation and make it unusable”, explains Bob.

At this scale, gathering money to support the project is difficult, yet on the other hand, taken in perspective, nothing is more expensive than natural disasters that cause human losses. If such a supercomputer could indeed simulate the Earth’s behaviour so well that we could predict or even prevent dangerous natural phenomena from occurring, that would certainly be one of the best deals of human history. “Nature works by integration rather than disintegration”, continues Bishop. “We have spent 200 years achieving a very deep understanding of single phenomena. Today we have a high degree of specialization everywhere throughout science. In the case of the Earth sciences and climate sciences, one can find 50 or 60 specialised areas of overlap. Geology, but also geography, atmospheric physics, clouds physics, solar physics and cosmology are all involved, and these areas need to be integrated. I think the 21st century will be a century of reintegration – making the pieces talk to each other again, as they do in Nature”.

by CERN Bulletin