WMOD00561

• Title: Snow Accumulation Algorithm for the WSR-80D Radar: Final Report
• Date: 7/1/1998
• Weather Modification - Doc Type: Report

It will be shown that this optimization scheme:, based on searching for the minimum summation of
absolute difference between radar estimates and surface observations of S, often resulted in reasonable a
and p values even when applied to individual measurement sites. As shown in table 3 the scheme always
produced reasonable values for grouped surface observations from fairly near the radar with the
exception of the southern Minnesota area. Of course, grouping provides greater sample size. The
importance of large sample size has been disclLlssed by Krajewski and Smith (1991). Reasons for the
failure of Minnesota data to optimize will be discussed.

Pairs of a and p are summarized for the many individual gage or snow board locations in table 2. Snow
board observations near Albany were grouped by range because the number of observations were usually
limited for individual sites. Frequently only portions of storms were sampled since volunteers made the
hourly Sand SD readings, many of whom held jobs. Late night observations were uncommon and most
Albany observers did not sample all storms that affected their sampling sites. Still, an excellent data set
resulted. In contrast to Albany snow board sampling, all gage and the Denver NWS FO snow board
samples (KFTG # 7) were usually available 24 hours per day. A couple of exceptions existed because of
gage malfunction (KMPX # 5) or lack of operator availability (KFTG # 6). However, almost all gage
operators were very conscientious and gages were reliable with the result that missing data were quite
limited.

All Ze values used in the calculations of table 3 and similar results to follow were from the lowest
available tilt. That was 0.5 deg with only 4 exceptions. The lowest tilt was blocked for gage KFTG # 5,
located on Mt. Evans where the 1.45 deg tilt was used. The lowest useable tilt over the 3 gages on top of
the Grand Mesa (KGJX) was 2.4 deg tilt. This finding was a surprise as terrain data suggest the 0.5 deg
beam should have been useable over the 2 gag(~s nearest the radar, at 6 and 11 km range, respectively,
and the 1.45 beam should have been useable for the gage at 21 km range. However, the radar Ze data
from several storms clearly showed almost complete blockage of the lowest tilt and partial blockage of
the second lowest tilt along the east-west axis of the Grand Mesa. The radar is located near the Mesa's
west edge. Fortunately, the greatest range was only 21 km so the center of the 2.4 deg tilt was not
excessively high above the gages (see table 2). However, the vertical separation is believed at least
partially responsible for the exceptionally small a values from each gage location.

The authors have previous experience working on the Grand Mesa where snow particle observations
were made by low-flying aircraft and on the Mesa top. Differences in snow particle concentration and
size were pronounced in this approximately 600 m vertical separation (Super and Boe, 1988), indicating
considerable growth in this layer which approximates the 2.4 deg beam separation from the Mesa top.
These differences might be expected as considerable water condensate was produced at low levels over
the Mesa by forced orographic lifting of the air during storms.

The primary optimization scheme has been used with individual surface observing sites and with
combinations of surface sites at similar ranges from the radar. However, the scheme frequently failed to
provide a minimum with equation (7) with some data sets within the reasonable p range of 1.2 to 2.6.
These data sets were usually of limited S range or had other shortcomings. For example, all the
Minneapolis (KMPX) area comparisons failed to reach a CTF minimum within the noted p range. This
is believed to be related to the limited range of hourly S accumulations sampled by the gages over the
course of an entire winter. Only 5 of the 889 gage-hours with detectable snowfall exceeded 0.100 inch
and 4 of the 6 gages never observed an hourly accumulation reaching 0.090 inch. It is difficult to fit an
optimized equation when almost all hours of snowfall are very light to light.

23