This document summarizes a study on improving food productivity in Sri Lanka's dry zone through conjunctive use of surface and groundwater. The study aimed to model the local groundwater system and analyze different operational policies for irrigation schemes. Key steps included selecting a study area, collecting field data, developing a mathematical model, calibrating the model, validating predictions, and analyzing scenarios like modified irrigation operations or boundary treatments. The calibrated model was able to predict future water levels with errors of -0.8% to 2.1%, allowing assessment of management options to optimize water use and agricultural productivity.

1. Operational Options of Water Resource to
Improve Food Productivity by
Conjunctive Use of Surface and
Groundwater in Dry Zone
Dr (Eng.) S.S.Sivakumar
Deputy Director of Irrigation
Mullaitivu
Seminar on “Operational options of water resource to improve food productivity by conjunctive use of surface water and groundwater
in dry zone” on 26th
February 2009 in the Northern provincial council conference hall Trincomalee Sri Lanka

2. “Growth of a Nation Depends on
Effective Economic and Equitable
Use of Water Resource”
2

3. 3
Problem in Water Resource
• Economically feasible surface water storage
sites are limited
• Unplanned utilization of water resource by
various stake holders
• Difficulty in analytical solution for groundwater
storage, due to non-homogeneous and
anisotropic nature of groundwater resource

4. 4
Objective
To improve the groundwater system in a restricted area
using modeling technique to spell out
 An economic policy in operating the minor and
medium Irrigation schemes
 A new technique of peripheral treatment by clay
or geotextile or a subsurface dam
To economize the cost of water for irrigation
and in turn improve the productivity

5. 5
Methodology
1. Study area selection
* Selection of observation wells.
* Polygonal network formulation.
* Polygonal parameter calculation.
2. Data collection
* Seasonal/Monthly field data
* Data from documents and publication
3. Model
* Formulation
* Calibration
* Validation
* Prediction
4. Analyzing predicted system response for various
* Operational policy of minor medium irrigation scheme
* Boundary treatment
5. Economic analysis

6. 6
Selection of Study Area
• Vavuniya District.
• 71.5 squire miles (185 sq.km)
• 41 Observation wells
• 6 Medium Irrigation schemes
• 40 Minor Irrigation schemes
• 3 Agrarian service centers
• 31 Grama Nilathary divisions

8. 8
Polygonal Network of the Study Area
The study are is divided into 41 Thissin
polygons based on the 41observation wells
Maximum polygonal area 8440 m2
(node 26)
Minimum polygonal area 1294 m2
(node 35)

9. 9

10. 10
Aquifer Characteristic
– Unconfined
– 10 to 15 m. thick
– Gravelly or decomposed material
– Bottom layer of this aquifer is a rarely fractured
crystalline rock having vertical transmissibility less
than one sq. meter per day.
– Darcy's law (Linear resistance to laminar flow) and
Dupuit's assumption (vertical flow can be
neglected) are applicable
– Two-dimensional flow system

11. 11
Hydraulic Assumptions
 The aquifer is treated as a two-dimensional flow
system
 Only one aquifer system is modelled with one
storage coefficient in vertical direction
 The aquifer is bounded at the bottom by an
impermeable layer(aqutard)
 The upper boundary of the aquifer is an
impermeable layer (confined aquifer) or a slightly
permeable layer (semi confined aquifer) or the free
water table (unconfined aquifer)

12. 12
Hydraulic Assumptions ctd.
 Darcy’s law (Linear resistance to laminar flow) and
Dupuit’s assumption (vertical flow can be
neglected) are applicable for the aquifer under
study.
 The processes of the infiltration and percolation of
rain and surface water and of capillary rise and
evapotranspiration, taking place in the unsaturated
zone(vadozone) of the aquifer (above the water
table) need not be simulated.
“This means the net recharge to the aquifer is calculated
manually and prescribed to the model”

13. 13
Operational Assumptions
 Same groundwater elevation within a polygonal
area.
 The area where the minor & medium irrigation
schemes are governing the water table, then
one meter below FSL of the tank can be taken as
the water table elevation
 The rain fall of Vavuniya can be used for entire
area under study as all the polygons are around
Vavuniya rainfall station and within 15 km radius.
 The irrigation efficiency 70 %

14. 14
Operational Assumptions ctd.
 Conveyance efficiency of the canal 80%
 As the irrigation canals within this study area are very
small and for simplicity recharge from canal and
irrigation field can be combined for calculation
 All the 6 medium and 40 minor irrigation schemes within
the study area have their maximum head of water less
than 3m. Hence the percolation of water is calculated
either as 0.005m/day/planearea or 0.5% of the volume
stored monthly
 For keeping 10% of the full capacity of the minor
medium irrigation scheme, 12% of the cultivation to be
forgone.

15. 15
Economic Assumptions
 Benefit cost ratio based on present value should exceed
unity
 Percaptia domestic water consumption - 160 litres/day
 Water requirement for OFC
Maha 1.5 to 2.0 ac.ft./ac.
Yala 2.0 to 2.5 ac.ft./ac.
 Net return from one acre of paddy cultivation - Rs. 3875/=
 One meter raise in water table will save 1.4 unit of
electricity for the pumping of 10 m3
of water
 One mile of peripheral treatment will cost Rs. 5.32million

16. 16
Data collection
 Field data
 Seasonal water levels collected from September 1997 to
May 2004
 Monthly water levels collected from April 2001 to May 2003
 Data from yearly publication
 Rain fall
 Population
 Paddy/OFC Cultivation
 Water stored in Irrigation schemes
 Pumping from production wells

17. 17
Processing Data
• Connecting water levels to MSL
• Converting data obtained from publications in to polygonal
seasonal data such as
• Capacity of water store in Irrigation schemes
• Water issued for cultivation from Irrigation schemes
• Rain fall volume
• Pumping from domestic wells
• Pumping from agro. wells
• Pumping from production wells
Note:-Discharging period 1st
June to 31st
Sept.- 122 days
Recharging period 1st
Oct. to 31st
May - 224 days.

18. 18
Modeling Technique
 Conceptually the modeling technique used for
system representation can be very simply
explained as below.
 Select or formulate a suitable model
 Assume the parameters approximately
 Adopt some error function to quantify the difference
between measured and predicted responses
 Minimize the error function
 Determine the parameters accurately
 Predict system response

19. 19
Modeling Technique
Actual System
Response
Model predicted
System Response
Mathematical
Model
Modeled Input
Non - Modeled
Input
Real Physical
System
Solution Strategy
(Optimization)
Schematic representation of the process of system modeling and optimization.

20. 20
Model Formulation
Observation well
Typical polygon for node B
hi - piezometric head of node i
hB - peizometric head at node B
YiB = (JiB/LiB) - conductance factor
TiB - transmissibility at mid point between node B and i
JiB - length of perpendicular bisector associated with node B and i.
LiB - distance between nodes i and B
AB - polygonal area of node B
SB - storage coefficient of node B
QB - volumetric flow rate per unit area at node B.
M - No of observation wells surrounding node B
∆t - time step between j and j+1

21. 21
Model Formulation ctd.
Integrated finite difference form of water balance for
a typical polygon can be written as
1111
1
)()( ++
∆
++
=
+−=−∑ j
BBB
j
B
j
Bt
S
iBiB
j
B
j
i
M
i
QAAhhTYhh B
11 ++
== j
BB
j
B QAAQowVerticalfl
iBiB
j
B
j
i
M
i
j
B TYhhQQflowSubsurface )( 11
1
1 ++
=
+
−== ∑
)(. 11 j
B
j
B
BBj
hh
t
SA
BSTORorageChangeinst −
∆
== ++
11111 +++++
−++== j
dB
j
irB
j
ifB
j
isB
j
BB QdQcQbQaQAowVerticalfl BBBB

22. 22
Model Formulation ctd.
Subsurface flow + Vertical flow = Change in storage
1111
1
1
)()( ++
∆
++
=
+
+−−−= ∑ j
BBB
j
B
j
Bt
S
iBiB
j
B
j
i
M
i
j
B QAAhhTYhhRES B
The first step of the model running is to find the following
1. QQB
j+1
/TiB
2 STORE.Bj+1
/SB
Where
1+
= j
BQQflowSubsurface
1
. +
= j
BSTORorageChangeinst

23. 23
Model Formulation in Spreadsheet
 Data entry
Seasonal Identification on one cell (D7)
Nodal connectivity
A 41 x 41 square symmetric matrix B11 to AP 51
Seasonal water level
Water level in MSL B61 to K101
Conductance factor (J/L)
Manually calculated and fed into a 41x 41 symmetric
matrix ( B 211 to AP 251)

24. 24
Model Formulation in Spreadsheet ctd.
 Water level matrix for the seasons
 41 x 41 matrix ( B111 to PP151) will give water levels
of the seasons
 Head difference matrix
 Will give the head difference for the particular
connection
 Denoted by row number and column number (B141 to
AP201)
Intermediate calculations

25. 25
Model Formulation in Spreadsheet ctd.
 Matrix of lateral flow divided by Transmissibility
41 x 41 square matrix B261 to AP301
Change in storage divided by Storage coefficient
Only a column 311 to B351
Final calculation

26. 26
Calibration of Model Using Optimization
Where, M - No of observation wells surrounding node B
N - No of seasons calibration is to be done
Calibration is done from October 1997 to September 2001 (eight seasons)
Subject to 0.001 < SB <0.01
15 < TiB <25
0.075 <a <0.15
0.05 <b <0.1
0.05 <c <0.1
0.9 <d <1.1

27. 27
Boundary sub surface flow component will be
replaced by another unknown parameter
"q"times the difference in head of water. In the
boundary node minimisation model, this "q"
also will come as an additional constraint.
Constrain for the boundary nodes of
study area
The range of this will be given by analysing the
topography of the area surrounding the boundary
and the seasonal water level contour.

28. 28
Four types of boundaries
1.Zero flow 2,3
2.Head controlled 4,5,6
3.Flow controlled 1
4.Free surface 7

29. 29
 123variables for polygonal recharge coefficients
 Rainfall
 Tank storage
 Irrigation
 41 variables for withdrawal
 100 variables for transmissibility
 41 variables for specific yield
 17 additional variables for boundary lateral flow
Constrains for the Entire Study Area
Found by error minimisation using
“MATCAD 2000”.
.

30. 30
Prediction Process
 M - No of observation wells surrounding node B
 hi - piezometric head of node i
 hB - peizometric head at node B
 YiB = (JiB/LiB) - conductance factor
 TiB - transmissibility at mid point between node B and i
 JiB - length of perpendicular bisector associated with node B and i.
 LiB - distance between nodes i and B
 AB - polygonal area of node B
 SB - storage coefficient of node B
 QB - volumetric flow rate per unit area at node B.
 j -time
[ ] BBiBiB
j
B
j
i
M
i
j
BB
j
B
j
B AStTYhhQAhh /.)(. 11
1
11
∆−++= ++
=
++
∑

31. 31
Prediction Model in Spreadsheet
For prediction the water balance equation has been re arranged to have hBj+1 in
LHS with RHS as function of hBj+1 . as below.
 Gauss-Seidal iteration method, hBj+1 could be found
utilizing the function of GOALSEEK on spreadsheet.
 As hBj+1 is connected to M surrounding nodes, while
finding the hBj+1 the already found hBj+1 will slightly vary
 After completing first iteration process for all the forty
one nodes, the same process to be repeated five to six
times to get accurate results.
[ ] BBiBiB
j
B
j
i
M
i
j
BB
j
B
j
B AStTYhhQAhh /.)(. 11
1
11
∆−++= ++
=
++
∑

32. 32
Prediction Model in Spreadsheet ctd.
Main matrixes formulated in spreadsheet
 Time step Δt
 Nodal connectivity matrix
 Water level matrix for the selected time step
 Head different matrix
 Conductance factor matrix
 Transmissible matrix
 Calculation of trial hBj+1
 Error in water level
 The function of GOALSEEK in spreadsheet with a small micro has been
used in this prediction to do all the iteration on few key strokes to predict
the water level with zero error.
 Even though the model is formulated season wise prediction is possible
monthly or even weekly by changing few cells formula and adding few
more cell formula.

33. 33
Model Validation
 To test the validity of the model, using the calibrated parameters
and using 8th
season (Sept. 2001) water level as initial water level
and for the rest of the inputs the 9th
season (May 2002) water level
was predicted using the prediction model
 Where ever the predicted values were not matching with the
observed value the stress parameters were systematically slightly
adjusted to get good response
 The same way, using 9th
season (May 2002) water level as initial
water level and for the rest of the inputs the 10th
season (Sept.
2002) water level was predicted using the prediction model
 The same way the water levels of May 2003, Sept. 2003 May 2004
and Sept. 2004 were predicted and compared with observed water
levels.

34. 34
Results of Validation
Boonstra and Ridder suggested
a refinement by introducing a
relax coefficient for all nodes.
Comparing the predicted results with the data used
for calibration an error of –0.8% to +2.1% is
observed. For a groundwater simulation model in
integrated finite difference method an error of this
magnitude is allowable depending on the scope of
the project

35. 35
Refining Results
While calibrating the actual model all hydro geological
stress parameters TiB , SB and Recharge coefficients
obtained from optimisation have to be systematically
slightly adjusted to get less error in water levels
111
Re*Re)(Pr)( +++
+= j
BB
j
B
j
B slaxedictedhModifiedh
1111
Re ++++
−+= j
B
j
B
j
B
j
B orageChangeinstowLatteralflQs
tSATY
lax
BBiBiB
B
∆+
=
∑ /.
1Re

36. 36
Peripheral Treatment Area Boundary

37. 37
Behavior of Aquifer with Decrease in
Transmissibility
 The T values between seventeen very extreme peripheral
nodes (18, 30, 41, 29, 28, 40, 39, 38, 37, 36, 24, 35, 22, 34, 33,
32, 31) and fourteen interior adjoining nodes (8, 17, 16, 15, 27,
26, 25, 13, 23, 12, 11, 21, 20, 19) were reduced in steps of 2 –
3 m2
/day and the water levels were predicted.
 T values were reduced in five steps
 It was found from the predicted water levels, there was not
much significant variation in water level change after 40 – 55 %
reduction in T.
 Even below this range also few nodes gained on the expense
of others in the close vicinity
 The maximum gain or lose occurred for the reduction of T by
39.8 % on average (i.e. step 4).
 This effect occurred only in recharge season (Oct. – May).

38. 38
Behavior of Aquifer with Decrease in
Transmissibility ctd.
Effect on Outer Boundary Nodes
Nodes 18, 41, 29, 35, 34, 33, 32 and 31 lost
their water level by 1.25 – 2.0 ft. in recharge
season (Oct – May). The nodes 28, 39, 36
and 24 gained their water level by 1.0 – 2.25
ft. in recharge season Oct – May, whereas
nodes 30, 40, 38, 37 and 22 showed only
very negligible variation.

39. 39
Behavior of Aquifer with Decrease in
Transmissibility ctd.
Effect on Nodes within Treated Boundary
The gainers of this boundary treatment are
mostly the nodes within the boundary. The
nodes 8, 19, 17, 16, 27, 26, 25, 23, 11, 21,
20 and 2 gained their water level by 1.5 –
2.75 ft. in season Oct – May, whereas nodes
13 and 14 lost their water level by 0.75 –
1.25 ft. in recharge season (Oct – May) and
other nodes did not show any significant
impact due to this boundary treatment

40. 40
Behavior of Aquifer with Various
Operational Policies of Irrigation Schemes
 The behavior of water table of this catchment was analyzed by keeping
10, 20, 30, 40 and 50 % of the full capacity of the irrigation scheme
during season June – Sept.
 The water table in almost all (except node 37 and 38) nodes, mainly
during discharging season (June – Sept.) considerably gained.
 During step 2 and step 3 the gain of water table was between 2.5 and
3.75 ft., scattered among study area during season June – Sept. and
1.75 to 3 ft., during season Oct – May.
 But this trend was changed in step 4 and step 5. In step 4 and step 5
the nodes 6, 7, 9, 17, 20, 21, 13, 25, 27, 39, 31and 19 showed only
little incremental change of 0.25 to 0.75ft. in water level.
 This positive progressive change observed during discharging season
(June – Sept) with change of operational policy of irrigation schemes
within catchment under study.

41. 41
Behavior of Aquifer with Various
Operational Policies of Irrigation Schemes
with Boundary Treatment
 During this analysis every steps first option was carried out
for all five steps of second option. By this method altogether
twenty five trials were carried out. Even though this was very
tiresome and very cumbersome exercise very interesting
shift was observed in the research finding.
 The result of option one was shifted positively almost one
step further. The step 4 of option one and two resulted an
average gain of water table during discharging season (June
– Sept) by 3.0 to 4.75 ft excluding nodes 37 and 38.
 This is almost a gain of water level of 60% to 70% of water
table loss in between two seasons in 95 % of the catchment
under study.

42. 42
Summary of Operational Research
The above operational research summarizing the following
out puts as policy alternatives as integrated conjunctive
operation and management of minor / medium irrigation
schemes in any restricted catchment.
 Peripheral boundary treatment up to the reduction of permeability by 35%
to 45% is giving water table raise of nodes closer to treated boundary by
1.5 ft. to 2.75 ft. during recharging season (Oct – May).
 Changing the operational policy of minor / medium irrigation schemes by
forgoing cultivation by 25% to 35% is giving water table gain in almost all
nodes except nodes 37 and 38 by 1.75 ft to 3.0 ft during discharging
season (June – Sept).
 Combining peripheral reduction in permeability by 35% to 45% and
forgoing cultivation of minor / medium irrigation scheme by 45% to 55%
result an average gain of water table during discharging season (June –
Sept) by 3.0 to 4.75 ft excluding node37 and 38.
 This is almost a water level gain of 60% to 70% of water table loss in
between two seasons in 95 % of the catchment under study.

43. 43
Options
Steps of each
option
Discharging
period
Recharging period
Nodes on which raise in
water level occurred
Operational policy
change
Forgoing 24% -
36% cultivation
0.762
to
1.143
0.533
to
0.914
Scattered except 37 and 38
Forgoing 48% -
60% cultivation
0.838
to
1.371
0.610
to
1.067
Scattered except 37 and 38
Boundary
treatment
Reduction of
permeability by
40%
-0.381 to
-0.610.
18, 41, 29,35,34,33,32 and
31
0.305 to 0.686
28, 39,36 and 24
-0.229 to
-0.457 13 and 14
0.457 to 0.838 8,19,17,16,27,26,25,23,11,21,
20 and 2
Combination of
policy change and
creation of artificial
boundary
Step 4 of both the
season
0.914 to 1.448
95 % of the nodes within the
catchment
Summary of maximum water level change in m for various options

44. 44
Economic Performance Indices
 There are mainly the following four methods
or indices that are considered conceptually
correct for comparing alternatives.
 The present worth method
 The rate of return method
 The benefit-cost ratio method
 The annual cost method.

45. 45
Summary of Economic Analysis of the Ope
Detail cost benefit analysis for all the three findings
of the operational research reveled
 The boundary treatment led almost 90% of the trial, the benefit cost
ratio became less than one. Only if the year of implementation
exceeded 20 years and interest rate 10% it showed positive results and
that too was for the last two trials
 The alternate policy on changing the operational policy of minor /
medium irrigation schemes by forgoing cultivation by 24% and 36%
gave the benefit cost ratio became greater than unity occurred almost
80% of the observation wells and the gain in water level was around
45% to 65% of the loss in water table between two consecutive
seasons.
 The combination of the above two alternatives did not yield much
economic results due to high cost of boundary treatment. Even though
for the last two steps it showed positive results those were also for the
implementation period of 20 years and interest rate of 10% and
permeability reduction by 47% to 50%.

46. 46
Option
Steps for each
season
Benefit cost ratio
Operational policy
change
Discharging season Recharging season
2 14.52 1.59
3 14.63 1.46
4 12.43 1.33
5 10.27 1.13
Boundary treatment
Year of
implementation
20 25
Interest rate 7.5% 10% 7.5% 10%
3 0.73 0.97 1.15 1.66
4 0.88 1.17 1.39 2.01
5 0.83 1.10 1.30 1.88
Combination of
policy change and
creation of artificial
boundary
Year of
implementation
20 25 20 25
Interest rate 7.5% 10% 7.5% 10% 7.5% 10% 7.5% 10%
3 0.97 1.13 1.28 1.78 0.82 1.09 1.17 1.75
4 1.09 1.19 1.49 2.23 1.01 1.13 1.44 2.18
5 1.04 1.13 1.42 2.22 0.97 1.15 1.37 2.02
Summary of benefit/cost ratio greater than unity option and steps

47. 47
Summery of the Research Finding
The best economically feasible option for implementation in any restricted area
“The change in operational policy of minor /
medium irrigation schemes by forgoing one
third of the cultivation under minor / medium
irrigation schemes or keeping one fourth of the
storage of minor / medium irrigation schemes
at any time will recover on an average of 45% to
65% of the loss of water table in any
consecutive seasons in almost 80% to 90% of
the area under consideration”

48. 48
Implementation of the Research Finding
 No time bound
 No area specific
 Additional financial resource not require to implement
But proper knowledge based awareness is
required to implement this policy in field among
the stake holders.
Even now the practice of alternate track cultivation
in different years depending on availability of water
in the irrigation shames are practiced

49. 49
Conclusion
“This research finding shows that 25% of the
storage of minor / medium irrigation scheme can
be kept at any time balance by forgoing one third
of the cultivation under the minor / medium
irrigation scheme for recharging the groundwater
and make the water table to gain on an average
of 45% to 65% of the loss of water table in any
consecutive seasons in almost 80% to 90% of the
area under consideration for the economic
utilization for domestic as well as agriculture
use”

50. Recommendation
“Construction of new or reconstruction of
abandoned minor /medium irrigation scheme
with 25% of storage exclusively for recharging
groundwater and changing the operational policy
to keep 25% of the present storage of existing
minor /medium irrigation scheme to recharging
groundwater will reduce considerably the
average cost of irrigation water due to less
energy cost and this in turn will increase the
extent of cultivation per unit of irrigation water
and led to increase in productivity”
50

51. 51
Area of Future Study
“Further study on policy alternative
towards conjunctive use water
management policy of major irrigation
schemes will be very useful in macro
development of water resource in
developing country like Sri Lanka”

52. 52
Advantages of Conjunctive use of Surface
and Groundwater
 Greater water conservation
Operation of both surface and groundwater reservoirs provides for
large water storage
 Smaller surface storage
Groundwater storage can provide for water requirements during a
series of dry years
 Smaller surface distribution system
Greater utilisation of groundwater from widely distributed wells
 Smaller drainage system
Pumping from wells aids in controlling the water table
 Reduced canal lining
Seepage from canals is an asset because it provides artificial
recharge to groundwater

53. 53
Advantages of Conjunctive use of Surface
and Groundwater ctd.
 Smaller evapotranspiration losses
Greater underground water storage with lowered groundwater level is
reduces losses.
 Greater control over outflow
Surface waste and subsurface outflow are reduced by conjunctive use,
thereby providing greater water conservation.
 Less danger from failure
The smaller the dam reservoir storage, the smaller the damage or risk.
 Reduction in weed seed distribution
With a smaller surface distribution system there is less opportunity for
spread of noxious weed seeds.
 Better timing of water distribution
An irrigator prefers to have water available when he wants it, as from a
pump, than to take water on schedule from surface conduits.

54. Thank U
“Growth of a Nation Depends on
Effective Economic and Equitable
Use of Water Resource”
54