You are on page 1of 131

INFORMATION TO USERS

This manuscript has been reproduced from the microfilm master. UMI
films the text directly from the original or copy submitted. Thus, some
thesis and dissertation copies are in typewriter face, while others may be
from any type o f computer printer.
The quality of this reproduction is dependent upon the quality o f the
copy subm itted.

Broken or indistinct print, colored or poor quality

illustrations and photographs, print bleedthrough, substandard margins,


and improper alignment can adversely affect reproduction.
In the unlikely event that the author did not send UMI a complete
manuscript and there are missing pages, these will be noted.

Also, if

unauthorized copyright material had to be removed, a note will indicate


the deletion.
Oversize materials (e.g., maps, drawings, charts) are reproduced by
sectioning the original, beginning at the upper left-hand comer and
continuing from left to right in equal sections with small overlaps. Each
original is also photographed in one exposure and is included in reduced
form at the back o f the book.
Photographs included in the original manuscript have been reproduced
xerographically in this copy. Higher quality 6 x 9 black and white
photographic prints are available for any photographs or illustrations
appearing in this copy for an additional charge. Contact UMI directly to
order.

UMI
I

A Bell & Howell Information Company


300 North Zeeb Road, Ann Arbor MI 48106-1346 USA
313/761-4700 800/521-0600

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w ith o u t perm ission.

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

ECONOMIC SIMULATION AND OPTIMIZATION OF IRRIGATION


WATER IN HUMID REGIONS

Ram N. Acharya

A Dissertation
Submitted to
the Graduate Faculty of
Auburn University
in Partial Fulfilment of the
Requirements for the
Degree of
Doctor of Philosophy

Auburn, Alabama
December 15, 1997

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w ith o u t perm ission.

UMI Number: 9811941

UMI Microform 9811941


Copyright 1998, by UMI Company. All rights reserved.
This microform edition is protected against unauthorized
copying under Title 17, United States Code.

UMI

300 North Zeeb Road


Ann Arbor, MI 48103

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

II
I

ECONOMIC SIMULATION AND OPTIMIZATION OF IRRIGATION


WATER IN HUMID REGIONS

Ram N. Acharya

Certificate o f Approval:

Neil R. Martin, Jr.


Professor, Agricultural
Economics and Rural Sociology

L. Upton Hatch, Chair


Professor, Agricultural
Economics and Rural Sociology

Gregorys. Traxler
Associate Professor, Agricultural
Economics and Rural Sociology

F. Pntchett
Dean, Graduate School

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

Style manual or journal used: American Journal of Agricultural Economics

Software used: Corel Word Perfect 7.0. GAMS. Lotus Release 5. SAS 6.11 for
Windows. EPIC__________________________________________________________

i
R eproduced w ith perm ission o f the copyright owner. F urth er reproduction prohibited w itho ut perm ission.

VITA

Ram N. Acharya, son of Janardan and Leelawati Acharya, was bom in a remote
village in Nepal, from where he started his journey to reach his ultimate goal- to achieve
a Ph.D. from the land o f opportunity, the United States of America. He finished his
schooling, and tried his hand at Engineering. Two years later, he decided that his career
was else where so he entered the college for business and obtained a degree. Upon
completion, he joined the working population in order to prepare himself for further
education. He then entered Tribhuvan University for his Masters degree in Economics
and graduated as a gold medalist in the faculty. After graduation, he taught economics in
the Shaker Dev Campus and the Kirtipur Campus of Tribhuban University, Nepal. He
was, then, granted scholarship by Winrock International to pursue his academic career in
Malaysia. He completed his Masters Degree in Natural Resources and then few months
later arrived at Auburn University, Alabama, as a doctoral student in Agricultural
Economics.

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w ith o u t perm ission.

DISSERTATION ABSTRACT
ECONOMIC SIMULATION AND OPTIMIZATION OF IRRIGATION
WATER IN HUMID REGIONS

Ram N. Acharya
Doctor of Philosophy, December 15, 1997
(M.S., University Pertanian Malaysia, 1993)
(M.A., Tribhuvan University, 1988)
(B.M., Tribhuvan University, 1985)
127 Typed Pages
Directed by: Dr. L. Upton Hatch

A combination of simulation, econometric, and optimization models were used to


develop a methodology that can be applied to examine the effect o f water scarcity on net
farm income. Since both quantity, as well as, timing of irrigation has important bearing
on crop yield, a biophysical simulation model, known as Erosion Productivity Impact
Calculator (EPIC), was used to simulate the relationship between water, from rainfall
and/or various irrigation management practices, and crop yields. The water yield
response functions for selected crops were estimated using EPIC simulated data. The
estimated parameters o f these yield response functions were used to develop a recursive
stochastic linear programming model.

iii

I
__

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

In the resulting model, rainfall was treated as a stochastic process, which follows
its own historical pattern randomly. This capability to deal with an erratic rainfall
process distinguishes the recursive stochastic linear programming model from other
multi-period programming models. The optimization model was designed to solve a
series of irrigation decisions problem faced by a model farm, which is growing com,
cotton, and peanuts. Different irrigation decision rules were derived for dry and normal
weather conditions. Using a preference scale and the flow data of Chattahoochee river
measured at Columbus, the residual flow that could be used for crop irrigation was
calculated. The results indicated that even if the historical flow could be maintained in
the future, it would not be enough to meet the total irrigation demand in many instances.
The aggregate optimal demand for irrigation water in the Middle Chattahoochee
Sub-Basin was estimated to be 3.211 million gallons per week. The contribution of this
optimal irrigation level to net farm income would be $ 1.175 million per year for dry
years and $ 0.711 million for normal years. Thus the aggregate impact of water shortage,
measured at the optimum use level, would be higher in dry years by $ 0.464 million as
compared to normal years. Since the impact of water scarcity on net farm income is
expected to be much higher in dry years than in normal years, two separate marginal
relationships were estimated. Once the existing supply and weather conditions are
known, these marginal functions can be used to derive the impact of reduced stream flow
on net farm income. For the Middle Chattahoochee Sub-Basin, the average impact of a
15 percent draw-down in downstream flow on net farm income was calculated to be less
than $3.10 per acre in dry years and $ 0.57 per acre in normal years.
iv

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w ith o u t perm ission.

ACKNOWLEDGMENTS

The author extends genuine appreciation to Dr. Upton H atch, Dr. Neil R. Martin,
Jr., and Dr. Greg Traxler for their guidance throughout the graduate years. It would not
have been possible to finish this work without their support and confidence. I would also
like to thank God for all the knowledge gained and used during this four year period.
Knowledge is light to the world and He is the creator of this light.
Above all, I would like to express earnest appreciation to my wife, Anita and son
Ajju for their understanding and encouragement throughout the stressful years. Without
their love, sacrifices and indefinite support, the work would never have been completed.
This dissertation is dedicated to my father, Janardan Acharya, who taught me to
respect education and believe in its strength.

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

TABLE OF CONTENTS
Page
LIST OF TABLES....................................................................................

ix

LIST OF FIGURES..................................................................................

xi

LIST OF M A PS........................................................................................

xiii

INTRODUCTION....................................................................................

Methods and Procedures...............................................................


ACF River Basin...........................................................................
The Study A rea.............................................................................
Objectives.....................................................................................
Rational and Significance.............................................................

3
4
7
11
12

LITERATURE REVIEW.........................................................................

13

Introduction..................................................................................
Water Scarcity and Sources......._.................
Basinwide Water Resource Management Models.......................
Plant Water Relationship..............................................................
Irrigation Demand Studies............................................................
Yield Response to W ater..............................................................
Optimal Allocation of Irrigation W ater.......................................

13
14
15
18
20
23
25

RESEARCH METHODOLOGY.............................................................

29

Conceptual Framework................................................................
Methods and Models....................................................................
The EPIC M odel..........................................................................
Recursive Programming..............................................................
Recursive Stochastic Linear Programming................................
Specification of the RSLP Model
________

29
31
32
33
36
41

VI

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

RESULTS AND DISCUSSION...............................................................

43

Simulation Results........................................................ ,..............


Yield Response Functions...........................................................
Marginal Physical Productivity of W ater....................................
Marginal Value of Irrigation Water..............................................
Dry and Normal Weather Conditions..........................................
Irrigation Decision Rules..............................................................
Optimal Demand for Irrigation Water..............................
Aggregate Irrigation Demand.......................................................
Historic Flow of Chattahoochee River........................................
Impact of Water Shortage on Net Farm Income..........................
WATER MANAGEMENT ISSUES...........................................

44
45
51
54
54
57
64
67
71
75
83

Land Allocation/Cropping Pattern................


Adoption of Water Saving Technologies........................
The Impact of Reduction in Stream Flow .......................

85
86
87

SUMMARY AND CONCLUSIONS......................................................

93

Limitations and Recommendations.............................................

99

REFERENCES..........................................................................................

101

APPENDIX- 1 ..........................................................................................

107

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

LIST OF TABLES

1. Summary o f Farming Activities by County in the Study A rea

2. Major Crops Grown in 1996......................

3. Population Distribution by County........................................................

10

4. Number of Farms, Total Land in Farms, and


Average Farm Size by County and Y ear..............................................

10

5. Average Crop Yield and Water Stress Levels......................................

46

6. Yield Response Functions for Com, Cotton and Peanut .....................

50

7. Average Marginal Physical Productivity o f Com, Cotton, and Peanut

52

8. Marginal Value of Irrigation Water by C rops.......................................

55

9. Ranking of Marginal Value of Irrigation Water by Crops....................

55

10. Optimal Allocation of Irrigation Under Water Scarcity


(Supply=l acre-inch)..........................................................................

58

11. Optimal Allocation of Irrigation Under Water Scarcity


(Supply=2 acre-inch). ..............................................................

58

12. Optimal Allocation of Irrigation Under Water Scarcity


(Supply=3 acre-inch)..........................................................................

59

13. Optimal Allocation of Irrigation Under Excess Water Supply


(Supply=6 acre-inch)..........................................................................

59

14. Optimal Demand for Irrigation Water by Crop and Growth Stage....

66

15. Total Optimal Demand for Irrigation and Rainfall by Growth Stage

66

viii
|

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

16. Total Soil Moisture Available to the Plants at Different Growth Stages

68

17. Irrigation Demand Estimates for the Middle Chattahoochee


Sub-Basin by C rop..............................................................................

68

18. Aggregate Irrigation Demand for the Middle Chattahoochee Sub-Basin

70

19. M&I and Irrigation Demand Projections for the ACF


Basin and the Study A re a ...................................................................

70

20. Incidence o f Water Shortage in Meeting Future M&I


and Irrigation Demand (r=0)...............................................................

74

21. Incidence o f Water Shortage in Meeting Future M&I


and Irrigation Demand (r= l) ...............................................................

74

22. Expected Incidence o f Water Shortage To Meet M&I


and Irrigation Demand (r=0)...............................................................

76

23. Expected Incidence of Water Shortage To Meet M&I


and Irrigation Demand (r = l) ...............................................................

76

24. Summary o f Results from the Optimization Model (RSLP)...............

78

25. Average Net Farm Income and Marginal Impact of Water Scarcity..

79

26. Aggregate Net Farm Income at Various Supply Levels.....................

79

27. Scenario: 1 Incidence of Water Shortage in Meeting Future


Irrigation Demand (r=0)......................................................................

91

28. Scenario: 1 Incidence of Water Shortage in Meeting Future


Irrigation Demand (r= 1 )......................................................................

91

29. Scenario: 2 Incidence of Water Shortage in Meeting Future


Irrigation Demand (r=0)......................................................................

92

30. Scenario: 2 Incidence of Water Shortage in Meeting Future


Irrigation Demand (r= l)......................................................................

92

ix

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

LIST OF FIGURES

1. The Recursive Programming M odel............................

35

2. Com Yield and Water Stress Levels.....................................................

46

3. Cotton Yield and Water Stress Levels..................................................

47

4. Peanut Yield and Water Stress Levels..................................................

47

5. AMPP of Irrigation at Various Plant Growth Stage (com )..................

52

6. AMPP o f Irrigation at Various Plant Growth Stage (cotton)...............

53

7. AMPP o f Irrigation at Various Plant Growth Stage (peanut)...............

53

8. Marginal Value of Irrigation for Normal Y ear....................................

56

9. Marginal Value of Irrigation for Dry Years.........................................

56

10. Optimal Allocation of Irrigation in Normal Years


(Suppiy=2 acre-inch)...........................................................................

60

11. Optimal Allocation of Irrigation in Dry Years


(Supply=2 acre-inch) ........................

60

12. Optimal Allocation of Irrigation in Normal Years


(Supply=3 acre-inch)..........................................................................

61

13. Optimal Allocation of Irrigation in Dry Years


(Supply=3 acre-inch)..........................................................................

61

14. Optimal Allocation of Irrigation in Normal Years


(Supply=6 acre-inch)..........................................................................

62

15. Optimal Allocation of Irrigation in Dry Years


(Supply=6 acre-inch)..........................................................................

62

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w ith o u t perm ission.

16. Optimal Allocation of Irrigation in Normal Years


(Supply=5 acre-inch)..........................................................................

65

17. Optimal Allocation o f Irrigation in Dry Years


(Supply=5 acre-inch)...........................................................................

65

18. Irrigation & Net Farm Income in Normal Years................................

77

19. Irrigation & Net Farm Income in Dry Years......................................

77

20. Net Farm Income in Dry and Normal Years.......................................

80

21. Marginal Impact of Irrigation on Farm Income.................................

80

22. Impact o f Irrigation & Weather on Farm Income (Standard Deviation)

82

23. Impact o f Water Scarcity on Net Farm Income.................................

82

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

LIST OF MAPS

1. Apalachicola-Chattahoochee-Flint River Basin.....................................

2. Location o f Counties Within the Study Area.........................................

r
I

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

INTRODUCTION

Water has always been an important factor in determining social and economic
development. Search for new sources of water supply and construction of large dams
have been the dominant water management practice in the past. As the demand for the
fixed water resource is increasing and the possibility of developing new sources are
becoming scarce, searches for efficient management systems are being initiated.- The
conflict among water user groups in the Alabama-Coosa-Tallapoosa and ApalachicolaChattahoochee-Flint (ACF) river basins presents a case o f such need for a change in
water sharing rules. A better understanding of temporal and spatial aspects o f water
allocation issues will provide a sound starting point for negotiation to resolve water
conflicts. This study examines optimal irrigation strategies under different water scarcity
levels and plausible weather scenarios and estimates the economic impact o f irrigation
shortage on net farm income.
The water flowing in a river can be put into different uses at various locations.
Uses that involve withdrawal of water from the stream such as municipal, industrial, and
irrigation reduce flow in the river. Both quantity as well as quality of water available to
downstream users is directly affected by upstream water management decisions. As
water becomes scarce, the conflict between upstream and downstream users and among
competing uses at a particular location within the watershed increases. This increase in
1
(

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w ith o u t perm ission.

competition among uses and users within and across the geographical location over time
makes it difficult to derive optimal allocation rules. The value of water applied to various
uses at different times and places within the river basin should be known before any
optimal allocation rules can be derived.
This difficulty in deriving an objective value of water, when applied to different
competing uses, has given rise to various alternative approaches such as conflict
resolution to resolve basin-wide water allocation problems. This method attempts to
resolve water allocation disputes through various processes which may include
mediation, negotiation, and bargaining (Dinar and Loehman). Basically, this approach
aims to resolve the conflict by bringing all stakeholders together, educating them about
each others economic interests, strengths, and weaknesses, and encouraging them to
settle the conflict through consensus (sometimes mediators are used to initiate this
process).
The role of economics in this process, whenever possible, is to provide the value
o f the resource applied to its competing uses and users at different times and locations.
This information, the marginal value of resource, provides the economic basis for
deriving optimal allocation rules. Wreck et al. (hereafter referred as Comprehensive
Study) examined the impact of changes in upstream water allocation rules, in the ACTACF river basins, on downstream uses by using long term planning models. The
methodology used in the above study is not capable of handling short term issues such as
drought and floods. Although it is important to know long term prospects for developing
existing resources, short term allocation issues cannot be ignored.

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

Other things remaining the same, the value of a resource applied to a production
process depends on its ability to contribute in producing output and the value of that
output in the market. Unlike in many industrial production processes, the value of
irrigation water depends on various factors including weather conditions and stages of
plant growth. For example, the marginal physical productivity of irrigation water would
be much higher in flowering and fruit development stage than in other periods (Bruce et
al.). On the other hand, the demand for water would be relatively higher in dry period as
compared to normal weather conditions. These aspects of water demand and productivity
play an important role in short term water allocation decisions.
The methodology, developed in this study, can be used by water use
administrators in making optimal allocation decisions under various weather conditions
and water scarcity levels. Since the value of irrigation applied at different stages of plant
growth is the economic basis for allocation decision making, availability o f this
information is expected to provide impetus in resolving the existing conflict among water
user groups in the ACF river basin.
The results obtained from the farm level optimization model are extrapolated to
calculate aggregate demand for irrigation water in the Middle Chattahoochee Sub-basin.
The incidence of water scarcity is estimated using basinwide demand and supply data and
optimal water allocation rules are derived.
Methods and Procedures
A combination of simulation, optimization, and econometric models are
developed to examine the optimal allocation of irrigation water under two plausible
i

I?
R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w ith o u t perm ission.

weather scenarios, normal and dry, among three competing crops (peanut, cotton, and
com) grown in the middle Chattahoochee sub-basin. Since both timing and amount of
irrigation have important bearing on crop yield, the relationship between the amount of
irrigation water applied at various plant growth stages and crop yield must be established
for each crop before the optimal allocation o f irrigation water can be determined.
The empirical estimation of such relationships requires actual experimental data,
which are not readily available. In the absence of real world data, a biophysical
simulation model, known as Erosion Productivity Impact Calculator (EPIC), is used to
simulate the relationship between various irrigation management practices and crop yield.
The relationship between the amount of soil moisture available to the plants at different
stages of plant growth and final crop yield is specified to be log linear and is estimated by
using the simulated data. The estimated parameters of the yield response function are
used to develop a Recursive Stochastic Linear Programming (RSLP) model. Then, the
RSLP model is used to determine the optimal irrigation strategies and to estimate the
impact of water shortage on net farm income.
ACF River Basin
The ACF basin originates in northern Georgia, covers parts of Alabama, Florida,
and Georgia and drains into the Gulf of Mexico (Map 1). The total area drained by this
basin is about 19,800 square miles. Approximately 2.636 million people were living in
this basin in 1990. On average, the daily water withdrawal rate for 1990 was
approximately 2,098 million gallons, o f which nearly 86 percent was withdrawn from
surface water sources. About 17 percent of the total withdrawal was consumptively used,
i

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

' I

* I

'*

5 S

f *>

1 W V j S - ' - v S.C.

iawoi/V

...

i Omtm

j ___

/ I

t Carrot /

fc -V S v
v

<

r^ -~ 1 "W

M Q R fM

!
i

h-.~<

KM

r \

U \i > A

-A

u-! A _

\otJm V

'

^ -;,i

Trap

/ \ A

*y

>

xx

-A l "" V r h v

T*W

A -f V

J \ "* \ ---r H
\ .Iew
v'
*p*/. I;
^
- iX r
T f ~ 'V RinMpn
j/w y ^
_3---|

Doo* \

2..L canon f c ,

,e *

~J

. i> r k

.T

' Jyr
-? r

..

f - x - - ' J

| . g
r

-5

'

V-

~ I S. Townr .

? ;

w-

l*nt
7

---,

I---------- --

- i fmm r

>

L.
* /f\

COQlM

.C )
IW y y |

V . Sadsoan

>

:3

L.GE^RGIAv

_{_

>.f l o r 2 > /

--------------- 1 , CHnm

r *

H /

'- W

m:smz A
7

c u t/- O f MEXICO
SO MILES

h r r f i i I 1

50 KILOMETERS

Map 1. Apalachicola-Chattahoochee-Flint River Basin

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

6
5 percent was exported out of the basin, and 78 percent was returned back to the system.
About 223 million gallon per day (mgd) of municipal wastewater was generated and
discharged in the ACF basin. The sectoral distribution of surface water withdrawal was
as follows: power generation 60 percent, public supply 24 percent, self-supplied
commercial-industrial uses 12 percent, and agricultural uses 4 percent. Alabama shares
about 2,800 square miles of land area (14 percent), 0.19 million people (7 percent), and
183 million gallon/day (9 percent) o f total water withdrawn from the basin (Marella,
Fanning, and Mooty).
Droughts and floods are the two extreme cases of water supply that impose severe
costs on the economy. During February and March 1990, record flooding occurred at 74
sites, and 46 sites had peak level discharges that exceeded or equaled the 100-year
recurrence interval discharge (Pearman et ai). The drought of 1980-81 caused a
reduction in power generation, curtailed navigation, increased water level drawdown in
recreational lakes, and imposed restrictions on lawn watering and other uses. During
mid-summer, a lowest flow on record occurred in many streams, much earlier than the
minimums that occurred during past droughts. Discharge measurements o f zero flow
were observed at 694 non-recording stream locations in Alabama, Georgia, South
Carolina, and eastern Tennessee during the 1986 drought period. Out of 370 continuousrecord gauging stations, 99 stations experienced new record minimum daily flows in
1986. Moreover, 27 stations had 7-day minimum average flows and 11 stations recorded
90-day minimum flows with a recurrence interval of more than 50 years (Hale, Hopkins,
and Carter).

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

Hatch et al. studied farmers response to droughts in the ChattahoocheeChoctawhatchee river basins of Alabama. They observed that water is a limiting resource
for agricultural production in Alabama. They report that supplemental irrigation
increases yield, reduces yield variability, improves output quality, allows double
cropping, and helps in reducing frost damage.
The Study Area
The study area comprises the middle Chattahoochee sub-basin of ACF river basin
(Cataloging Unit Number 03130003), which includes Barbour and Russell counties of
Alabama and Chattahoochee, Muscogee, Quitman, and Stewart counties of Georgia (Map
2). In 1996, the total farmland of these six counties was 361,182 acres and the total
cropped area was 152,120 acres (Table 1). About 37 percent of the total cropped area
was irrigated.
Based on total cropped area, the major crops grown in this sub-basin were
peanuts, cotton, hay, com, and wheat (Table 2). Complete information on irrigation
status of individual crops is not reported to avoid disclosing data of individual farms.
Based on available data, cotton was the major irrigated crop followed by peanut, com,
and hay. Wheat was not irrigated at all in the 1996 crop year.
There were 276,352 people living in the middle Chattahoochee sub-basin in 1990.
About 84 percent of the population was living in urban areas and the rest in rural areas.
Most of the people o f Muscogee county lived in urban area (96.8 percent), while in
Stewart and Quitman counties, there was no urban population (Table 3).

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w ith o u t perm ission.

8
84*

S.C.

33*

32*

I___

30*

GULF OF MEXICO
SO MILES
SO KILOMETERS
29*

Map 2. Location of Counties within the Study Area

i
i

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w ith o u t perm ission.

9
Table I . Summary o f Fanning Activities by County in the Study Area
Barbour

Russell

Musco
gee

Chatta
hoochee

Quit
man

Stewart

421

213

44

16

24

97

815

Acres

177189

112620

4870

5901

11559

49043

361182

Acres

80496

36513

2006

1677

5870

25558

152120

Irrigated

24654

14127

95

321

16850

56047

Description
Total
Farm

Crop
Land

No.

Total

38.69
4.74
5.47
%
30.63
0.00
36.84
65.93
Source: http://govinfo.kerr.orst.edu/cgi-bin/imagemap/agga2752,138, Table 1, 8 and
13 for respective counties.

Table 2. Major Crops Grown in 1996


Description
Total
Com

Irrigated
%
Total

Cotton

Irrigated
%
Total

Peanut

Irrigated

Barbour

Russell

Chatta
hoochee

Quitman

Stewart

4476.00

789.00

70.00

469.00

1782.00

7586.0

492.00

307.71*

23.45*

1176.12*

1999.3

10.99

39.00

66.00

26.4

5953.00

4823.00

5.00
d

1040.00

11816.0

925.00

1640.00

686.40*

3251.4

15.54

34.00

66.00

27.5

21994.0

2324.00

1390.00

5217.00

30925.0

993.00

906.36*

69.50*

1628.00

3596.9

Total

5.00
31.21
39.00
11.6
4.51
%
Source: http://govinfo.kerr.orst.edu/cgi-bin/imagemap/agga2752,138, Table 13 for
respective counties.
Notes a The ratio between total irrigated and total cropped area was used to estimate
the proportion o f irrigated acreage because the irrigation information was not
available for these crops.
d Indicates data withheld to avoid disclosing data for individual farms.

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

10
Table 3. Population Distribution by County
Description

Barbour

Russell

Chatta
hoochee

Musco
gee

25417

46860

16934

179278

2209

5654

276352

No.

12378

30318

14614

173541

230851

48.7

64.7

86.3

96.8

0.0

0.0

83.5

13039

16542

2320

5737

2209

5654

45501

Total

Quit
man

Stewart

Total

Urban
No.
Rural
13.7
35.3
3.2
51.3
100.0
100.0
%
Source: U.S. Department of Commerce, Bureau of the Census, 1990 Census o f
Population and Housing, Special Tabulations.

16.5

Table 4. Number o f Farms, Total Land in Farms, and Average Farm Size by County
Description

Barbour

Russell

Chatta
hoochee

Musco
gee

Quit
man

Stewart

Total

1992

421

213

16

44

24

97

815

1987

498

276

13

49

25

99

960

1982

587

314

18

49

34

111

1113

1992

177189

112620

5901

4870

11559

49043

361182

1987

207906

143568

4268

5304

17655

47913

426614

1982

222066

141048

5086

11879

23354

67679

471112

1992

421

529

369

111

482

506

2418

1987

417

520

328

108

706

484

2563

687
283
1982
449
242
378
Source: Agriculture Census - Table 1 County Summary Highlights
(http ://go vinfo.kerr.orst.edu/cgi-bin/ag-list701 -005.ale).

610

2649

Number
of Farm

Total
Area
(acres)
Average
Farm
Size

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

The trend of the total number of farms, farm acreage, and the average farm size
for the period 1982 to 1992 is reported in Table 4. The total number of farms in this area
consistently decreased for all counties. The total farm acreage has decreased, except for
Chattahoochee county.
The average farm size, however, has moved in both directions. In the case of
Barbour, Russell, and Chattahoochee counties, average farm size is increasing. The
average farm size of Muscogee and Steward counties decreased substantially in 1987
from its 1982 level and then increased again in 1992. While in the case of Quitman
County, the total area under farm, the number o f farms, and the average farm size are
decreasing over time.
Objectives
This study will attempt to develop optimal irrigation strategies for three
competing crops grown in middle Chattahoochee sub-basin and estimate the marginal
value of supplemental water applied to these crops at different crop growth stages and
weather conditions. In particular, it attempts to fulfill the following objectives:

simulate various irrigation management practices and associated crop yield for
com, cotton, and peanuts grown in the Middle Chattahoochee Sub-Basin using
EPIC,

estimate the functional relationship between irrigation water applied at various


growth stages and final crop yield,

determine optimal irrigation strategies at various levels of water shortages and


growth stages for each crop, and

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

derive policy conclusions that are helpful in addressing basin-wide water


management issues.

Rationale and Significance


During recent droughts in the Southeast U.S., conflict over the allocation o f water
became more contentious. Political and legal conflict is currently proceeding between the
U.S. Army Corps of Engineers and the states of Georgia, Florida and Alabama. Data on
demand and supply o f water are being collected and analyzed for eventual inclusion in a
basin-wide management model under an institutional framework yet to be determined. In
this respect, information on the optimal irrigation strategies and marginal value o f water
applied at various stages of crop growth and weather conditions would be helpful in
making basinwide water allocation decisions. This marginal relationship can also be used
to calculate the total value of irrigation water and compare it with other competing uses
under different supply scenarios.

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

LITERATURE REVIEW

Introduction
As the twenty-first century approaches, supply of enough water of required
quality to satisfy growing demand is becoming a major challenge for many communities.
The demand for water is increasing over time but the possibility of developing new
sources is becoming more scarce. The cause o f this problem can partly be attributed to
the faulty management approach adopted in the past. Search for new sources o f water
supply and construction of large dams have been the dominant strategy rather than
making better use of existing resources (Winpenny).
Water flowing in a river system is characterized by serial technical externality,
which means that upstream users can affect downstream users but downstream users
cannot impact those upstream (Brooks et al.). Both quantity as well as quality of stream
flow is mainly determined by upstream uses. For example, municipal and irrigation uses
involve withdrawal of water from the stream, which affects both water quality and stream
flow. On the other hand, recreation, hydro-power generation, and navigation are non
consumptive in-stream uses, which may not reduce the flow of water in the river but can
change the quality of water.
The major uses of water in the middle Chattahoochee sub-basin have been
classified as municipal and industrial, hydro-power generation, recreation, and irrigation
13

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

14
(Comprehensive Study). Among these uses, whenever a serious shortfall in water supply
is encountered, the municipal and industrial water demand receives first priority. Then,
the remaining water can be allocated among other competing uses based on marginal
conditions.
Water Scarcity and Sources
Water comes from two sources: surface water and groundwater. Surface water
consists of water from rivers, lakes, ponds, and reservoirs that is stored or flows on the
earth's surface. Ground water comes from the bodies of water stored beneath the earth
known as aquifers. Unlike surface water, ground water is a finite as well as a renewable
resource. It is a finite depletable resource, when the rate of withdrawal exceeds the
natural rate of recharge. For the US as a whole, an annual withdrawal of 400 trillion
gallons is considered to be a renewable resource, while a rate of withdrawal higher than
this amount makes it a finite depletable resource (Council on Environmental Quality).
Irrigation is the largest single user of groundwater resource in the nation, which
accounts for more than 70 percent of the total groundwater use. Use of groundwater is
increasing by double the rate of increase of surface water. About a half million wells are
drilled each year (Henderson et al.). In some regions, groundwater overdraft is
accompanied by increased aquifer salinity. The contamination of groundwater will also
affect the quality o f surface water because they are closely related. About 30 percent of
the stream flow in the US is supplied by ground water (Saliba). Thus, unrestricted
pumping from the aquifer could reduce the stream flow. In particular, the effect of

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w ith o u t perm ission.

15
excessive withdrawal of groundwater would be chaotic during the growing season when
the stream flow is lower than average.
The consequences of drawdown in stream flow could be serious when it is
accompanied by a severe drought. On the other hand, heavy rainfall may result in
flooding and damage crops. As discussed before, the ACF basin experienced a record
flooding at 74 sites, and 46 sites had peak level discharges that exceeded or equaled the
100-year recurrence interval discharge in 1990 (Pearman et al.).
A severe drought was observed in 1980, which caused a reduction in power
generation, curtailed navigation, increased water level drawdown in recreational lakes,
and imposed restrictions on lawn watering and other uses. A lowest record flow was
experienced in many streams (Hale, Hopkins, and Carter).
At 694 non-recording stream locations of Alabama, Georgia, South Carolina, and
eastern Tennessee, a discharge measurement of zero flow was observed during the 1986
drought period. While in 99 continuous-record gauging station, out of 370, new record
minimum daily flows were observed. Moreover, seven-day minimum average flows
were observed in 27 stations and 11 stations recorded ninety-day minimum flows with a
recurrence interval o f more than 50 years (Hale, Hopkins, and Carter).
Basinwide Water Resource Management Models
Since water flowing in a river system is subject to serial technical externalities, a
water allocation scheme that ignores this spatial relationship between upstream and
downstream users and uses is not capable of putting water to its optimal use. To

\I

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

address this issue o f basinwide water management, many studies have used a
combination of simulation and multiobjective programming models.
The Comprehensive Study, Basin-wide Management o f Water in the AlabamaCoosa-Tallapoosa and Apalachicola-Chattahoochee-Flint River Basins has developed a
simulation model, Shared Vision Model, to help decision makers in determining
optimal allocation rules. The data used in this study come from other demand forecast
studies and future supply estimates. Future demand estimates were made for municipal
and industrial (M&I), navigation, agriculture, electric power, recreation, water quality,
and environment by sources. Most agricultural withdrawal of water in the ACF basin
comes from groundwater. As there is a close relationship between the amount of water
withdrawn from an aquifer and the stream flow, this relationship should be quantified
before any effective demand and supply analysis can be started. The Comprehensive
Study used correlation coefficients to quantify this relationship (Wreck et al.).
The Shared Vision Model is expected to provide a broader view on basin-wide
water management issues to different water user groups and policy makers and help them
in resolving the existing conflict on water allocation. This model, however, is not
capable of addressing short term water allocation issues for two reasons. First, it is based
on monthly time step in estimating demand and supply scenarios, which might be very
long for many uses. For example, irrigation decisions, in most cases, can effectively be
made on weekly basis depending on weather condition and soil moisture content. On the
other hand, an hourly decision making might be more appropriate for hydro-power
generation.

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

17
Second, the methods used in estimating the value o f water are not capable of
capturing short-term effects o f water scarcity. For example, in case of com production,
the impact o f drought would be much higher during tasseling period which lasts less than
a month (ACES Circular ANR-165). On the other hand, a long drought of two to three
weeks might be enough to destroy crop production, if supplemental irrigation is not
applied.
Multiobjective programming models have also been used by many researchers to
determine the optimal allocation of resources at a watershed level. Chang et al. used
compromise programming techniques and multiobjective simplex method to examine the
impact o f various land use patterns on water quality in Tweng-Wen reservoir watershed
system o f Taiwan. Other uses of multiobjective programming models include watershed
management (Goicoechea and Duckstein), regional development, environmental quality
control, and industrial land use (Van and Nijkamp; Das and Haimes), constrained
optimization o f limited resources (Glover and Martinson), impact of land use pattern on
groundwater richarge rate (Ridgley and Giambelluca), and correlation between land use
pattern and lake alteration (Leon and Marini).
As the Shared Vision Model, these multiobjective programming studies are
more tuned to long-term resource planning and examine the watershed management
problem in general rather than addressing the short-term water allocation issues. While
the present study focuses on the short-term water allocation problems faced by an
individual production unit and relates them with the basinwide resource management
issues.

R eproduced w ith perm ission o f the copyright owner. F urth er reproduction prohibited w ith o u t perm ission.

18
Plant Water Relationship
Water is an essential element of any plant life support system. Plant roots absorb
water from the soil. On the average, a large tree uses 100 to 150 liters of water on a hot
sunny day (Kramer and Kozlowski).
The soil texture, which is determined by the relative percentage o f sand, silt, and
clay, affects the water retention capacity of soil, and thus the amount of water available
for plant use. However, not all moisture contained in the soil is available for plant use.
Plants can effectively absorb water from the soil when the level of moisture contained in
the soil is within the domain of the permanent wilting point and the field capacity of the
soil (Heady and Hexem).
The permanent wilting point is defined as the level of soil moisture at which the
leaves o f the plants become permanently wilted. While the field capacity is defined as
the amount o f water a soil can hold against the gravity when allowed to drain freely
(Heady and Hexem). A plant can easily absorb water from the soil with moisture at its
field capacity. As plants absorb water, soil moisture content decreases; whereas, soil
tension (the force at which soil particles hold water) increases. This increase in soil
tension makes it difficult for plants to extract water from the soil. At the permanent
wilting point, soil moisture content becomes very low, resulting in lower amount of water
absorbed than the amount lost due to transpiration. Therefore, it is necessary to apply
supplemental water so that a desired level of crop yield can be obtained (Heady and
Hexem).

R eproduced with perm ission o f the copyright owner. Further reproduction prohibited w ith o u t perm ission.

19
Factors such as deep percolation, stages of plant growth and coverage of soil by
the plant, leaf glossiness, geographical location, planting season, and temporary weather
fluctuation affect the total consumptive use of water. Moreover, irrigation can be used as
a means to control soil salinity, to cool plants during hot periods, and to reduce frost
damage during cold periods (James). Part of this water need is supplied through
precipitation. However, the amount of precipitation is erratic and not sufficient for most
crops. The difference between optimal water level and the water available through
precipitation gives the demand for supplemental water.
Although the daily water requirement of a plant depends on various factors
including locality, plant type, soil conditions, and weather conditions, the impact of a
water shortage on crop yield at some critical stages of plant growth can be particularly
serious. For example, the daily water requirement of a peanut plant increases with its age
up to 80 days and then starts to decline (Stansell et al.; Rochester et al.). While, in the
case of com, the daily water requirement reaches its peak during the period 50 to 100
days after planting. Moreover, as the amount of soil moisture (water) available to the
plants decreases from its field capacity, the demand for supplemental irrigation increases
until it reaches to the permanent wilting point.
The actual impact o f shortage in soil moisture on crop yield depends primarily on
the stage of plant growth. Field experiments conducted by the experiment stations
indicated that the impact o f drought on com at the tasseling period would be much more
serious than at any other period o f plant growth (ACES Circular ANR-165 and ANR-531;
Bryant et al.). Moreover, crops may also differ in their ability to withstand droughts.

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

20
Irrigation Demand Studies
The demand for irrigation arises because it enables cultivation in dry soils and
increases productivity. The demand for irrigation water, as the demand for other farm
inputs, is determined by the optimizing behavior of the farmers. A rational farmer will
use irrigation, as long as the marginal benefit o f irrigation water is higher than the cost of
irrigation. Since the marginal productivity o f water decreases as the application of water
increases, keeping all other factors constant, the profit maximizing level of water will be
lower than the yield maximizing level, unless the water is a free good.
As the price of water increases, in the short run, a rational farmer may respond by
reducing the amount of irrigation. Such reduction in irrigation will stress the plants and
reduce crop yield. As water becomes more scarce and costly, adoption of more efficient
irrigation systems or investment in more efficient irrigation technology may become a
more attractive strategy than allowing increased plant stress. For example, use of drip
irrigation in Arizona made it possible to double the cotton yield by using only half of the
amount of water used in conventional method (furrow irrigation systems).
Caswell and Zilberman reported that the cost of technology is an important factor
affecting growers decisions to adopt new water saving techniques. In particular, they
found that a) sprinkler as well as drip irrigation technologies are beneficial for tree crops
like almond and pistachios, b) ground water users are most likely to adopt new
technologies than the surface water users, c) locational factors have differential impact on
the adoption o f new technologies, and d) water pricing policies can be used to induce
farmers to adopt new water saving technologies.

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

21
A farmer can respond in many ways to water cost increases. Depending on
factors such as output price, cost of inputs, soil type, risk, slope, salinity, and climate, he
may shift to a more drought resistant crop, change crop mix, practice crop rotation.
Bryant et al. proposed a methodology to provide decision rules to allocate irrigation
water optimally among the competing crops. They report that under certain conditions,
permanent or temporary abandonment of low-valued crops to irrigate high-valued crops,
would be the best strategy to deal with the short-term water scarcity problems.
The empirical irrigation demand estimates are limited in various ways. Firstly,
unlike other marketable commodities, reliable data on price and quantity demanded of
irrigation water are not readily available. As a result, various indirect valuation methods
have been used in estimating the demand for irrigation water. Secondly, the value of
water depends on various factors such as crop variety, plant growth stage, soil fertility,
crop, weather, price o f the output and many other biological, environmental, and
geographical factors. Such variability in water values makes it difficult to measure the
value of irrigation water and to compare results from different studies.
Analytical tools such as water response functions, farm crop budgets, linear
programming, quadratic programming, dynamic programming, profit functions, and
many other direct and indirect market valuation methods have been used in estimating the
demand for irrigation water. Moreover, the value o f water reflects various measures such
as marginal or average value, individual or mixed crop value, short-run or long-run value,
on-farm or in-stream value.

R eproduced w ith perm ission o f the copyright owner. F urth er reproduction prohibited w ith o u t perm ission.

!i

22
Lacewell et al. estimated water value for Texas High Plains based on crop budgets
for 1974. Their estimates of net water value or in-stream value ranges between $15 to
$87 per acre-foot. Willitt et al. (cited in Gibbons) estimated average water values based
on crop budgets o f 1971-74 period for different crops grown in four counties of Arizona
and found it to range between $-7 to $67. Beattie found an in-use marginal value of
water to be $44 acre-foot for the northern and central Ogallala and $20 for the southern
area.
The value of water across various uses is often used as an economic justification
for the regional transfer of water resources. Robert et al. estimated the value of irrigation
water for northern and southern regions o f Arkansas. The value of water applied in rice,
soybean, and cotton production was estimated for loamy and clay soils by region.
Among these three crops, rice grown in clay soil o f northern region produced the highest
value o f water. On the average, the value o f water for the state was found to be $2.90 per
acre-foot at 1975 prices. This difference in water value was used as a justification for
regional water transfers.
Gibbons used a production function approach to estimate the value of irrigation
water. The marginal value of water at ten percent reduction from the yield maximizing
output level was estimated to range between $36 and $54 per acre-foot for low ($0.51)
and high ($0.76) crop price per pound of cotton grown in Arizona at 1975 and 1980 crop
prices. Using different efficiency levels of irrigation, he derived value of water to range
between $61 to $94 per acre-foot for low and high efficiency irrigation.

'4

- -

___

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w ith o u t perm ission.

23
Cotton and grain sorghum growers o f seven western counties of the High Plains
of Texas experienced a relatively fixed cost o f pumping during 1957-72 period and a
sharp increase during 1973-80. Nieswiadomy examined the response of cotton growers
under two scenarios of constant and increasing extraction costs of groundwater. He
reported that as the cost of pumping increased, farmers were more responsive to changes
in the price o f a relatively less water-intensive crop (cotton) than to a change in the price
of a water-intensive crop (grain sorghum). A profit function approach was used to derive
an indirect water demand function. The results of the study showed that pumpage
regulation was not beneficial in Texas High Plains. A sensitivity analysis was performed
to test the Gisser-Sanchez rule. The Gisser-Sanchez rule was found to be useful in
determining the divergence in the time paths of water uses and the percentage difference
in profits but not in determining the difference in nominal profit.
Most of these studies have estimated the value of water used to irrigate a
particular crop irrespective of irrigation timing, except for Bryant et al. The present
study, however, attempts to estimate the value of a fixed but a continuous flow of water
which is available for crop irrigation at some critical plant growth stage. Depending upon
weather condition, the water available for crop irrigation may or may not be used at a
particular irrigation decision period. However, water not used in a particular decision
period cannot be saved for the future.
Yield Response to Water
The water-yield relationship can be derived by observing the response o f output
as the use of irrigation water increases, keeping other things constant. However,

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

empirical estimation of such water production functions is limited because it is difficult to


obtain reliable data on output associated with different levels of water. Therefore, various
indirect ways of measuring water response functions are used.
Heady and Hexem used various experimental data to estimate water response
functions for different crops. Using a quadratic water response function, they obtained
the following yield-water-fertilizer relationship for com:
Yield = - 10586 + 688.4 W,* + 36.4 N ,* * -10.0 W,2 * - 0.08 N,2** + 0.41 W,N,*
R2 = 0.93,

F = 100.32

where Y, N, and W denote com yield, nitrogen, and amount of irrigation water,
respectively and **,* denote significant at one and five percent level, respectively.
For this yield response function, the optimal level of nitrogen and irrigation water
will be 347 pounds o f nitrogen and 41.4 acre-inch of water per acre, respectively. This
optimal combination of nitrogen and irrigation water yield 9,985 pounds of com per acre
of land. This procedure can be used to estimate the functional relationship between total
irrigation water and yield. Crop yield is not only a function of the total amount of
irrigation applied but also a function of timing o f irrigation.
Various crop-growth simulation models have been used to estimate water-yield
response functions. Using these simulated water-yield response functions, the optimal
allocation of irrigation water across a growing season is determined (Musser and Tew;
Boggess and Ritchie; Swaney et al.). Many other studies have used water-response
functions in dynamic programming models to find optimal irrigation scheduling.
Zavaleta, Lacewell, and Taylor; Harris and Mapp; Yaron and Dinar; McGuckin et al.

R eproduced w ith perm ission o f the copyright owner. F urth er reproduction prohibited w itho ut perm ission.

25
analyzed irrigation scheduling using dynamic programming models and found that
irrigation scheduling might produce higher outputs with less irrigation water.
A numerical simulation model, known as Agricultural Field Scale Irrigation
Requirements Simulation (AFSIRS), has been used to estimate irrigation requirements for .
Florida crops, soils, irrigation systems, growing seasons, climate conditions and irrigation
management practices. AFSIRS provides gross as well as net irrigation requirements for
various plant stress levels.
The relationship between the demand for supplemental irrigation and the age of
plant has been found to be nonlinear. For example, the daily water requirements of a
peanut plant increases with its age up to 80 days and then starts to decline (Stansell et al.;
Rochester et al.). In this case, quantity as well as timing of irrigation is important in
determining the value o f supplemental water.
Optimal Allocation of Irrigation Water
The future supply o f surface water is not significantly affected by current
consumption. For surface water, allocational efficiency requires a balance among the
competing users and an allowance for periodic fluctuation. In the case of ground water, if
the withdrawal rate exceeds the natural recharge rate current use affects future
availability. Therefore, an efficient allocation of groundwater delineates a distinct set of
problems.
First, since present consumption affects future availability of groundwater, it
gives rise to a marginal user cost or opportunity cost. Second, the marginal extraction or
pumping cost would rise over time as the water level in the aquifer declines. Third,
<

!
ii
|

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

26
groundwater pumping will stop when the aquifer dries, or the marginal extraction cost
becomes higher than the marginal benefit, or the marginal extraction cost becomes higher
than the cost of obtaining water from alternative sources. Finally, groundwater might be
over-utilized because it exhibits several common property features.
Factors such as the possibility of private bargaining solutions, hydrological
conductivity o f the aquifer, and Gisser-Sanchez rule; however, may mitigate the problem
of common property exploitation. As argued by Beattie, groundwater in the High Plains
is not seriously subject to depletion by actions o f neighboring pumpers because of limited
lateral movement of aquifer. The Gisser-Sanchez rule states that if the natural recharge
rate and the slope of the demand curve for groundwater are small relative to the area of
the aquifer times storativity, and the groundwater rights are exclusively assigned, then the
welfare loss due to the inter-temporal misallocation o f pumping effort is negligible
(Gisser and Sanchez).
Following Burt and Stauber, Bryant et al. (1993) defined crop output as a function
of soil moisture condition at various plant growth stages and used a dynamic optimization
procedure to allocate predetermined number of irrigations between com and sorghum.
Although the numerical optimization procedure, which is used to solve the dynamic
programming models, is ensured to attain global optimum, it becomes unmanageable
when a large number of state and decision variables is involved. In dynamic
programming literature, this problem is known as the curse of dimensionality (Bellman
and Dreyfus).

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

27
For this reason, many studies have used linear programming models to determine
the optimal allocation of irrigation water. For example, Bemado et al. and Hardin and
Lacewell used linear programming models to allocate a fixed amount o f irrigation water
available to a farm. Yaron and Dinar used a dynamic programming model to select
optimal irrigation strategies and a linear programming model to determine the optimal
number of acres to be allocated to each crop. The main limitation of these studies is that
they did not recognize the stochastic nature of plants need for supplemental water. The
rainfall and the amount of irrigation water applied are the only two sources of soil
moisture. Among these two sources, rainfall is a random process, which makes the
demand for irrigation water stochastic. This erratic nature of irrigation demand can be
modeled in linear programming framework by using a recursive stochastic linear
programming (RSLP) procedure.
Day (1963) defined recursive programming as a sequential programming problem
where the parameters of the current model are functionally related with the solution of
previous models in the sequence. In this process of sequential optimization, current
decisions are made based on past, present, and expected information. The current choices
are irreversibly constrained by past decisions and all future options are determined by
choices made in the past and present decision periods. Applications of RSLP type
models include crop analysis (Kolajo), farm planning (Smith), aquacultural investment
decisions (Tai), peanut policy (Lamb), and peanut production (Curtis).
Kolajo and Smith focused on the strategic nature of decision making under
uncertainty. They introduced stochastic elements in the model by incorporating a series

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w ith o u t perm ission.

of specific historical observations in which expectations were not equal to actual


outcomes for crop prices and yields. Tai, Lamb, and Curtis used probability distribution
functions simulated by biophysical simulation and historical data for yield and price,
respectively. A series of multiple draws were taken from the distribution functions and
optimal values associated with each combinations of yield and prices were estimated.
Since these models were designed to address the multiple year decision problems, the
strategic rather than stochastic characteristic was specified in the RSLP model description
and nomenclature.
This study examines the irrigation decision problem of a model farm faced with a
fixed but continuous flow of water. In this case, both irrigation decisions and irrigation
impacts are highly dependent on rainfall, which is a purely stochastic process. Therefore,
the RSLP modeling framework was adapted from previous studies and used to evaluate
weekly water allocation decisions. Both strategic, as well as, stochastic elements were
present in models used by Kolajo, Smith, Tai, Lamb, and Curtis. The present study,
however, mainly focuses on the stochastic nature o f rainfall and reduces the strategic
nature of sequential decision making to irrigation weeks within a single year.
Accordingly, the present study retains the RSLP identity, but it represents recursive
stochastic linear programming as opposed to recursive strategic linear programming.

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w ith o u t perm ission.

RESEARCH METHODOLOGY

Conceptual Framework
Water flowing in a stream can be put into different uses at various locations. Uses
such as municipal and irrigation water involve withdrawal of water from the stream,
which reduces stream flow. Other uses such as recreation, hydro-power generation, and
navigation are non-consumptive in-stream uses, which may not reduce the flow of water
in the river. Therefore, a basin-wide water allocation model should be flexible enough to
accommodate these features of water uses.
Surface water is a replenishable but depletable resource. An efficient allocation of
this scarce resource would result when the marginal net benefit is equalized for all uses.
Moreover, most sources of surface water have seasonal patterns. They are also observed
to go through various cycles of year-to-year fluctuations. Therefore, a mechanism should
be devised to deal with such seasonal and annual fluctuations in stream flows so that the
cost of abnormal supplies can be minimized (Tietenberg).
Both surface and groundwater sources are being used to supply irrigation water in
the ACF river basin. As discussed in preceding chapter, the conditions required for
optimal allocation o f groundwater are different from those required for surface water
allocation. However, studies have shown that ground and surface waters are closely
related (Saliba). Generally, ground and surface waters are assumed to be perfectly
29

!
i

Ii
4'

- -

-______

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w ith o u t perm ission.

i
30
correlated below the fall line. While above this line, the degree of correlation is expected
to be less than perfect. Since the fall line crosses the ACF basin, correlation coefficients
ranging from zero to one have been used by previous studies to examine the impact of
ground water withdrawal on downstream flow (Wreck et al.). Thus, while addressing
basin-wide water allocation issues, the relationship between ground and surface water
should be taken into account.
Crop plants absorb water from the soil moisture available in the root zone. As
plants withdraw water, soil moisture content decreases. On the other hand, a decrease in
moisture content increases soil tension, the force at which soil particles hold water. This
increase in soil tension makes it difficult for plants to extract water from the soil. At
permanent wilting point, the amount of water absorbed by plants from the soil becomes
less than the amount required for plant survival, and eventually plants die. Whereas at
field capacity, which is defined as the amount of water a soil can hold against gravity
when allowed to drain freely, plants can easily absorb water from the soil. Thus, for
optimal crop yield soil moisture condition must be maintained in between permanent
wilting point and field capacity. Rainfall and irrigation are the only two external sources
of soil moisture. In humid regions, such as the Middle Chattahoochee Sub-Basin,
supplemental irrigation is applied, whenever rainfall is not enough to maintain the desired
level of soil moisture.
Factors such as locality, plant type, growth stage, soil conditions, weather
conditions, etc. determine the actual demand for irrigation. Other things remaining the
same, the stage of plant growth and the amount of soil moisture available to the plant in

R eproduced with perm ission o f the copyright owner. Further reproduction prohibited w ith o u t perm ission.

its root zone are the main factors that determine the value of irrigation water. Various
empirical studies have shown that the daily water requirements of crop plants increases
initially, reaches the peak, and then starts to decline. Moreover, the actual impact of
shortage in soil moisture on crop yield depends primarily on the stage of plant growth
(Stansell et al; Bruce et al.; Rochester et al.).
Methods and Models
Although it is difficult to measure the actual demand, various attempts have been
made to quantify the requirements for irrigation water. Marella, Fanning, and Mooty
measured the demand for irrigation in the ACF basin by multiplying the number of
irrigated acres per crop type by a constant rate of application. In a study "ACT/ACF
River Basins Comprehensive Study: Agricultural Water Demand" conducted by USDASCS in 1994, the Blaney-Criddle method was used to measure consumptive use of water
by plants. Assuming that the amount of water used by crops during their normal growing
season is closely related with mean monthly temperatures and daylight hours, Blaney and
Criddle developed coefficients that can be used to estimate consumptive use of water in
areas for which climatological data are available.
Heady and Hexem provide various functional relationships between irrigation
water and crop yield and estimated demand for irrigation water. Their procedure can be
used to estimate the functional relationship between irrigation water and yield. In most
cases, however, the empirical data required to estimate yield response functions are not
readily available. Moreover, crop yield is not only a function of the amount o f irrigation

>
f
!

I
R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w ith o u t perm ission.

32
but also a function o f the timing of irrigation. In this case, Heady and Hexem procedure
is not of much help.
The biophysical simulation models have also been used to examine the impact of
various irrigation management practices on crop yield (Swaney et al.; Musser and Tew;
Lynne et al.; Boggess and Ritchie). A combination of simulation and optimization
models have been used by others to find optimal irrigation scheduling. For example,
Zavaleta, Lacewell, and Taylor; Harris and Mapp; Yaron and Dinar, McGuckin et al.; and
Bryant et al. used dynamic programming models to analyzed irrigation scheduling and
observed that higher crop yield could be obtained through irrigation scheduling.
The relationship between the demand for supplemental irrigation and the age of
plant has been found to be nonlinear. In other words, the daily water requirements of a
plant increases with its age up to certain growth stage and then starts to decline (Stansell
et al.; Bruce et al.; Rochester et al.; and Bryant et al.). In this case, quantity as well as
timing of irrigation is important in determining the value of supplemental water. Though
the estimation procedures discussed above enable us to quantify irrigation requirements
for optimal growth o f plants, none of them provide an appropriate methodology
demanded by the present study.
The EPIC Model
The EPIC model, which was originally developed to examine the impact o f tillage
practices on soil erosion, can be used to simulate the impact of various water
management practices on yield under different biophysical environments (Williams et

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

al.). This model is capable of generating daily weather and other biophysical parameters
required to estimate water response functions for major competing crops.
Bryant et al. (1992) used the EPIC model to simulate yield response of com to
soil water in the southern Texas High Plains. They used actual experimental data to
validate the simulation results and found that simulated yield explained up to 86 percent
of the variation in actual yield. Cabelguennne, Jones, and Williams used field data from
Toulouse, France, to calibrate the simulated results and concluded that EPIC can be used
to determine the optimal irrigation strategies.
Recursive Programming
In a classical-neoclassical theoretical framework, an economic agent is assumed to
behave rationally. Rationality implies that an agent is always capable of making the best
choice among available alternatives. In other words, given the level of resource
endowment, goals, and environment faced by the agent, his actions can be represented as
a solution to an optimization problem. This characterization of the economic man
imposes a strong restriction on his behavior. However, Simon observed that, by nature,
humans are not optimizing agents. At best, they are locally optimizing because resources
available to them, in terms of computational power, memory, etc., are limited. This
boundedness in each individuals resource endowment may result in imperfect behavior,
which gives rise to bounded rather than perfect rationality (Good; Simon; Doyle; Russell
and Wefald; Zilberstein).
An agents expectations may not be realized because perception and cognition are
not perfect. Moreover, due to bounded rationality, agents are likely to make inconsistent

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

34
decisions. Day (1983) observed that individuals are imperfect decision makers. He also
maintained that even if the external environment remains unchanged, agents consistently
attempt to improve their status by enhancing decisions. Inconsistent decisions that are
based on faulty anticipation are likely to lead to regrets rather than satisfaction. In this
case, conventional optimization procedures that are based on the assumption o f rationality
are not entirely adequate.
Day (1983) contended that these facts of reality should be modeled as a dynamic
adaptive process, where the agent responds to its own internal conditions and to the
changes in the external environment. In this dynamic system, actions taken by an agent
affect both external environment, as well as, his own internal conditions. Thus, both
agents and environment receive feedback from each other. In this sense, the economy is
assumed to be composed of interactive adaptive processes.
Day (1978) proposed a recursive programming procedure to model the dynamic
behavior o f boundedly rational agents, who interact with system through feedback
mechanisms. In this procedure, a dynamic multi-period decision problem is broken down
into a series of recursively connected local optimization problems. The solution o f each
individual optimization problem satisfies certain optimality properties. The sequence
as a whole need not and in general will not satisfy the principle of optimality (Day, p.
10).

Day (1978) defined recursive programming as a dynamic system, which is


composed o f three elements known as data (6,), optimization (Tj), and feedback (&,). He
named these components operators. The data operator, 6 specifies the relationship
t
!

t
f
* J('

____

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

35
between the parameters of objective function and the resource constraints. The
optimization component, t, defines the kinship among choice variables associated with
objective function, resource constraints, and other parameters. While the feedback
mechanism, 5) determines how the arguments of succeeding stages (Pt+I), such as state
(St+I), data (6t+1), and optimal decision variables (Yt+l), are functionally related with thenrespective counterparts in current period (P,). Day (1978) used the following diagram to
demonstrate the interaction among these three components o f recursive programming

Initial
conditions (S,)
and ex ogen ou s
variables (e,)

w
"

O fm p rate
*
^

/
Datum

^
.
Optimizing
O p t im iz e
Vector
C M ------------ -------------

Feedback
(&t)
Figure 1. The Recursive Programming Model.
S ource: D ay, 1978.

In this recursive programming framework, first of all, the data set required for
optimization is generated from the existing initial conditions and then the optimization
problem is solved. Since both the optimal solution obtained in the first stage, as well as,
f

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

36
other external factors affect the information transferring to the next stage, the data set is
regenerated by incorporating information from the last stage (feedback).
The information contained in Diagram 1 can also be utilized to compare and
contrast recursive programming methods with other dynamic models such as dynamic
programming and optimal control theory. Given the feedback mechanism (<b) and an
objective function, the objective o f dynamic programming would be to find the optimal
strategy that would yield optimal result for any data set generated by 3). In this case, the
set of results as a whole must satisfy the principle of optimality. While in the case of
optimal control theory, both the data (6) and optimal strategy (*F) are selected to satisfy
the optimality condition. In contrast, in a recursive programming model, both data (6)
and optimizing (T ) operators are specified for each stage of the recursively connected
system and the quantitative and qualitative results are examined for the system as a
whole. Since the specification o f the data and optimization operators may not reflect the
optimal strategy with respect to feedback mechanism, the sequence of optimization as a
whole need not satisfy the principle of optimality (Day, 1978). In this sense, recursive
programming can be viewed as a suboptimization or suboptimal control method. In this
sense, the term optimality or optimal used in this study refers to the local but not the
global optimum.
Recursive Stochastic Linear Programming
As discussed in previous chapter, the recursive stochastic optimization procedure
developed in this study is named as recursive stochastic linear programming to emphasize
the fact that the present study accounts for the stochastic nature of rainfall while

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w ith o u t perm ission.

37
determining optimal allocation of irrigation water. In terms of diagram 1, the current
RSLP model differs from recursive programming in two ways. First, the RSLP model
accounts for stochastic nature of rainfall while making irrigation decisions. Second, the
results of the RSLP model are based on multiple specifications of data and optimizing
operators rather than a single sequence of specification used in recursive programming.
The basic set up in which the optimization model was developed can be explained
as follows. A fixed but a continuous flow of water, at each irrigation decision period, is
available to a model farm to allocate among three competing crops - com, cotton, and
peanuts. Based on existing soil moisture condition and the agents expectation about the
future weather pattern, irrigation decisions are made at the beginning of each growth
period. However, the agents expectations may not be realized. For example, it may rain
heavily immediately after the crop is irrigated and vice versa. Therefore, the total amount
of soil moisture available to the plants at a particular growth stage may exceed the
amount that can be consumed by the plants. To account for the transfer of excess
amounts o f soil moisture from the current period to the next period, moisture transfer
activities are included in the model. In this process of transferring soil moisture from one
period to the next, however, a percentage of the moisture is lost in the environment. The
transfer mechanism may allow transfer of soil moisture from one stage to the next even if
the total amount of soil moisture available in the current period is less than the amount
that will be consumed by the plants. To prevent this process, the moisture transfer
activity is made highly expensive in the objective function and an accounting row is used
to obtain true net income data..

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

38
The information set (Sj) available to the decision maker at each decision stage
includes the total amount of irrigation water available to him (Wj), current soil moisture
condition (SM1J k), which is defined as the amount o f soil moisture transferred from the
last period (MThl lc) times the soil moisture transfer coefficient y, (i.e., SMi-J>lc=y*MTJ.u ),
total area under each crop (Ai.j.I J , and the marginal contribution of irrigation water
(Pi. j. J . The irrigation decision at each plant growth stage is assumed to be based on the
information contained in the information set (Sj) and the expectation that a normal
weather condition (E [R ^ n] = p, for n=l, 2,..., N, and p = normal weather condition) will
prevail in all future decision periods (i.e., Ij lc= f(Sj, E[Rj+n]).
This irrigation decision problem was developed as a recursive stochastic linear
programming model. The stochastic element is introduced in the model at each decision
period by making rainfall a series of random draws from an empirical probability
distribution function (estimated from the past rainfall observations). In the first period,
the irrigation decision is made with the expectation that rainfall will follow a normal
pattern in all decision periods. In the second decision period, the actual weather
condition, which prevailed in the first period, has occurred. Therefore, the actual weather
condition rather than normal expectation is imposed on the first period. However,
normal weather conditions are expected to prevail in the subsequent decision periods. In
the third decision period, the weather conditions for the first and the second decision
periods are known and imposed on the model. Rainfall in the third period is randomly
drawn from the same probability distribution function with normal weather conditions
L

still expected to prevail in all other future periods. This irrigation decision/random

i
f

|
R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

39
rainfall process is repeated until the last decision period when all weather conditions are
known and have been experienced in the model.
The RSLP model developed in this study estimates optimal net farm income
associated with different levels of irrigation water using yield response function
parameters and data on input and output prices. The price data were obtained from the
1996 crop budgets (ACES, 1996) and the yield response functions were estimated using
EPIC simulated data. The crop yield was defined as a log-linear function (CobbDouglas) of soil water available to the plants in their root zone at different stages of plant
growth (Burt and Stauber; Bryant et al.). Crop growth stages were defined as the weeks
after planting. The relationship between crop yield and the amount of soil moisture at
various stages of plant growth was specified as
j

lnYk = In a + 2 ( 2 pi>j>k In CDfj J for


Jl 1-1

i = 1, 2,..., I (state of soil moisture)


j = 1, 2,..., J (crop growth stage)
k = l , 2,..., K. (crops)

where In Yk = the natural log of crop yield,


In toy = the natural log of soil moisture level i at plant growth stage j, and
i, j, and k

= are subscripts that refer to the state o f soil moisture, stage o f plant
growth, and crop type, respectively.

The relevant domain of soil moisture, at each stage of plant growth, is assumed to
be between permanent wilting point and field capacity because soil moisture condition
beyond these two extreme points limits plant growth (since the cases where soil moisture
conditions are beyond these two extremes were not relevant to this study, they were
excluded even though they may occur in the real world). On the other hand, the scope of

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

40
this domain depends on the nature o f the soil available in the plant root zone. For
example, the maximum water holding capacity for the top two feet of soils found in
major crop lands of Alabama was reported to fall between 1.4 to 4.0 acre-inches (i.e.,
0<gjs 4; ACES Circular ANR-156).
In the recursive stochastic linear programming framework, the irrigation decision
problem, when a fixed but a continuous flow of water is available for irrigation, can be
set up as follows:
Maximize:

K
Z = Pk Yk
k=t

Subject to:

lnYk = In a + 2 2 (Pi j k In coj j k)


j-i i=i

fc

I lt s Wj.

k=l

Yk, Ijk 0

for

i = 1, 2,..., I (state of soil moisture)


j = 1, 2,..., J (crop growth stage)
k = l, 2,..., K (crops)

where Pk = net price of crop k,


Yk = yield or output of crop k,
Ij k = the amount o f irrigation water applied at stage j to crop k, and
Wj = the amount of irrigation water available at stage j.
In this simple model, only two of the relevant constraints - irrigation and crop
yield - are included. The complete model used in this study is presented in the next
section.

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

41
Specification of the RSLP Model
Using the estimated yield response functions, a recursive stochastic optimization
procedure was developed to allocate a fixed but a continuous flow of surface water
among three competing crops - com, cotton, and peanut. This procedure is capable of
allocating a unit o f irrigation water available at a particular stage of plant growth among
three competing crops optimally by assuming that the rainfall follows a stochastic
process. The optimization model can be summarized as follows.
3

Maximize:

2 PkYk
k=I

Subject to:

Land

Crop

6
2 Yi.j.k S 1J
i=l
6

2Yi.j-i.k i=t

2 Yi.j.k s 0;
i=l

Moisture

2 (2 ooi,j, k) - Ii.j, k - Rj - yMtj.j-i.k + Mtj. j-i. k = 0;


k=l i=I

Irrigation

2 Ij. k s W j;
k=l

Yield

2 2 Pi, j ,ic ojt.j.k= Yk;


j=i 1=1

Non-negativity

Pk, Y i, j , k, S M i, j , P i, j , I j , k, R j, Mtj , j +i

for

i = 1, 2,..., 6
j = 1 ,2 ,..., 9
k = 1, 2, 3;

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

0;

42
where y = soil moisture transfer coefficient (set to be 0.5), co = states o f soil moisture (a
continuous variable discretized to take values o f 0.0,0 .5,1.0, 1.5, 2.0, and 2.5), and as
before, i, j, and k refer to soil moisture state, crop growth stage, and crop type,
respectively.
In total, 116 equations were used to represent constraints on 253 different
activities. A fixed but a continuous flow of water is assumed to be available for irrigation
during the peak irrigation season. The maximum soil water level allowed in the model
was 2.5 acre inch per crop and nine plausible water supply scenarios, ranging from 0 to 8
acre inches, were considered. As explained in the methodology section, the RSLP
procedure used in this study allocates water among three competing crops so that the
expected net farm income is optimized.

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

RESULTS AND DISCUSSION

The research procedures followed in the study are briefly explained in this
section, and the results obtained from those procedures are also presented and discussed.
The discussion and results are mainly focused on the short-term water allocation issues.
First of all, optimal irrigation strategies available to a model farm, which is planted with
three major cropscom, cotton, and peanuts- grown in the Middle Chattahoochee sub
basin, are examined under dry and normal weather conditions. Then, using the results
obtained from the model farm, basinwide water management issues are examined.
Before looking at basin-wide water management issues, the fundamental question is
irrigation water really a limiting factor for crop production in the Middle Chattahoochee
sub-basin? is examined. Then, the optimal demand for irrigation water is estimated for
the sub-basin. Based on these estimates and historic flow data, other policy relevant
issues such as: a) what are the long-term implications of changes in cropping/land use
pattern on irrigation demand?, b) how much water can be saved by adopting water
saving technologies?, and c) what would be the impact of reduction in stream flow by
15 percent on net farm income?, are examined.
The EPIC model is used to simulate the impact of different irrigation management
practices on crop yield. Using these simulated data, yield response functions are estimated
for each crop included in the study. Since all other factors of production, except for
43
i

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

44
irrigation, are fixed at recommended levels, crop yield was specified as a log-linear
function o f soil moisture available to the plants at various stages of plant growth (Burt and
Stauber; Bryant et al.). The plant growth stages are defined as the number o f weeks after
planting. The soil moisture levels available to the plants at different plant growth stages
are measured as the sum of irrigation water applied and the amount o f rainfall received
within that week. From the estimated yield response functions, marginal impacts of water
applied at different stages of plant growth are calculated. Using these marginal values of
irrigation water as parameters, a recursive stochastic linear programming model is
developed to estimate the impact of changes in irrigation water supply on net farm income
under dry and normal weather conditions.
Simulation Results
The EPIC model is used to simulate com, cotton, and peanut yields for Dothan soil
under different weather conditions and irrigation scenarios. The weather data from the
Weather Station in Montgomery, Alabama are used to simulate rainfall and other weather
parameters used in the model. The tillage practices and input applications are specified as
recommended by extension experts (personal communications). The automatic irrigation
option available in the EPIC model is utilized to specify various irrigation decision rules.
The amount o f irrigation water per application is defined as a continuous variable within
the interval o f 0.25 inch to 3.0 inch. The impact of changes in irrigation strategies on crop
yield was simulated by specifying different irrigation decision rules that are based on
water stress levels. In total, eleven water stress levels ranging from 0 (no irrigation) to 1
(no water stress) are specified. For each water stress level, crop yields are simulated for
tfi
!i
j

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

45
25 years. Thus, 275 yield levels associated with different weather conditions and eleven
water stress levels are simulated. The average simulated crop yields associated with
different water stress levels are reported in Table 5 and graphed in Figures 2-4.
In all 275 cases, the simulated yield increased as the intensity of water stress is
reduced. In other words, in a 0.0 to 1.0 scale of water stress level, as we move up from no
irrigation (0.0) to full irrigation (1.0) at an equal interval of 0.1, the simulated yield
increased continuously and generally at a decreasing rate. This indicates that the
simulated yield levels are consistent with theoretical expectations.
Yield Response Functions
The relationship between daily water requirement of plants and their growth stage
is expected to be nonlinear. Experimental studies have shown that the daily water
requirement of crop plants -- com, cotton, and peanuts increases initially, reaches the
maximum, and eventually starts to decline over time (Bruce et al.). For example, the daily
water demand o f field com is expected to increase until 100 days after planting and then
starts to decline (ACES Circular ANR-531; Bruce et al.; Bryant et al.). While in the case
of peanuts, the daily water demand increases up to 80 days after planting and then it starts
to decline (ACES Circular ANR-207; Stansell et air, Rochester et al.). For this reason, the
log-linear functional form has been used by researchers to represent the relationship
between crop yield and the amount of soil moisture available to the plants at different
growth stages (Burt and Stauber; Bryant et al.).
The weekly sum of the amount of rainfall received and the amount of supplemental
irrigation applied was defined as the total amount of soil moisture available to the plants

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

46
Table S. EPIC Simulation Results: Average Crop Yield and Water Stress Levels
Yield
Stress Levels
Com (bushel)

Cotton (lbs.)

Peanut (lbs.)

0.000

72.775

575.496

2191.829

0.100

133.765

731.896

2766.584

0.200

141.867

792.872

2839.297

0.300

151.932

829.096

3040.048

0.400

156.464

880.688

3091.793

0.500

157.802

890.532

3210.578

0.600

161.343

916.140

3295.795

0.700

167.900

962.868

3379.478

0.800

169.199

976.712

3438.533

0.900

174.196

1007.436

3508.803

1.000

176.219

1019.948

3539.223

200

180

160

O
Q>
.C
3
.
S
.92

120

>

100

80

60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

W ater S tre ss Levels (0-1)

Figure 2: Corn Yield & W ater Stress L evels

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

47
1100
1000

goo

"ST
w
O
TO
2

800

>
700

600

500
0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0 .9

W a te r S tr e s s Levels (0-1)

Figure 3. Cotton Yield & W ater S tress L ev els

4000

3500

35
2 3000
32
.92

>

2500

2000
0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

W ater S tr e s s L evels (0-1)

Figure 4. P ean u t Yield & W ater S tress L ev els

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w ith o u t perm ission.

for that week. Thus, the functional relationship between the amount o f weekly soil
moisture (to;) available to the plants and the final crop yield (YJ can be expressed as
j

lnYk = 2 ( 2 Pj.j. k In a>UJc)


j*i i*i

for

i = 1, 2 , I
j 1j 2 , J
k = 1, 2 , K.

This study examines the impact of soil moisture shortage on crop yield at some
critical plant growth stages. The impact of water shortage at these critical growth stages is
expected to be more serious than at any other periods. Moreover, even if the existing
supply is more than enough to meet irrigation demand in normal years, it may not be
enough to meet the demand in dry years. In this case, the impact of drought on crop yield
would be much more serious if it occurs at the critical plant growth stage. For example,
the impact of drought on com at the tasseling period would be more serious than at any
other period (Bryant et al.\ ACES Circular ANR-165). The daily water use of field com at
this growth period may reach up to 0.34 inch (2.38 inch/week; ACES Circular ANR-531).
While in the case o f peanut, the critical irrigation demand period would range between 5585 days after planting. During this period, the daily water requirement o f peanut plants
might be as high as 0.26 inch per day (1.82 inch per week; ACES Circular ANR-158).
In the EPIC simulation model, com and peanuts were assumed to be planted on
April 1 and cotton on April 15. On May 27, com and peanut plants are 57 days (8 weeks)
old and cotton plants are 42 days (6 weeks) old. Therefore, the critical period for
irrigation was defined as the period beginning from May 27 to July 28. The yield response
function defined in terms of critical irrigation period would be

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

49
9

lnYk= 2 (2 Pi. j. k In co, j k)


j-i i-i

for

i = l , 2 , ...,6
j = 1 2 , 9
k = 1, 2, & 3.

As discussed before, the impact of water shortage on crop yield in a normal year
would be different from its impact on yield in a dry year, two different yield response
functions were estimated. For a normal year, the yield response function was estimated by
using the data set which included both dry and irrigated yields under various water stress
levels. Since plants are expected to experience more frequent water stresses in a dry year
than in a normal year, the yield levels associated with no water stress cannot be used to
examine the impact o f water shortage in dry years. Therefore, crop yields with different
levels of rainfall but without irrigation were simulated and used to estimate the yield
response function for dry years. The estimated response functions associated with dry and
normal weather conditions for each crop are reported in Table 6.
All of the estimated coefficients of the response functions are positive and most of
them are highly significant. The coefficient of determination (R2) is reasonably high, in all
cases, implying a good fit o f the model. The means o f the respective simulated crop yields
also fall within the expected ranges. The sum of the estimated coefficients, which can be
defined as the partial elasticity o f production with respect to water (ew= 2 9jH Pj), is less
than one for all three crops. This implies that the yield response functions are concave
(Beattie and Taylor). In other words, if the amount of water input is increased by one
unit, keeping all other factors of production constant, it will increase crop yield by less
than one unit, i.e., the yield functions are subject to the law o f decreasing returns to scale.

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

50
Table 6. Yield Response Functions for Com, Cotton, and Peanut
Growth

Com

Cotton
Dry

Stage

Normal

Constant

5.1154**

5.2093**

6.8755**

6.8728**

8.0989**

8.0538**

5/27-6/2

0.0159**

0.0347**

0.0095**

0.0308**

0.0071**

0.0001

6/3-6/9

0.0191**

0.0506**

0.0033

0.0308**

0.0041

0.0215**

6/10-6/16

0.0420**

0.0664**

0.0163**

0.0115*

0.0109**

0.0280**

6/17-6/23

0.0371**

0.0710**

0.0186**

0.0171**

0.0111**

0.0097*

6/24-6/30

0.0417**

0.0577**

0.0246**

0.0273**

0.0244**

0.0252**

7/1- 7/7

0.0497**

0.1000**

0.0231**

0.0360**

0.0490**

0.0499**

7/8-7/14

0.0456**

0.0815**

0.0260**

0.0360**

0.0246**

0.0352**

7/15-7/21

0.0211**

0.0407**

0.0154**

0.0277**

0.0165**

0.0367**

7/22-7/28

0.0182**

0.0104

0.0085**

0.0148*

0.0201**

0.0259**

0.1453

0.2320

0.1678

0.2322

e*
N
Mean(Y)
R2

0.2904

0.5130

Normal

Peanut
Dry

Normal

275.0000

100.0000

275.0000

100.0000

275.0000

151.2240

75.0000

871.2440

596.2910

3118.3600 2342.8130

0.4300

0.5990

0.4900

0.5980

0.4930

Note: **, * Denote significance at one and five percent level, respectively.

Dry

--------

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

100.0000

0.6780

The estimated coefficients differ across stages indicating that the impact o f water
on yield varies across stages. The results indicated that a unit of water applied to com or
peanut on the first week of July would produce more output as compared to the water
applied on any other stages of plant growth. While in the case o f cotton, which was
planted two weeks after com and peanut, the water applied on the second week of July,
yields higher output than on any other stages of plant growth. On the first week of July
com and peanut plants are 14 weeks old (98 days) and on the second week of July cotton
plants are 13 weeks old (90 days). These two weeks, the first and second weeks of July,
coincide with the expected peak water demand periods for the respective crops (ACES
Circular ANR-165, ANR-158, and ANR-531, and Circular 467).
Marginal Physical Productivity of Water
The average marginal physical productivity o f water (AMPP* J applied at stage j
to crop k was calculated by using sample averages o f respective crop yields and soil
moisture levels. The results are reported in Table 7 and graphed in Figure 5-7.
The results indicated that the impact o f water on crop yield in a dry year would be
much higher than in a normal year. For com, the maximum AMPP* k occurs earlier in
normal years than in dry years. In the case o f peanut, however, the maximum AMPP*. k
was obtained on the first week of July (14 weeks after planting) for both dry and normal
years. Although cotton was planted two weeks after com, for normal years, the period of
maximum AMPPwJk o f cotton coincides with that of com. However, in dry years, the
maximum AMPPwj>kof cotton occurs much earlier than in normal years.

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w ith o u t perm ission.

52
Table 7. Average Marginal Physical Productivity of Com, Cotton, & Peanut
Growth

Com

Stage

Cotton

Normal

Dry

Normal

Bu./Are-inch

Peanut
Dry

Normal

Dry

Lbs./Are-inch

Lbs./Are-inch

5/27- 6/2

1.530

2.406

10.992

23.582

19.303

0.299

6/3 - 6/9

1.920

5.218

3.447

25.894

12.646

58.040

6/10-6/16

3.764

6.559

6.056

6.809

27.657

89.133

6/17-6/23

3.094

7.091

11.230

17.080

23.309

30.645

6/24 - 6/30

4.197

6.872

17.078

18.418

52.079

82.008

7/1 - 7/7

4.053

7.096

12.539

18.779

87.917

108.421

7/8 -7/14

3.780

6.052

13.201

18.259

49.174

85.666

7/15 - 7/21

1.867

2.667

7.842

19.743

27.604

76.209

0.692
10.222
31.995
46.694
1.922
4.980
7/22 - 7/28
Note:The average marginal physical productivity (AMPPkw) was estimated as follows:
AMPPkw = Pj * (Yk/Wjk), for j = plant growth stages (1,2,..., 9), and k = crops (com,
cotton, and peanuts), where Yk = average crop yield and Wjk = average amount of soil
moisture in week j available to the crop k.
8
Normal

- Drought

0
1

W eeks

Figure 5. AMPP o f Irrigation at Various Plant Growth Stage

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

53
30

a Normal

Drought

25

20

2
^

15

8
10

W eeks

Figure 6. AMPP o f Irrigation at Various Plant Growth S tage

120

100

80

60

40

20
a Normal

0
1

Drought

W eeks

Figure 7. AMPP o f Irrigation at Various Plant Growth S ta g e


i

R eproduced w ith perm ission o f the copyright owner. F urth er reproduction prohibited w itho ut perm ission.

54
Marginal Value of Irrigation Water
The average marginal value of irrigation water (AMV* J applied at stage j to crop
k, was calculated by using average crop price (APJ and AMPP* k data (i.e., AMVwJk
=AMPPW
J>k * A PJ and the results are reported in Table 8 and graphed in Figures 8 and 9.
The AMV* k can be utilized to make static water allocation decisions. The
average marginal value of water was used to rank crops and the results are reported in
Table 9. The results of a simple ranking o f the average marginal value o f water indicated
that different water allocation rules would apply for dry and normal weather condition. In
particular, reallocation of water would occur in stages 2, 5, and 9. However, the marginal
water allocation rules derived from the model, which ignores the stochastic nature of
rainfall, are limiting. The importance of irrigation water is mainly determined by the
amount o f soil moisture available to the plants at a particular irrigation decision period.
Rainfall and the amount of irrigation water applied are the only two external sources of
soil moisture. Among these two sources of soil moisture, rainfall is a random process,
which makes the demand for supplemental irrigation stochastic. Therefore, the irrigation
decision problem should be modeled as a stochastic process.
Dry and Normal Weather Conditions
Two different rainfall scenarios pertaining to normal and dry years were defined
using 100 years daily rainfall data generated by the EPIC model. First, for each crop, two
separate 100 year simulations (with full irrigation and without irrigation) were run to
generate potential and actual (non-irrigated) yield levels. Then, using these simulated
t
i
i

R eproduced with perm ission o f the copyright owner. Further reproduction prohibited w ith o u t perm ission.

55
Table 8. Marginal Value o f Irrigation Water by Crops
Growth
Stage

Com
Normal

Cotton
Dry

Peanut
Dry

Normal

Normal

Dry

Dollars/acre-inch

Dollars/acre-inch

Dollars/acre-inch

5/27-6/2

2.65

4.17

4.67

10.02

2.36

0.04

6/3-6/9

3.33

9.05

1.47

11.01

1.54

7.08

6/10-6/16

6.53

11.37

2.57

2.89

3.37

10.87

6/17-6/23

5.37

12.30

4.77

7.26

2.84

3.74

6/24-6/30

7.28

11.92

7.26

7.83

6.35

10.01

7/1-7/7

7.03

12.31

5.33

7.98

10.73

13.23

7/8-7/14

6.56

10.50

5.61

7.76

6.00

10.45

7/15-7/21

3.24

4.63

3.33

8.39

3.37

9.30

7/22-7/28

3.33

1.20

2.12

4.34

3.90

5.70

Total

45.31

77.43

37.13

67.48

40.47

70.41

Table 9. Ranking of Marginal Value o f Irrigation Water by Crops


Stage

Cotton

Com

Growth
Normal

Dry

Normal

Peanut
Dry

Normal

Dry

5/27-6/2

6/3-6/9

6/10-6/16

6/17-6/23

6/24-6/30

7/1-7/7

7/8-7/14

7/15-7/21

7/22-7/28

j
I

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

56
12

10

a>
<0

^3

>

CO

c
s>

Maize

Peanut

Cotton

0
1

Weeks

Figure 8. M arginal V alu e o f Irrigation for Normal Years

15

10

J3
<0
>
C
cO
'e

a Maize

* Peanut

+ Cotton

0
1

Weeks

Figure 9. Marginal V alue o f Irrigation for Dry Years

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

57
yield data, a weighted average ratio (py) of actual (Yaty) and potential (Y ^ y) yield levels
was constructed, i.e.,
3

py = { S [YayJt /YPyjt] }/k.


k=l
The weekly rainfall data, for the nine critical weeks considered in this study, were
calculated for each year. Then using the weighted average ratio of actual and potential
yield (py) as a sorting variable, the 100 year weekly rainfall data were sorted in ascending
order. The weekly rainfall data for the first 10 years were then used to generate random
rainfall patterns for a dry year. Similarly, rainfall data that fall in the middle range
(between 45-55) were used to generate random rainfall patterns for a normal year. A
similar approach was adopted by Hook in defining drought in Costal Plains of Georgia.
Irrigation Decision rules
The optimal allocations of irrigation water among three competing crops were
examined at nine different water supply levels ranging from zero to eight acre inches by
using a recursive stochastic linear programming model. The recursive stochastic
optimization procedure was presented in a previous section. The results obtained from the
optimization model are reported in Tables 10-13 and graphed in Figure 10-15.
The irrigation decision rules obtained from the optimization model indicated that
when the total supply o f irrigation water was fixed at one acre-inch, the optimal allocation
would require irrigation of only one crop at each growth stage. In this case, the optimal
irrigation decision rules replicate the results from the static analysis in both dry and
normal weather conditions (compare the optimal irrigation rules reported in Tables 9 and

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

58
Table 10. Optimal Allocation of Irrigation Under Water Scarcity (Supply=l acre-inch)
Stage

Maize

Peanut

Cotton

Total Irrigation

Dry

Normal

Dry

Normal

Dry

Normal

Dry

Normal

Table 11.Optimal Allocation o f Irrigation Under Water Scarcity (Supply=2 acre-inch)


Maize
Stage

----
- ------ -

Cotton

Peanut

Total Irrigation

Dry

Normal

Dry

Normal

Dry

Normal

Dry

Normal

0.47

0.45

1.53

1.55

0.00

0.00

2.00

2.00

0.07

1.51

1.93

0.00

0.00

0.49

2.00

2.00

1.99

1.87

0.00

0.00

0.01

0.13

2.00

2.00

1.55

1.71

0.45

0.29

0.00

0.00

2.00

2.00

1.80

1.61

0.00

0.39

0.20

0.00

2.00

2.00

0.21

0.52

0.00

0.29

1.79

1.14

2.00

1.95

1.41

1.26

0.11

0.20

0.48

0.53

2.00

2.00

0.00

0.00

0.65

0.71

1.36

1.29

2.00

2.00

0.00

0.98

0.85

0.06

1.15

0.96

2.00

2.00

__ ____

--

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

59
Table 12.0ptimal Allocation of Irrigation Under Water Scarcity (Supply=3 acre-inch)
Stage

Maize

Cotton

Peanut

Total Irrigation

Dry

Normal

Dry

Normal

Dry

Normal

Dry

Normal

1.47

1.45

1.53

1.55

0.00

0.00

3.00

3.00

1.07

1.45

1.93

0.06

0.00

1.50

3.00

3.00

1.99

1.87

0.00

0.00

1.01

1.13

3.00

3.00

1.55

1.71

1.45

1.28

0.00

0.02

3.00

3.00

1.80

1.61

0.00

1.23

1.20

0.17

3.00

3.00

1.21

0.72

0.00

0.52

1.79

1.10

3.00

2.34

1.30

1.21

0.40

0.35

1.13

1.13

2.82

2.70

0.29

0.27

1.36

1.34

1.36

1.17

3.00

2.78

0.69

1.04

1.15

0.82

1.15

0.96

2.99

2.83

Table 13. Optimal Allocation Under Excess Supply of Irrigation (Supply=6 acre-inch)
Maize
Stage

Cotton

Peanut

Total Irrigation

Dry

Normal

Dry

Normal

Dry

Normal

Dry

Normal

1.53

1.55

1.53

1.55

0.00

1.55

3.06

4.65

1.93

1.43

1.93

1.43

1.99

1.43

5.85

4.29

1.99

1.87

1.99

1.87

1.99

1.87

5.97

5.62

1.55

1.71

1.55

1.71

1.55

1.71

4.64

5.12

1.80

1.61

1.80

1.61

1.80

1.61

5.40

4.83

1.78

0.72

1.78

0.72

1.78

0.72

5.34

2.16

1.13

1.21

1.13

1.21

1.13

1.21

3.38

3.64

1.36

1.15

1.36

1.15

1.36

1.15

4.07

3.45

1.15

0.96

1.15

0.96

1.15

0.96

3.46

2.89

if
R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

2 .5

Total Supply = 2.0

a Maize

* Peanuts

Cotton

Total

oto
c

&

0.5

10

G ro w th S ta g e

Figure 10. O ptim al A llocation Of Irrigation in Normal Y ears

2.5
Total Supply = 2.0

a Maize

Peanuts

* Cotton

Total

&
O
CO

0.5

10

Grow th Stage

Figure 11. Optimal A llocation Of Irrigation in Dry Years

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

61
4

Total Supply = 3.0

a Maize

Peanuts

Cotton

Total

0
0

10

Growth Stage

Figure 12. Optim al A llocation of Irrigation in Normal Y ears

4
Total Supply = 3.0

Peanuts

a Maize
Cotton

Total

0
0

Growth Stage

Figure 13. Optim al A llocation o f Irrigation in Dry Years

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

10

62
6
T otal Supply = 6.0

Maize

Peanuts

Cotton

Total

10

Growth Stage

Figure 14. Optimal A llocation o f Irrigation in Normal Years

8
Total Supply = 6.0

a Maize

* Peanuts

Cotton

Total

0
0

10

Growth Stage

Figure 15. Optimal A llocation o f Irrigation in Dry Years

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

63
10). As the total irrigation water supply increases from one acre inch, more than one crop
was irrigated at each decision period. When two acre inches of water was available for
irrigation, the crop which had the second highest marginal value received the remaining
amount of irrigation after irrigating the crop with highest marginal value in both dry and
normal years (Table 11, Figures 10 and 11).
As the total supply of irrigation water was increased to three acre-inch, the crop
with the lowest marginal value was irrigated (Table 12, Figures 12 and 13). In normal
years, all three crops received some amount of irrigation in seven out o f nine growth
stages. While in dry years, all three crops received supplemental irrigation only in three
growth stages. At this supply level, the total demand was less than supply in some growth
stages. When the total supply was fixed at six acre-inch, the optimal demand was always
less than total supply in both cases (Table 13, Figures 14 and 15). In this case, exactly the
same amount of irrigation was optimal for each crop at all growth stages in normal years
(optimal irrigation rules associated with other irrigation supply levels are reported in
Appendix 1). Even for dry years, the optimal amount of irrigation was identical for each
crops, except for peanuts at first growth stage (zero irrigation was optimal). Thus, a
continuous flow of irrigation water, which is more than or equal to six acre-inch, might not
have any impact on the allocation of irrigation water and on net farm income of the model
farm.
Thus, the optimal irrigation decision rule, when the supply of water is less than or
equal to one acre-inch, would be to irrigate only one crop, which has largest marginal
value, at each decision period. As the supply of water increases from one acre-inch,

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

64
irrigation is applied to more than one crop at a time. The total optimal irrigation demand
is less than six acre-inch per decision period. Therefore, irrigation is limiting factor for
crop production only when its supply is less than six acre-inch per decision period.
Optimal Demand for Irrigation Water
The results obtained from the recursive stochastic linear programming model
indicated that a continuous supply o f six acre-inch or more of irrigation water, at each
irrigation decision period, would be more than enough to meet the optimal irrigation
demand of our model farm (Table 14). A flow of water less than six acre-inches would
impose restriction on irrigation. For example, if the total supply is fixed at five acreinches per decision period, the optimal demand for irrigation would be higher than total
supply at the third and fourth plant growth stages in normal years (Figures 16 and 17; see
Table 32 in Appendix-1 for details). While in dry years, the total supply will fall short in
the second, third, fifth, and sixth decision periods.
On average, the aggregate demand would always be higher than three acre-inch in
dry years and greater than two acre-inch in normal years. The highest optimal irrigation
demand was observed at the third decision period under both dry (5.97 acre inches) and
normal (5.62 acre inches) weather conditions. Crop-wise average total demand for
irrigation in dry years would be 14.21, 14.21, and 12.74 acre-inch for com, cotton, and
peanuts, respectively. For normal years, the average demand was 12.22 acre-inch for all
three crops (Table 14). The average simulated rainfall was 10.99 inches for normal years
and 5.64 inches for dry years. As a result, the total optimal demand for irrigation water
was higher by 4.5 acre-inch for dry years as compared to the normal years (Table 15).

i
R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

65
6
Total Supply = 5.0

g Maize

* Peanuts
Total

Cotton
5

0
0

10

Growth Stage

Figure 16. Optimal A llocation o f Irrigation in Norm al Years

6
Total Supply = 5.0

Peanuts
Cotton

Total

Grow th Stage

Figure 17. Optimal A llocation o f Irrigation in Dry Y ears

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

10

I
i

66
Table 14. Optimal Demand for Irrigation Water by Crop and Growth Stage
Stage

Maize

Cotton

Peanut

Total Irrigation

Dry

Normal

Dry

Normal

Dry

Normal

Dry

Normal

1.53

1.55

1.53

1.55

0.00

1.55

3.06

4.65

1.93

1.43

1.93

1.43

1.99

1.43

5.85

4.29

1.99

1.87

1.99

1.87

1.99

1.87

5.97

5.62

1.55

1.71

1.55

1.71

1.55

1.71

4.64

5.12

1.80

1.61

1.80

1.61

1.80

1.61

5.40

4.83

1.78

0.72

1.78

0.72

1.78

0.72

5.34

2.16

1.13

1.21

1.13

1.21

1.13

1.21

3.38

3.64

1.36

1.15

1.36

1.15

1.36

1.15

4.07

3.45

1.15

0.96

1.15

0.96

1.15

0.96

3.46

2.89

Total 14.21
12.22
12.22
12.22
41.16
36.66
14.21
12.74
Note: These irrigation demand figures correspond to the optimizing behavior of a model
farm of three acres. The irrigation was measured in acre-inch.

Table 15. Total Optimal Demand for Irrigation and Rainfall by Growth Stage
Stage

Total Irrigation Demand (acre-inch)

Rainfall (inch)

Dry

Normal

Difference

Dry

Normal

Difference

3.06

4.65

-1.59

0.55

0.60

-0.05

5.85

4.29

1.55

0.05

1.02

-0.97

5.97

5.62

0.35

0.52

0.68

-0.16

4.64

5.12

-0.48

0.39

0.84

-0.45

5.40

4.83

0.57

0.27

1.55

-1.28

5.34

2.16

3.18

1.22

1.97

-0.75

3.38

3.64

-0.26

0.47

1.11

-0.63

4.07

3.45

0.62

0.85

0.87

-0.03

3.46

2.89

0.57

1.33

2.35

-1.03

Total

41.16

36.66

4.50

5.64

10.99

-5.35

i
j

f
i
i
i

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

67
As discussed before, irrigation decisions are assumed to be made at the beginning
o f each growth stage using information on existing soil moisture and expected future
weather conditions. Although the actual soil moisture condition was known with certainty
at the time of decision making, the amount and timing of rainfall is not known until the
end of the period. This uncertainty about rainfall is modeled in the recursive stochastic
linear programming model by assuming normal (average) weather conditions to prevail at
all future decision stages including current period. The actual rainfall, however, was
modeled as a stochastic process, which follows its own historic patterns randomly. The
erratic nature o f rainfall makes it difficult to maintain the exact level of soil moisture
(measured as rain plus irrigation). As a result, the amount of soil moisture available to the
plants exceeded its maximum (2.5 acre-inch) in five out of nine growth stages in normal
years (Table 16). In dry years, it exceeded only in two growth stages. The average
simulated soil moisture level was always higher than two inches in normal years. While in
dry years, it was less than two inches in three out o f nine cases. Due to the stochastic
nature of rainfall, excess irrigation was more likely in normal years as compared to the
dry years.
Aggregate Irrigation Demand
The aggregate demand for irrigation water under two weather scenarios by three
competing crops grown in four different counties o f Middle Chattahoochee Sub-basin was
also estimated by using results obtained from the optimization model. Among six
counties that lie in this sub-basin, none o f the crops considered in this study were grown in
Muscogee County in 1996. Com was grown in Chattahoochee County but it was not
II

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

68
Table 16. Total Soil Moisture Available to the Plants at Different Growth Stages
Maize

Stage

Peanut

Cotton

Dry

Normal

Dry

Normal

Dry

Normal

2.08

2.15

2.08

2.15

0.55

2.15

1.97

2.45

1.97

2.45

2.04

2.45

2.51

2.56

2.51

2.56

2.51

2.56

1.94

2.55

1.94

2.55

1.94

2.55

2.07

3.16

2.07

3.16

2.07

3.16

3.00

2.69

3.00

2.69

3.00

2.69

1.60

2.32

1.60

2.32

1.60

2.32

2.20

2.02

2.20

2.02

2.20

2.02

2.48

3.32

2.48

3.32

2.48

3.32

Total

19.85

23.21

19.85

23.21

18.38

23.21

Table 17. Irrigation Demand Estimates for the Middle Chattahoochee Sub-Basin by Crop
Maize
Stage

Cotton

Peanuts

Dry

Normal

Dry

Normal

Dry

Normal

0.578

0.585

0.940

0.952

0.000

1.053

0.729

0.540

1.186

0.878

1.352

0.972

0.752

0.706

1.222

1.149

1.352

1.271

0.585

0.646

0.952

1.050

1.053

1.162

0.680

0.608

1.106

0.989

1.223

1.094

0.672

0.272

1.093

0.442

1.210

0.489

0.427

0.457

0.694

0.743

0.768

0.822

0.514

0.434

0.835

0.706

0.924

0.781

0.434

0.363

0.706

0.590

0.781

0.652

8.296
6.16
Total
5.371
8.734
7.499
8.663
Note: The irrigation demand was measured in million gallons per week (mgw).

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

69
irrigated. Therefore, the demand for irrigation water was not calculated for these two
counties. The total optimal demand for irrigation water by individual crops grown in the
Middle Chattahoochee Sub-Basin under dry and normal weather conditions is reported in
Table 17. The aggregate optimal demand for irrigation of the three crops considered in the *
study was obtained by aggregating the total demand by individual crops (Table 18).
Since crops other than com, cotton, and peanut were grown in 1996 and
information on irrigation status of these crops was not available for the counties included
in the study, the aggregate total optimal demand for irrigation water could not be
estimated. Therefore, projections made by other studies were used to approximate the
total demand for the Middle Chattahoochee Sub-Basin. Wreck et al., in their study
Basinwide Management o f Water in the Alabama-Coosa-Tallapoosa and ApalachicolaChattahoochee-Flint River Basins have made projections for municipal and industrial
(M&I) and irrigation demand. For the year 1990, the total population and irrigated area in
ACF basin was estimated to be 2.636 million and 0.601 million acres, respectively
(Marella et al.). Using this information on total population and irrigated acreage in ACF
basin and Middle Chattahoochee Sub-Basin, two ratios - for population (rp=total
population in the study area/ total population in ACF basin) and irrigated acreage (rj=total
irrigated acreage in the study area/ total irrigated acreage in the ACF basin) - were
estimated. These ratios were then used to derive the total M&I and irrigation demand for
the study area from the projections made for the whole ACF basin. Both the demand
projections for ACF basin made by Wreck et al. and the projections derived for the study
area are reported in Table 19.
i
I
i
i

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

70
Table 18. Aggregate Irrigation Demand for the Middle Chattahoochee Sub-Basin
Stage

Irrigation Demand Estimates for 1992

Irrigation Demand Estimates for

Dry Years

Normal Years

Dry Years

Normal Years

1.518

2.591

6.342

10.826

3.267

2.390

13.650

9.988

3.326

3.126

13.899

13.061

2.591

2.858

10.826

11.943

3.009

2.691

12.572

11.245

2.975

1.204

12.432

5.029

1.889

2.023

7.892

8.451

2.273

1.922

9.499

8.032

1.922

1.605

8.032

6.705

22.77
20.41
95.144
85.28
Total
Note: As forecasted by the Comprehensive Study (Wreck et al), the aggregate demand for
2050 is expected to be 4.1784 time higher from its 1992 level.

Table 19. M&I and Irrigation Demand Projections for the ACF Basin and the Study Area
CS Projections for the Whole Basin
Description

M&I

Agri.

2010

2020

2050

2000

2010

2020

2050

Surface 5586.00

6132.00

6279.00

6468.00

649.71

713.21

730.31

752.29

Both

6328.00

6937.00

7105.00

7357.00

736.01

806.84

826.38

855.69

Surface 2492.00

3129.00

3402.00

4529.00

232.39

291.80

317.26

422.36

11228.00 13734.00 14700.00 18823.00

1047.08

1280.78

1370.86

1755.36

882.10

1005.01

1047.57

1174.65

Both

2000

Projections for Study Area

Surface 8078.00

9261.00

9681.00

10997.00

Total

17556.00 20671.00 21805.00 26180.00 1783.09 2087.62 2197.25 2611.05


Both
Note: The demands were measured in million gallons per week (mgw). The future
demands for Middle Chattahoochee Sub-Basin were calculated as a fraction of
projections made for the whole ACF basin using estimated ratios for population
(rp=0.116) and irrigated acreage (r( = 0.093).

..

_____ _______________ _

__________________

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

Other things remaining the same, these projections of M&I and irrigation demands
are consistent as long as the estimated ratios of population and irrigated acreage remain
stable over time. Both surface and ground water supply sources are being used to meet the
existing water demands. The relationship between groundwater withdrawal and stream
flow is expected to be positive (Saliba). Wreck et al. used correlation coefficients ranging
from zero to one to specify the relationship between groundwater withdrawal and stream
flow. This study calculates the demand for surface water using two extreme correlation
coefficients - zero and one.
When the correlation coefficient is one, a unit of water withdrawn from an aquifer
is equivalent to a unit o f water withdrawn from the stream in terms of its impact on
downstream flow. In this case, the total demand for surface water would be the sum of
surface and ground water demand. While a zero correlation coefficient would mean that
there would be no impact o f groundwater withdrawal on downstream flow and the total
withdrawal from the stream alone (it was assumed that only the stream flow is the source
of surface water) would represent the total demand for surface water.
Historic Flow of Chattahoochee River
The historic stream flow data of Chattahoochee river measured at Columbus,
Georgia were used to approximate the supply of surface water in the study area. The
weekly stream flows for the study period were calculated by summing the daily flows
recorded at Columbus station from 1943 to 1993. The agreement between the Mobile
District and the State o f Georgia requires that a minimum flow o f 1150 cfs (cubic feet per
second) be maintained at Columbus (see Table 35 in Appendix-1 for weekly stream flow

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

72
data). This flow was considered to be too low for navigational purposes and an alternative
flow of 1650 cfs was considered by the Comprehensive Study (Wreck et al.).
Since the contract between the Mobil District and the State of Georgia ensures the
minimum flow o f 1150 cfs at Columbus, it was assumed that there would be no water
available for other uses if the flow is less than or equal to this minimum. Only the portion
of stream flow, vhich is more than 1150 could be allocated among its competing uses.
The hydropower generation, recreation, M&I, and irrigation are the major competing uses
of water after maintaining the required level o f downstream flow. The hydropower
generation and recreation are not consumptive uses of water. Water that passes through a
turbine generates the hydropower and enters to the downstream flow immediately.
Therefore, unless the amount of water used to generate electricity is more than the amount
required to maintain the minimum stream flow, it does not impose any restriction on water
use. This study assumes that the water used in hydropower generation is less than the
amount required to maintain the minimum flow.
Among M&I, recreation, and irrigation demand, M&I receives the first priority in
periods of water shortage. Since there have been no reallocations of reservoir space for
recreation (Wreck et al.), it is assumed to receive the lowest preference. Thus, in the case
of water shortage, maintenance of downstream flow receives the first priority followed by
M&I, irrigation, and recreation. Using this scheme of priority scale and historic flow data,
the incidence o f water shortage in meeting future irrigation demands were examined and
results are reported in Tables 20 and 21.

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w ith o u t perm ission.

73
The incidence o f water shortage, measured as the number o f times the historic
stream flow of Chattahoochee river measured at Columbus would not be enough to meet
the future irrigation demands, indicated that even if the stream flow remains constant over
time, it would not be enough to meet the future demands on various occasions (Tables 20
and 21). However, if there is no relationship between the amount o f water withdrawn
from an aquifer and the stream flow, there would be enough water to meet all projected
irrigation demands until 2050 even after maintaining the minimum downstream flow of
1150 cfs. On the other hand, if the minimum flow requirement is increased to 1650 cfs,
there would not be enough water to meet future irrigation demands. In this case, the
probability of water shortage was highest for the seventh growth stage. At second, third,
and fourth growth stages, the probability of water scarcity was zero (Table 20; see Tables
36 and 37 in Appendix-1 for probability estimates).
Assuming that the relationship between the level of groundwater withdrawal and
the stream flow is perfect (one), the probability of water shortage when the minimum flow
requirement is 1150 cfs, is positive only for the eighth growth stage. While the incidence
o f water scarcity increases significantly as the minimum flow requirement is increased
from 1150 to 1650 cfs. In particular, the probability of experiencing water scarcity in the
last growth stage (stage 9) increased from 0.04 to 0.08 (Table 21).
Since the historic flow rate might not be sustainable in the future for various
reasons, the incidence of water scarcity was also examined by assuming a ten percent
draw-down in stream flow and the results are reported in Tables 22 and 23 (see Tables 38
and 39 in Appendix-l for probability estimates). In both cases, when the correlation
i

I
!

ft
i

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

74
Table 20. Incidence of Water Shortage in Meeting Future M&I and Irrigation Demand
Year

Min- Stage1 Stage2 Stage3 Stage4 Stage5 Stage6 Stage7 Stage8 Stage9

2000

1150
1650

2010

1150
1650

2020

1150
1650

1150
2050

2
1
2
1
1
3
1650
Note: The total surface water demand projections for municipal and industrial (M&I) and
irrigation were taken from: Basinwide Management o f Water in the ACT/ACF River
Basins, Draft Report, August 1996. The projections for the study area were calculated as
a fraction o f total demand in the watershed area.

Table 21. Incidence o f Water Shortage in Meeting Future M&I and Irrigation Demand
Year

Min- Stage1 Stage2 Stage3 Stage4 Stage5 Stage6 Stage7 Stage8 Stage9

1150
2
1
1
3
1650
3
3
1
1150
2010
2
1
4
1650
I
3
3
1
1150
2020
2
1
4
1650
1
3
3
1
1
1150
1
1
1
2050
1650
1
1
2
3
4
1
3
3
Note: Assuming that a unit withdrawal of groundwater reduces stream flow by one unit,
the future surface water demand was defined as the sum of projected withdrawal from
ground and surface water. The projections for total municipal and industrial (M&I) and
irrigation demand (surface and ground water) were taken from: Basinwide Management
o f Water in the ACT/ACF River Basins, Draft Report, August 1996. The projections for
the study area were calculated as a fraction o f total demand of the watershed area.
2000

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w ith o u t perm ission.

75
between the level of water withdrawal from an aquifer and the stream flow is one (perfect)
and when it is zero (no correlation), the chances o f having water shortage increased
substantially. For the first case, the probability of encountering water shortage at the ninth
growth stage reached up to 0.14 (Table 23).
Impact of Water Shortage on Net Farm Income
The optimal net farm income that could be generated under both dry and normal
weather conditions were estimated for all nine water supply scenarios. The average net
farm income associated with nine different water supply situations and weather conditions
are reported in Table 24 and graphed in Figures 18 and 19.
The results indicated that a consistently higher level of net farm income could be
generated by increasing the supply of irrigation water up to the level where the aggregate
optimal demand equals total supply. For example, as the total supply of irrigation water
was increased from zero acre-inch, a higher level of net farm income was obtained until
the total supply reached six acre-inch. The rate of increase in net farm income, however,
was decreasing. The increase in water supply beyond six acre-inch had no impact on net
farm income. Thus, a continuous supply o f a six acre-inch of irrigation water at each plant
growth stage would ensure optimal net farm income, under both dry as well as normal
weather conditions, to a farm of three acres, which is equally planted with three crops com, cotton, and peanuts.

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

76
Table 22. Expected Incidence o f Water Shortage To Meet M&I and Irrigation Demand
Year Min- Stagel Stage2 Stage3 Stage4 Stage5 Stage6 Stage7 Stage8 Stage9
2000

1150
1650

1150
2010
1650
1150
2020

1650
1150

2050

4
1650
1
2
3
3
1
Note: The future stream flows were expected to be 90 percent of the historical stream
flows measured at Columbus Georgia. The projections for municipal and industrial
(M&I) and irrigation demand were taken from: Basinwide Management o f Water in the
ACT/ACF River Basins, Draft Report, August 1996. The projections for the study area
were calculated as fractions o f total demand in the watershed area.

Table 23. Expected Incidence o f Water Shortage To Meet M&I and Irrigation Demand
Year
2000
2010
2020

Min- Stagel Stage2 Stage3 Stage4 Stage5 Stage6 Stage7 Stage8 Stage9
1150
1650
1150
1650
1150
1650
1150

1
1
1
1
1
2
1

1
3
2
3
2
3
2

1
4
1
5
1
5
1

1
3
2
3
2
3
3

2050

I
4
1
5
I
5
1

5
1
6
2
6
2

2
3
3
7
3
5
7
3
3
1650
Note: The future stream flows were expected to be 90 percent o f the historical stream
flows measured at Columbus Georgia. Assuming that a unit withdrawal of groundwater
reduces stream flow by one unit, the future surface water demand was defined as the sum
o f projected withdrawal from ground and surface water. The projections for total
municipal and industrial (M&I) and irrigation demand (surface and ground water) were
taken from: Basinwide Management o f Water in the ACT/ACF River Basins, Draft
Report, August 1996. The projections for the study area were calculated as a fraction of
total demand of the watershed area.

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

77
300

+ Std. Deviation

B hcom e
250

200

150

00

100

Supply (acre-inch)

Figure 18. Irrigation & Net Farm Incom e in Normal Y ears

500

Std. Deviation

Income

400

300

fl!

>

00

jjj 200

o
100

Supply (acre-inch)

Figure 19. Irrigation & Net Farm Income in Dry Years

R eproduced with perm ission o f the copyright owner. Further reproduction prohibited w ith o u t perm ission.

78
Table 24. Summary of Results from the Optimization Model (RSLP)
Dry Years

Normal Years

Difference

Income

Std.

Income

Std.

Income

Std.

Income
Ratio
(Dry/normal)

67.44

76.38

126.03

73.05

-58.59

3.33

0.54

142.71

85.81

160.59

65.83

-17.88

19.98

0.89

253.12

58.71

198.68

39.14

54.44

19.57

1.27

315.92

50.30

214.48

35.64

101.44

14.66

1.47

366.94

41.63

232.75

15.99

134.19

25.64

1.58

389.10

43.67

239.26

12.85

149.84

30.82

1.63

398.42

45.57

240.98

11.21

157.44

34.36

1.65

398.42

45.57

240.98

11.21

157.44

34.36

1.65

Supply

1.65
398.42
45.57
240.98
11.21
34.4
8
157.44
Note: The level of farm income reported in this Table was based on a typical optimization
problem o f a model farm, which grows com, cotton, and peanuts at one acre o f land each.
All inputs, except for irrigation, were fixed at recommended levels and supplemental
irrigation was applied optimally using a recursive stochastic linear programming model.

To generalize the results obtained from the model farm, both input levels (water)
and associated output levels (net farm income) were divided by farm size (three). The
results obtained from this transformation are presented in Table 25 and graphed in Figure
20 .

Thus, the results indicated that the contribution of irrigation to net farm income
would be optimal, when the total supply of water is set at two acre-inch per decision
period. As expected, higher increments to net farm income would be obtained in dry years
($ 132.81) as compared to normal years ($ 80.33). The difference between dry and normal
years return was highest ($52.48), when two acre-inch of water was available at each

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w ith o u t perm ission.

79
Table 25. Average Net Farm Income and Marginal Impact of Water Scarcity
Dry Years
Supply

Income

Normal Years

Marginal

Income

Difference

Marginal

Income

Marginal

Income
Ratio
fDrv/normal)
0.54

-19.53

11.52

-5.96

13.57

0.89

66.23

12.70

18.15

24.11

1.27

20.93

71.49

5.27

33.82

15.67

1.47

122.31

17.01

77.58

6.09

44.73

10.91

1.58

1.67

129.70

7.39

79.75

2.17

49.95

5.22

1.63

2.00

132.81

3.10

80.33

0.57

52.48

2.53

1.65

2.33

132.81

0.00

80.33

0.00

52.48

0.00

1.65

2.67

132.81

0.00

0.00

22.48

0.33

47.57

25.09

53.53

0.67

84.37

36.81

1.00

105.31

1.33

42.01

0.00
80.33
52.48
0.00
1.65
Note: The entries reported in this Table were derived from Table 24 by dividing inputs
(water) and output (net farm income) levels by farm size (3).

Table 26. Aggregate Net Farm Income at Various Supply Levels


Amount o f

Net Farm Income (million $)

Irrigation (mgw)

Dry Years

Normal Years

Difference

0.000

0.199

0.372

-0.173

0.535

0.421

0.474

-0.053

1.070

0.746

0.586

0.161

1.605

0.932

0.632

0.299

2.140

1.082

0.686

0.396

2.676

1.147

0.706

0.442

3.211

1.175

0.711

0.464

3.746

1.175

0.711

0.464

1.175
0.711
0.464
4.281
Note: The amount o f irrigation water and associated aggregate net farm income was
calculated by using information provided in previous Tables. The numbers printed in
bold face are optimal values.

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

80
145

Dry
Normal

120

Difference
95

45
201

-30
0

0.3

0 .5

0 .8

1.3

1.5

1.8

2.3

2 .5

2.8

Supply (acre-inch)

Figure 20. Net Farm in com e in Dry and Norm al Y ears

40
Incom e Variability b e tw ee n Normal & Dry Y ears
35

30

Normal

25

</

Io

0.25

0 .5

0.75

1.25

1.5

1.75

2 .2 5

2.5

2.75

Supply (acre-inch)

Figure 21. M arginal Im pact of Irrigation on Farm Incom e

X'

- -

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

81
decision period. Also, the ratio between dry and normal years net returns to irrigation
was also highest at this level o f water supply. On the average, the contribution o f
irrigation to net farm income was 1.65 time higher in dry years than in normal years
(Table 25).
The marginal contribution of irrigation water to net farm income was also
estimated (Table 25 and Figure 21). The results indicated that the average marginal
contribution of water to net farm income increases initially with the increase in the amount
of irrigation, reaches its peak (at 0.67 acre-inch), and then starts to decline. As expected,
the marginal contribution o f irrigation in dry years is always higher than in normal years.
The gap between dry and normal years marginal value of irrigation was largest, when the
average supply of irrigation water was 0.67 acre-inch. As the flow of irrigation water
increases, its marginal contribution to net farm income keeps on falling until it reaches to
zero.
The income stabilizing role of irrigation can be observed from the Figure 22. In
both dry and normal years, the standard deviation of net farm income decreases as the
supply of water increases until all crops are fully irrigated. Moreover, the net farm income
is more stable in normal years than in dry years. As it can be seen from the Figure, the
gap between dry and normal years standard deviation is much wider when the supply of
irrigation is more than enough to meet the optimal demand as compared to when it is
limiting.
Using information provided in Tables 17 and 25, the aggregate net farm income
associated with different water supply levels was estimated for the study area. The results

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

82
100

S t. Deviation o f Net Farm Incom e in Normal & Dry Y ears

o
o
S

40

Supply (acre-in ch )

Figure 22. Im pact o f Irrigation & W eather on Farm Incom e

/
J

0.8

0.6

g
o

0.4

k.

D ifference

0.2

0.2

0.5

1.5

2.5

3.5

4.5

Supplem ental irrigation (rrg w )

Figure 23. Im pact o f W ater Scarcity on Net Farm Incom e

tI
\f

|
R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

83
are presented in Table 26 and graphed in Figure 23. The aggregate optimal demand for
supplemental irrigation was estimated to be 3.211 million gallons per week (mgw). The
optimum net return to irrigation associated with this optimal irrigation level would be
$ 1.175 million per year for dry years and $0.711 million for normal years. Thus, the
aggregate impact of water shortage, measured at optimum level, would be higher in dry
years by $0,464 million as compared to normal years.
Water Management Issues
Water has many uses. First and foremost, it is an essential element of a life
support system. It is used directly to maintain the fluid composition of human body and
indirectly through other consumption goods which themselves need water for their
production. Second, it is also an important factor of production used in almost all
production activities. Among others, this dual role has made it difficult to treat water as a
purely economic resource.
Although The Global 2000 Report has estimated the total world water supply to be
3.5 times higher than the estimated demand for the year 2000, spatial and temporal
variation in water supply has created serious water scarcity problems in many parts of the
world (Tietenberg). Whenever it becomes scarce, depending upon the degree of scarcity,
it is allocated among its competing uses by using a combination of social priority schemes
and economic factors. In this priority scale, the role of water as a human life support
system receives the first priority. Once this demand is satisfied, the remaining amount of
water is generally allocated among other competing uses by applying a combination of
regulatory and market forces (Winpenny).

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

84
Moreover, water flowing in a river is subject to sequential technical externalities,
which means that upstream users can affect both quantity and quality o f water available to
downstream users but downstream users cannot impact those upstream. This
interdependence between upstream and downstream uses brings many non-economic
factors into play. In addition, when a water basin includes more than one political unit, the
allocation problem becomes more complex. The existing conflict among various water
user groups in the ACT-ACF basins presents a case where the water allocation problems
have brought various economic and non-economic forces into play.
Although non-economic forces are equally important players in framing water
allocation decision rules, a better resolution of most of these problems can only be
achieved by maintaining a balance among all competing forces. From an economic point
of view, efficient allocation o f surface water requires equality o f marginal net benefits
among competing uses. Thus, to determine the optimal allocation rules that are capable
of maintaining such a balance among competing forces, the marginal value of water
associated with different water supply levels and uses must be known. Unlike other
commodities, water markets are not well developed, which makes it difficult to estimate
the marginal value o f water.
To fill this information gap, various indirect methods have been used in the
literature. A combination o f simulation, econometric, and optimization models were
developed to estimate the value of water. The methodology developed in this paper can be
used to examine the impact of changes in irrigation technology and other market and
nonmarket forces on optimal irrigation decision rules. One of the aspects of technological

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

85
improvement would be to increase output without increasing the use of inputs. This will
shift the production function upward, which in turn would increase the average
productivity o f the input. This problem can be handled in the model by simulating the
yield response function that corresponds to the new relationship. The methodology can
also be used to examine the impact of an improved weather forecasting capability on
farming activities and net farm income.
The methodology developed in this study can also be used to examine the various
facets of market forces. One common application would be to examine the sensitivity of
irrigation decision rules with respect to a change in input and output prices. Depending
upon the nature o f price change, its impact would be to move the marginal value product.
For example, if output price increases, the marginal value product would be higher, and
would result in higher net farm income. The effect o f such changes in marginal value can
be examined by changing the parameter value in the optimization model.
For illustrative purposes, the following policy issues are analyzed based on the
results obtained from the optimization model and expectations about future conditions. As
mentioned before, the three basic questions of policy interest are: what are the long-term
implications o f changes in cropping/land use pattern on irrigation demand?, how much
water can be conserved by adopting water saving irrigation technologies?, and what
would be the impact of reduction in stream flow by 15 percent on net farm income?.
Land Allocation/Cropping Pattern
Changes in various market and non-market forces such as crop prices, availability
of improved crop varieties, and government policy on subsidy and regulation may affect

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

86
cropping pattern. For example, a policy change that increases (decreases) the relative
price o f peanuts will encourage producers to increase (decrease) the total peanut acreage.
Assuming that the total irrigable land is fixed, such reallocation may affect irrigation
demand in various ways. For example, if the crop replaced by peanut is relatively less
water-intensive, then this change in cropping pattern will increase the demand for
irrigation. On the other hand, if peanut is more water intensive than the crop that replaces
peanut, then it will reduce the overall irrigation demand. This issue can be incorporated in
the optimization model by adjusting water response function parameters of the crop in
question.
Adoption of Water Saving Technologies
As discussed above, a plants ability to absorb water from soil depends on the level
of moisture contained in the soil. When soil moisture is at the field capacity (when soil
particles are fully saturated) plants can easily extract water from the soil. As the level of
moisture decreases from the field capacity, soil tension increases, which makes it difficult
for plants to withdraw water from the soil. Plants start to wane out as the level of soil
moisture approaches permanent wilting point, a state when the amount of water withdrawn
by plants becomes less then required for plant survival. Supplemental irrigation is applied
to maintain soil moisture in the plants root zone so that plants can easily absorb water
from the soil.
Various methods such as traveling gun, center pivot (towable, low pressure, and
sprinkler), low energy precision application (LEPA), and irrigation scheduling are used in
the ACF basin to deliver water from its source to crop fields. As in most other uses,

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

87
technology plays an important role in determining irrigation efficiency. For example, in
terms o f water saving potential, the irrigation efficiency of a center pivot is higher than a
traveling gun irrigation system by ten percent. Based on the information about percentage
o f farmers using a particular technology, expectation about future adoption rate, and water
saving factor for different technologies, the Comprehensive Study (Agricultural Water
Demand, Technical Report) has estimated various water saving factors for the ACT/ACF
water basins. For the Middle ACF basin, the potential for water conservation through
improvement in row crop irrigation technology is estimated to be 8.6 percent for the year
2050. If this expectation is realized, about 276,146 gallons of water per week could be
saved just by enhancing the efficiency o f row crop (com, cotton, and peanuts) irrigation
technologies in the Middle Chattahoochee sub-basin.
While, for orchards, the potential water saving factor was estimated to be 3.2
percent. Assuming that an average of these two ratios (5.9 percent) can be used to
approximate the overall water conservation factor for the study area, about 103.57
(0.059*1755.35) million gallons of water per week could be saved in the Middle
Chattahoochee sub-basin. Moreover, if this amount of water is not used in other
consumptive activities or it is allowed to flow in the stream, it would increase downstream
flow approximately by 23 cfs during peak irrigation period.
The Im pact of Reduction in Stream Flow
Based on discussion with policy makers who are involved in the ongoing water
wars among Alabama, Florida, and Georgia, a drawdown of 10-15 percent in downstream
flow would be a most likely scenario for the future. The burden o f such a drawdown on

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

88
various uses can be examined by using different sharing rules. Out o f many possibilities,
the impact o f water scarcity on crop production is examined for two different cases - i) the
burden o f scarcity is shared equally by all sectors and ii) irrigation, which lies in the lower
priority scale, will receive water only after satisfying needs of those in higher priority
levels.
As discussed earlier, maintenance o f minimum flow would be the first priority,
followed by M&I and irrigation. Although hydropower generation, navigation, flood
control, recreation, and water quality were the other components of the Comprehensive
Study, they are not included. Using this priority scheme, the future supply o f irrigation
water was calculated from the historic flow o f Chattahoochee river measured at Columbus,
Georgia.
Scenario 1:
The downstream flow in the Chattahoochee river decreases by 15 percent and the
burden of shortfall in water supply is equally shared by all competing uses (i.e., each
sector reduces demand by 15 percent). In other words, the residual amount o f water after
maintaining the minimum flow and satisfying the demand for M&I (85 percent o f normal)
will be available for irrigation.
Scenario 2:
Despite the fact that stream flow will decrease by 15 percent, only the residual
amount of water, after satisfying the minimum flow requirement and the total M&I
demand, will be available for irrigation.

i.

. .

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w ith o u t perm ission.

89
Results
The incidence o f water shortage for both scenarios was examined by using two
extreme assumptions about the relationship between stream flow and groundwater
withdrawal rates (Tables 36-39). The results indicated that a shortfall of stream flow by
15 percent would increase the incidence of water scarcity substantially. If stream flow and
groundwater withdrawals are perfectly correlated, the historic stream flow would not be
enough to meet the projected demand for the year 2050 at the sixth crop growth stage in as
many as 23 cases out o f 50. This indicates the seriousness o f the water scarcity problem
faced by the agriculture sector.
The weekly shortfall in irrigation water supply may not have any economic
consequences if it does not persist for long and if it is not accompanied by a dry weather
condition. To examine the seriousness of this problem, the number of cases when stream
flow was not enough to meet the irrigation demand continuously for more than three
weeks were recorded. In the case o f perfect correlation, there were nineteen cases (out of
50), when stream flow was not enough to meet the total irrigation demand continuously
for more than three weeks.
When the burden of water scarcity was shared equally by M&I and irrigation
(scenario 1), both the extent of shortfall, as well as the incidence of irrigation water
scarcity were slightly lower. The actual impact of water scarcity on net farm income
depends on the extent o f shortfall in relation to its optimum level. For example, as shown
in Table 34, a decrease in water supply from its optimum level of 2" per decision period to
1.67" would reduce the incremental net farm income from $ 132.81 to $129.7 per acre in a

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

90
dry year. A shortfall from 0.67" to 0.33" per acre would reduce the incremental net farm
income from $ 84.37 to $ 47.57 per acre. Thus, the impact of water scarcity on net farm
income depends on the level o f scarcity faced by the farm. As hypothesized, if a 15
percent reduction in future stream flow is realized and the burden of this shortage is
equally shared by all existing users and uses, then its impact on net farm income would be
less than $3.10 per acre in dry years and $ 0.57 per acre in normal years (assuming that
the level of irrigation before this policy change was at optimum).

ti

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

91
Table 27. Scenario 1: Incidence of Water Shortage in Meeting Future Irrigation Demand
Year Min- Stage1 Stage2 Stage3 Stage4 Stage5 Stage6 Stage7 Stage8 Stage9

1150
2000
1650

1
7

1
8

3
11

~ 1150
2010
1650

1
7

1
8

3
11

1150
2020
1650

1
8

1
8

1150

3
13

5
18

3
7

4
7

5
9

3
13

5
18

3
7

5
7

6
9

3
11

3
13

5
18

3
7

5
7

6
9

2050

8
8
9
11
6
13
18
7
7
1650
Note: The correlation between groundwater withdrawal and stream flow was assumed to
be zero and a 15 percent decrease in stream flow was assumed.

Table 28. Scenario 1: Incidence of Water Shortage in Meeting Future Irrigation Demand
Year

Min- Stage1 Stage2 Stage3 Stage4 Stage5 Stage6 Stage7 Stage8 Stage9

2000

1150
1650

2
9

4
11

3
13

3
7

4
16

8
21

4
11

5
10

7
12

2010

1150
1650

3
10

4
14

4
15

3
8

4
16

9
23

4
12

5
11

8
12

2020

1150
1650

3
10

4
14

4
15

3
9

4
16

9
23

4
14

6
13

8
12

1150

11

2050

13
15
12
15
10
18
23
16
14
1650
Note: The correlation between groundwater withdrawal and stream flow was assumed to
be one and a 15 percent decrease in stream flow was assumed.

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

92
Table 29. Scenario 2: Incidence of Water Shortage in Meeting Future Irrigation Demand
Year Min- Stagel Stage2 Stage3 Stage4 Stage5 Stage6 Stage7 Stage8 Stage9
2000

1150
1650

1
7

1
8

3
11

2010

1150
1650

2
8

2
8

2020

1150
1650

2
8

1150

3
13

5
18

3
7

4
7

5
9

3
11

1
6

3
13

5
18

3
7

5
7

6
9

2
8

3
11

1
6

3
13

5
18

4
7

5
7

6
9

2050

9
6
15
11
19
8
7
8
9
1650
Note: The correlation between groundwater withdrawal and stream flow was assumed to
be zero and a 15 percent decrease in stream flow was assumed.

Table 30. Scenario 2: Incidence of Water Shortage in Meeting Future Irrigation Demand
Year Min-Flow

Stagel Stage2 Stage3 Stage4 Stage5 Stage6 Stage7 Stage8 Stage9

2000

1150
1650

3
9

4
11

3
14

3
7

4
15

9
21

4
11

4
10

7
12

2010

1150
1650

3
10

4
15

4
15

3
9

4
16

9
21

4
14

6
12

8
12

2020

1150
1650

3
11

4
15

5
15

4
10

4
16

10
23

4
16

6
13

8
12

1150

11

2050

10
14
16
15
17
13
18
23
1650
13
Note: The correlation between groundwater withdrawal and stream flow was assumed to
be one and a 15 percent decrease in stream flow was assumed.

i'

---------------------------------------------------------_

--------------------------------------------------------

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

SUMMARY AND CONCLUSIONS

The resolution of the ongoing water wars among Alabama, Florida, and Georgia
is likely to have some impact on downstream flow in the ACF river basin. As
anticipated, if it reduces stream flow, the effect of such changes will be encountered in
varying degrees by different water user groups. The Comprehensive Study has made an
attempt to examine the impact o f changes in stream flow on downstream users by looking
at various sectors such as power generation, municipal and industrial, environmental,
navigation, agriculture, recreation, and water quality. The two main issues focused by the
Comprehensive Study were will there be enough water?" in ACT and ACF river basins
and is there a better way to manage water in the basins?". This study attempts to find
answers to those questions for the Middle Chattahoochee river basin by examining the
optimal demand for irrigation water and the residual stream flow that could be available
for crop irrigation.
The existing legal and social setup prohibits unlimited withdrawal of water from
ACF river systems. For example, the contract between the Mobile District of the Army
Corps of Engineers and the State of Georgia requires that a minimum flow of 1150 cfs be
maintained in Chattahoochee river at Columbus. It means that only the residual amount,
actual flow minus minimum flow required by the contract, is available to the off-stream
users.
93

I
I
I

I
t
t

R eproduced w ith perm ission o f the copyright owner. F urth er reproduction prohibited w ith o u t perm ission.

94
All major reservoirs on the ACF river basins are owned and operated by the Army
Corps of Engineers. The U. S. Congress authorized the construction of these reservoirs
for specific purposes such as hydropower generation, navigation, and flood control. The
Corps can support other activities such as recreation and water quality releases to the
extent that authorized purposes are not significantly affected (Comprehensive Study).
The amount o f reservoir storage space allocated to a particular use can be used as
a measure o f relative weight assigned to that activity. The storage space is allocated and
approved by the Congress when the approval to construct the reservoir is granted. Once
the storage space is assigned to an activity, it cannot be reallocated without Congressional
approval. The Chief of Engineers, however, is authorized to reallocate up to 15 percent
of the total storage space or 50,000 acre-feet (whichever is less) to M&I at his discretion.
Although the Corps has made such reallocations in favor of M&I in the past, it has not
yet reassigned any reservoir space to recreational uses (CS). Thus, it may mean that
when water becomes scarce and M&I and recreational uses are competing with each other
for the same unit of water, M&I will be the winner.
Moreover, among hydropower, M&I, recreation, navigation, and irrigation
demands, M&I is likely to receive preferential treatment whenever water becomes scarce.
Since hydropower generation is not a consumptive use and the water that passes through
the turbine generates the electricity and enters immediately to the river system, it may not
impose any resource constraint if the total amount of water used in generating hydro
power is less than the flow required to ensure minimum downstream flow. In this study,
i

it is assumed that the water used in hydro-power generation is less than the amount

!
i

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w ith o u t perm ission.

95
required to maintain the minimum downstream flow. Thus, when water is scarce,
maintenance of downstream flow receives the first preference followed by M&I,
irrigation, and recreation. Although other uses such as water quality, navigation, and
environment are also important, this study uses this priority scale to calculate the supply
o f irrigation using historical flow data.
Since the impact o f water shortage is expected to be much higher in dry years than
normal years, separate yield response functions were estimated for dry and normal
weather conditions. Using these yield response functions, recursive stochastic linear
programming models were developed to examine the optimal behavior of a model farm
o f three acres which grows com, cotton, and peanut. The results obtained from the
optimization model were used to calculate the total optimal irrigation demand o f the
model farm and the Chattahoochee sub-basin.
A combination of simulation, econometric, and optimization models were
developed to estimate the impact o f water scarcity on net farm income. The functional
relationship between the amount o f irrigation applied at different stages o f plant growth
and final crop yield was assumed to be nonlinear. A Cobb-Douglas production function
was estimated by using the simulated data on crop yield, amount o f supplemental
irrigation, and rainfall. As expected, all the estimated coefficients o f the response
functions are positive and most of them are highly significant. These results indicate that
the relationship between the amount o f irrigation applied at different growth stages and
the final crop yield is positive. The partial elasticity of crop yield with respect to

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

96
irrigation water was less than one in all cases, indicating that the yield response functions
are concave.
The magnitude o f yield response function parameters differs across the stages.
This result is consistent with the expectation that the impact o f water stress on crop yield
depends on the stages o f plant growth when such scarcity is encountered. Thus, the
marginal physical productivity of water applied to a crop depends on the timing as well as
the amount o f irrigation. Since the marginal value is derived from the marginal physical
productivity, the marginal value of irrigation also differs across the stages.
The marginal value of water can be used to derive optimal irrigation decision
rules. These rules, however, are not capable o f handling the dynamic aspects of the
irrigation decision problem. Plants withdraw water from the soil moisture available in
the root zone. The amount of supplemental irrigation applied and rainfall are the only
two external sources o f soil moisture. Since rainfall is a random process, the irrigation
demand is stochastic.
Assuming that rainfall follows a stochastic process, an optimization model was
developed to estimate the impact of water scarcity encountered at different stages of plant
growth on crop yield under dry and normal weather conditions. The basic model was
designed to address the irrigation decision problem faced by a model farm which is
growing com, cotton, and peanut in one acre o f land each. Irrigation decision rules were
examined for nine different supply levels. The optimal irrigation demand of the model
farm is always less than six acre-inches. Thus, water scarcity may result when the supply
of irrigation water drops down from six acre-inches per decision period. A fixed flow of

{i
i

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

irrigation water, greater than or equal to six acre-inch, would not impose any restriction
on irrigation.
The historical flow data of Chattahoochee river measured at Columbus was used
to approximate the total supply of surface water in the study area. Using the preference
scale described above, the residual flow o f water that can be used to irrigate crops grown
in the Middle Chattahoochee sub-basin was derived. Since crops other than com, cotton,
and peanut are also grown and the irrigation demand is only a component of the total
agricultural demand for water, the agricultural use of water for the study area was derived
from the prediction made by the Comprehensive Study for the ACF basin. The incidence
of water scarcity, defined as the number o f times when historical flow of water would not
be enough to meet future irrigation demand, was estimated for the study area.
Both surface and ground water sources have been used in the ACF basin to
mitigate the demand for water. The relationship between the amount of water withdrawn
from an aquifer and the stream flow is expected to be positive. The Comprehensive
Study used different values o f correlation coefficients that vary within the range of zero
to one to reflect the relationship between groundwater withdrawal and downstream flow.
In this study, the two extreme values of correlation coefficients - zero and one - were used
to measure this relationship.
The comparison between the projected demand and the residual flow, which is
expected to be available for agricultural use, indicated that water scarcity could be
encountered on various occasions. However, if the relationship between the amount of
water withdrawn from an aquifer and the stream flow is zero and the minimum

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

98
downstream flow requirement is 1150 cfs, the historical flow would be more than enough
to meet the future agricultural demand for water until 2050. As the minimum flow
requirement is increased to 1650 cfs from 1150 cfs, the historical flow is not enough to
meet the projected demand in many instances. When a correlation coefficient o f one
(perfectly correlated) is used, the historical flow is not enough to meet agricultural
demand even when the minimum flow requirement is fixed at 1150 cfs.
Since the historical stream flow might not be sustainable in the future for various
reasons, the incidence of water scarcity was also examined by assuming a decline in
stream flow by ten and fifteen percent from its historical level. This substantially
increased the probability of encountering water scarcity in the future. In particular, when
minimum flow requirement was 1650, the chances of experiencing water shortage at sixth
and ninth growth stage reached as high as 14 percent. This result reflects the seriousness
of the water scarcity problem in the ACF basin.
The results obtained from the optimization model were also used to examine the
impact of water shortage encountered at different stages of plant growth on net farm
income. As expected, the impact o f water shortage on net farm income is much higher in
dry years as compared to normal years. Another role of irrigation, as indicated by the
results, was to stabilize net farm income. The fluctuation in net farm income, measured
by standard deviation of income, decreased substantially as the supply o f irrigation water
increased from zero. Even at the same supply level, the standard deviation is much
higher in dry years as compared to normal years.

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

99
The impact o f water scarcity on net farm income was examined both at the farm
and sub-basin level. For our model farm o f three acres, a continuous flow of six acreinches would ensure availability of optimal irrigation at all stages o f plant growth. The
optimal net farm income associated with this level of irrigation is $398.42 in dry years
and $240.98 in normal years. On the average, an availability o f two acre-inch of
irrigation water per decision period would ensure a net income of $132.81 from one acre
of land. Using this average relationship, the total agricultural demand for water in the
Middle Chattahoochee sub-basin was estimated to be 3.211 mgw. The optimal net farm
income associated with this irrigation demand level was $ 1.175 million in dry years and
$ 0.711 million in normal years.
Limitations and Recommendations
This study focuses only on three major crops grown in the Chattahoochee sub
basin. In 1996, other crops such as soybean, wheat, and hay were also grown. Moreover,
crop production is only one component of total use of water in the agriculture sector.
Therefore, the irrigation demand estimated by the study represents only a component of
total agricultural demand for water.
The gap between water withdrawn from the source and the actual amount of water
available to the plant at their root zone depends mainly on the irrigation technology.
Most center pivot technologies are said to attain an efficiency level of 90 percent. In this
study, only the actual amount of water that was available to the crops at its root zone was
measured as the demand, without accounting for conveyance loss.

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w itho ut perm ission.

100
The historical flow data o f the Chattahoochee river measured at Columbus were
used to approximate the total supply of irrigation without looking at other alternative
sources such as lakes, ponds, aquifer, and creeks from which irrigation water might be
used before it reaches the Chattahoochee river. Although it is difficult to maintain the
total inventory of water in a watershed basin, use of more complex modeling techniques
could provide flexibility in generating different supply scenarios. A simulation model
known as Soil and Water Assessment Tool (SWAT) has been used to simulate water flow
in a watershed basin. This simulation model has the capability to read various
topographical and land use data directly from a digital map. In particular, it can read GIS
(Geographical Information System) maps directly from the Geographical Resource
Analysis Support System (GRASS) and simulates water flow for the whole watershed
basin. This model is also capable o f generating various supply scenarios using user
specified what-if conditions.
The methodology developed in this study can be used to determine optimal
irrigation strategies and to examine the impact of irrigation water shortage on net farm
income. It is not designed, however, to deal with the problem of multiobjective resource
allocation. Nevertheless, the results obtained from this study can be used in developing a
multiobjective programming model which is capable of addressing basinwide water
allocation issues.

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w ith o u t perm ission.

REFERENCES

Alabama Cooperative Extension Service. Irrigated Com Production. Circular ANR165, Auburn University, Alabama, undated.
Alabama Cooperative Extension Service.. Scheduling Com Irrigation Using the Soil
Moisture Balance Method. Circular ANR-531, Auburn University, Alabama,
undated.
Alabama Cooperative Extension Service. 1996 Budgets for Major Row Crops in
Alabama. AEC BUD 1-1 January 1996, Department of Agricultural Economics
and Rural Sociology, Auburn University, Alabama, 1996.
Alabama Cooperative Extension Service. Scheduling Irrigation Using Soil Moisture
Tension. Circular ANR-158, Auburn University, Alabama, undated.
Alabama Cooperative Extension Service. Scheduling Peanut Irrigation Using the Soil
Moisture Balance Method. Circular 467, Auburn University, Alabama, undated .
Beattie, B. R. "Irrigated Agriculture and the Great Plains: Problems and Policy
Alternatives." West. J. Agr. Econ. 6( 1981 ):289-99.
Beattie, B. R. and C. R. Taylor. The Economics o f Production. John Wiley & Sons: New
York, 1985.
Bellman, R. E. and S. E. Dreyfus. Applied Dynamic Programming. Princeton:
Princeton University Press, 1962.
Bemado, D. J., N. K. Whittlesey, K. E. Saxton, and D. L. Bassett. An Irrigation
Model for Management of Limited Water Supplies." W. J. Agr. Econ.
12(December 1987): 164-73.
Boggess, W. G. and J. T. Ritchie. Economic and Risk Analysis of Irrigation
Decisions in Humid Regions." Journal o f Production Agriculture 1(February
1988): 116-122.

101

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w ith o u t perm ission.

102
Brooks, K. N., P. F. Ffolliott, H. M. Gregersen, and J. L. Thames. Hydrology and the
Management o f Watersheds. Ames: Iowa State University Press, 1991.
Bruce, R. R., J. S. Chesness, T. C. Keisling, J. E. Pallas, D. A. Smittle, J. R.
Stansell, and A. W. Thomas. Irrigation of Crops in the Southeastern United
States: Principles and Practice." U.S. Department o f Agriculture, Science and
Education Administration, Agricultural Reviews and Mannuals, ARM-S-9/May
1980.
Bryant, K. J, V. W. Benson, J. R. Kiniry, J. R. Williams, and R. D. Lacewell.
Simulating Com Yield Response to Irrigation Timings: Validation of the Epic
Model. Journal o f Production Agriculture 5(April-June 1992): 237-42.
Bryant, K. J., J. W. Mjeide, and R. D. Lacewell. An Intra-seasonal Dynamic
Optimization Model to Allocate Irrigation Water between Crops. Amr. J. Agr.
Econ. 75(November 1993): 1021-1029.
Burt, O. R. and M. S. Stauber. Economic Analysis of Irrigation in a Sub-humid
Climate."
Amr. J. Agr. Econ. 53(February 1971):33-46.
Cabelguenne, M ., C. A. Jones, and J. R. Williams. Strategies for Limited Irrigations
of Maize in Southwestern France - A Modeling Approach." Transactions o f the
ASAE 38(2, 1995):507-511.
Caswell, M. and D. Zilberman. "The Choices of Irrigation Technologies in California".
Amr. J. Agr. Econ. 67(Febmary 1985):224-34.
Chang, Ni-Bin, C.G. Wen, and S. L. Wu. Optimal Management of Environmental and
Land Resource in a Reservoir Watershed by Multiobjective Programming.
Journal o f Environmental Management 44 (1995):145-61.
Council on Environmental Quality. Environmental Trends. Washington, DC:
Government Printing Office, 1978.
Das, P. and Y. Y. Haimes. Multiobjective Optimization in Water Quality and Land
Management. Water Resource Research 15(6 1979): 1313-22.
Day, R. H. Recursive Programming. Workshop on the Firm and Market, Social
Systems Research Institute, University of Wisconsin, Workshop paper, 1963.
Day, R. H. Part I: General Concepts. Modeling Economic Change: The Recursive
Programming Approach. R. H. Day and A. Cigno, eds. New York: North-Holland
Publishing Company, 1978.

ji

i
|

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

103
Day, R. H. Farm Decisions, Adaptive Economics, and Complex Behavior in
Agriculture. Modeling Farm Decisions fo r Policy Analysis. K. H. Baum and L.
P. Schertz, eds. Boulder, Colorado: Westview Press, 1983.
Dinar, A. and E. T. Loehman. Water Quantity/Quality Management and Conflict
Resolution: Institutions, Processes, and Economic Analysis. Connecticut: Praeger,
West Port, 1995.
Doyle, J. Rationality and Its Roles in Reasoning. Proceedings of the Eighth National
Conference on Artificial Intelligence, Boston, Massachusetts, 1990.
Gibbons, D. C. The Economic Value o f Water. Washington, D C: Resources for the
Future, 1986.
Gisser, M. and D. Sanchez. "Competition Versus Optimal Control in Groundwater
Pumping." Water Resour. Res. 16(August 1980):638-42.
Glover, F. and S. Martinson. Multiple-use Land Planning and Conflict Resolution by
Multiple Objective Linear Programming. European Journal o f Operation
Research 28 (1987):343-50.
Goicoechea, A. and L. Duckstein. Multiobjective Programming in Watershed
Management: A Study of the Charleston Watershed. Water Resour. Res.
28(1976):343-50.
Good, I. J. Twenty-Seven Principles of Rationality." Foundations o f Statistical
Inference. V. P. Godambe and D. A. Sprott, eds. Toronto: Holt, Rinehart and
Winston, 1971.
Hale, Timothy W., Evelyn H. Hopkins, and Robert F. Carter. Effects of the 1986
Drought on Stream-flow in Alabama, Georgia, North Carolina, South Carolina,
Tennessee, and Virginia." U. S. Geological Survey, Water Resources
Investigation Report 89-4212, 1989.
Hardin, D. C. and R. D. Lacewell. Temporal Implications of Limitations on Annual
Irrigation Water Pumped from an Exhaustible Aquifer. W. J. Agr. Econ.
5(July l980):37-44.
Harrish, T. R., and H. P. Mapp. A Control Theory Approach to Optimal Irrigation
Scheduling in the Oklahoma Panhandle. Southern Journal o f Agricultural
Economics 12(1 1980): 165-71.
Hatch, Upton L., J. Harrington, R. Perritt, and M. Masucci. Agricultural

!!

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w ith o u t perm ission.

104
Adjustments to Drought in the Chattahoochee-Choctawhatchee River Basins of
Alabama: Survey Results." A paper submitted for publication consideration to
the Southern Business and Economics Journal, 1996.
Heady, Earl O., and Roger W. Hexem. Water Production Functions fo r Irrigated
Agriculture. Iowa: The Iowa State University Press, 1978.
Henderson, T. R. Groundwater: Strategies fo r State Action. Washington DC:
Environmental Law Institute, 1984.
Hook, J. E. Water Withdrawals fo r Irrigation in Drought Years. Environmental Resource
Center ERC 03-91, Georgia Institute of Technology, Atlanta, GA, 1984.
James, L. D. Economics o f Water Resources Planning. New York: MacGraw-Hill Inc,
1971.
Kolajo, E. K. Expectation Planning and Realization: A Recursive Strategic Linear
Programming Analysis of the Farm Firm Growth. Dissertation, Auburn
University, Auburn, AL, 1986.
Kramer, P. J. and T. T. Kozlowski. Physiology o f Woody Plants. Orlando: Academic
Press Inc, 1979.
Lacewell, R. D., J. M. Sprott and B. R. Beattie. "Value of Irrigation Water with
Alternative Input Prices and Yield Levels: Texas High Plains and Lower Rio
Grande Valley." Texas A and M University: Texas Water Resource Institute,
Technical Report 58, 1974.
Lamb, M. C. Economic Analysis of Planning, Management, and Marketing Peanuts in
the Southeast United States. Dissertation, Auburn University, Auburn, AL,
1995.
Lynne, G. D., P. J. d Almada, W. C. Martin, and R. S. Mansell. Area-wide
Agricultural Water Demand Projection Model for South Florida: Technical
Documentation, Version 2.2." Agricultural Experiment Station, Institute of
Food and Agricultural Sciences, University of Florida, Gainesville, 1987.
Marella, R. L. J. L. Fanning, and W. S. Mooty. Estimated Use of Water in the
Apalachicola-Chattahoochee-Flint River Basin During 1990 and Trends in
Water Use from 1970 to 1990. U. S. Geological Survey. Water Resources
Investigation Report 93-4084. Prepare in Cooperation with Geological Survey
of Alabama, Florida Department of Environmental Regulation, and the Georgia
Geological Survey, 1993.

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

105
McGuckin, J. T ., C. Mapel, R. R. Lansford, and T. W. Sammis. Optimal Control of
Irrigation Using a Random Time Frame." Amer. J. Agr. Econ. 69(January
1987): 123-33.
Musser, V. N. and B. V. Tew. Use of Biophysical Simulators in Production
Economics. South Journal o f Agricultural Economics 16(January 1984):74-86.
Nieswiadomy, M. "The Demand for Irrigation Water in the High Plains of Texas, 195780." Amer. J. Agr. Econ. 62(August 1985): 619-26.
Pearman, J. L., T. C. Stamey, G. W. Hess, and G. H. Nelson. Floods of February
and March 1990 in Alabama, Georgia, and Florida." U. S. Geological Survey,
Water Resources Investigation Report 91-4089, 1991.
Ridgley, M. A. and T. W. Giambelluca. Linking Water-Balance Simulation and
Multiobjective Programming: Land Use Plan Design in Hawaii. Environment
and Planning B: Planning and Design 19(1992):317-36.
Rochester, E. W ., P. A. Backman, J. A McGuire, L. M. Curtis, J. Starling, and H.
Ivey. Irrigation Schedules for Peanut Production. Bulletin 556, Alabama
Experiment Station, Auburn University, 1984.
Russell, S. J. and E. H. Wefald. Do the Right Thing: Studies in Limited Rationality.
Cambridge, Massachusetts: MIT Press, 1991.
Saliba, B. C. "Irrigated Agriculture and Groundwater Quality-A Framework for Policy
Development". Amr. J. Agr. Econ. 67(December 1985): 1231-37.
Shulstad, R., E. D. Cross, and R. D. May. The Estimated Value o f Irrigation Water in
Arkansas." Arkansas Farm Research 27(August 1982):2-3.
Simon, H. A. Models o f Bounded Rationality, Vol. 2. Cambridge, Massachusetts: MIT
Press, 1982.
Smith, N. B., E. F. Kolajo, and N. R. Martin, Jr. Developing a Method for the
Influence o f Conservation Practice on Farm Firm Growth. 8th International
Farm Management Congress, 1991.
Stansell, J. R., J. L. Shepherd, J. E. Pallas, R. R. Bruce, N. A. Minton, D. K. Bell,
and L. W. Morgan. Peanut Responses to Soil Water Variables in the
Southeast." Peanut Science 3(January 1976): 44-48.
Swaney, D. P., J. W. Jones, W. G. Boggess, G. G. Wilkerson, and J. W. Mishoe.

I|

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

106
Real-Time Irrigation Decision Analysis Using Simulation. Transactions o f the
ASEA 26(1983):562-68.

Tai, C.-F. An Integrating Analysis of Biological and Economical Aspects of Catfish


Culture. Dissertation, Auburn University, Auburn, AL, 1995.
The Comprehensive Study Technical Cooperation Group. 'Comprehensive Study:
Alabama-Coosa-Tallapoosa and Apalachicola Chattahoochee-Flint River
Basins." Vol. I, Plan of Study: Main Report, January, 1992.
Tietenberg, T. Environmental and Natural Resource Economics. 3rd ed. New York:
Harper Collins Publishers, 1992.
U. S. Department of Commerce, Bureau of the Census, 1990 Census o f Population and
Housing. Special Tabulations, 1991.
Van, D. A. and P. Nijkamp. A Multiobjective Decision Model for Regional
Development, Environmental Quality Control and Industrial Land Use. Papers
o f the Regional Science Association 36 (1976):35-57.
Werick, W. J., W. Whipple, Jr., and J. Lund. Basinwide Management of Water in the
Alabama-Coosa-Tallapoosa and Apalachicola-Chattahoochee-Flint River Basins.
Draft Report, ACT and ACF River Basin Comprehensive Study, August, 1996.
Williams, J. R., P. T. Dyke, W. W. Fuchs, V. W. Benson, D. W. Rice and E. D. Taylor.
EPIC: Erosion/Productivity Impact Calculator-2. User Manual. United States
Department o f Agriculture Technical Bulletin No. 1768, Washington, DC, 1990.
Winpenny, J. Measuring Water as an Economic Resource. Routledge: London, 1994.
Yaron, D. and A. Dinar. Optimal Allocation of Farm Irrigation Water During Peak
Seasons. Amer. J. Agr. Econ. 64(ApriI 1982):681-89.
Zavaleta, L. R., R. D. Lacewell, and C. R. Taylor. Economically Optimum Irrigation
Patterns for Grain Sorghum Production. Texas High Plains. Texas Water
Resources Institute, TR-100. College Station, Texas, 1979.
Zilberstein, S. Models o f Bounded Rationality: A Concept Paper. AAAI Fall
Symposium on Rational Agency, Cambridge, Massachusetts, November 1995.

R eproduced w ith perm ission o f the copyright owner. Further reproduction prohibited w ith o u t perm ission.

APPENDIX-1

Table 31 .Optimal Allocation o f Irrigation Under Water Scarcity (Supply = 4 acre-inch)


Maize

Stage

Cotton

Peanut

Total Irrigation

Dry

Normal

Dry

Normal

Dry

Normal

Dry

Normal

1.53

1.55

1.53

1.55

0.00

0.90

3.06

4.00

1.93

1.43

1.93

1.03

0.14

1.50

3.99

3.96

1.99

1.87

0.02

0.25

1.99

1.87

4.00

4.00

1.55

1.71

1.62

1.84

0.83

0.43

3.99

3.98

1.80

1.61

0.38

1.61

1.83

0.68

4.01

3.89

1.78

0.72

0.44

0.72

1.78

0.97

4.00

2.41

1.13

1.21

1.20

0.85

1.13

1.21

3.46

3.28

1.28

0.99

1.36

1.25

1.36

1.15

3.99

3.40

1.16

1.00

1.15

0.96

1.15

0.96

3.47

2.92

Table 32. Optimal Allocation o f Irrigation Under Water Scarcity (Supply=5 acre-inch)
Stage

Maize

Peanut

Cotton

Total Irrigation

Dry

Normal

Dry

Normal

Dry

Normal

Dry

Normal

1.53

1.55

1.53

1.55

0.00

1.55

3.06

4.65

1.93

1.43

1.93

1.43

1.14

1.43

4.99

4.29

1.99

1.87

1.02

1.23

1.99

1.87

5.00

4.98

1.55

1.71

1.62

1.78

1.55

1.33

4.71

4.81

1.80

1.61

1.41

1.61

1.80

1.34

5.01

4.56

1.78

0.72

1.44

0.72

1.78

0.82

5.00

2.27

1.13

1.21

1.24

1.21

1.13

1.21

3.49

3.64

1.36

1.15

1.36

1.15

1.36

1.15

4.07

3.45

1.15

0.96

1.15

0.96

1.15

0.96

3.46

2.89

107

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

108
Table 33. Optimal Allocation Under Excess Supply of Irrigation (Supply = 7 acre-inch)
Stage

Maize

Peanut

Cotton

Total Irrigation

Dry

Normal

Dry

Normal

Dry

Normal

Dry

Normal

1.53

1.55

1.53

1.55

0.00

1.55

3.06

4.65

1.93

1.43

1.93

1.43

1.99

1.43

5.85

4.29

1.99

1.87

1.99

1.87

1.99

1.87

5.97

5.62

1.55

1.71

1.55

1.71

1.55

1.71

4.64

5.12

1.80

1.61

1.80

1.61

1.80

1.61

5.40

4.83

1.78

0.72

1.78

0.72

1.78

0.72

5.34

2.16

1.13

1.21

1.13

1.21

1.13

1.21

3.38

3.64

1.36

1.15

1.36

1.15

1.36

1.15

4.07

3.45

1.15

0.96

1.15

0.96

1.15

0.96

3.46

2.89

Table 34. Optimal Allocation Under Excess Supply o f Irrigation(Supply = 8 acre-inch)


Maize
Stage

Cotton

Peanut

Total Irrigation

Dry

Normal

Dry

Normal

Dry

Normal

Dry

Normal

1.53

1.55

1.53

1.55

0.00

1.55

3.06

4.65

1.93

1.43

1.93

1.43

1.99

1.43

5.85

4.29

1.99

1.87

1.99

1.87

1.99

1.87

5.97

5.62

1.55

1.71

1.55

1.71

1.55

1.71

4.64

5.12

1.80

1.61

1.80

1.61

1.80

1.61

5.40

4.83

1.78

0.72

1.78

0.72

1.78

0.72

5.34

2.16

1.13

1.21

1.13

1.21

1.13

1.21

3.38

3.64

1.36

1.15

1.36

1.15

1.36

1.15

4.07

3.45

1.15

0.96

1.15

0.96

1.15

0.96

3.46

2.89

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

109
Table 35. Historic Stream Flow of Chattahoochee River at Columbus, Georgia (mgw)
Year

Weekl

Week2

Week3

Week4

Week5

Week6

Week7

Week8

Week9

1944
1945
1946
1947

28023.5
17125.5
43630.8
24728.9
31208 3
34632.1
17093.2
15885.1
48488.8
20859.3
16150.0
20407 1
11511.7
12280.5
13275.3
54167.1
21582.9
23501 5
16376.1

21046.7
16040.2
33999.0
16641.0
25491.2
26815.5

32429.2
12771.4
28669.5
223322
15084.1
272224
32965.4
15284.4
173322
16854.1
13566.0
13566.0
10264.9
12028.5
9864.4
23747.0
19476.9
20568.6
27041.6
136823
21356.8
50058.5
31259.9
29321.9
21931.7

25071.3
22325.8
28049.3
35930.5
17648.7
48165.8
16847.7
16027.3
18294.7
19341.2
16796.0
10587.9
14095.7
10943.2
19380.0
17577.7
16169.4
40381.5
25045.4
35484.8
18146.1
32571.3
24980.8
25646.2
22926.5
15768.9
19373.5
18598.3
35672.1
47623.1

16460.1
205622
23230.2
34147.6
18016.9
34250.9
12700.4
152327
17474.3
21363.2
13411.0
12403.2
101228
12803.7

14858.0
11279.2
27429.2
17564.7
149226
31944.7
10148.7
20846.4
9160.3
23223.7
8714.5
76228
10387 7
12991.1
13210.7
16931.7
13566.0
28566.1
22597.1
40284.6
24703.0
156720

16511.8
15239.1
28191.4
15898.1
138218.2
25555.8

19218.5
30607 5
26918 8
24599.7
70795 1
972876
267121
10523.3
9948.4
26886 5
11879.9

17693 9
220609
21188 8
17177 1

1945
1949
1950
1951
1952
1953
1954
1955
1956
1957
1955
1959
1960
1961
1962
1963
1964
1965
1966
1967
1965
1969
1970
1971
1972
1973
1974

34929.2
34754.8
16466.5
49276.9
24586.8
32183.7
30278.0
13520.8
17964.0
25710.8
91344.4
18669.4
25232.8
45271.7
14993.7
17836.1
29774.1
42642.5
33921.5
25284.4
30820.7
35407.3
7493.6

35807.8
14883.8
22681.1
15290.8
13372.2
12267.5
10032.4
15878.7
10646.1
731918
17287 0
18914.9
21634.5
18876.1
31247 0
21699.1
28766.4
39283.3
19257.3
20484.7
48301.4
13169.4
14870.9
52313.1
241927
31111.4
68540.6
15839.9
254524
36673.4

22086.7
22907.2
21079.0

1986
1989

10620.2
18927.8

1990
1991
1992
1993

26066.1
47481.0
22965.3
31247.0

327522
20168.1
28430.5

15633.2
54134.8
229524
44767.8
35116.6
15691.3
19961.4
27790.9
17713.3
10226.2
13553.1
244123
227521
24993.7
11640.9
22157 8
182624
21828.3
21298.6
25297.4
14089.3
25420.1

Mean

27112.6

247923

22288.9

1975
1976
1977
1975
1979
1950
1951
1982
1983
1984
1955
1986
1987

14341.2
14690.0

24967.9
15846.4
26737.9
25743.1
34044.2
14044.0
10704.2
18333.5
11356.7
167120
27771.5

177521
27855.5
20071.2
39457.7
16918.7
48798.8
17661.8
23314.1
16214.8
27455.0
19308.9
18191.4

17629.3
20555.7
11996.2
30620.4
14018.2
93114.4
187728
33417.6
17855.4
22203.0

7383.8
149420
37946.0
29070.0
7887.7
21363.2
281721
12196.5
14967.8
13255.9
44761.3
144123
217121
26440.8
12971.7
18889.0
11311.5
41163.1
12286.9
474229
21048.7
55556.0
20788.3
25478.2

239225

223321

16983.3
28940.8
31395.6
17726.2
15426.5
17435.5
19328.3
10775.3
17364.5
25097.1

27603.6
26163.0
18650.0
13656.4
10758.5
16763.7
25226.3
27946.0
11976.8
231268
26389.1
13966.5
14909.7
14470.4
26486.0
15691.3
20840.0
24425.3
12319.2
130557
10859.3
434823

16085.4
14877.4
11221.0
25006.7
80427
160725
15911.0
8378.6
34987.4
19619.0
20529.9
31524.8
28701.8
27209.5
21369 7
23378.7
15904.5
41344 0
24399.4
18449.8
14354.1
22584.2
18075.1
28029.9
13436.8
40549 4
49451.3
21796.0
15491.1
32170.8
19903.3
24896 8
25071.3
23236.6
15193.9

15943.3
15148.7
9935.5
32836 2
21356.8
15361.9
33475.7
13759.8
34489.9
24793.5
186823
16292.1
37791.0
231526
17939.4
15736.6
31104.9
20730.1
26692.7
9741 7
20122.9
27571.3
21621.6
20116.4
17054.4

11460.0
81014.7

18165.5
13650.0
29489.9
7887.4
36951.2

27635.9
29257.3
27894.3
23837.1

25219.8
29457 6
24173.3
243025

15478.2
20633.2
25626.8
23385.2
21763.7
21305.1
17829.6
15943.3
7170.6
50194 2
18746.9
40743.2
27455.0
18766.3

20325.8

24719.3

244374

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

27681 1
40788.4
21163 0
10077 6
80039
50000 4
10420 0
34496 4
9386 4
8701 6
28449.8
17771.5
15213.3
24354 2
116022
35678 6
37952.5
16892.9
151616
25194 0
21589 3
22965.3
41654 1
23540 2
16628 0
23507 9
13178 4
25820 6
23966.6
21660 4
22177 2
24612.6
18100 9
201294
28223 7
26008.0
30265.1
20181.0
11033.7
20329 6
8204.2
52119.3
15226 2
31918.9
27797 4
19954 9
22758 1

110
Table 36. Probability of Water Supply Being Less than Total M&I and Irrigation Demand
Year
2000

2010

2020

Min- Stage1 Stage2 Stage3 Stage4 Stage5 Stage6 Stage7 Stage8 Stage9
1150

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

1650

0.02

0.00

0.00

0.00

0.04

0.02

0.04

0.02

0.04

1150

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

1650

0.02

0.00

0.00

0.00

0.04

0.02

0.06

0.02

0.04

1150

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

1650

0.02

0.00

0.00

0.00

0.04

0.02

0.06

0.02

0.04

1150

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

2050

0.00
0.00
1650 0.02
0.00
0.04
0.02
0.06
0.02' 0.04
Note: Probabilities were calculated from the Table 20. Since these probabilities were
estimated using historical data for the period 1943 to 1993, some o f the entries are zero.

Table 37. Probability of Water Supply Being Less than Total M&I and Irrigation Demand
Year
2000

2010

2020

Min- Stage1 Stage2 Stage3 Stage4 Stage5 Stage6 Stage7 Stage8 Stage9
1150

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

1650

0.02

0.00

0.00

0.00

0.04

0.06

0.06

0.02

0.06

1150

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.02

0.00

1650

0.02

0.00

0.00

0.00

0.04

0.06

0.06

0.02

0.08

1150

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.02

0.00

1650

0.02

0.00

0.00

0.00

0.04

0.06

0.06

0.02

0.08

1150

0.02

0.00

0.00

0.00

0.02

0.02

0.02

0.02

0.00

2050

0.02
0.02
0.04
0.06
1650 0.02
0.00
0.06
0.06
0.08
Note: Probabilities were calculated from the Table 21. Since these probabilities were
estimated using historical data for the period 1943 to 1993, some of the entries are zero.

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

Ill
Table 38. Probability o f Water Supply Being Less than Total M&I and Irrigation Demand
Year
2000

Min- Stage1 Stage2 Stage3 Stage4 Stage5 Stage6 Stage7 Stage8 Stage9
1150

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

1650

0.02

0.00

0.00

0.00

0.04

0.06

0.06

0.02

0.06

1150

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

1650

0.02

0.00

0.00

0.00

0.04

0.06

0.06

0.02

0.08

1150

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

1650

0.02

0.00

0.00

0.00

0.04

0.06

0.06

0.02

0.08

1150

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

2010

2020

2050

0.04
0.00
0.00
0.06
0.00
1650 0.02
0.06
0.02- 0.08
Note: Probabilities were calculated from the Table 22. Since these probabilities were
estimated using historical data for the period 1943 to 1993, some of the entries are zero.

Table 39. Probability of Water Supply Being Less than Total M&I and Irrigation Demand
Year Min- Stagel Stage2 Stage3 Stage4 Stage5 Stage6 Stage7 Stage8 Stage9
2000

2010

2020

2050

1150

0.02

0.00

0.00

0.00

0.02

0.02

0.02

0.02

0.00

1650

0.02

0.02

0.06

0.00

0.06

0.08

0.06

0.08

0.10

1150

0.02

0.00

0.00

0.00

0.04

0.02

0.04

0.02

0.02

1650

0.02

0.02

0.06

0.02

0.06

0.10

0.06

0.10

0.12

1150

0.02

0.00

0.00

0.00

0.04

0.02

0.04

0.02

0.04

1650

0.04

0.06

0.04

0.02

0.06

0.10

0.06

0.10

0.12

1150

0.02

0.00

0.00

0.00

0.04

0.02

0.06

0.02

0.04

0.14
0.06
0.14
0.06
0.06
0.10
1650 0.04
0.06
0.06
Note: Probabilities were calculated from the Table 23. Since these probabilities were
estimated using historical data for the period 1943 to 1993, some of the entries are zero.

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

Table 40. Irrigation Demand o f Com in Dry Years by County and Growth Stage
Stages

Barbour

Russell

Quitman

Stewart

Total

752.76

472.77

35.19

1799.28

3060.00

948.58

595.75

44.34

2267.33

3856.00

979.08

614.91

45.77

2340.24

3980.00

760.63

477.71

35.56

1818.10

3092.00

885.60

556.20

41.40

2116.80

3600.00

875.76

550.02

40.94

2093.28

3560.00

553.99

347.93

25.90

1324.18

2252.00

666.66

418.70

31.17

1593.48

2710.00

567.77

356.59

26.54

1357.10

2308.00

6990.83
4390.58
326.81
16709.78
Total
28418.00
Note: Chattahoochee and Muscoggee counties were excluded because there was no
irrigated com grown in the year 1996.

Table 41. Irrigation Demand o f Com in Normal Years by County and Growth Stage
Stages

Barbour

Russell

Quitman

Stewart

Total

762.60

477.40

35.65

1822.80

3098.45

704.05

440.75

32.91

1682.86

2860.57

922.01

577.19

43.10

2203.82

3746.13

839.84

525.76

39.26

2007.43

3412.29

791.63

495.57

37.01

1892.18

3216.39

354.73

222.07

16.58

847.90

1441.28

597.29

373.91

27.92

1427.66

2426.79

565.80

354.20

26.45

1352.40

2298.85

474.29

296.91

22.17

1133.66

1927.04

3763.76
6012.24
281.06
14370.72
24427.78
Total
Note: Chattahoochee and Muscoggee counties were excluded because there was no
irrigated com grown in the year 1996.

R eproduced with perm ission o f the copyright owner. F urther reproduction prohibited w ith o u t perm ission.

Table 42. Irrigation Demand o f Cotton in Dry Years by County and Growth Stage
Stages

Barbour

Russell

Quitman

Stewart

Total

1415.25

2509.20

0.00

1049.58

4974.03

1783.40

3161.92

0.00

1322.61

6267.93

1840.75

3263.60

0.00

1365.14

6469.49

1430.05

2535.44

0.00

1060.56

5026.05

1665.00

2952.00

0.00

1234.80

5851.80

1646.50

2919.20

0.00

1221.08

5786.78

1041.55

1846.64

0.00

772.44

3660.63

1253.38

2222.20

0.00

929.53

4405.11

1067.45

1892.56

0.00

791.64

3751.65

9747.37
Total
23302.76
13143.33
0.00
46193.46
Note: Chattahoochee and Muscoggee counties were excluded because there was no
irrigated cotton grown in the year 1996. Entries for Quitman county are zero because
com was not grown in 1996.
Table 43. Irrigation Demand of Cotton in Normal Years by County and Growth Stage
Stages

Barbour

Russell

Quitman

Stewart

Total

1433.75

2542.00

0.00

1063.30

5039.05

1323.68

2346.84

0.00

981.67

4652.18

1733.45

3073.36

0.00

1285.56

6092.37

1578.98

2799.48

0.00

1171.00

5549.46

1488.33

2638.76

0.00

1103.77

5230.86

666.93

1182.44

0.00

494.61

2343.97

1122.95

1990.96

0.00

832.80

3946.71

1063.75

1886.00

0.00

788.90

3738.65

891.70

1580.96

0.00

661.30

3133.96

8382.92
39727.22
11303.50
20040.80
0.00
Total
Note: Chattahoochee and Muscoggee counties were excluded because there was no
irrigated cotton grown in the year 1996. Entries for Quitman county are zero because
com was not grown in 1996.

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

114
Table 44. Irrigation Demand of Peanuts in Dry Years by County and Growth Stage
Stages

Barbour

Russell

Quitman

Stewart

Total

0.00

0.00

0.00

0.00

0.00

1976.07

1802.94

139.30

3239.72

7158.03

1976.07

1802.94

139.30

3239.72

7158.03

1535.18

1400.68

108.22

2516.89

5560.96

1787.40

1630.80

126.00

2930.40

6474.60

1767.54

1612.68

124.60

2897.84

6402.66

1118.12

1020.16

78.82

1833.13

4050.22

1345.52

1227.63

94.85

2205.94

4873.94

1145.92

1045.52

80.78

1878.71

4150.94

11543.35
891.87
Total
20742.35
12651.81
45829.38
Note: Chattahoochee and Muscoggee counties were excluded because there was no
irrigated peanut grown in the year 1996.

Table 45. Irrigation Demand of Peanuts in Normal Years by County & Growth Stage
Stages

Barbour

Russell

Quitman

Stewart

Total

1539.15

1404.30

108.50

2523.40

5575.35

1420.98

1296.49

100.17

2329.67

5147.31

1860.88

1697.84

131.18

3050.87

6740.78

1695.05

1546.54

119.49

2779.00

6140.08

1597.74

1457.75

112.63

2619.45

5787.57

715.95

653.23

50.47

1173.79

2593.44

1205.50

1099.88

84.98

1976.39

4366.76

1141.95

1041.90

80.50

1872.20

4136.55

957.25

873.38

67.48

1569.39

3467.51

855.40
43955.34
11071.32
19894.16
12134.46
Total
Note: Chattahoochee and Muscoggee counties were excluded because there was no
irrigated peanut grown in the year 1996.
i

R eproduced w ith perm ission o f the copyright owner. F urther reproduction prohibited w itho ut perm ission.

You might also like