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

5

than predicted. The apparent increase for tho combined 12 cases was 104 per cent.
Application of Students T test grwe a 0.995 confidence level. The actual increment
was 1200 acre-feet (1.48 x 106 cubic metres). .

21. The more usual circumstance is that storm flows are too difficult to
separate accurately, requiring' longer time ]Jeriods to be chosen for flow averaging.
The week and month are usually too short to provide a stable number of storm occur-
rences, but in some places seasonal precipitation can be used. This is most suitable
where the precipitation which provides the bulk of the annual water supply occurs in
the winter months as in Mediterranean climates. Where significant amounts of
precipitation occur throughout the year, the annual floH is best suited for use.
The usual practice is to average the flow over the water year which runs from 1
October to 30 September (for many areas, th(~ auturllJl period offers a low point in
hydrological activity so that dividing a major river rise into 2 parts for distri-
bution into 2 water years is not a frequent problem).

22. Al though annual flovl has usuall~f 'been used as the ba:o,is for evaluating
non-randomized experiments, it is possible '\;0 rand,omize seeding by years, The
resul ting eval ua tion ,1ill be credible if the seeded years exhi bi t a marked departure
of runoff compared "\Vi th that of the unseedec, yeaTS without benefitting from a lareer
number, or lillusually heavy, storm events. Grant, et al., (1971), reported use of
this approach in evaluating seeding in the \'Tolf Creek Pass region of Colorado. The
experiment ran from 1965 to 1969, with seeding of all storm events eluring the years
1965, 1967 and 1969. No seeding was done in 1966 or 1968. The indicated inc:r;ement
of flow from the three target river basins was 228,000 acre-feet (2.8 x 108 m').
Mielke, et a1., (1977), have recen-t1y published a cor:rection stating that 1970 Has
an unseeded year.

23. The estimates of increase derive froll! a statistical regression relating
the streamflow in the target area (San Juan River gauged at Pagosa Springs, drainage
area 772 km2; South ]'ork Rio Grande Hi vel' near South ]'ork, drainage area 559 km2;
and the Alamosa Creek above "J~errace Reservoir, drainage area 277 km2), to that from
the control area (Piedra River near Piedra, drainage area 961 km2; Animas River at
Howardsville, drainage area 145 km2; and Lake Fork of the Gunnison River at Gateview,
drainage area 865 km2), for a historical pel:'iod of 26 years. The correlation
coefficient of target area streamflo11 to control area streamflovl was 0.97. 'rhe
probability of the comb:i.ned seeded streamflo"\V being observed by chance is 0.005.

24. Of the various statistics 'v,'hich might serve to differentiate between
effective and. ineffective seeding, Narkovic (1966) identifies the mean a.nd the
variance as the statistics of choice in testing for significance. Effective seeding
would apply to seeding associated ,ii th an increase in streamflo"\V; seeding wh.ich
produced no increase 'Iiould be considered ineffective.

25. There have been hlO basiC approaches used to estimate the effectiveness
of cloud seeding in increasing the mean flow from the target area. The first
invol ves comparisons of target area flow following seeding wi th target area flo"l in
the absence of seedinG" Thi~: type of comparison has usually been made by comparing
a few years of seeded experience with the flo\'l observed over the years beforl~ seeding
was undertaken. The validi t;y of such a comparison ,.is heavily dependent on the vari-
abili ty of the annual floHS which go to make 1..11) the record. Examination of the
stability of past records 'Iiill quickly reveal v]hether or not there have been groups
of years in the past .,hich, if they had been seeded, liould have g1ven indications
of seeding effectiveness that would be entirel;y' ficti tious. "'hen this is the case,
a long period of seeded record is required before much confidence can be placed in
the seed/no seed ratio. The number of years required to constitute a "long period
of seeded record" ,'lill depend. on the variability of streamflo,'lfrom year to '(ear
. before and. afte:c seeding is undertaken. Five years of seed.ing could easily ;~oincide
a fortui taus wet l)eriod. Ten-year seeding programmes 'liould be much le:3s to bo so
easily confounded.