Title: On the Evaluation of Operational Cloud Seeding Projects
Full Citation
Permanent Link: http://ufdc.ufl.edu/WL00002536/00001
 Material Information
Title: On the Evaluation of Operational Cloud Seeding Projects
Physical Description: Book
Language: English
Publisher: American Water Resources Association
Spatial Coverage: North America -- United States of America -- Florida
Abstract: On the Evaluation of Operational Cloud Seeding Projects, Aug 1983, Vol 19, No 4
General Note: Box 10, Folder 22 ( SF Water Modification - 1981-83 ), Item 6
Funding: Digitized by the Legal Technology Institute in the Levin College of Law at the University of Florida.
 Record Information
Bibliographic ID: WL00002536
Volume ID: VID00001
Source Institution: Levin College of Law, University of Florida
Holding Location: Levin College of Law, University of Florida
Rights Management: All rights reserved by the source institution and holding location.

Full Text

''L. 19, NO. 4


C.-F. Hsu and S. A. Changnon, Jr.2

\BSTRACT: Operational cloud seeding projects, those designed to
iuce a desired change in the weather and that are nonexperimental
ii nature, continue to be pursued widely in the United States. A re-
zrring question by scientists, project sponsors, and cloud seeders has
rn, "was the weather altered and if so, by how much?" Evaluation
of such project ts is now recognized as having scientific benefits, and a
,fr-year study has addressed various techniques and statistical methods
u perform evaluations and to learn more about how to modify the
weather. Most such evaluations hinge on some type of space-time com-
prisons, but valid comparisons can be obtained only avoiding biases
in the project design and operation. Through simulated changes in
weather conditions, it was determined that the principal component
session n techniques were used to evaluate selected rain and hail modi-
fication projects, revealing modification in certain projects and none in
:;ers. Various relevant issues have been examined such as use of other
their r variables (covariates) to increase detection power, the validity
of using historical data as controls for discrete operational periods,
-sible randomization options during cloud seeding operations, and
...lyses of individual rain events versus that based on monthly or sea-
o nits.
TERMS: weather modification; cloud seeding; precipitation
.iange; statistical evaluation.)

Since the historical experiments of Langmuir's CIRRUS
oject (1 48), it has been recognized that modifying weather
cloud seeding may be a viable option for supplementing
deficient natural precipitation, and hence may be a useful tool
water resources planning. The potential beneficiaries of in-
creased precipitation cover a wide spectrum of activities such
agriculture, power generation, urban water supply, irriga-
..*n, forest -fire control, and recreation. For example, the
mporal distribution of rainfall is critical to certain farming
.,vities and has led to much interest in cloud seeding by
agricultural interests.
The existence and growth of commercial weather modifica-
ion operations are generally based on a premise that weather
modification works. Consequently, evaluation of such pro-
ects has not always been addressed properly or believably. Un-
fortunately, our knowledge of the effectiveness of weather
modification at the present time is largely still a major un-

More than 30 years of scientific research on weather modi-
fication has not produced enough convincing evidence that we
can successfully modify precipitation (Changnon, 1980).
Factors responsible for this lack of evidence include poor pro-
ject design and administration, poor conduct of the projects,
bad draws in randomization, inadequacies in data collection,
poor leadership, and most critically, insufficient funding both
at the federal and state levels (Changnon, 1980). Neverthe-
less, many operational cloud seeding projects have been con-
ducted over the years covering a vast portion of the U.S.
(Figure 1).
As discussed in detail by Changnon, et al. (1)79), the
evaluation of operational cloud seeding projects can provide
useful scientific information. Another reason for the evalua-
tion of operational cloud seeding projects is to provide infor-
mation needed by the operational community. This com-
munity is calling for the standardization, development, and
adoption of more stringent operational criteria and instru-
mentation, which in turn will make it possible to provide bet-
ter evaluations. A third reason is that the user group wants to
know the cost-effectiveness of their paid seeding options
The Secretary of Commerce's Weather Modification Advisory
Board (1978a) and its Statistical Task Force (1978)) iccom-
mended that considerable attention be given to the evaluation
of selected operational projects.
The lack of predictability of natural precipitation is a ke)
factor which hinders evaluation of operational cloud seeding
projects. Operational cloud seeding projects are sufficiently
complex to require a sizable scientific and technical effort for
a thorough assessment. A multi-year research project currently
in progress (Operational Seeding Evaluation Techniques) is
aimed at study of means of evaluating operational weather
modification projects. Its primary goal is to devise innovative
approaches for the evaluation of weather modification'opera-
tional projects and to focus on statistical-physical techniques.
Objectives of the project include evaluation of selected opera-
tional projects to test the techniques developed, development
of criteria needed for evaluation criteria, and recommenda-
tions for planning of future operational projects to facilitate
their evaluation.

'Paper No. 82137 of the Water Resources Bulletin.
'Respectively, Professional Scientist and Chief; Illinois State Water Survey, 605 E. Springfield, P.O. Box 5050, Station A, Champaign, Illinois 61820.




I J.~ ~A



Hsu and Changnon

Figure 1. Locations of Research and Operational Weather Modification Projects in the United States, 1973-1977.
(Two operational projects were in south-central Alaska and no project in Hawaii; from WMAB, 1978a.)

This paper presents a review of the primary results of this
pioneering project. Such a review is considered useful to wa-
ter resources interests who must consider and decide on the
use of operational weather modification projects and how to
evaluate them. The interested reader can obtain more details
on methodology and other results from the cited publications.
The paper first treats the development and testing of statis-
tical techniques, and then the results from the application of
the techniques to selected cloud seeding projects are described.
The uses of meteorological covariates to aid in evaluation are
then presented, followed by findings relating to the validity
of use of historical data as the control for evaluation. Records
of operations and data collection in projects are critical and
criteria developed for this are described. Two important ap-
proaches to better evaluation of future projects, randomiza-
tion, and evaluation of weather events are then presented.

A number of statistical-physical techniques to evaluate
nonrandomized weather modification projects were com-
pared, primarily through extensive simulation testing, by

superimposing assumed seeding effects upon natural precipiti-
tion distributions. The techniques selected for comparison
included multiple regression (MR), two simple regression
(2Reg), principal component regression (PCR), double rati
(DR), and sum of rank power test (SRP). Detailed description,
and relevant issues regarding these techniques can be found ,
Hsu, et al. (1981a). The test statistic used in MR and PCR
was the mean differences between the observed and predict
target values in the seeded period. Statistical power (Hsli
1979) was the main index used for the comparison.
Five data sets from four areas were selected for simulatta.,
(Figure 1),. They included monthly (May-September) and se-
sonal rain data for southwestern Kansas and east central I
linois; annual hail (insurance) loss-cost values in central M!,
tana; and 48-hour and storm total rain data for southwestern
Illinois. The techniques were all applied to the simulated pre
cipitation changes.
Findings from the simulation studies revealed that in tiI
Kansas rain simulation the principal component regressio.
retaining the first component (PCR[1]) was one of the moi
powerful techniques for various summer months and tareg
control designs. In the east-central Illinois (ILL-EC) rain;
study, PCR[1] and double ratios were generally the mo'


1 ( __~__~II1

SOnlt valuation of Operational Cloud Seeding Projects f

powerful; and the two regressions method was the next most
B ful.. he results of the ILL-EC simulation, when com-
with those of the Kansas simulation, indicated that
PCRl] had high powers in both simulations in every month
except June, when only the double ratio had high power (in
the averaged-target simulation). In the Montana hail suppres-
sion simulation, for the largest target area, the principal com-
ponent regression with 3 components (PCR[3]) was the most
powerful in both 3-seeded-year and 6-seeded-year simulations.
For smaller targets PCR[3] worked well only in the 3-seeded-
year study, while DR was most powerful in the 6-seeded-year
study, followed closely by PCR[3] and the sum of rank power
test. SRP had high powers in the 3-seeded-year study only
when the assumed seeding effect was large. In the Illinois-
storm simulation, results from using the constant seeding-
induced increases revealed that PCR[I] and MR were the two
most powerful techniques. SRP, MR, or PCR[1] were the
more powerful techniques when the nonconstant seeding ef-
fect models were assumed (Hsu, et aL, 1981a). In the Illinois
48-hour rain period simulations, MR was the most powerful
technique and PCR[1] was a close second. The double ratio
hid high powers in some tests.
The results of these tests are summarized in Table 1. In
general, PCR([1 was the best with convective rainfall, and
PCR[3] worked best with hail.

TABLE 1. Results of Weather Modification Simulations
Sand Tests of Statistical Techniques.
Most Powerful
Seeding Simulation Areas Statistical Techniques*

Kansas Rain (monthly and seasonal) PCR(1)
Eastern Illinois Rain (monthly and seasonal) PCR(1) and DR
Montana Hail (seasonal) PCR(3)
southern Illinois Storms
Constant Increases PCR(1) and MR
Varying Increases PCR(l), MR, and SRP
Southern Illinois 48-Hour Rains MR and PCR(I)



Principal component regression.
Retaining the first component.
With three components.
Multiple regression.
Double ratio.
Sum of rank power test.

Past commercial seeding projects were reviewed to deter-
mine a few suitable for testing using the better statistical-
physical techniques. Suitability was based upon local climate,
Il of the project, goal of seeding (rain enhancement and/or
hai suppression), and adequacy and availability of data. Rain
enhancement projects selected for testing included several
small-scale airborne-seeding projects operated in Illinois within

the past seven years, and one project using ground-based
generators in northwestern Oklahoma.
For testing the techniques developed in the Montana hail
study, a hail suppression project carried out in the Texas Pan-
handle during 1970-1976 was selected. In addition, a com-
bined hail Suppression and rain enhancement project in south-
western Kansas during the warm seasons of 1975-1979 (the
Muddy Road Project) has been selected for evaluation. This
relatively large-scale project encompassed 15 counties (21,200
For the five small-scale rain enhancement projects in Illinois
(Changnon iand Hsu, 1981); Hsu and Changnon, 1981), results
generally reflected mixed outcomes. Results for two of the
projects (years) indicated increases in the target rainfall and/or
radar echoes (39 percent in 1976 rated as nonsignificant using
a target-control comparison; 39 percent in 1979 which was
found to be significant at the 10 percent level using a target-
control comparison involving historical data). For the 1979
seeding results, see Figure 2 as well as the results of Hsu and
Changnon (1981). A project in another year (1978) showed
a 29 percent rain decrease found to be statistically (nonsignifi-
cant). The radar echo results were also mixed. In all instances,
regardless Of the apparent increases or decreases in rainfall or
echoes in the target areas, the 1-year (2-month duration) pro-
jects were too short to draw any conclusions that have statisti-
cal or physical significance when taken alone.
The Mitddy Road Project in southwestern Kansas was
evaluated using both monthly rain and annual crop-hail loss-
cost data (evaluation using daily and storm event data is des-
cribed in a later section). The evaluation of the hail suppres-
sion data indicated that there was a reduction of annual hail
loss-cost values during the 1975-1979 seeded period compared
to the 1948-1971 period; however, only the reduction in the
eastern subiarget was significant at the 10 percent significance
level (Figure 3). This example also demonstrated that th:
principal component regression is a better technique for
evaluating hail suppression than the multiple regression (Hsu,
et al., 198 a). On the other hand, the statistical evaluation on
the monthly and seasonal rainfall observations indicated that
there was a minor nonsignificant reduction of rainfall in the
target area during the seeded period.
For the evaluation of a 1970-1976 operational hail suppres-
sion project in the Texas Panhandle, a technique of rotated
factor analysis was applied to a 12-county data set for the
period of 1947-1976 (Hsu, et al., 1981a). Seven factors ex-
plained 91 percent of the total hail loss-cost variance. From
the loading matrix it was found that the two target (seeded)
counties were both heavily loaded on one of the factors, Fac-
tor 4. Furthermore, both Welch's t-test and the Mann-Whitney
test showed that only Factor 2 (south controls) and Factor 4
(target) displayed decreases which were significant at the 5 per-
cent significance level between the historical and seeded factor
scores. The two target counties together showed a significant
reduction of more than 40 percent in hail loss-cost.



Figure 2. Total Rainfall (inches) in 1979 Operational Period of Weather Modification, Southeastern Illinois.

Figure 3. Ratio of Hail Loss-Cost, 1975-1979 Average
to 1948-1971 Average, Muddy Road Project.

The use of surface meteorological covariates to aid evalua-
tion has long been advocated. Basically, the hope is to combine

those atmospheric variables such as pressure changes, cloudi-
ness, and winds with surface rain (or hail) data to improve Lth
detection of significant changes from seeding. Chadges de-
tected under certain weather conditions also have porren-
for improving our understanding of the physical processes alt
Gabriel (1981) illustrated how the significance level (P-
value) of the test changes with increased correlation when
using one covariate. He showed that at the 5 percent level, the
P-value increases 0.04 and 0.06, respectively, when the corre-
lation coefficient between the response variable and covariate
is 0.2 or 0.3.
We collected and studied a number of appropriate meteore-
logical covariates (Achtemeier, 1981; Westcott, 1979). These
covariates consisted of 24 meteorological variables, each cal
culated for a mesh of 7 by 9 grid points centered around a
dense raingage network in the St. Louis area (Hsu, 1981a).
Rainfall variables observed in the raingage network. during
five summers (1971-1975) were used as response variables in
simulation studies for testing whether the inclusion of the
covariates can improve power of the techniques in evaluating
cloud seeding.
Findings indicated that, by using a two-stage stepwise re-
gression to screen out spurious covariates, powers were high'"
than when no covariates were used. For instance, the improve-
ment of powers was in the range of 0.01 to 0.10 at a 5 percent
nominal significance level, and the larger the seeding effect,


il_ ,;~ (

11 I ~ I J L 1 1 I .I'd.i.. ib

On valuation of Operational Cloud Seeding Projects

tht greater the improvement (Hsu, et al., 1981a). The co-
es found to be useful were the geostrophic and observed
..d directions, divergence, geostrophic moisture advection,
moisture advection and divergence, and sky cover. It was
also found that using the principal components derived from
each meteorological variable before the screening yielded
slightly higher power values (1-2 percent) than using the point-
value variables. For the meteorological variables to be useful
covariates in evaluation, we concluded they have to be screened
and the spurious variables removed; otherwise, no improve-
ment, or even reduction of power values may result.

One of the major conclusions of the statistical testing was
the absolute need to use historical data, defined as that col-
lected prior to seeded periods, as part of the evaluation. We
studied variations in historical data to ensure the validity of
using temporal comparisons (seeded period data versus histori-
cal data). The question, put simply, is whether a.given x-year
long cloud seeding project might sample a period unrepresenta-
tive of conditions in the historical data period. Eight precipita-
tion stations with 100-year records were analyzed by com-
paring the fluctuations in the 3-, 4-, 5-, and 10-year periods
within their records.
A number of statistical tests were applied, and the result-
j distributions of P-values were compared between pairs of
Sns (Hsu, etal., 1981a); Gabriel and Petrondas, 1981). No
striking evidence of departures from uniformity was apparent.
At most, there was a very vague suggestion of a tendency to
smaller P-values than expected from uniformity. On the other
hand, a high significance was discovered for large P-values with
excessive dispersion. The comparison of 5-year values sug-
gested that the standard tests were slightly radical, whereas
10-year comparisons yielded apparently conservative tests. We
concluded that the standard statistical techniques of comparing
operational period precipitation with historical precipitation
are not blatantly invalid.

Assessment of the assumptions in the statistical technique
tested and investigations of several operational projects,
coupled with discussions with several statisticians, led to the
conclusion that widely accepted evaluations of operational
weather modification projects rested on the quality and avail-
ability of records regarding the project. This included a need
for preproject statements as to the goals and specific objectives,
plus the modification hypotheses to be followed, and the ex-
plicit reasons and timing for initiating and terminating all
cloud seeding periods. During the project there is an absolute
ad to record such decisions, and to collect detailed data on
SIerations (aircraft flights, radar calibrations, seeding times,
etc.). We assessed these needs and developed criteria for future
projects (Huff and Changnon, 1980). This provides guidance
to modifiers, sponsors, and evaluators about the records needed

for achieving effective, reliable evaluation of seeding results
(and, consequently, establishing credibility in these evalua-
tions), and for providing scientific information leading to bet-
ter understanding and greater skills in future weather modifi-
cation operations. Needs for four tasks are discussed in the
report design, selection of seeding criteria, conduct of seed-
ing mission, and collection and recording of data. A number
of key issues and recommendations covered include personnel
required for operation, seeding criteria, requirements for
operations for different types of seeding, needs for radar and
other field instrumentation, and requirements for detailed
documentation of operational activities.

The addition of scientific measurements to operational
weather modification projects has been referred to as "piggy-
back" science (Weather Modification Advisory Board, 1978b).
Because of the potential of piggyback efforts as a viable means
of conducting scientific experiments in conjunction with com-
nmrcial operations, efforts were exerted to obtain a better
understanding of their feasibility in future projects (Gabriel
and Changnon, 1982).
Some form of randomization was considered essential in
the envisioned piggyback-type experiment. Inventive ways to
incorporate randomized research in future projects were
studied. Two types of piggyback projects appear scientifically
possible and potentially acceptable to the users of operational
weather modification. The first type utilizes partial ran-
domization on some (rainfall) occasions before, during, and
after the designated operational seeding period. The weather
and climate conditions before or after would have to be very
similar to those of the designated seeding period. Also, mini-
mal randomizations would have to be undertaken, if at all
possible, during the period of user need for modification.
The second type of potential piggyback research would
employ randomization only during seeding operations, with
sufficient frequency (percentage of seedable situations) to
provide adequate statistical data for reliable evaluations. For
instance, different seeding agents or different seeding rates
could be compared. Combination of both the first and second
lype of piggyback approaches is also feasible. Two examples
of possible piggyback experiments and outlines of the experi-
mental designs were explicitly given by Gabriel and Changnon

Another major effort of the project involves the evaluation
of projects based on use of daily and storm-by-storm synoptic
weather information, as opposed to monthly or seasonal data.
This event approach offers a more meaningful scientific assess-
In our project we have used the daily and storm event data
to assess the Muddy Road Project in western Kansas (1975-


-A06, ,,


Hsu and Changnon

1979) and the northwestern Oklahoma ground-based-proiect
(1972-1976). For each seeded occasion, storm motions) and
synoptic weather type(s) were assigned by study of radar echo
characteristics and surface weather maps. Low level wind
directions were determined for the Oklahoma seeded events to
track the seeded material. We studied approximately 150 days
from the Muddy Road Project, and about 110 days (40 per-
cent randomly selected days) from the Oklahoma project.
In the Muddy RoadProject, control areas similar in size to
the target area were chosen along radials of direction from the
center of the target according to the storm motion. The mean
target rain was found to be larger than that in the mean con-
trol rain in storms moving from the WSW (73 percent), SW
(63 percent), SSW (38 percent), NW (38 percent), and WNW
(19 percent); but the target less in storms moving from W
(-7 percent). The target had more rain than the control in
storms with certain synoptic types including air mass (111
percent), squall line (101 percent), stationary front (79 per-
cent), and cold front (4 percent). The target receiving less
rain in squall zone storms (-10 percent) and post-front storms
(-26 percent). Overall, the rain increase in the target (com-
pared to the control) was 27 percent, Furthermore, a 30-year
rainfall average (1941-1970) was used to normalize the target
and control area rain to adjust for the climatic rainfall gradient.
The percentage differences between the normalized target
rain and the normalized control rain.is slightly smaller than
the unnormalized rain. They were, respectively, 55 percent for
SW rain motions, +43 percent for SSW, +38 percent for WSW,
S +8 percent for NW, -11 percent for WNW, and -31 percent
for W. The corresponding rain differences for the synoptic
weather types were +97 percent for stationary front, +63 per-
cent for air mass, +58 percent for squall line, -20 percent for
cold front, -23 percent for squall zone, and -46 percent for
post-front. Overall, the differences between the normalized
target and the control rain was +14 percent.
Isohyetal patterns of the Oklahoma project revealed a note-
worthy feature: a relatively strong high was centered approxi-
*mately 90 km east of the target area. This high was pro-
nounced in rain events moving from the west, events associated
with stationary fronts, events in which the low-level winds
(plume winds) were from the SE and NE, and for rain events
moving from the W with SE plume winds. The differences
between the total rainfall. in the target area and the easterly
high rain center. suggest the possibility of a downwind seeding
relationship between seeding agent release at the ground and
cloud reaction to the seeding input. However, it was not pos-
sible to prove this supposition unequivocally with the available
data, and the high degree of natural variability found in warm
season rainfall in this climatic region. For the majority of the
stratifications, there is no indication of increased rainfall in the
Oklahoma target area with respect to the surrounding area.
Further study of the Oklahoma data plus data from other
ground seeding projects would be desirable and necessary to
ascertain with a higher degree of reliability the efficacy of
ground seeding and possible downwind effects.
This approach holds great promise for use in future opera-
tional projects. It requires careful documentation of the

operational decisions, actual field operations, and collection of
all possible meteorological data.

The fundamental principle of any evaluation of seeding
effects is that of "comparisons" be they temporal, spatial,
or both (Hsu, et al., 1981b). Factors relevant to an evaluation
include choice of response variables, sampling unit, project
type (rain, snow, or hail), period used for data formation, tar-
get control design, sample size (seeded and unseeded), co-
variates, sizes of (target and controls areas), assumed seeding
effect (additive, multiplicative, or varying), and statistical
techniques employed. Rerandomization testing was recom-
mended for the actual evaluation of operational cloud seeding
projects (Gabriel, 1979; Hsu and Changnon, 1981; Hsu, et a,
Valid comparisons can be obtained only if one avoids cer-
tain obvious biases, advertent or inadvertent. Elements essen-
tial to better evaluations are: using uniformly defined observa-
tions for seeded and nonseeded occasions; using a uniform
method in measuring response variables; selecting target con:
trol design and statistical technique a prior; adhering rigorous
ly to a preordained protocol; and documenting the seedir;
operation as completely as possible, including seeding criteria~
used, methods of collecting and recording data. seeding ageni
amount and timing, instrumentation, and seeding rate. It ap-
pears that useful evaluations of operational projects can be

This work was supported by grants ATM 79-05007 and ATM S1-
07027 from the National Science Foundation. The opinions and cr -
clusions expressed here are those of the authors and do no: necessanl.
reflect the views of NSF. Appreciation is extended to Floyd A. HuI-
and Ruben Gabriel for their many useful suggestions and good ad% w'-.
to Julia Chen, who provided computer programming assistince:-and ,
Bob Scott, who performed the synoptic analyses of the Muddy Ro-.t
and Oklahoma projects.

Achtemeier, G. L., 1981. Relationships Between Rainfall and te le'; -
logical Surface Variables: Applications to Midwestern Operational
Weather Modification. Final Report, NSF ATM79-05007, Ilhnor
State Water Survey, Champaign, Illinois, 69 pp.
Changnon, S. A., Jr., 1980. The Rationale for Future Weather Modifi-
cation Program. Bull. Amer. Meteor. Soc. 61(6):546-551.
Changnon, S. A., Jr. and C. F. Hsu, 1981. Evaluation of Illir, is Wealher
Modification Projects of 1976-1980: A Summary. Circular 148,
Illinois State Water Survey, Champaign, Illinois, 31 pp.
Changnon, S. A., Jr., F. A. Huff, and C. F. Hsu, 1979. On the Need to
Evaluate Operational Weather Modification Projects. Bull. Amer.
Meteor. Soc. 60(7):77-774.
Gabriel, K. R., 1979. Re-randomization Tests on Weather Experiments.
Sixth Conf. on Planned and Inadv. Wea. Mod., Amer. Meteor. Soe.,
Champaign, Illinois, 10-13 October, pp. 15-J6.


ii.a. I

_AL4AA,-.L,1- 1-1- Al

'* f

the Evaluation of Operational Cloud Seeding Projen"

1 Gabriel, K. R., 1981. On the Roles of Physicists and Statisticians in
ShU.,'ler Modification Experimentation. Bull. Amer. Meteor. Soc.
Gabriel, K. R. and S. A. Changnon, Jr., 1982. Piggyback Weather Ex-
periments: Superimposing Randomized Treatment Comparisons.
J. Wea. Mod. 14(1):7-10.
ing Gabriel, K. R. and D. Petrondas, 1981. On Using Historical Compari-
ai, sons in Evaluating Cloud Seeding Operations. Tech. Rept. 81/12,
n Dept. of Statistics, University of Rochester, 6 pp.
S Hu, C. F., 1979. Two Methods of Computing Statistical Powers With
'ct Application to Weather Modification. Proc. Statist. Comp. Sec.,
Amer. Statist. Assoc., Washington, D.C., pp. 243-246.
co-- Hsu C. F., 1981a. An Annotated Bibliography on the Evaluation and
Statistical Issues of Weather Modification. Contract Report 265,
Illinois State Water Survey, Champaign, minois, 93 pp.
ibu, C F., 1981b. Using Surface Meteorological Covariates to Predict
M11 Storm Area Rainfall 1 to 3 Hours in Advance. Proceedings 7th Con-
ing ference on Probability and Statistics in Atmospheric Sciences, Amer.
Meteor. Soc., Boston, Massachusetts, pp. 217-221.
hru, C. F., and S. A. Changnon, Jr., 1981. Assessment of Summer 1979
Weather Modification Effort in Southeastern Illinois. J. Wea. Mod.
er- 13(1):132-140.
i- h, C. F., S. A. Changnon, Jr., F. A. Huff, and K. R. Gabriel, 1981a.
Iva- The Assessment of Statistical-Physical Techniques for the Evalua-
.rm tion of Weather Modification Operations. Contract Report 286,
Illinois State Water Survey, Champaign, Illinois, 90 pp.
I Hm C F., K. R. Gabriel, and S. A. Changnon, Jr., 1981b. Statistical
Techniques and Key Issues for the Evaluation of Operational Weather
ing Modification. J. Wea. Mod. 13(1):195-199.
ria Haft. F. A. and S. A. Changnon, Jr., 1980. Criteria for Weather Modifi-
: cation Operations and Effective Evaluation. Contract Report, NSF
SATM 79-05007, Illinois State Water Survey, Champaign, Illinois,
22 pp.
LIagmuir, L, 1948. The Production of Rain by a Chain Reaction in
Cumulus Clouds at Temperatures Above Freezing. J. Meteor. 5(5):
weather Modification Advisory Board, 1978a. The Management of
Weather Resources. Volume I: "Proposals for a National Policy and
S 3 3~Re.-" : D-. f C:'r-

R*- esar.s '?r'L-r rI-'- --C. Ssa' T;i F::e.
Dept. of Commerce, Washington, D.C.
WNestcott, N., 1979. Annotated Bibliography of Predictor Variables for
Ad Weather Modification Applications. Technical Report, Illinois State
Water Survey, Champaign, Illinois, 117 pp.


l -



University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs