WMOD00519

• Project Name: Weather Modification Programme - Precipitation Enhancement Project
• Title: A Review of the Hydrological Aspects of Evaluation of Precipitation Enhancement
• Date: 5/1/1977
• Weather Modification - Doc Type: Report

6

26. Another form of comparison, which involves only flow records ta]wn from
the target area, results when the decision to seed an area is made periodically at
random before the nature of the flow for an ensuing water year has been influenced
by observed events. All storms exhibiting pre',-iously specified temperature, moistuTe,
and .dnd characteristics could be seeded in the years desienated for seedine, and
left untreated in the years designated for no seeding. Comparison of target area
flow during years selected for seeding at random with the flows observed in the
intervening years i"hen no seeding occurred will at 1east be free of the complication
of fortuitous runs of several i'let or dry years in a row, but care must be taken that
the assumptions of normal distribution that underlie some tests of significance are
adequately met before they are applied (Thorn, 1957). It is important that successive
year flows do not exhibit large carryover effects.

27. The second, general type of hydrological evaluation involves comparison of
each period of seeded target flow with an estimate of what the flow i'lould have been
in the absence of seeding. These estimates have usually been made by employing
prediction equations developed by regression analysis inVOlving the past records of
the target basin and one or more neighbouring basins. The name "target-control"
has usually been applied to this approach.

28. The evaluation of
is an example of the use of
ring basins as predictors.

the seeding of the Kines River Basin (Henderson, 1975)
a prediction equation employing the fIoH of two neighbou-
The equation:

Ke

1.85 CI + 1.72 O2 - 124.4,

where X is the predicted floi'! (in acre-feet) of the Kings River and 01 ancl C2
represegt the flow of the Merced and Kern Rivers respectively, has be~n used to
assess the seeding effectiveness during the period 1954 to 1974. In the 20 years
evaluated, 14 shovred evidence of positive effect. A significance level of 0.01 is
reported to apply.

29. Where good records of streamflow and precipitation are ava:Llable" the
estimates of target area flo.r can be based on controls that employ both precipit>ction
and streamflow values as precUctors. An outstanding example of this app:J:oach is the
evaluation applied to the Skagit River in the Cascade Hountains of the State of
"'ashington (Hastay and Gladvlell, 1968). The data bank: available for use in develop-
ing the prediction equations of target area streamflow consisted of 34 years of
record for 26 precipitation stations, 21 river gauginG stations and 13 temperature
stations. Eliminat,ion of 15 r;i.ver gauging stat:i.ons because of distance from the
target area, and elimination of certain precipitation records in which inhomogeniety
of observing circumstances seemed to be a problem, reduced the number of possible
control values to 29. In making the regression analysis, some very high correlations
were computed betiVeen the streamflows from the Skagit River target areas and neigh-
bourine basins. For example, the correlation coefficient between the records for
the Cascade River, gauged, at l1,-'],rblemount (d.I'ainage area 470 km2) and the Skagit
River, gauged at Concrete (drainage area 7000 km2) ivaS 0.98637.

30. Applica tion of prinCipal component analysis, .li th a requirement that a
variable be significant at the 0.10 level, produced prediction equations of the fO:l.'"JlI

Yt Bo + BIKit + ,......" + BkXkt + Ut

i.here Yt is the preclicted tax'get streamfl01,v, the Bs arc real constants, the Ks are
the explanatory variables and Ut is a nornnlly independently d.istributed. random e~Tor
with a mean of zero.