washington

Still grounded: A computer-generated image of the EOS AM-1 spacecraft orbiting Earth. Credit: S. SUZUKI & E. M. DE JONG/JPL

Software problems will delay the long-awaited launch of the first major satellite in the US space agency NASA's Earth Observing System (EOS), a network of at least three large spacecraft and many smaller ones designed to study the Earth as an integrated system.

The AM-1 spacecraft — which carries US, Japanese and Canadian sensors for monitoring clouds, ice, land and water — was to have been launched in June, but will now remain grounded until at least December.

The problem is with a segment of the troubled EOS data system known as the EOSDIS, one of the most complex software engineering projects ever attempted. The $2 billion EOSDIS is designed to handle unprecedented volumes of data from a variety of orbiting sensors over a decade or more, and to format the data for a wide range of users, from scientists to the general public.

Although the system passed an important test of its science data processing segment last summer (see Nature 389, 108; 1997), other difficulties have continued. One problem area is the software's Flight Operations Segment (FOS), designed to control and monitor the health of all EOS spacecraft and to schedule onboard activities such as instrument pointing.

NASA had hoped that bugs in earlier versions of the FOS would be corrected in a new version which was delivered last month by the Lockheed Martin Corporation. But some of the old problems remain, and new ones have appeared, according to EOSDIS project manager Rick Obenschain of the NASA Goddard Space Flight Center in Maryland.

If the FOS cannot be fixed in time for the AM-1 launch, the space agency is considering adapting commercially available software similar to that used for the recently launched Tropical Rainfall Measuring Mission and the forthcoming Landsat 7 — which also has slipped from a July launch to at least February, due to a power supply problem with its main instrument.

Any off-the-shelf alternative would not be a full replacement for the FOS, says Obenschain, and would be a one-time-only fix. It is also unlikely that it would be ready by December, he says.

The EOSDIS is “meant to be everything to everybody”, he says, which may be its biggest problem. In retrospect, says Obenschain, it was “probably not a very good intent” to design a single, massive data system to handle all the needs of all EOS spacecraft (more than two million lines of computer code have been written already).

The project has suffered major delays, attrition rates as high as 38 per cent among its software engineers, and perhaps now an insurmountable problem with the FOS for AM-1.

Scientifically, the expected slip in the schedule will have little impact. Byron Tapley of the University of Texas, who will take over next week as chair of the science executive committee for the EOS Investigators Working Group, says there will be “no major Earth-shaking loss in science opportunity” if the delay lasts only a few months.

But the future of the EOSDIS may be on the line. Several outside advisory groups have recommended that NASA should switch to a distributed or ‘federated’ system, whereby many different organizations would supply the research community with data from one or two EOS instruments.

In December the space agency picked two dozen Earth Science Information Partners, who will have three to five years to develop and distribute prototype data products for scientists and other users.

Meanwhile the seven large, centralized data archives organized under the current EOSDIS architecture are being reviewed by a committee of the National Academy of Sciences. The committee, which is chaired by Francis Bretherton of the University of Wisconsin, meets in Washington this week, and hopes to report by the end of the summer.