WMOD00399

• Title: WMO Training Workshop on Weather Modification for Meteorologists - Lecture Notes
• Date: 12/1/1979
• Weather Modification - Doc Type: Report

- 72 -

The poorer radar performance for light showers is readily ex-
plained. For light showers even a small absolute error produces a rather
large error in terms of FD. If the mean results had been presented in
terms of volumetric differences, then the mean error would be less for
light showers than for heavy showers.

The better agreement between gage and radar for the day than ,
for individual showers was expected. Random errors are inherent in the.
analysis procedures. Further, the Z-R relation is known to vary within'a
shower and with time as well as among showers. Thus random errors com-'
bined with a varying Z-R might result in a radar overestimate for one
shower and an underestimate for another. Upon shower combination in the
formation of the daily comparison, however, these errors tend to compen~
sate resulting in better gage-radar correspondence.

The individual and daily gage and radar comparisons for 1972
and 1973 are presented in FD format in Figure 6; its interpretation is 1
analagous to that in the gaging section. One can see that the radar es~
timates are within a factor of two of the true individual shower and
daily rainfall in about 66 and 80 percent of the comparisons, respec- ,
tively. Comparison of the plots in Figure 6 with those in Figure 3 for'
raingages permits an evaluation of radar performance in terms of an equi-
valent gage network covering 200 to 250 mi2. The accuracies of radar :
measurements of individual showers ,in 1972 and 1973. are equivalent to
that which one might obtain with a gage network having a density of abo~t
50 mi2jgage (135 km2jgage). The equivalent accuracy of radar measure- i
ments of all showers combined over the course of. the day (the daily I
comparison) is roughly 25 mi2jgage (68 km2/gage).

"

This result suggests that if one is concerned with the measure-
ment of convective events over a fixed gage network, such as the mesonet
covering 200-250 mi2, then a rather small number (10 to 20) of gages will
do better than the radar.

Gage Adjustment of Radar-Rain Estimates

I
Because of the problems in the radar estimation of rainfall, it
is unwise to accept radar estimates of rainfall without. some check of
their accuracy. Having documented radar performance, one might ask whet
ther it is possible to improve the accuracy of the radar-rain estimates!

The value of our method of gage adjustment 9f the radar esti-
mates of rainfall was tested using the gage and radar estimates of rain-
fall in the clusters and the mesonet. Thirty-nine days in 1973, when ail
systems were operating properly without obstacles such as anomalous
propagation; were used in this test. The adjustment of the radar esti- i
mate of rainfall in the mesonet for each day was accomplished by: 1) ,
determining the ratio of daily gage to radar rainfall for the clusters :
2) applying the cluster ratio to adjust the radar estimate of rainfall ~n
the mesonet and 3) comparing radar performance for mesonet rain estima- I
tion before and after adjustment with the cluster ratio. If gage adjust-
ment has any validity, the radar measurements should be more accurate ,
after adjustment. The results are summarized in Table 6. Results are i
also plotted in the usual FD format in Figure 7.

~1It

Adjustment using the cluster ratios produced a statistically
significant improvement (better than one percent level with two-tailed
"t" test) in radar accuracy in this test, decreasing the mean daily fac-