A physically based model for

scour prediction at Srisailam dam

E.F.R. Bollaert, AquaVision Engineering Ltd, Switzerland
P.J. Mason, Black and Veatch Ltd, UK
This paper presents a comprehensive evaluation of 20 years of plunge pool scour development at Srisailam
dam in Andhra Pradesh, India. A comparison is made between one of the most widely used scour
expressions, and a new physically based engineering model for scour prediction. While the former only
predicts the ultimate scour, the latter also allows for estimating the time evolution of scour by incorporating
the geomechanical characteristics of the fractured bedrock. Detailed data on bathymetric surveys and gate
operations since 1984 allowed for a sound comparison between the methods.

A 1. Background and data

comparison is made here between one of the
most widely used scour expressions, [Mason Srisailam dam is on the river Krishna, in the Kurnool
and Arumugam [19851], and a new and com- District of Andhra Pradesh, India. It is about 110 km
prehensive model to evaluate scour of rock occurring upstream of the Nagarjunasagar dam and about 200 km
downstream of high-head dams and in plunge pools from the town of Hyderabad. The project consists of a
[Bollaert, 20042]. Scour of rock caused by high-veloc- 143 m-high masonry dam and a right bank power station
ity jets has been of concern to practicing engineers for equipped with seven 110 MW units. The plant is locat-
a long time. Today’s most popular evaluation methods ed about 450 m downstream of the dam. The dam reser-
are of an empirical and semi-empirical nature, and do voir has a full storage level of el. 269.75 for a capacity
not fully describe the physical background of the phe- of 8720 × 106m3. The spillway has been constructed in
nomenon. Therefore, a new model has been devel- the deep river portion with non-overflow blocks on
oped, based on a parametric description of the main either flank (see Fig. 1). The spillway has an overall
physical processes responsible for scour. operating crest length of 266.4 m at el. 252.98. It com-
prises 12 spans of 18.3 m clear width, equipped with
radial crest gates of 18.30 × 21.49 m, as shown in Fig. 1.
The maximum design discharge is 38 370 m3/s at a
maximum reservoir level of el. 271.88. A ski-jump
bucket with an invert at el. 188.98 has been provided
for energy dissipation at the downstream toe of the
dam. Two sluices provided in the right flank non-over-
flow portion are designed to pass a maximum dis-
charge of 1500 m3/s.
Dam construction started in 1963 and was complet-
ed in 1984, including erection of the spillway gates.
During construction, deep scour was observed near
the toe of the dam. The toe was protected from being
undermined by the progressive construction of a 36
m-wide concrete apron in front of block Nos. 8 to 18
Fig. 1. Cross section
(Fig. 1) between 1977 and 1981. During subsequent
of the Srisailam dam years, undermining of the concrete apron persisted
[Swamy, 19793]. and the necessary measures were taken for protection.
Because of concerns about intensive spray affecting
Table 1: History of scour at Srisailam dam the power station complex and the roads along the
banks, the end spans of the spillway have never been
Year Water levels Peak Gates Scour operated for flood relief.
discharge operating elevation A history of scour development is presented in Table
Upstream Downstream Difference Q 1, based on information provided by the Government
(el.) (el.) (el.). (m3/s) (-) (el.) of Andhra Pradesh, India. Furthermore, correspond-
1984 261.8 177.7 84.1 9301 1 to 12 140.8 ing flows and heads up- and downstream are given,
1985 264.9 173.1 91.8 5665 4 to 9 141.1 based on information supplied by the I&CAD Dept.
1986 268.2 175.3 92.9 9033 2 to 10 137.8 Swamy [19793] stated that the bedrock at the dam
1987 - - - - - 137.8 site is characterized by many joints, intercalations of
1988 269.4 179.8 89.6 12327 4 to 11 132.9 soft shales, weak pockets, sheared and fractured
1989 269.7 178.0 91.7 10801 4 to 11 133.2 zones, cavities and so on. He highlighted the presence
1990 268.1 180.1 87.9 12533 2 to 11 132.0 of horizontally disposed shear zones between els. 145
1991 268.3 182.3 86.0 13708 2 to 11 132.0 and 154, as shown in Fig. 2.
1992 269.2 178.0 91.2 10582 2 to 11 134.1
1993 269.6 182.6 87.1 14430 2 to 11 132.0
1994 265.4 181.5 83.9 16008 2 to 11 137.5 2. Mason and Arumugam formula
1995 - - - - - 137.5 The scour expression developed by Mason and
1996 269.7 182.3 87.4 13501 2 to 11 132.0 Arumugam [19851] is one that is widely used in prac-
1997 269.0 182.0 87.0 15150 2 to 11 131.0 tice. The authors compared more than 30 formulae for
1998 269.5 187.5 82.1 24471 2 to 11 125.0
estimating scour depth with data from 47 models and
Design 271.9 193.5 78.3 38369 all - 26 prototypes. Experience showed that scour depth is

2 Hydropower & Dams Issue Three, 2006

broadly proportional to the unit flow q0.6. This formed
the basis for the M & A expression:
⎛ q 0.60 ⋅ H 0.05 ⋅ h 0.15 ⎞
D = 3.27 ⋅ ⎜ ⎟
⎝ g 0.30 ⋅ d 0.10 ⎠
... (1)
where:
D = scour depth below tailwater level (m)
q = unit flow (m3/m)
H = head difference between reservoir level and tail-
water levels (m)
h = tailwater depth (m)
g = gravitational constant (m/s2)
d = characteristic size of bed material (m)
To compare the formula above with recorded data
for Srisailam dam, computed scour depths were plot-
ted against associated flows in the form of q0.6. The
results are shown in Fig. 3, together with a plot of Eq.
(1), which can be seen to fit the data points broadly.
In the absence of detailed geotechnical information Fig. 2. Rock at the
3. Comprehensive scour model dam site [Swamy,
in the area of the plunge pool, it has been shown that
assuming a particle size of 0.25 m or 0.30 m in the A comprehensive scour model was developed based 19793].
Mason and Arumugam expression can give a reason- on experimental and numerical investigations of
able prediction of scour depth. A value of 0.3 m was dynamic water pressures in rock joints [Bollaert,
therefore used initially in the case of Srisailam and, as 20024]. The model comprises two methods which
can be seen from Fig. 3, a reasonable approximation describe failure of jointed rock. The first, the compre-
was obtained. hensive fracture mechanics (CFM) method, deter-
However, three defining scour events occurred at mines the ultimate scour depth by expressing instanta-
Srisailam during the period of record. These occurred
in 1986, 1988 and 1998. In all three years, scour
increased as a result of the associated flow, as shown
in Fig. 4. It was therefore assumed that these scours,
and the associated flows, could be taken as defining
events and, for the convenience of calibration, the par-
ticle size was adjusted accordingly to obtain the best
fit for these three years. It was found that a particle
size of 0.65 m gave the best calibration and this is also
shown in Fig. 3. This relationship was then used to
predict a possible deepest scour elevation for the
design flood.
It should be noted that modifying the particle size in
this way is simply a calibration convenience. Equally
the main constant of 3.27 in the formula could be
adjusted. In fact the nature of any rocky riverbed is like-
ly to be quite complicated, both in terms of the nature
and the type of rock. Various rock cores were inspected
at the dam site, and although many of the formations at
the site appeared massive, hardly any pieces of core Fig. 3. Measured scour compared with the M&DA expression.
were longer than 0.3 m. This suggests a high degree of
internal fracturing or weakness which will yield to dis-
turbance by, for example, drilling or dynamic water
fluctuation. It is therefore considered that, as an initial
‘broad-brush’ approach, the use of a single particle
diameter may be both convenient and representative.
Furthermore, application of Eq. (1) to both the 1998
flood and the dam design flood resulted in the scour
depths shown below in Table 2.
Hence, the design flood would further increase the
actual scour formation by 6 to 7 m.

Table 2: Ultimate scour depths based on Eq. 1

Flood Measured Computed Computed

(m) (d= 0.30 m) (d= 0.65 m)
(m) (m)
1984 140.8 144.7 147.2
1998 125.0 119.0 124.1
Design - 112.0 118.0 Fig. 4. Scour elevation measured as a function of time.

Hydropower & Dams Issue Three, 2006 3

 Table 3: Spillway chute flow conditions
pressures at the water-rock interface. The plunge pool
water depth Y is essential. For almost vertically
Qtot Q/pass q Hup hdown Vdown Hdown impacting jets, it is defined as the difference between
(m3/s) (m3/s) (m3/s/m) (m) (m) (m/s) (m) the water level and the bedrock level at the point of
impact. The water depth increases with discharge and
7800 650 35.5 266.6 1.10 32.3 174.0
9600 800 43.7 267.6 1.35 33.0 177.5
scour formation. Initially, Y equals the tailwater depth
11040 920 50.3 268.3 1.50 33.4 178.5 t (see Fig. 5). During scour formation, Y has to be
13176 1098 60.0 269.4 1.80 33.8 181.5 increased with the depth of the formed scour, h.
14400 1200 65.6 270.0 1.95 34.0 182.0 The water depth Y and jet diameter at impact Dj deter-
15840 1320 72.1 270.7 2.15 34.2 182.5 mine the ratio of water depth to jet diameter at impact
24288 2024 110.6 274.3 3.20 35.2 187.5 Y/Dj. This ratio is directly related to jet diffusion.
31620 2635 144.0 276.5 4.00 36.0 192.0 The most relevant pressures are the mean dynamic
pressure coefficient Cpa and the root-mean-square
neous or time-dependent joint propagation caused by (rms) coefficient of the fluctuating dynamic pressures
water pressures inside the joint. The second, the C'pa, both measured directly beneath the centreline of
dynamic impulsion (DI) method, describes the ejec- the jet. These coefficients correspond to the ratio of
tion of rock blocks from their mass caused by sudden pressure head, in metres, to incoming kinetic energy
uplift pressures. of the jet (V2/2g) and are defined in Bollaert [20042].
The structure of the comprehensive scour model
consists of three modules: the falling jet, the plunge 3.3 The module of the rock mass
pool and the rock mass. The third module implements The pressures defined at the bottom of the pool are
the two previously mentioned failure criteria. More used to determine the transient pressures inside open-
details on the model equations can be found in a pre- end or closed-end rock joints. The parameters are:
vious paper by Bollaert [20042].
• maximum dynamic pressure coefficient Cmaxp;
3.1 The module of the falling jet • characteristic amplitude of pressure cycles Δpc;
• characteristic frequency of pressure cycles fc; and,
This module describes how the hydraulic and geomet- • maximum dynamic impulsion coefficient CmaxI.
ric characteristics of the jet are transformed from
issuance at the dam down to the plunge pool (Fig. 1). The first parameter is relevant to brittle propagation
Three parameters characterize the jet at issuance: the of closed-end rock joints. The second and third para-
velocity Vi, the diameter (or width) Di and the initial meters express time-dependent propagation of closed-
turbulence intensity Tu, defined as the ratio of veloci- end rock joints. The fourth parameter is used to define
ty fluctuations to the mean velocity (2). The jet trajec- dynamic uplift of rock blocks formed by open-end
tory is based on ballistics and air drag and will not be rock joints.
discussed further here. The jet module computes the The maximum dynamic pressure Cmaxp is obtained by
longitudinal location of impact, the total trajectory multiplication of the rms pressure C'pa with an ampli-
length L and the velocity and diameter at impact Vj fication factor Γ+, and by superposition with the mean
and Dj. dynamic pressure Cpa. Γ+ expresses the ratio of the
peak value inside the rock joint to the rms value of
3.2 The module of the plunge pool pressures at the pool bottom and has been determined
Fig. 5. Parameters based on prototype-scaled experiments [Bollaert,
of the scour model This module describes the characteristics of the jet 20042].
[Bollaert, 20042]. when traversing the plunge pool and defines the water The product of C'pa times Γ+ results in a maximum
pressure, written as:
Vj2 V2
Pmax [ Pa ] = γ ⋅ C max
p ⋅ = γ ⋅ (C pa + Γ + ⋅ C©pa ) ⋅ j
2g 2g
... (2)
The main uncertainty of Eq. (2) lies in the Γ+ factor.
It is interesting to note that, based on Eq.(2), maxi-
mum pressures inside joints occur for Y/Dj ratios of
between 8 and 10. This means that the most critical
flood situation may not be the PMF but rather the flow
that results in a critical Y/Dj ratio.
The characteristic amplitude of the pressure cycles,
Δpc, is determined by the maximum and minimum
pressures of the cycles. The characteristic frequency
of pressure cycles fc follows the assumption of a per-
fect resonator system and depends on the air concen-
tration in the joint αi and on the length of the joint Lf.
Beside the dynamic pressure inside the rock joints,
the resistance of the rock has to be determined. The
cyclic character of the pressures generated by the
impact of a high velocity jet makes it possible to
describe joint propagation by fatigue stresses occur-
ring at the tip of the joint. This can be described by
linear elastic fracture mechanics (LEFM), which
assumes a perfectly linear elastic, homogeneous and
isotropic material.

4 Hydropower & Dams Issue Three, 2006

 A simplified methodology is used [Bollaert, 20042].
The method attempts to describe the main principles
such that engineering applications become plausible.
It is called the comprehensive fracture mechanics
(CFM) method and is applicable to any partially joint-
ed rock. Pure tensile pressure loading inside the rock
joints is described by the stress intensity factor KI.
This parameter represents the amplitude of the rock
mass stresses that are generated by the water pressures
at the tip of the joint. The corresponding resistance of
the rock mass against joint propagation is expressed
by its fracture toughness, KIc.
The issue is to obtain a comprehensive and physical-
ly correct implementation of the complex and dynam-
ic situation encountered in fractured rock. Joint prop-
agation distinguishes between brittle (or instanta-
neous) joint propagation and time-dependent joint
propagation. The former occurs for a stress intensity
factor equal to or higher than the fracture toughness of tially sustained by the surrounding rock mass in two Fig. 6. Rock joint
horizontal directions. Corresponding stress intensity configuration of the
the material. The latter occurrs when the maximum CSM model.
possible water pressure results in a stress intensity that factors should be used in the case of low to moderate-
is inferior to the material’s resistance. Joints may then ly jointed rock. The second crack is single-edge
be propagated by fatigue. Failure by fatigue depends notched two-dimensional. Support from the surround-
on the frequency and the amplitude of the load cycles. ing rock mass is only exerted perpendicular to the
The fracture mechanics implementation of the hydro- plane of the notch and, as a result, stress intensity fac-
dynamic loading consists of a transformation of the tors will be substantially higher. Thus, it is appropriate
water pressures in the joints into stresses in the rock. for significantly to highly jointed rock.
These stresses are characterized by KI as follows: In practice, F values of 0.5 or higher are considered
to correspond to completely broken-up rock, that
K I = Pmax ⋅ F ⋅ π ⋅ L f means, the DI method becomes more applicable than
the CFM method. For values of 0.1 or less, a tensile
... (3) strength approach is more appropriate. However, most
in which KI is in MPa √m and Pmax (2) in MPa. of the values in practice can be considered to be
The boundary correction factor F depends on the between 0.20 and 0.40, depending on the type and
type of crack and on its persistency, that is, its degree number of joint sets, the degree of weathering, dis-
of cracking, defined as a/B or b/W in Fig. 6. This tances between the joints, and so on.
Figure presents three basic configurations for partial- The fracture toughness KIc has been related to the
ly jointed rock. The choice of the most relevant geo- mineralogical type of rock and to the unconfined com-
metry depends on the type and the degree of jointing pressive strength (UCS). Furthermore, corrections are
of the rock. made to account for the loading rate and the in-situ
The first crack has a semi-elliptical shape and is par- stress field. The corrected fracture toughness is
defined as the in-situ fracture toughness KI,ins and is
based on a linear regression of data from available lit-
erature. More detailed equations, as a function of the
Table 4: Parameters of falling jet and plunge pool modules mineralogical rock composition, can be found in a
Jet issuance from flip bucket of spillway chute paper by Bollaert [20024].
Unit discharge m2/s 62.0
Total discharge Q m3/s 13600 KIins, UCS= (0.008 to 0.010)·UCS+(0.054·σc)+0.42 ... (4)
Issuance velocity Vi m/s 34.0
Issuance width bi m 18.3
where σc represents the confinement horizontal in-situ
Issuance equivalent diameter Di m 6.7 stress and T, UCS and σc are in MPa. Instantaneous
Angle θi ˚ 45 joint propagation will occur if KI ≥ KI,ins. If this is not
Turbulence intensity Tui % 4 the case, joint propagation needs a certain time to
occur. This is expressed by an equation as originally
Jet impact in plunge pool proposed to describe fatigue growth in metals:
Velocity Vj m/s 39.0
dL f
Core diameter Dj m 3.6 = C r ⋅( ΔK I / K Ic )m r
Outer diameter Dout m 13.8 dN
Angle of jet θj ˚ -40.7 ... (5)
Distance from dam toe Xult m 120
Air drag reduction K - 0.9 where: N is the number of pressure cycles; Cr and mr
Trajectory length Lj m 128 are material parameters determined by fatigue tests,
Break-up length Lb m 60 and ΔKI is the difference in the maximum and mini-
Degree of break-up Lb/Lj - 2.13 mum stress intensity factors. To implement time-
Jet impact at rock bed
dependent joint propagation into the model, mr and Cr
have to be known. They represent the vulnerability of
Initial water depth Y m 38 rock to fatigue, and can be derived from available lit-
Static pressure Cstat - 0.52
erature data on quasi-steady break-up by water pres-
Mean dynamic pressure Cpa - 0.10
sures in joints [Atkinson, 19875]. A first-hand calibra-
RMS dynamic pressure C’pa - 0.11
Amplification factor Γ+ - 4
tion for granite (at Cahora Bassa dam) [Bollaert,
20024] resulted in Cr = 1E-8 for mr = 10.

Hydropower & Dams Issue Three, 2006 5

 a first approach, they can be disregarded. The pressure
field over the block is governed by jet diffusion. The
pressure field under the block corresponds to transient
pressure waves.
The first step is to define the maximum net impul-
sion Imax. This is defined as the product of a net force
and a time period. The corresponding pressure is made
non-dimensional by the jet’s kinetic energy V2/2g.
This results in a net uplift pressure coefficient, Cup.
The time period is non-dimensionalized by the travel
period that is characteristic for pressure waves inside
rock joints, that is, T = 2*Lf/c. This results in a time
coefficient, Tup. Hence, the non-dimensional impul-
sion coefficient CI is defined by the product Cup*Tup =
V2*L/g*c (m*s). The maximum net impulsion Imax is
obtained by multiplying CI by V2*L/g*c. Prototype-
scaled analysis of uplift pressures resulted in the fol-
lowing expression for CI:
2
⎛Y⎞ ⎛Y⎞
CI = 0.0035 ⋅ ⎜ ⎟ − 0.119⎜ ⎟ + 1.22
⎝ j⎠
D ⎝ Dj ⎠
Fig. 7. Jet trajectory The fourth dynamic loading parameter is the maxi- ... (7)
and plunge pool at mum dynamic impulsion CmaxI in an open-ended rock
Srisailam dam. joint (underneath a single rock block), obtained by Failure of a block is expressed by the displacement it
time integration of the net forces on the block undergoes as a result of the net impulsion CI. This is
(Newton): obtained by the transformation of VΔtpulse in Eq. (6) into
Δtpulse a net uplift displacement hup. The net uplift displace-
ment that is necessary to eject a rock block from its
I= ∫
0
( Fu −Fo − G b − Fsh ) ⋅ dt = m ⋅ VDtpulse
matrix is difficult to define. It depends on the protru-
... (6) sion and the degree of interlocking of the blocks. A
first-hand calibration on Cahora Bassa dam [Bollaert,
in which Fu and Fo are the forces under and over the 20024] resulted in a critical net uplift displacement of
block, Gb is the immerged weight of the block, and Fsh 0.20.
represents the shear and interlocking forces. The
shape of the block and the type of rock define the 4. Scour at Srisailam dam
immerged weight. Shear and interlocking forces The comprehensive scour model (CSM) has been
depend on the joint pattern and the in-situ stresses. As applied to the Srisailam dam scour formation. The
computed ultimate scour depth is compared with the
Mason and Arumugam [19851] expression. Also, the
Table 5: Parameters of the rock mass layers
computed scour evolution as a function of time of
Parametric assumptions floods is compared with the available prototype obser-
Property Symbol Conserv. Average Benef. Unity vations.
Unconfined compressive strength UCS 80 120 150 MPa
Density of rock γr 2600 2700 2800 kg/m3 4.1. Jet and plunge pool modules of the
Ratio horizontal/vertical stresses K0 1 1 1 - CSM model
Typical maximum joint length L 1 1 1 m Fig. 7 presents the jet trajectory and the plunge pool
Vertical persistence of joint P 0.15 0.25 0.6 -
geometry for a unit flow discharge of about 50 m3/s/m.
Form of rock joint - single-edge elliptical circular -
The flow characteristics on the spillway chute have
Quartzite

Tightness of joints - tight tight tight -

been computed from the dam crest to the lip of the flip
Total number of joint sets Nj multiple 3 2+ -
Typical rock block length Ib 1 1 1 m
bucket by a 1D two-phase numerical model for steep
Typical rock block width bb 1 1 1 m slopes. This made it possible to define the thickness
Typical rock block height Zb 0.5 0.75 1 m and velocity of the jet at the bucket lip for different
Joint wave speed c 150 125 100 m/s unit flows (Table 3).
Fatigue sensibility m 6 7 8 - It can be noticed that the jet has a very flat rectangu-
Fatigue coefficient C 1.00E-07 1.00E-07 1.00E-07 - lar shape, with a width-to-depth ratio of between 5 and
18, with an average value of about 9. As such, the jet
Unconfined compressive strength UCS 20 40 60 MPa issuing from the bucket lip will probably not com-
Density rock γr 2200 2400 2600 kg/m3
Ratio horizontal/vertical stresses K0 1 1 1 -
pletely transform into a quasi-circular shape, but
Typical maximum joint length L 1 1 1 m rather stay rectangular.
Vertical persistence of joint P 0.5 0.4 0.3 - Next, based on the available data on gate operations
Form of rock joint - single-edge single-edge single-edge - and unit flows, an average unit flow of 62 m3/s/m and
Tightness of joints - tight tight tight - average head and tailwater levels of els. 268 and 180
Shale

Total number of joint sets Nj multiple 4 3+ - have been defined. Based on these values, the main
Typical rock block length Ib 0.1 0.1 0.1 m parameters of the falling jet and plunge pool modules
Typical rock block width bb 0.1 0.1 0.1 m have been computed. They are summarized in Table 4.
Typical rock block height Zb 0.020 0.050 0.075 m
Joint wave speed c 150 125 100 m/s 4.2. Rock mass module of the CSM model
Fatigue sensibility m 8 9 10 -
Fatigue coefficient C 2.00E-07 2.00E-07 2.00E-07 -
The main rock mass characteristics are summarized in
Table 5. Based on Swamy [19793] and Mason [20006],

6 Hydropower & Dams Issue Three, 2006

the bedrock is considered to be a mixture of quartzite Fig. 8. Comparison
and shales. As presented in Fig. 2, two distinct rock of measured and
layers can be distinguished at the dam site: quartzite computed scour
formation:
and shale; they have different geomechanical charac-
teristics.
Quartzite is located near the surface, while near-hor-
izontal plane shale intercalations are present in the
form of 0.2 to 0.9 m bands between the crushed
quartzite at elevations of 130 m and deeper.
No direct information could be obtained related to
the UCS strength of the intact rock at the site.
However, based on Bollaert [20024], the UCS
strengths of quartzite are directly related to their frac-
ture toughness KIc. Based on typical KIc values avail-
able in literature for quartzites between 2 and 2.5, high (a) for an average
UCS values of 210-270 MPa are obtained. These val- unit flow of 62
ues are not site-specific, however. m3/s/m;
Gowda et al. [19997] carried out a seismic survey of
the plunge pool and determined the seismic wave
velocity of the quartzite bedrock at Srisailam as
between 4000 and 5500 m/s, indicating an essentially
non-weathered rock. A fair correlation between frac-
ture toughness and seismic velocity has been estab-
lished by Huang and Wang [19858]. Use of the above
seismic velocities results in KIc values of between 0.9
and 1.8 MPa √ m, or UCS values of between 80 and
180 MPa. Hence, in the following, this range of values
has been used as input to the CSM model. The lower
bound represents a conservative value, while the
upper bound represents a favourable assumption. An
average UCS value is defined as 130 MPa. Second, for
shales, much lower UCS values are generally valid, in
the range of 20 to 60 MPa.
The fatigue parameters of the rock layers, as well as (b) for a unit flow
the other parameters in Table 5, have been chosen of 110 m3/s/m;
based on previous calibrations made for granitic rock
at Cahora Bassa dam [Bollaert and Schleiss, 20059]
and at Kariba dam [Bollaert 200510].
The CSM scour computations have been carried out
for the quartzite rock layer. A more detailed analysis
would require additional computations accounting for
both shale and quartzite layers.

4.3. Results of the CSM model

Figs. 8a to 8d present the scour evolution as a function
of the time of flood spillage. The times of flood dura-
tion between 1984 and 1998 have been computed as
an average of the four most used gates, that is, at
blocks 8, 9, 10 11 and 12. As such, since 1984, the
spillway has been functioning on average for about (c) for the design
12 000 hours. unit flow of 144
m3/s/m applied after
Both the fracture mechanics (CFM) and the dynam-
1998; and,
ic impulsion (DI) methods have been applied. The for-
mer provides scour formation as a function of flood
duration, while the latter only provides an ultimate
state of scour formation.
Fig. 8a shows the computational results for an aver-
age unit flow of 62 m3/s/m, or a total flow of about 13
500 m3/s. This flow is more or less representative for
almost all years of flooding between 1984 and 1998,
except 1998, for which a unit flow of 110 m3/s has
been noticed. Based on Gowda et al. [19997], the ini-
tial bedrock at the point of impact of the jet, that is,
about 160 m downstream of the dam toe, is at about el.
135. Hence, this level has been used as the initial level
for the computations. It can be seen in Fig. 8a that no
scour is computed by either model, which seems to
(d) for the design
agree fairly well with the prototype observations.
unit flow of 144
Gowda et al. [19997] concluded that no significant m3/s/m applied after
scour could be observed in the middle of the plunge 2000.

Hydropower & Dams Issue Three, 2006 7

 pool. Only along the left bank, local scour of 2-4 m good agreement when it comes to prediction of the
deep has formed. 1998 event. For this event, the former model predicts
Second, the 1998 event produced about 10 m of el. 126.1 (average value) and the latter el. 119.
scour in the middle of the plunge pool. This was main- Measured scour was at el. 125.
ly because this event had a unit flow of about twice the For the design flood, the CFM method predicts el.
average of the previous years. By using the 1997 state 124 (average value) and the M&A expression el. 112.
of the plunge pool bottom as a reference for the com- No in-situ measurements are available for this flood.
putations, Fig. 8b shows that scour is computed by the The DI method, however, is in less agreement with
CFM method for average and conservative parametric the M&A expression. For the 1998 event, the former
assumptions on the rock mass quality. predicts deeper scour, with an elevation of 114.2 for
For beneficial assumptions on the rock mass quality, average parametric assumptions. For the design flood,
computed scour formation is negligible. For average the DI method only predicts scour until el. 121.6, as a
assumptions, the computed and measured scour for- result of the high tailwater protection of the bedrock.
mation as a function of flood durations are in very As such, the CFM and DI methods are consistent with
good agreement. This demonstrates the adequacy of each other when it comes to the design flood.
the CFM model to predict further scour formation in It should be underlined that the DI method represents
the plunge pool. Also, the DI method predicts ultimate the end-state of scour, assuming that the rock joint net-
scour depths that are somewhat deeper than the CFM work is formed completely (single block uplift) and
method. This is logical because the DI model takes that all joints have been opened. The DI method does
into account the ultimate state of scour. not account for the time that would be needed for this
Third, the design flood event has been applied fol- to happen.
lowing the 1997 event, that is, instead of the 1998 The relevance of each of the above methods depends
event. As shown in Fig. 8c, the computed scour for- on the degree of fracturing of the rock mass. For com-
mation is less restrictive than for the 1998 flood event. pletely fractured rock, the DI method is most appro-
This, at first sight rather contradictory, statement can priate. For partially fractured rock, the CFM method
be explained by the fact that jet diffusion and energy should be considered instead.
dissipation through the pool depth before impacting The most significant discrepancy between the M&A
the bedrock is more important for the design flood expression and the CFM/DI methods is the prediction
because of a higher tailwater level (el. 193 instead of for the design flood event. The former predicts the
187). Hence, the corresponding maximum pressure deepest possible scour for this event, while the latter
fluctuations inside the rock joints are less. methods predict lower scour than for the 1998 event.
Finally, the design flood has also been applied after
the year 2000, for which a bedrock level at el. 125 was 5. Conclusions
observed. The results are shown in Fig. 8d and are 20 years of scour surveying in the plunge pool down-
very similar to the previous computations. stream of Srisailam dam in Andhra Pradesh, India, has
been used to compare the Mason & Arumugam
4.4. Comparison of models [19851] scour expression with the physically based
As a summary, Table 6 compares the measured scour Comprehensive Scour Model (CSM) developed by
formation with both the CFM/DI [Bollaert, 20042] Bollaert [20042].
methods and the Mason and Arumugam [19851] The comprehensive scour model evaluates the ulti-
expression. The reference bedrock level for the design mate scour depth and the scour evolution in any type
flood is taken as el. 125, and for the 1998 event as el. of jointed rock. The model is based on near-prototype
131. measurements of dynamic pressure fluctuations at
Based on the results in section 4.3 and in Table 6, it plunge pool bottoms and inside artificially created
can be seen that the 1998 event is more restrictive to rock joints. Pressures in closed-end joints were found
scour than the design flood event. Effectively, the to be of cyclic character and have been assessed by
combination of jet diffusion and tailwater depth gen- their characteristic amplitude and frequency of occur-
erates higher pressures and thus more scour for the rence. Pressures in open-ended joints (underneath
former event. rock blocks) have been related to the corresponding
As a summary, it can be concluded that the CFM rock surface pressures to define the net uplift impul-
method and the M&A expression are in reasonably sion on the blocks.
The model represents a comprehensive assessment
of instantaneous and time-dependent (fatigue) fractur-
Table 6: Comparison of both models with in-situ ing of closed-end rock joints (CFM method) and of
measured scour formation dynamic uplift of the formed rock blocks (DI method).
The CFM method not only estimates the ultimate
Flood event 1998 Design scour depth, but also the time evolution of the scour
Flood duration (h) ~ 2000 ~ 2000 formation.
Unit flow (m3/s/m) 110 114 Both methods can also be applied outside the jet’s
Measured in-situ 125.0 - centreline at impact, which defines the spatial extent
M&A (d= 0.30 m) 119.0 112.0 of the scour hole. In general, emphasis is given to the
M&A (d= 0.65 m) 124.1 118.0 physical parameters that are necessary to describe the
CFM (average) 126.1 124.0 different phenomena accurately. These parameters are
DI (average) ultimate 114.2 121.6 defined in view of practical applications. This guaran-
CFM (conservative) 120.5 121.2
tees the comprehensive character of the model, with-
DI (conservative) ultimate 106.2 121.6 out neglecting the underlying physical principles.
The presented application and comparison with the
CFM (Beneficial) 130.8 124.7
DI (Beneficial) ultimate 122.0 125.0
well known Mason and Arumugam [19851] expression
shows the promising and comprehensive character of
the new model. Especially when past scour informa-

8 Hydropower & Dams Issue Three, 2006

tion is available, or when the relevant rock and
hydraulic characteristics at the site are well known, it
is believed that the model is able to predict future
scour formation as a function of time. Its physical
nature also makes it applicable to closely related engi-
neering issues, such as uplift and/or cracking of con-
crete slabs of stilling slabs, or break-up of fractured
coastal structures by violent wave impact. ◊
E. Bollaert

References
1. Mason, P.J. and Arumugam, K., “Free Jet Scour below
Dams and Flip Buckets,” Journal of Hydraulic Engineering,
Vol. 111, N° 2, 1985.
2. Bollaert, E.F.R., “A comprehensive model to evaluate scour
formation in plunge pools.” Hydropower & Dams, 2004.
3. Swamy, ????
4. Bollaert, E.F.R., “Transient water pressures in joints and
formation of scour due to high-velocity jet impact.”
Communication 13, Laboratory of Hydraulic Constructions, P.J. Mason
EPFL, Lausanne, Switzerland; 2002.
5. Atkinson, B.K., “Fracture Mechanics of Rock,” Academic
Press Inc., London, UK; 1987.
6. Mason, P., ????

7. Gowda, ????? Erik Bollaert is President of AquaVision Engineering Ltd.,

a company specialized in river and dam engineering. He
holds an MSc in Civil Engineering and a PhD in Hydraulics
8. Huang + Way????? and Rock Mechanics from the Swiss Federal institute of
Technology (EPFL), Lausanne, Switzerland. Since 1996 he
has participated on several international dam engineering
9. Bollaert, E.F.R. and Schleiss, A.J., “A physically-based projects (in the USA, Iran, Iceland, France, Bolivia, and
model for evaluation of rock scour due to high-velocity jet Switzerland). He developed his PhD in the field of rock
impact”, Journal of Hydraulic Engineering; March 2005. scour caused by high-velocity jet impact. He co-organized
10 Bollaert 2005???? the “International Workshop on Rock Scour” in Lausanne
(EPFL) and is one of the editors of the book “Rock Scour
due to falling high-velocity jets”, published in 2002 (Swets
& Zeitlinger, The Netherlands). He is also Senior Research
Bibliography Associate at the Laboratory of Hydraulic Constructions of
Bollaert, E.F.R., “The influence of joint aeration on dynamic the Swiss Federal Institute of Technology in Lausanne, and
uplift of concrete slabs of plunge pool linings.” Accepted for is responsible for research in the fields of rock scour and
presentation at the XXX IAHR Conference, Thessaloniki, reservoir sedimentation.
Greece; August 2003. AquaVision Engineering Ltd, Chemin des Champs-Courbes
Bollaert, E.F.R., Falvey, H.T., and Schleiss, A., “Turbulent jet 1, CH-1024 Ecublens (VD), Switzerland.
impingement in plunge pools: the particular characteristics
of a near-prototype physical model study.” Proceedings, Peter John Mason is a Divisional Director of Black &
Riverflow 2002, Louvain-la-Neuve, Belgium, 2002. Veatch Ltd, UK. He holds a BSc in civil engineering and
Ervine, D.A. and Falvey, H.R., “Behavior of turbulent jets in MSc and PhD degrees in applied hydraulics from the City
the atmosphere and in plunge pools.” Proceedings,
University, London, UK. His career spans more than 36
Institution of Civil Engineers. Part 2, Vol. 83, 1987.
years, with the last 34 spent exclusively on dams and
Ervine, D.A., Falvey, H.T., and Withers, W., “Pressure
fluctuations on plunge pool floors”, Journal of Hydraulic hydropower schemes. He began working with Sir Alexander
Research, Vol. 35, 2, 1997. Gibb & Partners and was involved in design work on major
Mason P.J., “Effects of Air Entrainment on Plunge Pool Scour,” hydro projects in southern Africa, Pakistan and Sudan. From
American Society of Civil Engineers, Journal of Hydraulic 1977 to 1980 he was based in Uruguay, advising on the
Division, 115, March 1989. design and construction of the Palmar (Constitucion) dam.
Mason, P.J., “Practical Guidelines for the Design of Flip He subsequently worked on schemes in Argentina, Jordan,
Buckets and Plunge Pools,” Water Power & Dam Botswana, Seychelles, Turkey and Ethiopia, and was Project
Construction, September/ October 1993. Engineer for the 122 m-high Victoria arch dam in Sri Lanka.
Mason P.J., “Review of plunge pool rock scour downstream of From 1987 to 1989 he was based in Lesotho as Chief
Srisailam Dam”, International Workshop on Rock Scour due Engineer for the Muela hydropower project. Subsequently,
to High Velocity Jets, Lausanne, Switzerland; September he worked on many international projects, including the
2002.
direction of projects in Uganda, Zimbabwe and Zambia and
Whittaker, B.N., Singh, R.N. and Sun, G., , “Rock Fracture
Mechanics,” Published by Elsevier, The Netherlands; 1992. gaveexpert advice on projects ranging from Canada and
Costa Rica, to Australia and New Zealand, as well as more
projects in Africa and Europe. In 1997 he joined Black &
Veatch as Head of International Dams & Hydropower and
with Board of Management responsibilities on the 1450
MW Ghazi Barotha hydropower project in Pakistan. He has
published extensively on all aspects of dam and hydropower
design and construction.
Black & Veatch Ltd, Grosvenor House, 69 London Road,
Redhill, Surrey RH1 1LQ, UK.

Hydropower & Dams Issue Two, 2006 9