washington

EOS AM-1: project's data system now back on track? Credit: NASA

The US National Aeronautics and Space Administration (NASA) was this week optimistic that its troubled computer system for handling vast amounts of data from a planned constellation of Earth-observation satellites will be up and running for the scheduled launch of the first major satellite, AM1, next June. The system has just come through a critical test.

Past problems with the system, which will underpin the world's largest Earth-observation programme — the US Earth Observing System (EOS) — had left it 18 months behind schedule. The $6-billion EOS is intended to provide information on the atmosphere, oceans and solid Earth for at least a decade.

The data-handling system — EOS Data and Information System (EOSDIS) — is vitally important, because, once operational, the satellites will generate about 1 terabyte of data daily. In six months they will produce the equivalent of NASA's entire 25-year archive of remote-sensing data. The EOSDIS ‘core system’, which will process raw science data from the EOS satellites, failed in tests last January. This, combined with management and budgetary problems, resulted in the delay to the project. Even now there is “very little, if any, slack” in the development schedule, says Rick Obenschain, who manages the EOSDIS at the space agency's Goddard Space Flight Center in Maryland.

The recent tests were designed by an outside review group established by NASA and chaired by Sara Graves, a computer scientist at the University of Alabama in Huntsville. Checks included verifying that the system could ingest raw data from Landsat and other satellites, for example, and that scientists could log on and retrieve data. The system accomplished all 42 of the tasks it was given.

Nevertheless, many scientists working on EOS are still pressing for EOSDIS to be replaced by a decentralized system giving greater control of data processing to the teams that designed the instruments. Such an approach has been endorsed by NASA advisory groups (see Nature 386, 203; 1997). But there seems to be a growing acceptance that NASA has gone so far down the EOSDIS road that it must try to get the most out of its investment. EOSDIS, including archive centres, will cost $1.8 billion over ten years. The core system alone — being developed by the Hughes Information Technology Company of Landover, Maryland — will cost $680 million.

Obenschain also warns that any alternative system could be at least as expensive as EOSDIS. NASA has agreed, on the advice of two outside advisory groups, to run the system at a quarter-capacity in the first year of operation. This will save money and make it easier to resolve teething problems. But it means that scientists will have to compete for reduced resources. A committee chaired by Eric Barron, an EOS interdisciplinary investigator from Pennsylvania State University, has the task of recommending which projects should be given priority during this period.

Any decision to change EOSDIS radically will probably have a political component. Despite its problems with the core system, Hughes has a home-state ally in Senator Barbara Mikulski (Democrat, Maryland), who is an influential member of the Senate appropriations committee that controls NASA's budget. Indeed, the committee advised NASA this summer, in a report accompanying the agency's spending bill, that it expects the Hughes contract to continue “in its current structure” provided the company meets set targets for delivering successive versions of the core system software.