WMOD00459

• Title: Snow Accumulation Algorithm for the WSR-88D Radar, Version 1
• Date: 6/1/1996
• Weather Modification - Doc Type: Report

time to avoid confusing the operators, the program also listed both local and UTC (universal
time coordinated) date and hours because radar data are recorded in UTC time.

Some data losses occasionally occurred, usually because of clock stoppages or gage
malfunctions. The only major data loss was the previously discussed Black Forest gage south
of Denver. Overall, the gage record is considered to be of very good quality, both because of
careful locating of gages so they were protected from the wind and because of considerable
care taken to calibrate and maintain the gages and in chart reduction.

4.2 Processing of Level II Radar Tapes

Copies of original Level II B-mm Exabyte tapes from the Albany, Cleveland, and Denver
radars have been, or will be, obtained from the NCDC (National Climatic Data Center) in
Asheville, North Carolina, for all periods with significant snowfalls. Tape processing was
done with a Sun Sparc 20 workstation under the Solaris 2.4 operating system. The first step
was to use a routine that extracted each file's start and stop times and size and then closed
the file. The resulting file directory from this initial processing provides the means to
determine exactly which files are needed from each tape to match snowstorm periods in later
processmg.

File number Ion Level II tapes contains header information and numbers 2 to 401 are data
files unless recording is interrupted before the last file, 401, is written. One volume scan is
usually equivalent to one file although occasionally two volume scans are erroneously written.
to a single file. Software has not yet been developed to extract the second file in such cases.
Occasionally, a short file is written which may contain little but header information and
perhaps part of a volume scan. All files less than 4 minutes in duration were ignored in later
processing.

File size indicates scan strategy. For example, a typical precipitation mode volume scan (see
Federal Meteorological Handbook No. 11, 1991) during snowfall requires about 5.75 minutes
and produces a 9.B-megabyte file. Clear air volume scans take almost 10 minutes and
produce somewhat smaller files.

The second step in processing is to open only the desired files on each tape and save the
fraction of the total information needed for comparison with snow observations. The Ze
values, recorded to the nearest 0.5 dBZ, where dBZ = 10 LogZe, were extracted for two arrays
of range bins over each snow measurement site from the lowest (0.50) beam. One array was
centered directly over the site (so-called "vertical" array) as though snow fell vertically
without advection. The other array was advected by the wind (so-called "upwind" array) in
an attempt to make a more realistic match between gage position and the region of the 0.50
radar beam from which the snow actually fell. .

For each type of array and each snow measurement site, dBZ values were extracted over
semi-equal areas. Each array consisted of a "box" exactly 3 km in range (depth) by at least
3 km in azimuth (width). A minimum of 30 of azimuth was always used so the azimuthal
width was greater than 3 km at ranges exceeding 60 km where 10 equals 1 km in width (e.g.,
three 10 radials result in a 6-km width at a 120-km range). The smallest array was,
therefore, 3 x 3 = 9 range bins. For ranges nearer to the radar than about 50 km, additional
radials were added to the array to maintain an approximate 3-km width. Radials were added
in steps of two to keep an odd number (5,7,9, . . .) of degrees azimuth by a 3-km range in

14