IAU Earth Collision Asteroid

On March 11, 1998 a report came out about an asteroid that appeared to be on an earth collision course. Well - close anyway, 30,000 mile and a margin of error to be within 180,000 miles - so a possible collision course. Yet, the later data that JPL gave put the asteroid at 600,000 miles. It's the margin of error that bothers me. It should be based on the reliability of the observations made and should be very quantifiable. The JPL data should have refined the margin of error and adjusted the course to some new value within the original margin of error - that's the nature of the thing, afer all. What mistake did the original scientists make? How could they have made a 500,000 mile error in their margin?

here's the two articles that have the numbers on them:

Bob Braeunig (writer of the orbital mechanics article elsewhere on my page replied with this explanation:

The original IAU Circular released on March 11 by Brian Marsden does not give a specific error. It says "This nominal orbit indicates that the object will pass only 0.00031 AU from earth on 2028 Oct. 26.73 UT! Error estimates suggest that passage within 0.002 AU is virtually certain". The circular also requested further observations to allow refinement of the 2028 miss distance. (For your information, 0.00031 AU is about 46,000 km and 0.002 AU is 300,000 km.)
Immediately following the release, two astronomers from JPL calculated the uncertainty ellipse as being 2.8 million km long and 2,500 km wide. The closest this ellipse came to the center of the Earth was only 30,000 km. Later their estimate of the most likely miss distance was about 86,000 km.
On March 12 pre-discovery images of the asteroid dating back to 1990 were found. With this new data the JPL astronomers re-calculated the miss distance placing it at 950,000 km from Earth. The new uncertainty ellipse had shrunk to just 175,000 km long by 1,000 km wide, but within their original larger uncertainty ellipse.
Marsden later admitted that it was erroneous to say that a close approach was "virtually certain". It seems that the only problem with his notice was that it lacked a detailed statement of the actual uncertainty ellipsoid in 2028.

Rudy Moore