WMOD00283

• Sponsor Name: USBR, Technical Service Center, River Systems 7 Meteorology Group
• Project Name: Use of NEXRAD WSR-88D Radar Snow Accumulations Algorith in the GCIP LSA-NC
• Title: Use of NEXRAD WSR-88D Radar Snow Accumulations Algorith in the GCIP LSA-NC
• Prepared For: USBR Technical Service Center
• Prepared By: Arlin B. Super
• Date: 1/1/1998
• State: MN
• Weather Modification - Doc Type: Report

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

VPS and rawinsonde data indicated "bright band" contamination caused by a warm
layer aloft. The method used to define storm periods ensured that dry snow was
falling in the general vicinity of the radar when VPS calculations were made.

The average VPS for each storm is plotted in figure 1, with a linear least squares
regression equation fitted to each set of 5 data points. Standard atmospheric
refraction was assumed in calculating beam centers above the radar. The noted
storm numbers correspond to those listed in table 1. The yPS for two of the storms
(3 and 6) have almost identical regression equations, which overlap in the plot. In
each case, the greatest average SWE rate for the storm period was estimated at the
lowest tilt sampled by the radar, with beam center located 380 m above the radar's
elevation. The VPS plots indicate that Minnesota snowfall rates generally continue
to increase as the snow particles fall all the way to ground level. Furthermore, the
lowest practical radar sampling should be most representative of surface snowfall
accumulations.

Examination of figure 1 and the results of table 1 both reveal that a linear least
squares regression equation usually provides a very good fit to the 5 plotted points.
The heights of these points above the radar correspond to the center of the 5 lowest
beam tilts at 33 through 37 km range. Each plotted point usually represents the
average of about a few hundred thousand individual range bin samples (i.e., 5 bins
x 360 deg = 1800 samples per volume scan per tilt; usually 10 to 11 volume scans
per hour; many hours per storm). The variance explained by the linear least
squares fits shown on the plots ranged from 93 to 100 percent with the single
exception of the storm of January 4-5, 1997, where only 68 percent of the variance
was explained. These generally large variance values indicate that no reason exists
to seek a more complex relation than linear regression to the VPS of Minnesota
snow storms.

The reduced variance of the January 4-5, 1997, storm is believed related to the
sharp reflectivity gradients around the approximately 35-km-radius circle from
which VPR samples were obtained. The SWE accumulation pattern from this same
storm will be portrayed in figure 4 of the next section. Although most storms of
table 1 were widespread around the radar, this storm produced only light
reflectivity values from northeast to south of the radar as compared to other sectors.
Moreover, a linear band of snowfall with marked gradients extended north-
northeast of the radar. Analysis of the VPS for 45-degree sectors (not shown)
revealed profiles similar to .other storms over most of the west half of the radar's
coverage, but unusual profiles to the north and northeast, likely related to the
intense snowfall core. These profiles had maximum values about 1500 m above the
radar, typical reductions in reflectivity above that level, and little change in
reflectivity below. When the north and northeast sectors are averaged with all
other sectors, the unusually "noisy" profile of storm 5 on figure 1 results. Review of
temperature and relative humidity profiles from rawinsondes released near the

5