WMOD00459

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

the array. Because the recorded radial azimuth values vary from scan to scan, the exact
azimuthal position of the array over a gage also varied slightly. In contrast, range was
always fixed with the center of each range bin exactly at 0.5, 1.5, 2.5, . .. kIn from the radar.
In any event, all vertical arrays were chosen so that the gage was always positioned directly
below the center range bin of the array.

For the purpose of extracting upwind arrays, VAD (vertical azimuth display) winds were
calculated at 1000-ft intervals for each v,olume scan (file) as discussed in section 10.2. This
process simulates the standard WSR-88D V AD product, which portrays winds in 1000-ft
intervals. The wind information was used to advect the falling particles. Snowflake fall
speed observations are not made with the WSR-88Ds. Although the Doppler shift could be
used to estimate particle fall speeds, a different scan strategy would be required, including
. an almost vertical antenna tilt. In the absence of fall speed measurements, the larger
snowflakes, which produce almost all the returned signal (Ze)' were assumed to fall at exactly
1.0 m S-l. Of course, graupel (snow pellets) may fall more than twice that fast and snowfall
consisting of individual, large, non-aggrl~gated ice crystals may fall half that fast. So the
constant fall speed used is only an approximation that mayor may not improve the Ze-S
relationship in general. Comparisons between radar-estimated snowfall accumulations from
both vertical and upwind arrays and undElrlying gages will be used to judge the improvement
provided by the advection scheme. Some improvement might be anticipated because snow
particles often can be advected for tens of kilometers between the lowest tilt radar beam and
the ground, especially at long ranges where the beam is high above the ground.

The method of advection calculation starts at each gage location and elevation and then steps
upward at 1000-ft intervals, calculating the horizontal motion of a particle falling at 1.0 m S.l
within each 1000-ft layer using the V.ADwind velocity for that layer. This process is
continued upward until the center of the 0.50 radar beam is reached. That point in space
then becomes the center of the upwind array.

5. Comparison of Cleveland and Denver Snowfalls

The da~a from gages No.1 to 5 in both tables 3 (Denver) and 5 (Cleveland) were summarized
for general information. Radar and gage observations from all Denver area storms listed in
Table 4 from November 1, 1995, through ~ranuary 30, 1996 (3 months), are considered in this
,report. Similarly, all Cleveland area s.torrns listed in table 6 from November 3,1995, through
January 3, 1996 (2 months), are considered. The later storms listed in tables 4 and 6 have
yet to be analyzed.

Table 7 shows the summation of all observed snow water equivalent for the five gages in each
area. Records were complete or almost complete at most gages. The two exceptions were
Cleveland gages No.1 (39 hours missing) and 4 (30 hours missing), which missed the season's
first storm simply because they had yet to be installed. More hours with snowfall are
considered in table 7 than in tables 8 or 10 because hours with missing radar data are.
included in table 7.

Significantly more snow fell in the Cleveland gages in 2 months than in the Denver gages
over 3 months. Gage totals east-northeast of Cleveland ranged between 2.86 and 4.88 inches.
In contrast, the gages located within 50 km of Denver received only 1.23 to 1.84 inches, the
mountain valley gage (No.4) only 0.52 Jinch and the mountain gage (No.5) 3.01 inches.
Although the 1995-96 winter had record snowfalls northeast of Cleveland, the Denver area

15