WMOD00425

• Title: Demonstration of Improved Operational Water Resources Management Through Use of Better Snow Water Equivalent Information
• Date: 9/1/2002
• Weather Modification - Doc Type: Report

I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I

found to be adequate for most radars in the GCIP LSA-NC (north-central U.S.) area. In
mountainous terrain where it is necessary to use higher tilt data or where gauges are well below
the sampling beam, relatively weak Ze values are measured. Therefore, a small a. value is
required to match actual surface snowfalls. Data from Grand Junction (KGJX), for example,
pointed to a. = 40. Further inspection, however, showed that the 0.5-degree (0) beam was largely
blocked in the sector where the gauges were located. Therefore, results at KGJX are
questionable.

Hourly data generally produced more scatter when comparing radar and gauge measurements of
snow accumulation than when comparing longer-period data. Storm-total estimates exhibited
much better agreement with actual measurements, generally within 0.20 inches and with much
better correlation coefficients. We found that surface observations of Sand SD were closely
associated, with correlation coefficients r = 0.71-0.75 for the Denver and Albany hourly data sets
and 0.85 for approximately daily values from Cleveland. The best SAA estimations of SD
resulted from dividing the accumulated S by the median fresh snow density for each locale.
Median hourly or daily snowfall densities observed at Albany, Denver, and Cleveland for the
entire 1995-96 winter were near 0.08,0.07, and 0.06 gram (g) centimeter-3 (cm), respectively.
This represents snow-to-liquid ratios of about 12: 1 to 17: 1, considerably drier snow than the
commonly used 10:1 ratio.

The WSR-88D PPS has an occultation file for every radar site. This file is based on terrain data
and specifies how much to add to Ze values as compensation for any partial (< 60%) beam
blockage by terrain. Based on the occultation file, the PPS uses a "hybrid scan" construction for
Ze in precipitation estimation. The original hybrid scan used decreasing elevation ("tilt") scans
with increasing radar range, from 3.40 at 20-km range to 0.50 beyond 50 km. Steps upward or
downward in the nominal construction, unfortunately, resulted in spurious discontinuities in
shallow snowfall, in which the vertical profile of Ze is normally maximized near the surface.
Later, the WSR-88D and SAA initially used a terrain-based hybrid scan (O'Bannon, 1997),
which attempts to use the lowest tilt at all ranges beyond 2 km, unless the bottom of that tilt fails
to clear the terrain by 500 feet. Even this hybrid scan caused frequent data collection from the
1.50 beam to about 45- through 60-km range, with a discontinuity at these ranges where the 0.50
beam usage commenced. Moreover, S underestimates sometimes still occurred along particular
azimuths, apparently caused by beam blockage by objects (buildings, towers, trees) not described
in the terrain elevation files. To remedy these problems, an empirical scheme was developed that
involves hand editing and specifies the lowest practical radar scan that can be used for each range
bin while still avoiding serious ground clutter. This scheme was successfully tested with actual
snowstorm data. For terrain with limited relief, such as near Minneapolis, it was usually possible
to employ the 0.50 tilt beam within 20 km of the radar. Such low scanning would often be
impractical with rain because of anomalous propagation (AP) echoes, but it is apparently not a
serious problem with snowfall.

11