Vous êtes sur la page 1sur 14

Received: 8 August 2019 Revised: 15 April 2020 Accepted: 16 April 2020

DOI: 10.1002/ird.2474

RESEARCH ARTICLE

Analysis of seepage loss from concrete lined irrigation


canals in Punjab, Pakistan†

Zulqarnain Shah | Hamza Gabriel | Sajjad Haider | Turab Jafri

National University of Sciences and


Technology - NUST Institute of Civil
ABSTRACT/RÉSUMÉ (ABSTRACT for English and RÉSUMÉ for French)
engineering, Islamabad, Pakistan A significant amount of water is lost in canals due to seepage. This could be
saved to irrigate barren land. This article presents the measured seepage data
Correspondence
Mr. Zulqarnain Shah. National University by the inflow-outflow method on seven so-called concrete-lined distributary
of Sciences and Technology - NUST canals in Punjab, Pakistan. A special feature of the inflow-outflow method is
Institute of Civil engineering, sector H 12.
the relatively long lengths of the canal reaches (9 to 5 km), which is the longest
Islamabad 44000, Pakistan.
Email: zulqarnain@nice.nust.edu.pk one could find in the literature so far. The seepage in canals is also calculated
using five empirical formulae. The results show that the Swamee and the
Punmia formulae performed admirably with mean percent errors of −20%
and + 19% respectively. Moritz formula stood at +53% while Kostiakov scored
+158%. The research also proposed a slight modification in the Swamee for-
mula whereby, instead of the canal width, the surface water width was used in
the calculation of the dimensionless variable F. This led to a beneficial effect of
bringing down the percent mean error from −42% to −20%.

KEYWORDS
empirical formulae for seepage estimation, inflow-outflow method, lined distributary canals,
seepage estimation, seepage loss

RĖSUMĖ
Une quantité importante d'eau est perdue dans les canaux en raison de l'infil-
tration. Cela pourrait être économisé pour irriguer les terres stériles. Cet article
présente les données d'infiltration mesurées par la méthode entrées/sorties sur
sept canaux de distribution dits en béton au Punjab, au Pakistan. Une par-
ticularité de la méthode entrées/sorties est la longueur relativement longue des
tronçons du canal (9 à 5 km), qui est la plus longue que l'on puisse trouver
jusqu'à présent dans la littérature. L'infiltration dans les canaux est également
calculée à l'aide de cinq formules empiriques. Les résultats montrent que les
formules de Swamee et de Punmia se sont admirablement comportées avec
des pourcentages d'erreur moyens de −20% et + 19% respectivement. La
formule de Moritz s'établit à + 53% tandis que Kostiakov marque +158%. La
recherche a également proposé une légère modification de la formule de
Swamee selon laquelle, au lieu de la largeur du canal, la largeur de l'eau de
surface a été utilisée dans le calcul de la variable sans dimension F. Cela a

Irrig. and Drain.. 2020;1–14. wileyonlinelibrary.com/journal/ird © 2020 John Wiley & Sons, Ltd. 1
2 SHAH ET AL.

conduit à un effet bénéfique de la réduction du pourcentage d'erreur moyenne


de - 42% à −20%.

MOTS CLÉS
perte par infiltration, canaux de distribution bordés, estimation de l'infiltration, méthode
entrées/sorties, formules empiriques pour l'estimation des infiltrations

1 | INTRODUCTION table, velocity of the flowing water, shear stress (force of


moving water on bed), slope between surface of water in
Pakistan is an agriculture-based country with a vast net- canal and beneath groundwater surface of an aquifer,
work of irrigation canals, inter-basin transfer canals, bar- constrains on groundwater flow e.g. drains, rivers and
rages, and two large dams. Agriculture is the backbone of streams (International Commission on Irrigation and
the economy. Being mostly, an arid and semi-arid land, Drainage (ICID), 1968; Swamee et al., 2000; Alam &
the need for irrigating the crops through man-made Bhutta, 2004; Martin & Gates, 2014). Less significant are
structures has been largely felt. The British colonialists in viscosity of canal water, canal water salinity, sediment
the 19th and early 20th century invested heavily in setting load and distribution of size, canal age and plants
up the irrigation network, chiefly, in the upper and lower (Swamee et al., 2000; Alam & Bhutta, 2004).
Indus river basin encompassing the present-day, Pakistan Permeability is governed by the physical properties of
provinces of Punjab and Sindh. Each year, on average, soil. The soil classification (based on grain size) influ-
about 130 BCM (109 m3) of river water is diverted into ences the pore size and the percentage of pore space and
the irrigation canals at its head from the barrages the following forces that act upon the water molecules
(Latif, 2007). A large part of this water is lost during its namely adhesion, cohesion and gravity (Kraatz, 1977;
travel through the canal system, mainly, due to seepage Eisenhauer & Hoffman, 2008). The larger the voids
which also given rise to the twin menace of salinity and between soil particles, the greater the amount of water
waterlogging (Soothar et al., 2015). Seepage losses in will be lost to gravity (Eisenhauer & Hoffman, 2008). A
Pakistan have been estimated to be about 45.5 and 66%, mixture of clay and gravel act as an impervious layer
for the lined and unlined watercourses respectively while Coarser textured soils (e.g. sand) will have higher
(Arshad et al., 2009; William Young et al., 2019). infiltration and permeability rates (Kraatz, 1977). How-
Due to a rise in the groundwater table and saliniza- ever, soil types of canals are never completely uniform
tion more than one-third of the world's irrigated land has and will change with distance so will permeability rates
been affected. In India, alone, 5.5 Mha (million hectares) (Leigh, 2014).
of land is waterlogged which is affected by salinization The depth of water in a canal can be regarded as a
problems (Singh, 2019). Waterlogging and salinity measure of energy placed on the canal's wetted perimeter
affected 4.5 Mha of irrigated land and reduced agricul- (Eisenhauer & Hoffman, 2008). Pressure and gravity are
tural production by 25% in Pakistan. What is more the driving forces that move water. If these forces are
alarming, is the fact that, across the Indus Basin, approxi- greater than the adhesion and cohesion forces in the soil
mately, one-third of seepage from canals goes into profile, then the water will move to the lowest energy
recharging the saline aquifers, in other words, perma- level. Therefore, as the water depth increases in the
nently lost unless huge money is spent to reclaim it. Last canal, seepage losses and infiltration rates will increase
but not the least canal seepage is objectionable, as even (ICID, 1968).
in areas having sweet water aquifer pumping adds an Seepage losses increase as the difference between the
extra cost (William Young et al., 2019). water level in the canal and groundwater table increases
The negative effects from canal seepage have been and losses reach the upper limit when difference is
variously described as: freshwater depletion, contamina- 5 times or greater than the surface width of the canal.
tion of groundwater, causing waterlogging and salinity, The losses from channel sides are greater than that from
rise in groundwater table leading to waste of useful land, the bed for a shallow groundwater table while in the case
reduction in irrigation efficiency and increase in opera- of deep groundwater table, those from the bed predomi-
tional cost (Swamee et al., 2000; Cakmak et al., 2004). nate over that from the sides (Kraatz, 1977).
Canal seepage primarily depends upon the soil per- The cost incurred on the lining is economically justi-
meability, water depth in the canal, length of wetted fied if it is equal to about 5% of the value of the conveyed
perimeter, channel geometry, location of groundwater water (Kraatz, 1977). The canal lining should only be
SHAH ET AL. 3

introduced in areas where conveyance efficiency is low. Skoperboe et al. (1999) measured flow losses in sev-
The soils, commonly found, in the plains of Punjab, are eral irrigation canals in southern Pakistan, using the
sandy loam and clay loam, having conveyance efficiency Inflow–Outflow and Ponding method in the pre and post
of 67.3 and 95.8% respectively, thus, making lining a lining phases. They concluded that the ponding test is
costly venture for the country (Sepaskhah & suitable when the conveyance losses are very small. Fur-
Salemi, 2004). ther, the Inflow-Outflow method gives more accurate
Seepage can be reduced up to 30 to 40% with lining results than the Ponding method, provided the discharge
but seepage cannot be controlled completely (Wachyan & can be accurately measured.
Rushton, 1987; Swamee et al., 2000; Sepaskhah & Alam & Bhutta (2004) concluded that the ponding
Salemi, 2004). The presence of cracks in the lining greatly test produced more accurate results than the inflow–
reduces its effectiveness (Plusquellec, 2019). Seepage can outflow method test but it was unfeasible during normal
be reduced by reducing the wetted perimeter and increas- canal operation.
ing flow velocity (Soothar et al., 2015). Optimization Salemi & Sepaskhah (2006) reported that the empiri-
models based on a genetic algorithm reduces water trans- cal formulae under-estimated the seepage losses com-
portation losses in canals. This is carried out by providing pared to the Inflow-Outflow method. The results
a reasonable arrangement of canal water allocation times suggested that Ingham and Moritz's equations were the
and discharges (Yao et al., 2019). most suitable for the measurement of seepage losses in
For all types of soil, in situ compaction with a vibra- different types of soils.
tory roller reduced seepage. The reduction in seepage was Akkuzu et al. (2007) estimated the seepage losses in
significant from 86 to 90% when both the sides and bot- canals using the Inflow –outflow method and observed
tom of canals were compacted (Burt et al., 2009). that the seepage rate depended on the canal shape.
The complicating factors in the canal seepage mea- Arshad et al. (2009) measured the conveyance losses
surement are the non-uniform soil properties, water qual- in canals through the Inflow-Outflow method. According
ity, sedimentation, fluctuation of the groundwater table, to them, the conveyance losses for the lined and unlined
the water level in a canal and periodic filling and drying watercourses stood at 45.5 and 66% of the inflow, respec-
up of canal (Akkuzu, 2011). tively. The water courses were poorly maintained with a
Seepage is measured in the field, through the Inflow- lot of silt.
Outflow, Ponding, Point measurement, and Permeameter Akkuzu (2011) used the Moritz and Davis-Wilson
measurement method. Also, Double ring infiltration test, equations for estimating the seepage rate in concrete-
ADCP (Acoustic Doppler Current Profiler) and Electrical lined canals and compared the results with the Inflow-
resistivity tests are also carried out for seepage estimation Outflow method. The two equations underestimated the
(Zhang et al., 2016). seepage rate hence these were not appropriate for esti-
The Inflow-Outflow is advantageous as it can be car- mating the seepage rate for canals lined with concrete in
ried out during the normal canal operation. However, which suitable repair and maintenance could not be
this method requires the prior establishment of a steady performed.
flow condition which is not easy for long reaches. Addi- Pognant et al. (2013) used the Electromagnetic Induc-
tionally, it also requires a long test reaches for accurate tion (EM) based system to verify the occurrence of seep-
measurement which means smaller areas of high and low age in canals. Results indicated that EM was successful in
seepage can be missed. A further pitfall is that seepage detecting the seepage losses.
quantity is very small compared to the magnitude of Sultan et al. (2014) compared the seepage losses,
inflow and outflow, hence, a small error in flow measure- before and after the introduction of lining in tertiary
ment translates into a large error in seepage canals in Pakistan. His results showed that the lining of
(Warnick, 1951; Alam & Bhutta, 2004). tertiary canals helped increase the conveyance efficiency
by 22.5%.
Zhang et al. (2016) used different techniques to esti-
2 | PREVIOUS R ESEARCH I N mate the seepage losses, i.e., measurement in the field,
CANAL SEEPAGE ESTIMATION empirical formulae and numerical simulations. The seep-
age losses calculated by the field methods and empirical
Bakry & Awad (1997) developed four equations to esti- formulae tallied to a large extent. However, the results of
mate the seepage in earthen canals using the Inflow- the Kostiakov formula differed, significantly, from the
Outflow method. The equations are only valid for earthen field methods.
canals having discharge ranging from 2 m3s−1 to Luo et al. (2016) estimated time-dependent canal
20 m3s−1. seepage using stage-seepage relationship. The results
4 SHAH ET AL.

showed that the estimated average seepage rates were empirical formulas, i.e. Swamee, Moritz, Davis-Wilson,
consistent with the results from ADCP (Acoustic Doppler Punmia, and Kostiakov.
Current Profiler) tests. The utility of the present research is of immense
Andrade et al. (2017) used tracer as a tool to importance at the planning and design stage. If the plan-
investigate canal seepage. It helps identify the maximum ners know beforehand the formulae, most effective in
seepage zones and probable seepage path, qualitatively estimating seepage from the canal(s) in a certain
and quantitatively, without effecting the canal operation. geopgraphical region, then this is a great help in design-
Tracer technique is a definite, reliable and ing efficient irrigation system.
economical tool.
This article presents the measured seepage data by
the inflow-outflow method on seven so-called concrete-
lined distributary canals in Punjab, Pakistan. A 2.1 | Hierarchy of canals in the irrigation
novelty of the research is the long lengths of the canal system of Punjab, Pakistan
reaches (9 to 5 km) for the inflow-outflow method
which is the longest one could find in the literature so The schematic diagram of the hierarchy of canals is given
far. The seepage in canals is also calculated using five in Figure 1.

F I G U R E 1 Hierarchy of
canals in the Punjab irrigation
network
SHAH ET AL. 5

3 | MATERIAL A ND METHOD was measured using the Price current meter


(Buchanan & Somers, 2016). The flow rate was measured
3.1 | Study area at the inflow and all outflow sections by the velocity-area
method. The canal width was measured using cable
The study area comprised of distributary canals in two while depth was measured by wading rod. Length of
different geographical areas of the Punjab province, reach was measured using a measuring chain.
Pakistan. The latter is the most populous province of The inflow into the test canal reach selected for the
Pakistan with a population of 110 million according to seepage measurement was kept constant, at least,
the 2017 census. The first area included six distributaries 24 hours before the start of the experiment. Then, the
situated in the command of the Lower Chenab Canal start and end of the reach were marked. A reference
(LCC), Faisalabad zone in central Punjab. The name of point was established to measure the depth at the start
the distributaries are, Nasrana 1, Nasrana 2, Lagar1, and end of the reach. The cross-section of the canal was
Lagar2, Khikhi, and Sehti Wala. The second group com- divided into 8 to 10 subsections. The flow depth and
prised of one minor, known as, Arain minor-I, situated in width in each subsection was measured by wading rod
the command of the river Sutlej in the Bahawalnagar cir- and cable. The mean depth was taken as an average of
cle of the Bahawalpur zone in southeastern Punjab. All the depths of two consecutive subsections.
distributaries have a trapezoidal cross-section and were The calibration of the current meter was performed
concrete-lined during 2014–2015. The study area is before using it for velocity measurement. One point
shown in Figures 2 to 4. method (V0.6d) for velocity measurement was used for
Usual crops harvested in the Faisalabad irrigation depth less than 0.6 m and two-point method (vertical
zone are cotton, rice, sugarcane, wheat, variety of fruit average of velocity at V0.2d and V0.8d) was used for depths
and vegetables. Major crops of Bahawalnagar are sugar- between 0.6 and 2 m.
cane, cotton, wheat, rice, tobacco, mustard seed, variety
of fruits and vegetables.
3.3 | Measurement of wetted perimeter

3.2 | Methods The depth and width of the canal were measured at 4 dif-
ferent locations of a reach from which an average depth
The Punjab Irrigation research Institute used the inflow- and width were calculated. The wetted perimeter was
outflow method for seepage measurement. The flow rate measured using average depth and width.

FIGURE 2 Study area


6 SHAH ET AL.

FIGURE 3 Faisalabad irrigation zone

3.4 | Soil classification

Standard penetration tests were conducted at 7 different


locations to obtain undisturbed samples at a depth of 1.5
to 2 m below the canal bank. The following tests were
performed: Sieve analysis, Hydrometer analysis, Liquid
limit and plastic limit. The soil was then classified
according to the Unified Soil Classification System
(ASTM D2487–11). The results show the soil type of area
can be categorized as a silt loam. The silt loam is a soil
having 50% or more silt (0.005 to 0.075 mm) content and
12 to 27% clay (0.001 to 0.005 mm), or 50 to 80% silt and
less than 12% clay.

3.5 | Seepage estimation using an


empirical formulae

In the Inflow-outflow seepage measurement method, the


FIGURE 4 Bahawalpur irrigation zone seepage in the reach is obtained by a water budget
SHAH ET AL.

TABLE I Distributary and minor channels details included in inflow-outflow seepage measurement campaign

Average Mean
depth velocity
Location Test Q1 start Q2 end Average Bed width of flow at start
Total of test reach of of Qnet wetted at start of Water surface at start of reach Mean
Name of length reach length reach reach (m3/ Qoff-takes perimeter reach b width at start of reach Vmean bed
distributary (m) (RD, m) (m) (m3/s) (m3/s) s) (m3/s) Pwetted (m) (m) of reach B (m) (m) (m/s) slope
Lagar 1 18971 0 + 336 7010 1.11 0.849 0.26 0.248 4.38 1.60 3.66 1.000 0.522 21 × 10–5
7 + 346
Lagar 2 18971 8 + 919 9098 0.51 0.209 0.304 0.290 2.94 1.55 2.90 0.716 0.371 21 × 10−5
18 + 017
Khikhi 40133 29 + 876 7105 1.19 0.145 1.04 1.03 2.91 2.26 3.91 0.774 0.613 41 × 10−5
36 + 981
Nasrana 1 54649 39 + 372 5048 1.27 0.979 0.287 0.278 3.77 1.42 3.41 0.951 0.690 26 × 10−5
44 + 420
Nasrana 2 54649 47 + 610 5303 0.51 0.266 0.248 0.239 2.57 0.97 2.67 0.796 0.459 26 × 10−5
52 + 913
Sehti Wala 7346 0 + 108 6059 0.44 0.173 0.266 0.256 2.49 0.84 2.36 0.753 0.442 29 × 10−5
minor 6 + 167
Arain 67500 6 + 678 8717 1.13 0.369 0.759 0.741 3.41 1.45 3.35 0.930 0.630 28 × 10−5
Minor-1 15 + 395
7
8 SHAH ET AL.

equation in which all the inputs are known except seep- F: function of canal geometry (dimensionless) (Swamee
age (Alam & Bhutta, 2004). et al., 2000)

S = Q1 −Q2 −Q f − F −U −E ð1Þ h 0:77 + 0:462m i1:31 ++0:6m


0:6m
1 + 0:6m
F= ðπð4 −πÞÞ1:3 + ð2mÞ1:3 1:3 + 0:6m
+ ðb=yÞ1:3 + 0:6m
Each term has units (m3/s) where:
S: canal seepage b: width of canal y: depth of flow m: side slope (1 in this
Q1: inflow at upstream end of reach study)
Q2: Outflow at downstream end of reach
Qf: Flow diverted into off-taking canals For normal cast in place concrete lining, we used a
F: losses in off-taking canals (0 in this study)
permeability value as: k = 1×10−7 m s−1 (Zhang
E: evaporation losses, less rainfall
et al., 2016) while for cracked concrete lining, k = 1×10−6
U: water losses through unmeasured orifices (Ignored in
m s−1 has been used (Swamee et al., 2000).
this study)

Evaporation losses were estimated as 0.5% of the Kostiakov formula (Zhang et al., 2016)
inflow discharge (Mistry et al., 2016).
a 1−b
Q= × Qnetflow ð5Þ
100
3.6 | Empirical formulae
where:
Seepage was estimated for the seven selected canals using Q: seepage quantity (m3 s−1)
5 different empirical formulae which are Moritz equation, Qnetflow: net flow through the canal (m3 s−1) a: permeabil-
Swamee equation, Punmia equation, Kostiakov formula ity coefficient of the soil in canal bed (coefficient of pro-
and Davis-Wilson equation for concrete-lined canals. portionality between flow rate of fluid and hydraulic
head, it is function of void ratio and degree of saturation
Moritz equation (Kraatz, 1977) (Cai et al., 2014))
b: permeability index of the soil in canal bed (defines the
rffiffiffiffi
Q grain size of soil (Hartley, 1992))
S = 0:2 × C × ð2Þ
V The Kostiakov formula is used for seepage estima-
tion both for lined and unlined canals (Zhang
where: et al., 2016). As the soil of the study area is silt loam
S: seepage losses (cusecs per mile length of canal) or light clay loam at a depth of 1.5 to 2 m from the
Q: discharge (m3/s) banks, the values of permeability coefficient a and per-
V: velocity (m/s) C: constant value depending on soil meability index b used in Equ. 5, are 2.65 and 0.45
types(m./s) respectively.
C: equal to 0.1 for concrete lined canal (Akkuzu, 2011) The selection of a and b is very important for seepage
estimation as soil conditions are subject to variation and
Punmia formula (Punmia, 2009) the choice of a representative set of values over the
domain of interest goes a long way in getting good esti-
K = 0:349 × Q0:68 ð3Þ mates of seepage.
Davis and Wilson equation (Akkuzu, 2011)
where:
K: seepage losses (m3/s per million square meters) Pw × L 1
S = 0:45 × C × × H 3w ð6Þ
Q: discharge (m3/s) (at start of the reach) 4 × 10 + 3650 × √V
6

Swamee equation (Uchdadiya & Patel, 2014) where:


S: seepage losses (m3 per length of canal per day) L:
qs = KyF ð4Þ length of canal (m)
Pw: wetted perimeter (m)
where: Hw: water depth in the canal (m)
qs: seepage discharge per unit length of canal (m2/s) V: velocity of flow in the canal (m s−1)
K: hydraulic conductivity of the porous medium (m/s) y: C: constant value depending upon lining (for concrete-
depth of water in the canal (m) lined canals, C is equal to 1 (Akkuzu, 2011))
SHAH ET AL. 9

4 | R E S UL T S A ND D I S CUS S I O N

computed/
measured
Table II and Figure 5 shows the comparison of measured

+1.58
1.96
1.20
4.48
5.00
1.91
1.64

1.84
2.58
1.51
and estimated seepage using empirical formulae both in

qr
tabular and graphical form. Figure 6 shows a pictorial
view of the canals included in the present study.

×10−6 m3/m2s
Kostiakov
5 | DISCUSSION

0.411
0.514
1.312
0.701
0.903
0.848

0.767
5.1 | Measured seepage

computed/
measured
Figure 7 shows the seepage loss rate for the 7 selected
canals. The loss rate varies from 1.4 × 10−7 m3/m2s s to

+0.53
1.83
1.08
1.89
2.96
1.00
0.89

1.09
1.53
0.75
5.1 × 10−7 m3/m2s with an average of 3.5 × 10−7 m3/

qr
m2s. The variation in seepage loss emanates from the

×10−6 m3/m2s
condition of the canal lining with cracks or spaces
between blocks promoting seepage.

Moritz
It can be seen that 4 of the 7 canals are exceeding the

0.384
0.461
0.553
0.415
0.476
0.462

0.454
suggested value by a good margin. It is an indirect obser-
vation about the less than optimal state of maintenance

computed/
of the canals. The photos of the canals show grass grow-

measured
ing at various locations in the lining which is an indica-

+0.19
tion of seepage through the root zone of the vegetation.
1.78
0.52
1.34
2.93
0.47
0.39

0.91
1.19
0.92
qr

The physical conditions of the canals as depicted in the


photos seem to confirm the findings e.g. Nasrana 1 shows
Comparison of measured seepage and estimated seepage using empirical formulae

×10−6 m3/m2s

no grass overgrowth and it is the best among the lot


while Sethi wala is the one most leaky showing an abun-
Punmia

dance of overgrowth (Figures 6 and 7). Further, the losses


0.374
0.222
0.393
0.410
0.222
0.200

0.379
in poorly maintained canals can increase dramatically
high, i.e. 45.5% of the inflow discharge (Arshad
computed/

et al., 2009).
measured

The issue of seepage from a well-maintained

−0.42
concrete-lined canal has been debated in the literature.
0.68
0.40
0.72
1.11
0.38
0.33

0.41
0.58
0.28
qr

Kraatz (1977) stated that a well-maintained concrete-


lined canal should have a seepage loss less than
3 × 10−7 m3/m2s while another more stringent figure is
×10−6 m3/m2s

1 × 10−8 m3/m2s (Land et al., 2002). It seems that the


Swamme

suggested upper limit of 3 × 10−7 m3/m2s is a more realis-


0.144
0.170
0.210
0.156
0.180
0.172

0.171

tic value in terms of conditions of maintenance prevailing


in developing countries.
The test reaches for all the canals favor no special
location. They are located in the upstream reaches
×10−6 m3/m2s

(Lagar1 and Arain-Minor 1); near tail (Khikhi and


Measured

Nasrana2); second half (Lagar 2); Sethiwala minor is


almost as long as the entire length of the distributary
0.210
0.427
0.293
0.140
0.472
0.517

0.416

canal. In Pakistan, tail end of the small canals convey-


ing water to the farmland is often complaining of
water shortage, thus, it is a good practice to select
Arain Minor-1
distributary
TABLE II

Mean error

canal test reaches in different parts of a canal to


Sehti Wala
Nasrana 1
Nasrana 2

St. dev. qr


Name of

(qr −1Þ
minor

Mean qr
Lagar 1
Lagar 2

remove bias.
Khikhi

The depth to the groundwater table in the neigh-


borhood of the distributaries is, roughly, about 20 m
10 SHAH ET AL.

TABLE III Results of seepage computation with a modified calculation procedure for F in the Swamee equation

qr qr
Name of Measured Swamee standard computed/ Swamee modified computed/
distributary ×10−6 m3/m2 s ×10−6 m3/m2 s measured ×10−6 m3/m2 s measured
Lagar 1 0.210 0.144 0.68 0.200 0.95
Lagar 2 0.427 0.170 0.40 0.224 0.52
Khikhi 0.293 0.210 0.72 0.275 0.94
Nasrana 1 0.140 0.156 1.11 0.218 1.56
Nasrana 2 0.472 0.180 0.38 0.260 0.55
Sehti Wala minor 0.517 0.172 0.33 0.246 0.48
Arain Minor-1 0.416 0.171 0.41 0.237 0.57
Mean qr 0.58 0.80
St. dev. qr 0.28 0.39
Mean error −0.42 −0.20
(qr −1Þ

F I G U R E 5 Comparison of
measured seepage and estimated
seepage using empirical formulae

while the maximum surface water width of the canals longest of 9 km (Table I). This is more than two times the
is 3.91 m. This means that the groundwater table lies reach length of 2245 m used by (Akkuzu, 2011).
at a depth greater than 5 times the width of the
canals, thus, the canals are operating under conditions
of maximum seepage rate (Kraatz, 1977). Further, it 5.2 | Seepage rate calculated using various
was reported that in Punjab, when the depth to the empirical formulae
groundwater table below the canal bed exceeds 3 ft,
the seepage rate is no longer influenced by the For comparison with the equation, the measured dis-
groundwater table (Dhillon, 1968). charge was normalized with the length of the test reach
This method is more sensitive to the errors in and wetted parameter, giving seepage in units of m3/s.m2
discharge measurement than other methods as in the or simply, m/s.
case of short canal reaches, the uncertainty in flow
measurement may be of the same order as the volume
lost to seepage (Alam & Bhutta, 2004). Therefore, in 5.3 | Swamee equation
order to overcome such errors, long reach lengths
are recommended. A strong point of the present The Swamee equation was derived as a design guideline
field experimental study has been the long canal for efficient minimum loss canal (Swamee et al., 2000).
reaches with the shortest reach of about 5 km and the However, it has also been used to calculate seepage. The
SHAH ET AL. 11

FIGURE 6 Pictorial view of distributaries under study

is most relevant as it corresponds to the mode of con-


struction for the canals in question (Zhang et al., 2016).
In Figure 8, the results show that the Swamee eq under-
estimates for all the canals except for Nasrana1. How-
ever, for Lagar1 and Khikhi, the error does not surpass
50%. It may be recalled that these three canals showed
smaller measured seepage losses and fulfilled the crite-
rion for a well-maintained canal lining. At the same time,
the other four canals (Lagar2, Nasrana2, Sethiwala, and
Arain) showed increased losses. It is reasonable to
assume that, the main uncertain parameter is k, the per-
meability of the concrete, increased by the presence of
cracks. In a cracked state, the k for concrete can increase
manifold, by as much as an order of magnitude,
i.e. 1 × 10−6 m/s (Swamee et al., 2000). Figure 8 shows
FIGURE 7 A comparison of the measured seepage loss, Δqs that with k = 1 × 10−6 m/s, the pendulum swings the
for the 7 distributary canals with line signifying the upper limit of
other way, and the Swamee eq, over predicts the seepage.
seepage loss rate for a well-maintained canal (Kraatz, 1977)
This shows that the selection of a representative value for
concrete lining permeability can provide reasonably
seepage is calculated as a product of geometric variable accurate results for the seepage through the Swamee
F, depth of flow and the permeability of the concrete lin- empirical formula.
ing. The variable F has a closed-form solution proposed In a majority of the canals, the Swamee equation has
by Swamee et al. (2000). Its value does not vary much underestimated the seepage (Table II). Barring Nasrana1
(Equation 4). The key variable, surely, is k, the perme- distributary, which has a very small measured seepage
ability of the concrete. The literature provides two values loss at 0.14 × 10−6 m3/s.m2, Swamee equation has under-
for k, with that of 1 × 10−7 m/s for in-situ placed concrete estimated the seepage in all the other canals. The mean
12 SHAH ET AL.

coefficient C value equal to 0.1 was used for the concrete-


lined canals following (Akkuzu, 2011).
The results of the seepage rate loss for the Moritz
equation show that if the Nasran1 canal is considered an
anomaly with an overestimation factor of about 3, the
maximum overestimation has been less than 2 for the
rest. Overall, the mean error is +0.53 with a standard
deviation of 0.75 (Table II).

5.6 | Kostiakov formula

The Kostakov formula (Equation 5) expresses the seepage


in terms of a power-law form, as follows:

FIGURE 8 Seepage calculations for cracked and uncracked q = cQd ð7Þ


values of concrete lining

where c and d depend on permeability coefficient and


error stands at −0.42 with a standard deviation of 0.28 permeability index, respectively and Q denotes the net
which is the least among all other empirical formulae. volume flow rate for the canal reach. The parameters
To improve the results of the Swamee equation, with c and d are expressed in terms of the soil texture only
a large majority showing underestimation, a change was and are selected from a table given in (Zhang
proposed in the calculation of the parameter F in Equa- et al., 2016). As there are only a few characterizations
tion 4 and in the ratio b/y, for canal bed width b, the for the soil texture, the parameters are described in a
water surface top width was used. This led to about 38% rather, generalized way. This formula does not use any
increase in the mean value of parameter F. Further, the measure for the permeability of concrete and instead,
mean error reduced to −0.20 which amounts to about uses the soil properties considering a steady seepage
52% reduction in mean error. state. No doubt, the foregoing limitations, partly
explain the rather large mean error of +1.58. Zhang
et al. (2016) used the Kostiakov formula on a lined
5.4 | Punmia equation canal and obtained over-estimated results as large as
10–2.5 times the measured seepage.
Surprisingly, the performance of the Punmia equation is The Kostiakov formula has a similar power-law form
found to be very good with a mean error of +0.19. How- as the Punmia equation (Equation 3). However, the
ever, the scatter in the results is relatively large at 0.92 Punmia equation is a function of flow volume rate Q only
which compares unfavorably with 0.28 of Swamee and where the Kostiakov formula depends on the soil perme-
0.75 of the Moritz formula. The formula relates seepage ability, in addition, to flow volume rate Q. The Kostiakov
quantity q to discharge carried by the canal in power-law formula relatively poor performance may also be due to
form as q = cQd. The coefficient c and d reflect the quality the non-attainment of the steady-state seepage conditions
of the database of measured seepage rates. The equation in the field due to long test reaches.
does not try to mimic any physical process for seepage.
The reason for its success is that it is closer to the local
conditions being developed for very similar Indo-Pak 5.7 | Davis-Wilson equation
conditions. The regional seepage loss estimation built
into it makes it a good predictor. Davis-Wilson equation results provide insignificantly
small seepage volumes (~10−6) compared to the measured
results and were, thus, not shown. The probable reason
5.5 | Moritz equation for its poor performance may be related to the fact that
pffiffiffiffiffiffiffiffiffiffi seepage varies as the depth of water raised to the power
This formula relates the seepage to Q=V ; dimension- 1/3 whereas, in Moritz and Swamee, the exponent is
ally, it is the same as the Swamee formula which relates 1. Another reason maybe, its unsuitability to the local
seepage to the depth of flow in the canal. Further, the conditions. Other researchers have reported similar
SHAH ET AL. 13

problems with the Davis-Wilson formula (Salemi & and empirical formulae results are chiefly, uncertainty in
Sepaskhah, 2006; Akkuzu, 2011; Mohsen & the discharge measurement, presence of cracks in con-
Mohammed, 2016; Momenzadeh, 2017). crete lining and uncertainty in the values of parameters
appearing in the empirical formulae. The long reach
lengths also had to contend with more diversions which
6 | C ON C L US I ON S probably, increased the uncertainty as at such places,
gauge rating curves were used to estimate the outflow.
In this research, the age-old problem of seepage loss The field experiment was carried out under uniform sur-
from the free surface distributary canals was explored. face flow conditions as 24 hours prior to seepage mea-
The so-called distributary canals are typical man-made surement, the canal discharge was kept constant.
water courses which serve to convey water from the However, steady-state seepage conditions require a much
main canal to the farms in the agrarian heartland of longer time period to establish which was not possible in
Pakistan, the Punjab province. Seepage measurement field canals delivering water to farmlands. This may have
data was obtained from the Punjab Irrigation Research led to greater error in seepage estimates by the kostiakov
Institute, who carried out the tests using the inflow- formula.
outflow method. The distributaries were in all 7 in The real contribution of the research is in helping the
number; 6 of which were from the central Punjab divi- engineer accurately estimate seepage from canal using
sion of Faisalabad and 1 from the south-eastern division empirical formulas. This will help the engineers engaged
of Bahawalpur. in planning, design and operation of canal system to
The seepage is measured in the field by the inflow- make informed choice about what are the best formulas
outflow method. This method is more sensitive to the to use in Punjab, Pakistan.
errors in discharge measurement than other methods as
in the case of short canal reaches, the uncertainty in flow ACKNOWLEDGEMENT
measurement may be of the same order as the volume We extend our thanks to Engr. Ghulam Zakir Hassan
lost to seepage (Alam & Bhutta, 2004) rendering the Sial, Ghulam Shabir and Adnan Hasan of the Punjab Irri-
results unreliable. Therefore, to overcome such errors, gation Research Institute, Lahore for sharing seepage
long reach lengths are recommended. A strong point of data for research purposes with the School of Civil and
the present field experimental study has been the long Environmental Engineering, NUST, Islamabad.
canal reaches with the shortest reach of about 5 km and
the longest of 9 km (Table I). To the best of our knowl-
edge, the occurrence of such long reach lengths have not RE FER EN CES
been reported in the scientific literature. Akkuzu, E. (2011) Usefulness of empirical equations in assessing
canal losses through seepage in concrete-lined canal. Journal of
The field seepage data was used to carry out a com-
Irrigation and Drainage Engineering, 138(5), 455–460.
parative study of seepage estimation using five empirical
Akkuzu, E., Ünal, H.B. and Karataş, B.S. (2007) Determination of
seepage formulae, i.e. Davis-Wilson, Kostiakov, Moritz, water conveyance loss in the Menemen open canal irrigation
Punmia and Swamee. network. Turkish Journal of Agriculture and Forestry, 31(1),
The results showed that the Swamee and the 11–22.
Punmia formulae performed admirably with mean per- Alam, M. and Bhutta, M. (2004) Comparative evaluation of canal
cent errors of −20% and + 19% respectively. Moritz for- seepage investigation techniques. Agricultural Water Manage-
mula stood at +53% while Kostiakov scored +158%. ment, 66(1), 65–76.
Andrade, R., Bhowmick, S. and Pund, A. (2017) Tracer an effective
The Davis-Wilson formula, overwhelmingly under-
tool in detecting canal seepage: Case studies from northern
estimated the seepage. A proposed a slight modification India. Sustainable Water Resources Management, 3(1), 25–31.
was proposed in the Swamee formula whereby, instead Arshad, M., Ahmad, N., Usman, M. and Shabbir, A. (2009) Com-
of the canal width, the surface water width was used in parison of water losses between unlined and lined watercourse
the calculation of the dimensionless variable F. This in Indus Basin of Pakistan. Pak. J. Agri. Sci, 46(4), 2076–0906.
has the beneficial effect of bringing down the percent Bakry, M.F. and Awad, A.A.E.-M. (1997) Practical estimation of
mean error from −42% to −20%. seepage losses along earthen canals in Egypt. Water Resources
The measured seepage data failed to show any trend Management, 11(3), 197–206.
Buchanan, T.J. and Somers, W.P. (2016) Techniques of Water-
for which, surely, more field tests are required. Visual
Resources Investigations of the United States Geological Survey.
evidence from the field suggested that canal with intact USA: US Geological Survey, city.
and fewer cracks in lining leaked less while the one with Burt, C.M., Orvis, S. and Alexander, N. (2009) Canal seepage reduc-
more vegetative growth sprouting out of lining, leaked tion by soil compaction. Journal of Irrigation and Drainage
more. The reasons for the divergence in the measured Engineering, 136(7), 479–485.
14 SHAH ET AL.

Cai, G., Zhou, A. and Sheng, D. (2014) Permeability function for Punmia. (2009) Irrigation and water power engineering, 16th edition.
unsaturated soils with different initial densities. Canadian Geo- India: Laxmi Publications. city.
technical Journal, 51(12), 1456–1467. Salemi, H. and Sepaskhah, A. (2006) Estimation of canal seepage
Cakmak, B., Betribey, M., Yildirim, Y.E. and Kodal, S. (2004) loss in Rudasht region of Isfahan. JWSS-Isfahan University of
Benchmarking performance of irrigation schemes: A case study Technology, 10(1), 29–43.
from Turkey. Irrigation and Drainage, 53(2), 155–163. Sepaskhah, A. and Salemi, H. (2004) An empirical model for predic-
Dhillon, G.S. (1968) Estimation of seepage losses from unlined tion of conveyance efficiency for small earth canals. Iranian
channels. Indian Journal of Power and River Valley Journal of Sciences and Technology, 28, 623–628.
Development.. Singh, A. (2019) An overview of drainage and salinization problems
Eisenhauer, D. and Hoffman, G. (2008) Irrigation principles and of irrigated lands. Irrigation and Drainage..
management. 13–50. In: Biological Sytems Engineering Depart- Skoperboe GU, Aslam M, Mahmoal K, Mahmmd S, Khan AW.
ment. Lincoln, Nebraska, USA: University of Nebraska. 1999. Inflow-Outflow channel losses and canal linning cost
Hartley, D. (1992) Interpretation of Kostiakov infiltration parame- effectiveness in the Fordwah Eastern Sadiqia irrigation and
ters for borders. Journal of Irrigation and Drainage Engineering, drainage project. city, country.
118(1), 156–165. Soothar, R., Mirjat, M., Mangrio, M., Chandio, A. and Leghari, N.
International Commission on Irrigation and Drainage (ICID). (2015) Estimating seepage seepage losses in different size of
(1968) Controlling seepage losses from irrigation canals: world- earthen watercourses at farm level. Pakistan Journal of Agricul-
wide survey, Vol. 1967. New Delhi, India.. ture, Agricultural Engineering and Veterinary Sciences, 31(1),
Kraatz, D.B. (1977) Irrigation canal lining: Food and Agriculture 81–92.
Organization of the United Nations. Italy: Rome. Sultan, T., Latif, A., Shakir, A., Kheder, K. and Rashid, M. (2014)
Land and Water Resources Research Development Corporation Comparison of water conveyance losses in unlined and lined
(LWRRDC). 2002. Guidelines to good practice for the construc- watercourses in developing countries. University of Engineering
tion and refurbishment of earthen irrigation channel banks. and Technology Taxila. Technical Journal, 19(2), 23.
city, country. Swamee, P.K., Mishra, G.C. and Chahar, B.R. (2000) Design of min-
Latif, M. (2007) Spatial productivity along a canal irrigation system imum seepage loss canal sections. Journal of Irrigation and
in Pakistan. Irrigation and Drainage, 56(5), 509–521. Drainage Engineering, 126(1), 28–32.
Leigh EW. 2014. Evaluation of method for predicting seepage loss Uchdadiya, K. and Patel, J. (2014) Seepage losses through unlined
rates for the hard lined irrigation canals of lower rio grande val- and lined canals. Int. J. Adv. Appl. Math. And Mech, 2(2),
ley of texas. city, USA 88–91.
Luo, Y., Bai, K., Traore, S., Gu, H., Sun, Y., Xiong, Y. and Fipps, G. Wachyan, E. and Rushton, K. (1987) Water losses from irrigation
(2016) Determining seasonal canal seepage using stage–seepage canals. Journal of Hydrology, 92(3–4), 275–288.
relationship. Irrigation and Drainage, 65(3), 334–340. Warnick. (1951) Methods of measuring seepage loss in irrigation
Martin, C.A. and Gates, T.K. (2014) Uncertainty of canal seepage canals. Publisher, City, Country..
losses estimated using flowing water balance with acoustic Yao, W., Ma, X. and Chen, Y. (2019) Optimization of canal water in
Doppler devices. Journal of Hydrology, 517, 746–761. an irrigation network based on a genetic algorithm: A case
Mistry KV, Khasiya DRB, Patel DJN. 2016. Economic Analysis of study of the North China Plain Canal system. Irrigation and
Canal Lining. Drainage..
Mohsen, M. and Mohammed, O. (2016) Compared between the Young W, Garthwaite III MG, Leb C, Lytton L. 2019. World Bank
measured seepage losses and estimation and evaluated convey- report: Pakistan getting more from water. city, country.
ance efficiency for part of the hilla main canal and three dis- Zhang, Q., Chai, J., Xu, Z. and Qin, Y. (2016) Investigation of irriga-
tributory canals (HC 4R, HC 5R and HC 6R) of hilla-kifil tion canal seepage losses through use of four different methods
irrigation project. Civil and Environmental Research, 8(2). in Hetao Irrigation District, China. Journal of Hydrologic Engi-
Momenzadeh SB. 2017. The evaluation of water leakage in the irri- neering, 22(3), 05016035.
gation channels of Kazeroon Plain using the input method of
discharge and experimental formulas. city, country.
Plusquellec, H. (2019) Overestimation of benefits of canal irrigation
projects: Decline of performance over time caused by deteriora- How to cite this article: Shah Z, Gabriel H,
tion of concrete canal linning. Irrigation and Drainage., 68(3), Haider S, Jafri T. Analysis of seepage loss from
383–388. https://doi.org/10.1002/ird.2341. concrete lined irrigation canals in Punjab,
Pognant, D., Canone, D., Previati, M. and Ferraris, S. (2013) Using Pakistan. Irrig. and Drain. 2020;1–14. https://doi.
EM equipment to verify the presence of seepage losses in irriga-
org/10.1002/ird.2474
tion canals. Procedia Environmental Sciences, 19, 836–845.

Vous aimerez peut-être aussi