Вы находитесь на странице: 1из 22

WATER RESOURCES RESEARCH, VOL. 44, W00D03, doi:10.

1029/2007WR006525, 2008
Click
Here
for
Full
Article
Rapid drawdown in slopes and embankments
Núria M. Pinyol,1 Eduardo E. Alonso,1 and Sebastià Olivella1
Received 13 September 2007; revised 3 January 2008; accepted 14 May 2008; published 27 September 2008.

[1] The rapid drawdown condition arises when submerged slopes experience a rapid
reduction of the external water level. Classical procedures developed to determine the flow
regime within the slope and the resulting stability conditions are reviewed in the paper.
They are grouped in two classes: the ‘‘stress-based’’ undrained approach, recommended
for impervious materials, and the flow approach, which is specified for rigid pervious
materials (typically a granular soil). Field conditions often depart significantly from these
simplified cases and involve materials of different permeability and compressibility
arranged in a complex geometry. The drawdown problem is presented in the paper as a
fully coupled flow-deformation problem for saturated/unsaturated conditions. Some
fundamental concepts are first discussed in a qualitative manner and, later, explored in
more detail in synthetic examples, solved under different hypotheses, including the
classical approaches. Some design rules, which include a few fundamental parameters for
the drawdown problem, have also been solved in a rigorous manner to illustrate the
limitations of simplified procedures. A significant portion of the paper is devoted to the
discussion of a comprehensive case history. In Shira, earth dam pore pressures were
recorded at different points inside the embankment during a controlled drawdown.
Predictions of four calculation procedures (instantaneous drawdown, pure flow, coupled
flow-elastic, and coupled flow-elastoplastic, all of them for saturated/unsaturated
conditions) are compared with measured pressure records. Only the coupled analysis
provides a consistent and reasonable solution. The role of the different soil properties in
explaining the phenomena taking place during drawdown is finally discussed.
Citation: Pinyol, N. M., E. E. Alonso, and S. Olivella (2008), Rapid drawdown in slopes and embankments, Water Resour. Res., 44,
W00D03, doi:10.1029/2007WR006525.

1. Introduction associated with total or partial failure of the upstream


[2] The drawdown condition is a classical scenario in slope. International Committee on Large Dams [1980]
slope stability, which arises when totally or partially sub- and Lawrence Von Thun [1985] provide further informa-
merged slopes experience a reduction of the external water tion on drawdown-induced failures.
level. This is a common situation in riverbanks, subjected to [5] Current approaches to analyze drawdown are classi-
changing river levels. Flooding conditions are critical in this fied into two different groups: Flow methods, which should
case because river levels reach peak values and the velocity of be applied in relatively pervious slopes and undrained
decreasing water level tends to reach maximum values also. methods, which find applications in impervious soil slopes.
[3] Operation of dams requires changes in water level, Methods from the first group concentrate on the solution of
which modify the safety factor against sliding of the the flow problem in a situation that involves changes in
upstream slope of earth dams. When the reservoir level is boundary conditions and a modification of the initial free
high, hydrostatic pressures help to stabilize the slope. A surface. These methods implicitly assume that the soil
reduction of water level has two effects: a reduction of the skeleton is rigid and therefore they do not consider any
stabilizing external hydrostatic pressure and a modification modification of the initial water pressure because of the
of the internal pore water pressures. The second effect has change in total boundary stresses imposed by the draw-
traditionally received considerable attention in dam design down. Methods developed to handle this problem include
because it may lead to critical conditions of the slope. The flow net analysis [Reinius, 1954; Cedergren, 1967]; methods
subject has been approached from different perspectives, based on ad hoc hypothesis (typically Dupuit-type of
which have been largely dictated by current advances in soil assumptions) [Brahma and Harr, 1962; Stephenson, 1978];
mechanics. finite element analysis of flow in saturated soil [Desai, 1972,
[4] Sherard et al. [1963] discuss the practical implica- 1977; Cividini and Gioda, 1984] and finite element analysis
tions of rapid drawdown and a number of case histories for saturated-unsaturated flow [Neumann, 1973; Hromadka
and Guymon, 1980; Pauls et al., 1999].
[6] The second group considers only the instantaneous
1
Department of Geotechnical Engineering and Geosciences, Universitat change in pore pressure induced by an instantaneous draw-
Politècnica de Catalunya, Barcelona, Spain. down. This is the undrained case in which flow is not
considered. Key early references for this approach are
Copyright 2008 by the American Geophysical Union. works by Skempton [1954], Bishop [1954] and Morgenstern
0043-1397/08/2007WR006525$09.00

W00D03 1 of 22
W00D03 PINYOL ET AL.: RAPID DRAWDOWN W00D03

Figure 1. Drawdown scenario. (a) Hydrostatic stresses acting against the exposed slope surface.
(b) Change in applied stresses on the exposed boundaries induced by a drawdown HD.

[1963] and more recent work has been published by Lowe stress distribution in its upper part (O0O). The bottom
and Karafiath [1980], Baker et al. [1993] and Lane and horizontal surface CB experiences a uniform decrease of
Griffiths [2000]. In a recent contribution, Berilgen [2007] stress of intensity, HDg w. In addition, there is a change in
uses two commercial programs for transient/flow and de- hydraulic boundary conditions. In its new state, water
formation analysis respectively and reports a sensitivity pressures against the slope are given by the hydrostatic
analysis involving simple slope geometry. distribution O0 A0 B on the slope face and by the uniform
[7] In dam engineering practice neither one of the two water pressure value pw = (H  HD)g w on the horizontal
mentioned approaches can reliably approximate the field lower surface.
situation because compacted soils are far from being rigid [12] The change in boundary total stresses result in a new
and pure undrained conditions, even in the case of fairly stress distribution within the slope. This stress change will
impervious soils, are too conservative for common draw- induce, in general, a change in pore pressure. The sign and
down rates, which fall in the range 0.1 to 1 m/d. intensity of these pore pressures depend on the constitutive
[8] In this paper the term ‘‘coupled’’ analysis refers to the (stress-strain) behavior of the soil skeleton. An elastic soil
joint consideration of flow and stress deformation analysis. skeleton will result in a change of pore pressure equal to the
In the general formulation applied in this paper balance change in mean (octahedral) stress. If dilatancy (of positive
equations of fluid and gas and the equilibrium equations are or negative sign) is present, shear effects will generate
solved simultaneously. However, when only flow problem additional pore water pressures. Changes in total stress-
is solved the mechanical equations are not considered. We induced pore pressures are, in fact, simultaneous with the
refer to this case as the ‘‘uncoupled’’ analysis. The soil is dissipation process owing to the new unbalanced hydraulic
now assumed rigid. The undrained case is solved by means boundary conditions. A transient flow will establish. In this
of the fully coupled formulation. case it is necessary to apply a fully coupled hydromechan-
[9] As an introduction to the remaining of the paper, ical approach to take into account the simultaneous stress
consider, in qualitative terms, the nature of the drawdown and flow phenomena. However, if the soil permeability is
problem in connection with Figures 1a and 1b. large enough, pore pressures may dissipate fast enough so
[10] The position of the water level MO (height H) that the effect of stress-induced pore pressures apparently
provides the initial conditions of the slope CBO. Pore water disappears. Otherwise, in a pure ‘‘undrained’’ condition
pressures in the slope are positive below a zero pressure line (high-speed of water level changes or very low permeabil-
(pw = 0). Above this line, pore water pressures are negative ity) changes in pore pressure will be exclusively induced by
and suction is defined as s = pw. A drawdown of intensity total stress changes.
HD takes the free water to a new level M0 N0 O0 during a time [13] It is sometimes stated that in cases of rigid materials
interval tDD. the flow-based analysis is sufficiently accurate, implying
[11] This change in level implies a change in total stress that no stress-related changes in pore pressures are gener-
conditions against the slope. Initial hydrostatic stresses ated. It is clear that this is never the case in practice since it
(OAB against the slope surface; M N B C against the is required that the effective soil volumetric modulus
horizontal lower surface) change to O0 A0 B and M0 N0 B C. becomes significantly higher than the water modulus. Only
The stress difference is plotted in Figure 1b. The slope OB if the ‘‘rigid’’ material happens to be pervious and for a
is subjected to a stress relaxation of constant intensity different reason, the stress coupling seems to be absent.
(Ds = HDg w) in the lower part (BO0) and a linearly varying
2 of 22
W00D03 PINYOL ET AL.: RAPID DRAWDOWN W00D03

found. The upper points in the slope may develop negative


pore water pressures (suction).
[16] At the other extreme of the slope, point P3, the slope
face BO is far away. Because of the one-dimensional nature
of this situation, it is well-known that pore pressures in the
soil, at any depth, will follow the changing water level.
However, in order to reproduce this elementary result with a
computational tool, it is necessary to use a fully coupled
hydromechanical approach or an ‘‘undrained’’ analysis.
Otherwise, a change in water level will trigger a transient
flow condition because no information on the instantaneous
change in pore water pressure is available in an uncoupled
model.
[17] Predicting the behavior of point P2, near the toe of
the slope is more difficult. Mean and shear stresses are high
and they experience significant gradients. New pore pres-
sures generated after unloading are far from being in
equilibrium among them and with respect to the new
hydraulic head imposed at the boundary. In fact, in a fully
coupled approach, the transient process of pore pressure
dissipation has several origins. They are: the rate of water
lowering (this is a boundary condition), the heterogeneous
distribution of ‘‘instantaneous’’ pore water pressures after
drawdown and the ‘‘source’’ or ‘‘storage’’ terms provided
by both, the changing saturation in some parts of the
domain and the deformation of the soil skeleton. Figure 2b
shows that the response of point P2 in a coupled analysis
will depend on the permeability of the soil. The problem
has, however, an additional difficulty because soil stiffness,
which controls the storage term associated with changes in
effective stress, will also dictate the rate of the process.
[18] Difficulties for the development of consistent, fully
coupled hydromechanical codes for saturated/unsaturated
soils, hampered by the issue of the effective stress principle
and the development of consistent constitutive equations for
unsaturated conditions, have probably prevented a more
advanced and realistic analysis of the classical drawdown
problem. This paper relies on one of the existing complete
formulations in this regard. The solved cases use the finite
element program CODE_BRIGHT [Department of Geo-
technical Engineering and Geosciences, 2002] developed
at the Department of Geotechnical Engineering and Geo-
sciences of UPC. The code solves in a fully coupled manner
thermal, mechanical and flow (air and water) problems in
porous media. It may handle a variety of mechanical
Figure 2. Change in pore water pressures in points (a) P1, constitutive laws but the results presented here correspond
(b) P2, and (c) P3 for coupled or uncoupled analysis and either to elastic conditions or to elastoplastic constitutive
pervious or impervious fill. models (BBM [Alonso et al., 1990]; Rockfill model [Oldecop
and Alonso, 2001]). These types of models go beyond
previous known attempts to analyze drawdown effects.
[14] Consider three representative points (P1, P2 and P3) of Some relevant aspects of the formulation used in CODE_
the slope sketched in Figure 1 and their expected evolution of BRIGHT are briefly described in Appendix A.
pore pressures in qualitative terms in Figure 2. A given [19] Some of the qualitative descriptions offered above
time, tDD, in the t axis marks the end of the drawdown will be made more precise by solving the drawdown
operation. problem in a simple slope. The cases of instantaneous and
[15] A point P1, close to the upper part of the slope, will progressive drawdown will be compared. Then a review of
experience a limited change in stress due to the unloading some existing rules to estimate drawdown effects on slopes
represented in Figure 1. Therefore, no major differences will be performed. Despite the long list of developments
should be found when comparing coupled or uncoupled and publications associated with drawdown analysis, almost
analysis, even if the soil is impervious. In a pervious case, it no comparison between field measurements and calcula-
has already been argued, no differences in practice will be tions exists. For this reason, it was appropriate to perform an
analysis of an interesting published field case (the response

3 of 22
W00D03 PINYOL ET AL.: RAPID DRAWDOWN W00D03

Figure 3. Geometry of the slope.

of Glen Shira Dam against a very rapid drawdown). Model flow when water pressure reaches a positive value. All the
results and measurements will be compared. remaining boundary surfaces remain impervious.
[23] Changes in pore water pressure developed immedi-
2. Drawdown in a Simple Slope ately after the drawdown will be exclusively owing to total
stress changes. Therefore, if an uncoupled analysis is run, the
[20] The geometry of the slope analyzed is given in pore water pressures inside the slope will maintain their
Figure 3. Some calculated results are given below for points initial values immediately after drawdown. In a coupled
PA and PB (midslope and slope toe respectively). Critical analysis, the magnitude of pore pressure changes depends
failures surfaces obtained in drawdown stability analysis are on the stress-strain behavior of the soil skeleton. In the
typically close to these two points. Two cases are consid- analysis presented here several elastic soil moduli are consid-
ered, either an instantaneous drawdown or a drawdown at a ered (E = 10,000 MPa, 1000 MPa and 100 MPa). The first
rate often found in dam engineering applications: 0.5 m/d. case corresponds to a stiff material (a soft clayey rock, for
[21] An elastic constitutive law will characterize the instance). The second case is an upper limit for a very rigid
soil. The retention curve has been defined by means of a compacted and low-porosity material. The third case is a
van Genuchten model and the relative permeability varies reasonable assumption for a well-compacted well-graded soil.
with the degree of saturation following a cubic law (krel = [24] Figure 4 shows the calculated evolution of pore
S3w). A constant saturated permeability ksat = 1010 m/s was pressure after the instantaneous drawdown in point PB
also used in all calculations. This is a low value, typical of an (Figure 3). At day 1, instantaneous drawdown is simulated.
impervious material in engineering applications. In the case of uncoupled analysis, no immediate effect of the
2.1. Instantaneous Drawdown drawdown is obtained, as expected. In the coupled analysis,
[22] The initial water level in the slope is horizontal and it the instantaneous pore pressure drop depends on the com-
is located at the maximum level. The initial pore pressures pressibility of the soil skeleton.
in the soil follow a hydrostatic pattern. Drawdown is then [25] The stiffer the soil, the more limited the stress-
simulated by removing instantaneously all the water in the induced change in pore water pressure. Immediately after
reservoir. The water level (pw = 0) is maintained at the level of drawdown a dissipation process begins. The rate of pore
the toe of the slope. The inclined slope surface is provided pressure dissipation is controlled by the initial conditions
with a ‘‘seepage’’ hydraulic condition, which allows water after drawdown but also by the permeability and stiffness of

Figure 4. Pore water pressure evolution after instantaneous drawdown in the point PB (see Figure 3).
4 of 22
W00D03 PINYOL ET AL.: RAPID DRAWDOWN W00D03

Figure 5. Pore water pressure immediately after instantaneous drawdown along vertical profile through
point PB (see Figure 3).

the soil. In an uncoupled analysis the calculated dissipation [27] Figures 5 shows the pore pressure distribution along
rates are higher, because the implicit assumption is an the vertical profile through PB immediately after the instan-
infinitely rigid soil. Eventually, all cases result in the same taneous drawdown. The toe of the slope has a more
long-term solution. complex stress distribution and this is reflected in a more
[26] The coupled analysis leads systematically to lower irregular distribution of pore pressures after drawdown,
water pressures than the uncoupled (pure flow) approxima- especially in points close to the slope boundary. For a
tion during the first stages of the dissipation. This is due to compacted soil, typical of earth dam materials (E =
the effect of the initial state after drawdown, controlled by the 100 MPa), the uncoupled, pure flow analysis provides an
change in stress. However, since pressures dissipate faster extremely unrealistic answer.
the stiffer the soil, this situation changes after some time and 2.2. Progressive Drawdown (v = 0.5 m/d)
the water pressure records may cross at some particular time,
which depends on the position of the considered point in the [28] Conditions of the analysis remain unchanged except
slope. Note also that full steady state conditions were not for the drawdown rate. Figure 6 provides the calculated pore
reached at the end of the simulation period. pressures in point PA. The pure flow analysis leads to high

Figure 6. Pore water pressure evolution after progressive drawdown in the point PA (see Figure 3).
5 of 22
W00D03 PINYOL ET AL.: RAPID DRAWDOWN W00D03

spond to the three points marked in Figure 7 for the slope angle
b = cot(b) = 1.8. A set of soil properties, matching the
conditions of these three points are given in Table 1.
[33] Drawdown velocity was fixed at 0.5 m/d. A com-
mon soil porosity n = 0.3 was also selected. Three values
of saturated soil permeability, ksat = 5  108 m/s, ksat =
106 m/s and ksat = 104 m/s correspond to a relatively
impervious shell (typically a mixture of gravel, sand, silt
and some clay), a partially draining material (typically a
compacted well graded mixture) and a free draining material
(typically a gravely sand). The effective porosity for these
three cases is indicated in Table 1. The ne values selected
reflect the type of soil associated with the three cases analyzed.
The additional soil property, not considered in Figure 7, is the
water retention of the soil. The effect of a reasonable variation
Figure 7. Chart for determining height of saturation at of this property was investigated. To do so, a van Genuchten
core face. Free draining upstream shell subject to drawdown representation of the water retention curve is selected. By
[U.S. Corps of Engineers, 1970]. changing parameter p0, associated with the air entry value,
different soil retention capabilities are simulated. The second
parameter of the retention curve, l, was kept constant at the
pore pressures, if compared with the more accurate coupled value given in Table 1. All the calculations have been
case, during a first stage. Later, the higher rate of pore performed in a coupled mode, using an elastic soil modulus
pressure dissipation implied by the flow model leads to E = 100 MPa.
pressures lower than the calculated values for the coupled [34] Figures 8 –10 indicate the calculated distribution of
case. In the latter case the effect of soil stiffness can be seen water pressures, below the saturation line, for the extreme
although it has a relatively minor influence for the range of cases analyzed. These plots provide the possibility of
moduli considered. If the soil permeability is increased, the calculating DHD and the range of calculated values has
differences between coupled and uncoupled analysis reduce been indicated in the caption of each figure. These values
and eventually they provide the same answer because the are also plotted on Figure 7.
high dissipation rates mask the stress-induced response of [35] At first sight, the results in Figures 8 – 10 may look
pore pressure change. However, it is by no means easy to contrary to expectations, since the height of the phreatic
decide ‘‘a priori’’ which is the threshold permeability which surface decreases when the air entry value increases. This is
justifies the use of an uncoupled analysis. a result valid for the particular permeability selected when
comparing the effect of alternative water retention curves.
3. Some Design Rules Revisited Therefore, it makes sense only when the range of water
[29] Let us consider now the case of relatively pervious retention curves analyzed is limited since all of them should
materials. Mechanical coupling in these cases is not relevant provide essentially the same saturated permeability. The
for the reasons mentioned before and the common recom- calculated result is better explained if one considers also the
mendation is to base the analysis on the determination of distribution of degree of saturation within the slope. In
flow nets by means of numerical, analytical or graphical Figure 11, the degree of saturation along a vertical profile
procedures. However, the drawdown implies that an initially in the middle of the slope is represented for the two
saturated soil becomes progressively unsaturated. The dis- extreme cases having a common saturated permeability
tribution of pore water pressures in the slope depends now ksat = 106 m/s (case 2).
on some key properties of the unsaturated soil, and, in [36] For a given soil permeability, the amount of water to
particular, on the water retention characteristics. be drained during drawdown is similar for both cases.
[30] Some approximate procedures have been proposed Above the phreatic line (s > 0), if p0 is low, even for low
to estimate the pore pressures in a slope during drawdown. suction (close to the value of p0) the degree of saturation
Figure 7 illustrates one case, in connection with the stability decreases significantly (this is determined by the retention
analysis of earth dams [U.S. Army Corps of Engineers, curve) and the amount of drained water from the unsaturated
1970]. The idea of the chart is to facilitate a procedure to zone is higher. In the other case (higher p0), the zone above
locate the position of the free surface after drawdown. This
is achieved by providing the lowering of the seepage line at Table 1. Soil Properties for the Analysis of the Drawdown of an
the interface between the impervious dam core and the Upstream Shell
upstream shell (DHD). Value
[31] This distance is a function of the total drawdown
Symbol Case 1 Case 2 Case 3 Unit
drop (HD), the soil permeability, k, the velocity of draw-
down, v, the effective, or ‘‘drained,’’ porosity, ne, and the E 100 100 100 MPa
slope geometry, given by the slope angle, b. ksat 5108 106 104 m/s
[32] With the purpose of showing the effect of correctly p0 2 – 0.05 0.2 – 0.007 0.05 – 0.005 MPa
l 0.3 0.3 0.3 -
modeling the saturated-unsaturated transition, a few cases Sw max 1 1 1 -
directly inspired in the geometry and conditions considered in Sw min 0 0 0 -
this design plot have been calculated. The cases run corre- ne 0.09 0.18 0.24 -

6 of 22
W00D03 PINYOL ET AL.: RAPID DRAWDOWN W00D03

Figure 8. Pore pressure distribution after total drawdown for case 1: (a) p0 = 2 MPa, DHD = 23 m and
(b) p0 = 0.05 MPa, DHD = 6 m.

the phreatic line is almost saturated (Figure 11) although pore failure criterion (strength parameters f0 = 28° and c0 = 0) has
water pressures remain negative. Then the phreatic line may been considered. For the case of p0 = 0.2 MPa, a safety factor
reach a lower elevation for the same amount of drained equal to 1.35 is obtained. If p0 is reduced to 0.007 MPa, the
water. Therefore, if p0 decreases, the phreatic line (pw = s = 0) calculated safety factor is 1.48. However, this conclusion
remains at higher elevation (the saturated zone of the slope is may change if a more comprehensive description of soil
larger). strength, valid for saturated and unsaturated conditions is
[37] If only positive pore water pressures are considered introduced in the analysis, a subject that is outside the
in stability calculations, higher p0 may lead to higher safety purpose of this paper.
factors against slope failure than the case of a lower air [38] The results obtained have been included in Figure 7.
entry value. For the particular case of ksat = 106 m/s again, The largest discrepancies with the Manual recommenda-
the safety factor calculated by means of a Morgenstern- tions are obtained for low values of the index PD (more
Price method against an imposed failure surface through the impervious materials, always with respect to drawdown
middle of the slope has been calculated. A Mohr-Coulomb velocity). Recommendations are too conservative in these

Figure 9. Pore pressure distribution after total drawdown for case 2: (a) p0 = 0.2 MPa, DHD = 29 m and
(b) p0 = 0.007 MPa, DHD = 15 m.
7 of 22
W00D03 PINYOL ET AL.: RAPID DRAWDOWN W00D03

Figure 10. Pore pressure distribution after total drawdown for case 3: (a) p0 = 0.05 MPa, DHD = 47 m
and (b) p0 = 0.005 MPa, DHD = 45 m.

cases. The fact to be stressed is that the set of parameters deformations will take place and additional local sources of
included in the design procedure implied in Figure 7 is water will develop. They will modify the pore pressure
incomplete, even if couplings effects are disregarded. Only response of the slope.
in the case of very pervious materials, drawdown predictions [40] In order to show some aspects of the elastoplastic
of the chart reproduced in Figure 7 seem to be accurate. response of the soil during drawdown, the dam geometry
analyzed in the previous section was considered again. Dam
4. Coupled Elastoplastic Analysis of Drawdown materials (core and shell) are now simulated by means of
elastoplastic models. To facilitate the selection of parame-
[39] All the coupled analyses reported so far describe the ters and to reproduce, as much as possible, a real situation,
soil by means of an elastic constitutive law. In principle, the mathematical description of the two materials involved
drawdown leads to a reduction in mean stress. However, the were borrowed from previous work by the authors on
particular geometry of the problem and the nonuniformity Beliche Dam. The shell material is equivalent to the ‘‘inner
of applied boundary stresses may result in significant rockfill’’ of Beliche whereas the clay core of the example
shearing. In addition, the progressive reduction in pore analyzed here reproduces also Beliche’s core. The shell will
water pressures implies a parallel increase in effective be described by a ‘‘rockfill model’’ presented by Oldecop
confining stresses. If yielding conditions are reached, plastic and Alonso [2001] and Alonso et al. [2005]. The clay core is

Figure 11. Degree of saturation after total drawdown along the profile indicated in the plot for case 2
(ksat = 106 m/s).
8 of 22
W00D03 PINYOL ET AL.: RAPID DRAWDOWN W00D03

Table 2. Mechanical Parameters for Rockfill which will be dissipated at the expense of an increase in
Symbol Units Beliche Dam Inner Shell pore water pressure. In the case analyzed this is a minor
effect because yielding at the final drawdown stage is very
I. Elastic Behavior limited. The next discussion on a case record (Shira Dam)
E MPa 150
n - 0.3 will provide additional insight into these phenomena.

II. Plastic Behavior


li  k - 0.025
5. Glen Shira Dam Case History
l0d - 0.028 [44] Glen Shira Lower Dam is part of a pumping storage
as - 0.010 scheme in Northern Scotland. The reservoir was expected to
Mdry - 1.75
Msat - 1.30
experience fast drawdown rates and this situation prompted
ks - 0 the field experience reported by Paton and Semple [1961].
py MPa 0.01 Probably this is one of the best-documented case histories
a - 0.3 concerning the effect of drawdown on earth dams. The
maximum cross section of the dam is presented in Figure 13.
III. Initial State for Dam Model
s0 MPa 20 The 16-m-high embankment has a centered thin reinforced
p*o MPa 0.02 concrete wall. The homogeneous embankment is made of
compacted moraine soil. A rockfill shell covers the upstream
slope of the compacted moraine to increase stability. Pub-
lished grain size distributions of the moraine soil indicate a
well-graded material having a maximum size of 15 cm.
described by means of the BBM [Alonso et al., 1990]. Plasticity is not reported for this soil. It was apparently
CODE_BRIGHT handles both models. Material parameters compacted wet of optimum at an average water content
were derived from the backanalysis of large-scale laboratory w = 15%. The attained average dry density was 19.8 kN/m3,
tests and are given in Tables 2 and 3. which is a relatively high value for a granular mixture. A
[41] The analysis performed reproduces construction, friction angle f0 = 36° is reported. For the rockfill a porosity
impoundment and drawdown stages. Figure 12 shows the of n = 0.4, a dry density of 16.7 kN/m3 and a friction angle
stress-suction path followed by a representative point locat- f0 = 45° are mentioned in the paper.
ed inside of the upstream rockfill shoulder. For the simula- [45] Five porous stone piezometer disks, previously cal-
tion of dam construction, the weight of the whole dam is ibrated against mercury columns, were located in the places
applied, in a single stage, in a ramp manner. A low initial shown in Figure 13. They were connected to Bourdon
isotropic yield stress, p0*, is assumed for the compacted gauges through thin polyethylene tubing. The authors
materials. Therefore, the weight load applied induces im- conclude in their paper that the possibility of instrumental
mediately the yielding of the dam. Plastic deformations will error are ‘‘of minor order and can be neglected.’’ No
accumulate during the construction stage (step 0 – 1 in significant pore water pressures were recorded during
Figure 12). During construction, suction decreases owing construction. Positive pore pressures were measured only
to the reduction of porosity (from s = 0.5 MPa, initial value, after reservoir filling.
to the calculated value, s = 0.36 MPa). [46] A total water level drawdown of 9.1 m in 4 days was
[42] During the impoundment step 1– 2, total stresses and applied to Glen Shira Dam. This maximum drawdown was
pore pressures change. Because of saturation a compressive imposed in four stages of rapid (7.2 m/d) water lowering
strain (collapse) develops in the rockfill and additional followed by short periods of constant water level. Details of
irreversible volumetric deformation are accumulated. The changing water level in the reservoir and the measured pore
final size of the yield envelope is determined by the water pressures are indicated in the set of figures prepared to
isotropic yield stress reached at zero suction. Path 1– 2 analyze this case.
essentially implies an elastic unloading in the deviatoric
plane. Mean and deviatoric stresses reduce simultaneously,
following a path parallel to the initial construction path.
Water pressures change from negative values (soil under Table 3. Parameters for the Mechanical Models Used for the Clay
suction) to positive ones. Note also that the strength Core
parameter (M) is not constant during this path. In fact
Symbol Units Beliche Dam Clay Core
strength envelopes in the rockfill model depend on the
current suction and they are defined in terms of two extreme I. Elastic Behavior
values (Mdry and Msat) given in Table 2. E MPa 100
[43] During drawdown (at a velocity of 0.5 m/d) point P n - 0.4
(l(0)  k) - 0.020
experiences a sudden reversal in its stress path (Figure 12a). r - 0.7
Both mean and the deviatoric stresses increase again simul- b MPa1 1.2
taneously and follow a path parallel to the initial construc- pc MPa 0.02
tion path (Figure 12a). The shape of this path depends on M - 0.88
ks - 0.1
the permeability and the compressibility of the material. In a - 0.3
the case represented in Figure 12, when the end of the
drawdown is close (point 3), the current yield surface is III. Initial State for Dam Model
reached and new plastic deformations take place. The s0 MPa 0.5
plastic reduction of the porosity will release some water, p*o MPa 0.02

9 of 22
W00D03 PINYOL ET AL.: RAPID DRAWDOWN W00D03

Figure 12. Stress-suction path followed by point P (indicated in the inset) during: 0 – 1, construction;
1– 2, impoundment; and 2 – 3, drawdown. Yield surfaces reached at the end of each stage are plotted.
(a) Effective mean stress versus deviatoric stress path. (b) Effective mean stress versus suction path.

[47] Measured pore pressures have been compared with does not correspond strictly to Bishop’s method because in
calculated values in Figures 14– 18. The following hypoth- the analyses reported here the correct change in total
eses, ordered in the sense of increasing complexity, were stresses is actually applied. The soil was simulated as an
made to perform calculations: elastic material (properties are given in Table 4).
[48] 1. A pure flow analysis for saturated/unsaturated [50] 3. A coupled analysis (saturated/unsaturated), fol-
conditions that follows the changing hydraulic boundary lowing the applied upstream changes in hydrostatic pore
conditions actually applied to the upstream slope. Table 5 pressures. The soil is considered elastic (properties are
provides the hydraulic parameters used in calculations. given in Table 4).
These parameters are common to the remaining analyses [51] 4. A coupled analysis (saturated/unsaturated) follow-
described below. ing the applied upstream changes in hydrostatic pore
[49] 2. An instantaneous drawdown of the maximum pressures. The soil is considered elastoplastic following
intensity, followed by pore water pressure dissipation. This the BBM model [Alonso et al., 1990] (properties are given
is a coupled analysis, which attempts to reproduce the in Table 4). The elastic parameters of this model are taken
classical hypothesis behind the undrained methods, briefly from the previous elastic model.
described in the introduction of the paper. The procedure

10 of 22
W00D03 PINYOL ET AL.: RAPID DRAWDOWN W00D03

Figure 13. Maximum cross section of Shira Dam. The positions of piezometers 1 to 5 are indicated.

[52] The case of Shira Dam is especially interesting properties were assumed to be similar to the compacted
because the permeability of the compacted moraine fill moraine, except for the higher friction angle.
(around 108 m/s; see below) is an intermediate value [54] The dam was built in a single step. A more detailed
between impervious clay and a free draining material. representation of dam construction plays a minor role in the
One may wonder to what extent the classical hypothesis analysis of drawdown. The following ‘‘as compacted’’
for drawdown analysis (undrained or pure flow) approxi- initial suction and saturated yield stress were imposed: s0 =
mates the actual behavior. This aspect will be discussed later. 0.01 MPa and po* = 0.01 MPa. Given the low value of p*, o
[53] The following ideas have guided the selection of which reflects the isotropic yield state after compaction,
parameters. The elastic (unloading-reloading) elastic moduli dam conditions at the end of construction correspond to a
of compacted moraine and rockfill are typical of a stiff soil. normally consolidated state. The dam was then impounded
In fact, well graded granular mixtures become rather stiff until steady state conditions were reached. The presence of
when compacted. The virgin compressibility, l(0)  k, is the impervious concrete membrane results in a simple initial
approximately 1 order of magnitude higher than the elastic state: all points upstream of the concrete wall maintain
compressibility. Parameters r and b controls the shape of the hydrostatic water pressure conditions. This initial state
yield LC curve of BBM. The moraine soil is assumed to correspond to day 5 in the plots presented later.
gain limited stiffness as suction increases (parameter r). [55] The information given in the original paper provided
Also, the increase in stiffness with suction is fast for data to approximate hydraulic parameters. Two saturated
relatively low values of suction and remains fairly constant values of permeability are mentioned for compacted speci-
thereafter (parameter b). The slope of the critical state mens in the laboratory (1.6  108 m/s, when compacted at
strength line reflects the friction angles provided in the optimum water content and 1.6  107 m/s when com-
paper. Zero cohesion is assumed throughout the analysis, pacted wet of optimum). However, the dry densities reached
irrespective of suction (parameter ks). A small reference in the field (19.8 kN/m3) are higher than the optimum
stress (pc) is assumed. Associated flow conditions were laboratory B.S. compaction (19.3 kN/m3) and this leads to
assumed in both materials (parameter a = 1). Rockfill a reduction in permeability. A saturated permeability value

Figure 14. Measured and calculated pore pressures in piezometer 1 (coupled saturated/unsaturated
analysis; elastic soil properties).
11 of 22
W00D03 PINYOL ET AL.: RAPID DRAWDOWN W00D03

Figure 15. Measured and calculated pore pressures in piezometer 2 (coupled saturated/unsaturated
analysis; elastic soil properties).

ksat = 1.6  108 m/s was therefore selected for field


ever, the pore diameter is a fraction of the equivalent grain
conditions. Relative permeability varies with the degree of
size. In the sand reported by Alonso and Romero [2003] this
saturation krel = S w3 . fraction is approximately 0.25. It is probably lower in a
[56] Water retention properties for the moraine were well-graded material although this ratio was accepted to
derived following a simplified procedure, which makes derive the pore size distribution from the known average
use of the grain size distribution. Since the moraine soil is value of the grading curve for the moraine soil. The next
a granular material, capillary effects will dominate the water step is to use Laplace equation to derive the suction
retention properties. On the other hand, pore size distribu- emptying a given pore size. This leads immediately to the
tions may be approximated if grain size distributions are water retention curve. The van Genuchten expression fitted
known. An example is given, for a beach sand, by Alonso to the derived water retention curve corresponds to param-
and Romero [2003]. The idea is that the pore size distribu- eters (see also Table 5): p0 = 0.05 MPa and l = 0.2. The
tion follows the shape of the grain size distribution. How- rockfill retention curve was approximated with a signifi-

Figure 16. Measured and calculated pore pressures in piezometer 3 (coupled saturated/unsaturated
analysis; elastic soil properties).
12 of 22
W00D03 PINYOL ET AL.: RAPID DRAWDOWN W00D03

Figure 17. Measured and calculated pore pressures in piezometer 4 (coupled saturated/unsaturated
analysis; elastic soil properties).

cantly lower air entry value (lower p0) and an increased conditions. The agreement is satisfactory. The pattern of
facility to desaturate (higher l) when suction is applied. recorded pore pressures and the smoothing effect introduced
Finally, a cubic law, in terms of the degree of saturation, by the soil stiffness and permeability (specially noteworthy
defined the relative permeability. in piezometers 1 and 2) are well captured by the model. A
[57] The known history of the final stages of reservoir better agreement between measurements and calculations
filling and drawdown history of the reservoir levels is probably requires the consideration of certain field hetero-
indicated in Figures 14– 18. geneity in permeability and/or soil stiffness.
[58] Figures 14– 18 include a comparison between the [59] Paton and Semple [1961] plotted also contours of
calculated evolution of pore pressures and the corresponding piezometric head during drawdown. Two examples are
measurements of the five piezometers. The analysis corre- given in Figures 19b and 20b. They correspond to draw-
sponds to case 3 of the list of four cases described above: a down drops of 4.85 and 8.8 m. The reservoir level reaches
coupled flow-elastic deformation for saturated/unsaturated 9.15 and 5.2 m respectively (with respect to the zero

Figure 18. Measured and calculated pore pressures in piezometer 5 (coupled saturated/unsaturated
analysis; elastic soil properties).
13 of 22
W00D03 PINYOL ET AL.: RAPID DRAWDOWN W00D03

Table 4. Parameters for the Mechanical Models Used for the that the predicted pore pressures are the lowest ones if
Analysis of Shira Dam compared with the remaining methods of analysis. Calcu-
Type of Soil lated water pressures follow closely the history of reservoir
levels. The ‘‘damping’’ effect associated with soil com-
Symbol Units Moraine Rockfill pressibility is absent. When the water level is increased, at
I. Elastic Behavior the end of the drawdown test, the pure flow analysis
E MPa 100 100 indicates, against the observed behavior, a fast recovery of
n - 0.3 0.3 pore pressures within the embankment.
[63] Coupled analyses are closer to actual measurements.
II. Plastic Behavior
l(0)  k - 0.020 0.020 This is true in absolute terms but also in the trends observed
r - 0.8 0.8 when boundary conditions (changes in reservoir level) are
b MPa1 6.5 6.5 modified.
pc MPa 0.01 0.01 [64] Construction of Shira Dam leaves most of the
M - 1.4 (f0 = 35°) 1.85 (f0 = 45°)
ks - 0 0 embankment under normally consolidated conditions. This
a - 1 1 is a consequence of the low initial yield stress, p*,o adopted
in the analysis. p*o is related to the energy of compaction,
III. Initial State for Dam Model but a detailed discussion of this topic is outside the limits of
s0 MPa 0.01 0.01
po* MPa 0.01 0.01
this paper. Granular materials, and certainly rockfill, tend to
yield under low stresses after compaction. Therefore, the
accumulation of layers over a given point will induce plastic
straining. The stress paths in points relatively away from the
reference level which in this paper is placed at the dam base: slope surfaces follow K0-type of conditions. Figure 26
point 0 in Figures 19 and 20). The authors used the data indicates the stress path of points located in the position
recorded on the five piezometers to interpolate the curves of piezometers 1 and 3, respectively. Plotted in Figure 26
shown in Figures 19b and 20b. They made the hypothesis are also the yield surfaces at the end of construction. The
of a zero water pressure at the shell-rockfill interphase. The maximum size of the yield surface corresponds to these
computed distribution of heads inside the dam shell, for the construction stages. Once the dam is completed, reservoir
same amount of drawdown, is also plotted in Figures 19a impoundment leads to a reversal of the stress path, which
and 20a. The agreement is quite acceptable, although some enters into the elastic zone. Drawdown leads to a new sharp
discrepancies exist, which, in part could be attributed to the reversal in the stress path and the increase in deviatoric
limited accuracy of the interpolation made. stresses. However, the end of the drawdown path remains
[60] There was also an interest in comparing the perfor- inside the elastic locus in the two cases represented in
mance of the different methods of analysis (1. to 4.) listed Figure 26. The possibility of inducing additional plastic
above. Figures 21– 25 illustrate this comparison. Consider straining during drawdown depends on the geometry of the
first the hypothesis of instantaneous drawdown (9.5 m of dam cross section and on the constitutive behavior of the
water level drawdown, instantaneously). The calculated materials involved. Shira Dam has a stable geometry
pressure drop is indicated in Figures 21– 25 by means of because of the low upstream slope (3 to 1) and shear
a vertical bar. A (coupled) dissipation process is then stresses inside the dam are relatively small. In addition,
calculated and the progressive decay in pore pressures is the granular shell material has a high friction angle (35°).
also plotted. If compared with the actual pore pressures However, under different circumstances, plastic straining
measured at the end of the real drawdown period, the may develop during drawdown, and, in this case, pore
hypothesis of instantaneous drawdown leads obviously to pressures will probably increase because the yield point,
an extremely pessimistic and unrealistic situation. (The end located in the ‘‘wet’’ (compression) side of the yield locus
point of the instantaneous drawdown at t = 9 d is to be (see Figure 26) implies that additional local sources of local
compared with the pore pressure recorded at the end of the are available for dissipation. Note also the differences in
drawdown period at t = 12.4 d). calculated stress paths for piezometers 1 and 3 during
[61] It is also interesting to compare the results of the drawdown. Piezometer 3 is located deep inside the embank-
fully coupled analysis of the instantaneous drawdown with ment, at a high elevation and therefore pore pressure
the approximated method of analysis suggested by Skempton/
Bishop. Table 6 shows the comparison. The change in vertical
stress (DsV) has two contributions: the change in free water Table 5. Hydraulic Parameters Used for the Analysis of Shira
elevation above a given point and the decrease in total Dam
specific weight of the rockfill material covering the moraine Type of Soil
shell. An effective saturated porosity of 0.3, after drainage,
was assumed to calculate the drop in total specific weight. Symbol Units Moraine Rockfill
Bishop hypothesis leads systematically to a higher pore I. Permeability
pressure drop than the more accurate analysis. This is ksat m/s 1.6.  108 1.0.104
specially the case for the piezometers located deep inside krel - Sw3 Sw3
the fill. Discrepancies are due to the simplified stress II. Water Retention
distribution assumed in the approximate method. p0 MPa 0.05 0.01
[62] Consider now the opposite calculation method: a l - 0.2 0.4
pure flow analysis. In this case, Figures 21 –25 indicate

14 of 22
W00D03 PINYOL ET AL.: RAPID DRAWDOWN W00D03

Figure 19. Distribution of pore pressures inside the shell for a drawdown 14 to 9.15 m. (a) Computed
results (coupled analysis). (b) Interpolated values plotted by Paton and Semple [1961].

Figure 20. Distribution of pore pressures inside the shell for a drawdown 14 to 9.15 m. (a) Computed
results (coupled analysis). (b) Interpolated values plotted by Paton and Semple [1961].
15 of 22
W00D03 PINYOL ET AL.: RAPID DRAWDOWN W00D03

Figure 21. Comparison of measured pore pressures in piezometer 1 and different calculation
procedures.

changes are small: the effective mean stress remains con- [65] There is, however, an additional effect, which leads
stant and the stress path moves vertically upward. However, to a different drawdown behavior when comparing elastic
the change in deviatoric stresses is also small and the final and elastoplastic modeling approaches. If permeability is
stress point is far from reaching critical state conditions. made dependent on void ratio, the construction of the dam
Piezometer 3, on the contrary, is close to the upstream shell, will lead to lower values of permeability (distributed in a
at a lower elevation. Changes in pore pressure and devia- heterogeneous manner). If the dam compacted material
toric stress are large in this position and the stress path yields during construction, plastic volumetric compaction
moves approximately parallel to the initial construction path will add to the elastic strains. In addition, collapse phenom-
and approaches yielding conditions in compression. ena upon impounding will reduce further the porosity.

Figure 22. Comparison of measured pore pressures in piezometer 2 and different calculation
procedures.

16 of 22
W00D03 PINYOL ET AL.: RAPID DRAWDOWN W00D03

Figure 23. Comparison of measured pore pressures in piezometer 3 and different calculation
procedures.

These effect has been also explored in the case of Shira that the classical methods of analysis are far from explaining
Dam. Permeability was made dependent on void ratio, e, the recorded behavior. The ‘‘instantaneous’’ or undrained
following a Kozeny type of relationship (permeability method is conservative, but very unrealistic. A fully coupled
depends on e3/(1 + e)). The calculated records of pore analysis of the instantaneous drawdown results in higher
pressure evolution during drawdown are also shown in pressure drops than the classical Bishop proposal. At the
Figures 21– 25. The reduction in permeability, if compared opposite extreme, the pure flow analysis leads to a system-
with the coupled elastic case, leads to a systematic increase atic and unsafe underestimation of fill pressures during
in pore pressures. The agreement with measurements is now drawdown. Coupled analysis captures well the actual meas-
better in some piezometers (1, 3 and 4). urements. In the case of Shira Dam, plastification during
[66] The conclusion, for the particular embankment ma- drawdown was probably nonexistent, and the simpler elastic
terial of Shira Dam and its overall geometry and design, is approach provides a good approximation to recorded pore

Figure 24. Comparison of measured pore pressures in piezometer 4 and different calculation
procedures.
17 of 22
W00D03 PINYOL ET AL.: RAPID DRAWDOWN W00D03

Figure 25. Comparison of measured pore pressures in piezometer 5 and different calculation
procedures.

water pressures. However, the full elastoplastic simulation stiffness and water compression modulus. In most situa-
offers a better understanding of the phenomena taking place tions, this ratio is small and the influence of soil stiffness is
during construction and impounding. This is shown in the negligible. This implies a maximum response of the satu-
stress paths calculated, in the occurrence of yielding during rated material to stress changes. Without this coupling, the
construction, and in the effect of permeability reduction on initial pore pressures do not change during fast unloading
the drawdown response. (as an illustration, pure flow models are unable to detect that
all points in the porous media instantaneously feel a change
6. Conclusions in water level in a large submerged area).
[69] Permeability and soil stiffness controls coupled flow.
[67] Pore water pressures in an initially submerged slope The uncoupled analysis implicitly assumes a rigid soil and
and later subjected to drawdown depend on several soil therefore it leads to a maximum dissipation rate. Both
parameters and ‘‘external’’ conditions: soil permeability effects (the initial change in pore pressure and the subse-
(saturated and unsaturated), soil water retention properties, quent dissipation) should be jointly considered for a better
mechanical soil constitutive behavior, rate of water level understanding of the evolution of pore pressures. In addi-
lowering and boundary conditions. The paper stresses that a tion, the rate of change of boundary conditions is a key
proper consideration of these aspects is only possible if a information to interpret the results. No simple rules can be
fully coupled flow – mechanical analysis, valid for saturated given to estimate the pore pressures in the slope. This is
and unsaturated conditions is employed. A review of the even more certain if due consideration is given to the
literature on the subject reveals that the published proce- unsaturated flow regime. In this regard, some design rules
dures refer usually to limiting cases (impervious or rigid for earth dam stability calculations, which provide the
materials), which prevent often its use in real problems and position of the phreatic surface in relatively ‘‘free draining’’
make it difficult to judge the degree of conservatism, if any,
introduced.
[68] Leaving apart for the moment the issue of the
transition from saturated to unsaturated conditions which
takes place during drawdown, there are two fundamental Table 6. Shira Dam Instantaneous Drawdown: Comparison of
mechanisms controlling the resulting pore water pressure: Coupled and Simplified (Bishop) Analysis
the change in pore pressure induced by boundary changes in Bishop
Calculated
stress and the new flow regime generated. Both of them Instantaneous Hypothesis
require a coupled analysis for a proper interpretation and Initial Pressure Pressure Drop (Du = B. DsV)
consistency of results. In particular, pure flow models are (Horizontal Water (CODE_BRIGHT) B=1
unable to consider the initial changes in pore pressure Piezometer Table) (kPa) (kPa) (kPa)
associated with stress unloading. The intensity of pore 1 96 42 42
pressure changes induced by a stress modification is con- 2 106 22 12
trolled by the soil mechanical constitutive equation. In a 3 67 10 1
simplified situation, under elastic hypothesis for the soil 4 56 17 12
5 23 6 0
skeleton, the pore pressure depends on the ratio of soil bulk
18 of 22
W00D03 PINYOL ET AL.: RAPID DRAWDOWN W00D03

Figure 26. Stress path in a (p0, q0) triaxial space during 0 – 1, construction; 1 – 2, impoundment; and 2 –
3, drawdown of points located in the position of (a) piezometer 1 and (b) piezometer 3. Also plotted are
the yield surfaces at the end of construction.

materials, have been reviewed with the help of the fully the phreatic line does not provide enough information to
coupled, complete formulation used in this paper. An calculate safety factors against slope failure if due consid-
interesting result is that, other parameters of the problem eration is given to the strength for positive suctions, above
being equal, the average height of the phreatic line the zero-suction surface. Therefore, methods for drawdown
increases as the air entry value of the water retention curve analysis, which concentrate on the determination of the
decreases. This is a paradoxical result at first sight, but it position of the phreatic line, using formulations for satu-
may be explained if one considers the amount of drained rated flow, may lead to significant errors. The evaluation
water induced by the drawdown. In addition, the position of of the design chart for dams subjected to drawdown,

19 of 22
W00D03 PINYOL ET AL.: RAPID DRAWDOWN W00D03

performed in one of the sections of this paper, is a good of the reduced formulation for hydromechanical problems is
example. presented in this appendix.
[70] The elastoplastic analysis performed on a synthetic [75] Mass balance of solid present in the medium is
example (an earth and rockfill dam whose parameters written as
correspond to a real case: Beliche Dam) has provided
additional information on the stress paths that develop @
ðr ð1  nÞÞ þ r  ðjs Þ ¼ 0; ðA1Þ
inside the dam. Points inside the embankment, except for @t s
shallow positions, follow a K0-type of stress path during
construction. Impounding and drawdown imply strong where rs is the mass of solid per unit volume of solid and js
stress reversals. Drawdown, in particular, is characterized is the flux of solid. From this equation, an expression for
by a parallel increase in effective mean stress and deviatoric porosity variation can be obtained if the flux of solid is
stress. Yield conditions may be approached although it is written as the velocity of the solid multiplied by volumetric
believed that the drawdown paths tend to remain in the fraction occupied by the solid phase and the density, i.e.,
elastic domain. js = rs (1  f) du
dt :
[71] A well-documented case history (Shira Dam) was
analyzed to provide further insight into the drawdown Ds n ð1  nÞ Ds rs du
¼ þ ð1  nÞr  : ðA2Þ
problem. The case is very interesting because the soil Dt rs Dt dt
involved (a compacted moraine) has an intermediate per-
meability between impervious clays and free draining The material derivative with respect to the solid is defined
granular materials. It should be added that materials with as
this intermediate permeability are very common in dam
engineering. Therefore, the two classical procedures to Ds ð
Þ @ du
analyze drawdown effects (undrained analysis for clays ¼ þ  rð
Þ: ðA3Þ
Dt @t dt
and pure flow for granular materials) will meet difficulties.
In fact, these two methods proved quite unrealistic when Equation (A2) expresses the variation of porosity caused by
compared with actual records of pore water pressures in volumetric deformation and solid density variation.
different points of the dam. In particular, the pure flow [76] In the formulation required for the analyses in this
(uncoupled) analysis leads to faster dissipation of pore paper, water component and liquid phase are the same. The
pressures and this is an unsafe result in terms of stability total mass balance of water is expressed as
calculations. The fully coupled analysis (elastic or elasto-
plastic) provides consistent results. @
[72] The elastoplastic analysis allows a proper consider- ðr Sw nÞ þ r  ðjw Þ ¼ f w ; ðA4Þ
@t w
ation of the entire history of dam construction, impound-
ment and drawdown. Since embankment dams experience where Sw is the degree of saturation of water, rw is the water
significant yielding during construction, this is an important density, jw is the flux of water, and f w is an external supply
consideration. It has been shown also that the reduction in of water. Water flux is a combination of a Darcy flux and an
permeability associated with material volumetric compres- advection caused by the solid motion
sion has a significant effect on the subsequent drawdown
behavior: it leads to higher pore water pressures being  
@ du
maintained inside the slope. ðrw Sw nÞ þ r  rw qw þ rw nSx ¼ f w: ðA5Þ
@t dt

Appendix A: Coupled Flow-Deformation The use of the material derivative leads to


Formulation for Saturated/Unsaturated Conditions  
[73] This appendix summarizes the balance equations Ds ðrw Sw Þ Ds n du
n þ rw Sw þ rw Sw nr  þ r  ðrw qw Þ ¼ f w :
required for coupled flow-deformation for saturated and Dt Dt dt
unsaturated conditions. In what follows, it will be consid- ðA6Þ
ered that the state variables (unknowns) are: solid displace-
ments, u (three spatial directions) and liquid pressure, Pl. The mass balance of solid is introduced in the mass balance
Balance of momentum for the medium as a whole is of water to obtain, after some algebra,
reduced to the equation of stress equilibrium together with a
 
mechanical constitutive model which relates stresses with nSw Ds rw Ds Sw ð1  nÞ Ds rs du
strains. Strains are defined in terms of displacements. Small þn þ Sw þ Sw r 
rw Dt Dt rs Dt dt
strains and small strain rates for solid deformation are 1 1 w
assumed. Advective terms due to solid displacement are þ r  ðr w q w Þ ¼ f : ðA7Þ
rw rw
neglected once the formulation is written in terms of
material derivatives (in fact, material derivatives are
This equation has four storage terms, related to (1) water
approximated as Eulerian time derivatives). In this way,
compressibility since r1 dr w
= K1w is the volumetric compres-
volumetric strain is properly considered. w dpw

[74] The governing equations for nonisothermal multi- sibility of water, (2) retention curve storativity since dS dpw is
w

phase flow of water and gas through porous deformable obtained from the retention curve, (3) solid compressibility
saline media have been presented by Olivella et al. [1994]. since r1 dr s
= K1s is the compressibility of the soil particles,
s dp
A detailed derivation is given there, and only a description and (4) soil skeleton compressibility since the divergence of
20 of 22
W00D03 PINYOL ET AL.: RAPID DRAWDOWN W00D03

solid velocity can0 be transformed into D  (du d


dt ) = dt (r  u) =
where suction can be defined as s = max (pa  pw, 0) and
dev ðs ;pw Þ dev ds
dev dev
dt , and dt = dt
dev dpw
= ds dt + dpw dt is the volumetric strain effective or net stress as s0 = s + max (pa, pw) . This is valid
rate that should be calculated with a corresponding for saturated and unsaturated conditions, and considers
constitutive model for the soil. The mechanical model stresses in compression as negatives. The model parameters
may include effective or net stress terms (volumetric or are included in D which is the stiffness tensor (6  6) or
deviatoric) or suction terms. Effective or net stress has to be constitutive matrix for changes in net or effective stress and
considered here as the total stress minus the water pressure h which is the constitutive vector for changes in suction.
or the air pressure, respectively, for saturated or unsaturated Both are nonlinear functions.
conditions. The final terms are left as a function of total [81] Note that, the derivatives of volumetric deformation
stress. needed in (A9) can be obtained in the following way @e@s =
v
@ev @s0 @ev @ev @s0 @ev @s
[77 ] The relative importance of the different terms @s0 @s and @pw = @s0 @pw + @s @pw . By comparison with
depends on the conditions of the soil. For instance, for equation (A10), the following terms are obtained: @e @s0 =
v

saturated conditions the second term disappears. When the t 1 @ev t 1


m D and @s = m D h.
compressibility of the skeleton is large, the compressibility [82] The nonlinear elastic part of the BBM model, gives
of the particles is negligible. The compressibility of the the following volumetric deformation:
water may be negligible in some cases but it is not possible
to neglect it in general for hard soils. k dp0 ks ds
[78] The final objective is to find the unknowns from the dev ¼ þ ; ðA11Þ
1 þ e p0 1 þ e s þ 0:1
governing equations. Therefore, the dependent variables
will have to be related to the unknowns in some way. Doing where, e is void ratio, k and ks are material parameters, p0
this in the last equation leads to is the mean net or effective stress which is defined as p0 =
  (sx + sy + sz)/3 + max (pa, pw), and s is suction.
nSw dpw dSw dpw @ev ds @ev dpw 1
þn þ Sw þ þ r
Kw dt dpw dt @s dt @pw dt rw
Notation
 ðrw qw Þ ¼ 0
nSw dpw dSw dpw @ev ds @ev dpw 1 c0 effective cohesion, MPa.
þn þ Sw þ Sw ¼  r  ðrw qw Þ; D stiffness tensor, MPa.
Kw dt dpw dt @s dt @pw dt rw
e void ratio.
ðA8Þ E Young modulus, MPa.
fw external supply of water, kg/m2s.
where the compressibility of the solid particles has been h constitutive vector for changes in suction, MPa.
neglected and the source/sink is assumed to be 0. The H height of reservoir, m.
material derivatives have been approximated as Eulerian. HD drawdown drop, m.
[79] This equation permits to calculate the pressure de- js flux of solid, kg/m2s.
velopment for a soil subjected to changes in total stress in jw flux of water, kg/m2s.
the following way: k permeability, m/s.
@ev dt K0 coefficient of earth pressure at rest, MPa.
Sw ds  r  ðrw qw Þ krel relative permeability.
@s rw
dpw ¼ : ðA9Þ Ks bulk modulus of the solid particles, MPa.
nSw dSw @ev
þn þ Sw ks parameter that controls the increase in cohesion
Kw dpw @pw
with suction (BBM).
Deformation is assumed negative in compression from these ksat saturated permeability, m/s.
equations, and stress is also negative in compression. This Kw bulk modulus of water, MPa.
implies that ds is negative in compression (loading) and M slope of critical state strength line, BBM model.
produces positive pressure increments. Note that the general Mdry slope of critical state strength envelope for
stress tensor is maintained because volumetric deformations dry conditions (Rockfill model).
can be caused by any stress variation (not only isotropic), Msat slope of critical state strength envelope for
and that depends on the response of the soil. For instance, saturated conditions (Rockfill model).
dilatancy is a volumetric expansion induced by shear. n porosity.
[80] In equation (A9), the volumetric deformation deriv- ne effective porosity.
atives @e @ev
@s and @pw should be calculated with and appropriate
v p net mean stress, MPa.
constitutive model. These are volumetric deformation terms p0 mean effective stress, MPa.
and can be obtained from a model for unsaturated soils such pa air pressure, MPa.
as the elastoplastic model BBM [Alonso et al., 1990]. A pc reference stress, BBM model, MPa.
general equation, including the effect of effective or net p0 parameter of water retention curve of
stresses and the suction, is written as van Genuchten model, MPa.
p0* initial mean yield stress, MPa.
0
ds ¼ Dde þ hds;
Pl liquid pressure, MPa.
de ¼ D1 ds0  D1 hds;
pw pore water pressure, MPa.
ðA10Þ py threshold yield mean stress for the onset of
dev ¼ mt de ¼ mt D1 ds0  mt D1 hds;
mt ¼ ½1 1 1 0 0 0 ; clastic phenomena (Rockfill model), MPa.
qw Darcy flux, m/s.

21 of 22
W00D03 PINYOL ET AL.: RAPID DRAWDOWN W00D03

r parameter that establishes the minimum value Bishop, A. W. (1954), The use of pore pressure coefficients in practice,
Geotechnique, 4, 148 – 152.
of the compressibility coefficient for high Brahma, S. P., and M. E. Harr (1962), Transient development of the free
values of suction (BBM). surface in a homogeneous earth dam, Geotechnique, 12, 283 – 302.
s suction, MPa. Cedergren, H. R. (1967), Seepage, Drainage and Flow Nets, John Wiley,
s0 initial suction, MPa. Hoboken, N. J.
Sw degree of saturation. Cividini, A., and G. Gioda (1984), Approximate F. E. analysis of seepage
with a free surface, Int. J. Numer. Anal. Methods Geomech., 8, 549 – 566,
Sw max maximum degree of saturation. doi:10.1002/nag.1610080605.
Sw min minimum degree of saturation. Desai, C. S. (1972), Seepage analysis of earth banks under drawdown,
t time, days. J. Soil Mech. Found. Div. Am. Soc. Civ. Eng., SM11, 1143 – 1162.
tDD time of drawdown, days. Desai, C. S. (1977), Drawdown analysis of slopes by numerical method,
J. Soil Mech. Found. Div. Am. Soc. Civ. Eng., GT7, 667 – 676.
u solid displacement, m. Department of Geotechnical Engineering and Geosciences (2002),
v velocity of drawdown, m/d. CODE_BRIGHT. A 3-D program for thermo-hydro-mechanical analysis
a parameter that defines the nonassociativeness in geological media. USER’S GUIDE, Cent. Int. de Métodos Numéricos
of plastic potential. en Ing., Univ. Politécnica de Catalunya, Barcelona, Spain.
as parameter to describe the rate of change of Hromadka, T. V., and G. L. Guymon (1980), Some effects of linearizing the
unsaturated soil moisture transfer diffusivity model, Water Resour. Res.,
clastic compressibility with total suction 16(4), 643 – 650, doi:10.1029/WR016i004p00643.
(Rockfill model). International Committee on Large Dams (1980), Deterioration of Dams
b parameter that controls the rate of increase in and Reservoirs. Examples and Their Analysis, Balkema, Rotterdam,
stiffness with suction, MPa1. Netherlands.
Lane, P. A., and D. V. Griffiths (2000), Assessment of stability of slopes
DHD lowering of the seepage line at the interface under drawdown conditions, J. Geotech. Geoenviron. Eng., 126(5),
between the dam core and upstream shell, m. 443 – 450, doi:10.1061/(ASCE)1090-0241(2000)126:5(443).
DsV changes in vertical net stress, MPa. Lawrence Von Thun, J. (1985), San Luis Dam upstream slide, Int. Conf. on
v volumetric strain. Soil Mech. Found. Eng., 11, 2593 – 2598.
f0 effective frictional angle, deg. Lowe, J., and L. Karafiath (1980), Effect of anisotropic consolidation on the
undrained shear strength of compacted clays, paper presented at Research
g w water specific weight, MN/m3. Conference on Shear Strength of Cohesive Soils, Am. Soc. of Civ. Eng.,
l parameter of water retention curve, Boulder, Colo.
van Genuchten model. Morgenstern, N. R. (1963), Stability charts for earth slopes during rapid
l(0)  k virgin plastic compressibility for saturated drawdown, Geotechnique, 13, 121 – 131.
Neumann, S. P. (1973), Saturated-unsaturated seepage by finite elements,
conditions. J. Hydraul. Div. Am. Soc. Civ. Eng., 99, 2233 – 2250.
li  k virgin plastic instantaneous compressibility Oldecop, L., and E. E. Alonso (2001), A model for rockfill compressibility,
(Rockfill model). Geotechnique, 51, 127 – 139, doi:10.1680/geot.51.2.127.40283.
ld0 virgin elastic compressibility for saturated Olivella, S., J. Carrera, A. Gens, and E. E. Alonso (1994), Nonisothermal
multiphase flow of brine and gas through saline media, Transp. Porous
conditions (Rockfill model). Media, 15, 271 – 293, doi:10.1007/BF00613282.
n Poisson’s ratio. Paton, J., and N. G. Semple (1961), Investigation of the stability of an earth
rs solid density, kg/m3. dam subjected to rapid drawdown including details of pore pressure
rw water density, kg/m3. recorded during a controlled drawdown test, in Pore Pressure and Suc-
s net stress tensor, MPa. tion in Soils, pp. 85 – 90, Butterworth, London.
Pauls, G. J., E. Karlsauer, E. A. Christiansen, and R. A. Wigder (1999), A
s0 effective stress tensor, MPa. transient analysis of slope stability following drawdown after flooding of
sx, sy, sz stresses in direction x, y and z. highly plastic clay, Can. Geotech. J., 36, 1151 – 1171, doi:10.1139/cgj-
36-6-1151.
Reinius, E. (1954), The stability of the slopes of earth dams, Geotechnique,
References 5, 181 – 189.
Alonso, E. E., and E. Romero (2003), Collapse behaviour of sand, paper Sherard, J. L., R. J. Woodward, S. F. Gizienski, and W. A. Clevenger
presented at 2nd Asian Conference on Unsaturated Soils, Org. Comm. of (1963), Earth and Earth-Rock Dams, John Wiley, Hoboken, N. J.
UNSAT-ASIA 2003, Osaka, Japan. Skempton, A. W. (1954), The pore pressure coefficients A and B, Geotech-
Alonso, E., A. Gens, and A. Josa (1990), A constitutive model for partially nique, 4, 143 – 147.
saturated soil, Geotechnique, 40, 405 – 430. Stephenson, D. (1978), Drawdown in embankments, Geotechnique, 28,
Alonso, E. E., S. Olivella, and N. M. Pinyol (2005), A review of Beliche 273 – 280.
Dam, Geotechnique, 55, 267 – 285, doi:10.1680/geot.55.4.267.65493. U.S. Army Corps of Engineers (1970), Engineering and design stability of
Baker, R., S. Rydman, and M. Talesnick (1993), Slope stability analysis for earth and rock-fill dams, EM 1110-2-1902, Washington, D. C.
undrained loading conditions, Int. J. Numer. Anal. Methods Geomech.,
17, 15 – 43, doi:10.1002/nag.1610170103.
Berilgen, M. (2007), Investigation of stability of slopes under drawdown 

conditions, Comput. Geotech., 34, 81 – 91, doi:10.1016/j.compgeo. E. E. Alonso, S. Olivella, and N. M. Pinyol, Department of Geotechnical
2006.10.004. Engineering and Geosciences, Edificio D-2, Campus Nord, UPC, E-08034
Barcelona, Spain. (eduardo.alonso@upc.edu)

22 of 22