G. Walters
                                                                2001Oct31
                NCEP SOUNDING TYPE IRREGULARITY


In the course of preparing our "super" inventories, my software which
distinguishes between raobs and pibals was confused by the data in a
period during 1987.  From 6Z on March 25 through 6Z on June 23, about
200 pibals at each 6-hourly time were being classified as raobs due to
an unusual pressure value that appeared in the significant level data
(ON29 Category 02).  These soundings, due to the appearance of a
temperature at the only pressure level in the sounding, were thus
being misidentified.  Being the first and only level in the Category
02 section of the report, the documentation indicates that this would
be the surface level, but in these particular soundings, they have a
surface pressure greater than 9000mb.  We decided to ignore the
temperatures on significant levels with bad pressures.  Another
intriguing fact about these particular soundings: the entire world
was involved, except North and South America, and China!  We suspect
this situation may be due to an irregularity in reporting practises
or in NCEP's processing - where perhaps surface reports somehow got
mixed into the upper air reports, and then perhaps NCEP altered the
pressure (multiplied by 10?) as a signal to ignore the level.

Ignoring the temperatures on significant levels with bad pressures
does not affect the total number of soundings.  It merely shifts
misclassified soundings from raobs to pibals, thereby maintaining
consistent counts.  As programmed, the wind at this level will still
be enumerated in the pibal records in the "super" inventories.

--------------------------------------------------------------------------------

                                                                G. Walters
                                                                D. Joseph
                                                                2000May02
                                                            rev.2001Oct31
        BAD CATEGORY/COUNTER GROUP

A user recently reported numerous decode errors appearing in the
output from our readupa.f software while processing the 1998Jun-1999Oct
files.  Investigation found that NCEP had flooded a field in the
category/counter group section of the ID for certain upper air significant
level reports.

Specifically, character locations 48 - 50, which are supposed to contain
the total number of characters in the current category.  For a few obs,
about 3 per day on average, coming from the tropical eastern Pacific,
there are so many category 2 and/or category 4 levels, that the total
number of characters exceeds the maximum which this field can show.

Our access software was sensitive to this, hitting the field of asterisks
and printing a decode error.  However, the software actually calculates
the correct value from the number of levels specified in characters
46 - 47 by multiplying by the known fixed length of the appropriate
category.

Nevertheless, we needed to increase the size of the arrays into which
the mandatory levels and significant levels are unpacked and merged.

Users should therefore obtain the latest copy of readupa2.f in order to
properly handle this condition.