Oct. and Nov. 1985 Tape Problems

Oct. and Nov. 1985 will be rerun. Hardware problems in the tape drives would randomly flip bits in writing and perhaps reading the tapes. We thought that we caught all the bad tapes but our test for bad tapes was not sensitive enough (read the tapes, un-tar the files, display reasonable looking fields and compute various means and variances). A more sensitive test recently found 15 bad grb2d fields out of a total of 100K/year. All these bad records were traced to Oct and Nov 1985.

We now believe that we can catch bad tapes. Tapes are now verified after being written. Unfortunately this procedure was not in place when the Oct. and Nov. 1985 tapes were written.

The BAMS CD-ROM may be checked to see if it was affected by the Oct. and Nov. 1985 tape problems. Send comments to wd23ek@sun1.ncep.noaa.gov.

News: we were not able to reproduce the Oct. and Nov. analyses exactly. Any changes in the compiler or hardware can introduces very small numerical differences. (Perhaps it was the change from the Cray Y-MP to the J16.) These differences get into the forecast model and amplify. Rather than have a discontinuity at the end of November 1985, the Bob Kistler has rerun October through December 1985. As a result, the discontinuity will be at the end of the year which is more anticipated.

Privacy Policy