EFFECTS OF RESERVOIR FLUCTUATIONS ON SETTLEMENT PATTERN OF JATILUHUR EARTH-CORE

ROCKFILL DAM

By: Engr. M. Ibrahim Samoon*

ABSTRACT: Applying basic soil mechanics concepts a method has been developed to compute the settlement-time history of an earth core rockfill dam. The method utilises consolidation characteristics of core material and effective stress changes in core due to reservoir variations. The method was applied to explain the settlement mechanism of the 108 m high Jatiluhur dam, located in West Java, Indonesia which continues to settle even after 33 years of construction. The settlement rate of the dam is accelerated during events of major reservoir drawdown. The results of consolidation tests on undisturbed samples from core and water level records of Jatiluhur reservoir spanning three decades were utilized to calculate the settlement-time history. A very close match was observed between the observed and calculated values. The method even explains the minor as well as major kinks in the observed time-settlement curve when reservoir level drops.

INTRODUCTION

The crest of the 108m high Jatiluhur earthcore rockfill dam located in West Java, Indonesia, continues to settle even 33 years after its construction. The observation records have shown that, while the downstream portion of the crest has practically stopped settling, the portion of the crest between the upstream edge and the interface of the core with a rockfill shell continues to settle at an appreciable rate. The settlement rate is accelerated whenever there is a major reservoir drawdown. The amount of accelerated settlement depends on the amount of reservoir drawdown. Significant settlements were observed during major drawdown events of 1972, 1982 and recently in 1997-98. The differential settlement in all these events has resulted in the development of longitudinal vertical cracks at

the contact between the core and the downstream rockfill over a length of about 300 to 400m.

This paper presents a review of the recorded movement data and probable explanation of the dam behavior during variations in reservoir levels.

FEATURES AND PAST HISTORY OF THE DAM

The Jatiluhur dam, constructed in 1965, in West Java Indonesia on the river Citarum, is a rockfill dam with a sloping clay core resting on a 200m thick claystone bed. The dam is 108m high at its deepest section, 10m wide and 1200m long at its crest. A typical cross section of the dam is shown in Figure 1.

3

22

221

4

4

5

3

Normal water level 107 El. 115.01. Impervious fill2. Rockfill3. Clay core4. Random berms5. Filters and transitions

5

Sanstone/ claystone foundation

Figure 1. Main dam - Typical cross section

_______________________________________________________________________________________* Chief Geotechnical Engineer, ACE Ltd, Pakistan and a member of the teams of Safety

Inspections and Design of Remedial Works for the dam and dykes of Jatiluhur Dam Project.

ACE/SEJD/Tech. Rep.

1

The central core is constructed of high plasticity clay (average LL = 65 %, PI = 38 %), derived from in situ weathering of claystone, rolled in layers of 0.2m thickness to an average moisture content 2% wet of the optimum. The transition zones between core and rockfill shells comprise 4m thick random filter upstream and zones of fine and coarse filter each 3m thick downstream. The rockfill of andesite stones was constructed in 5 to 10m high dumped sluiced lifts. The stabilizing berm is made of predominantly broken claystone compacted in 0.6m thick layers. A further description of the design and construction of the project is given in Bohn and Hamon (1967).

Wide longitudinal cracks, about 0.5km long, appeared at the dam crest during the construction

stage in early 1965 when the dam was built to El. 103m (Sherard, 1973).

The severity of cracks becomes appreciable and the amount of differential settlement becomes fairly large during events of major drawdown. Two such events were recorded in 1972 and 1982 when the reservoir dropped to El. 78.1m and El. 77.6m respectively. Recently a similar event was observed in 1997-98 when remedial works were under progress and, due to an unusual drought the reservoir level dropped to El. 76.6m in January 1998. The vertical longitudinal crack developed on the crest in the reach between St. 60R to 250 L at a distance varying from 1 to 2m from the downstream edge. The maximum measured sag in the crest was about 19cm at the highest section of the embankment.

1965 1970 1975 1980 1985 1990 1995 2000-2000

-1500

-1000

-500

0

50

65

80

95

110

Year

Set

tlem

ent (

mm

)

Res

ervo

ir Le

vel (

m )

U/S edge settlement-C7b

D/S edge settlement-C7bReservoir Level

D/S edge settlement-C8

D/S edge settlement-C10

D/S edge settlement-C5

U/S edge settlement-C8

U/S edge settlement-C5

U/S edge settlement-C10

Completion of Dam9 - Sep - 1965 Repair Crest Road

February 1992

Remedial WorkJune 1996 to February 1998

Chainages : C5 - 225L C7b - 75 L C8 - 25L C10 - 75RSection height (m ) : C5 - 80m C7b - 102m C8 - 92m C10 - 68m

Figure 2. Settlement of upstream and downstream crest-markers

ACE/SEJD/Tech. Rep.

2

MOVEMENT OBSERVATION AT THE DAM The Jatiluhur main dam has 41 surface movement markers located at the crest, upstream shoulder and downstream slope to observe vertical movements. A number of selected markers are also observed for horizontal movements. All these points are observed on more or less a monthly basis since their installation. POJ, the project authority, maintains comprehensive movement data both in tabulated and graphic forms.

The settlement data of four selected markers C5, C7b, C8 and C10 installed on the crest at St. 225L, 75L, 25L and 75R respectively are shown graphically in Figure 2.

It can be noted that settlement of the downstream edge markers has a diminishing trend. They also show minor effects of major drawdown events. The upstream edge markers show that settlement rates are decreasing with the passage of time and the amount of settlement during a drawdown event decreases generally compared with the

preceding event. Moreover, after every rapid settlement event there is an appreciable drop in settlement rate for normal reservoir operation.

The settlement records show that the amount of settlement is directly proportional to the embankment height on a log-log scale, as is evident from the graph shown in Figure 3.

The horizontal as well as vertical movements are observed to be maximum at the highest section of the dam which is located at a distance of 75m left of the powerhouse-tailrace axis. The crest settlements at this section are as under:

Total settlement:

Upstream edge: 1836mmDownstream edge: 349mm

Settlements during events of major drawdown:

During 1972 drawdown : 97mmDuring 1982 drawdown : 81mm

During 1997-98 drawdown: 96mm

10 1001

10

100

1000

10000

Section Height ( m )

Set

tlem

ent (

mm

)

D/S edge Settlement U/S edge Settlement

Correlation f orU/Sedge settlement

Correlation f orD/Sedge settlement

Figure 3. Settlement versus embankment height (log-log graph)

ACE/SEJD/Tech. Rep.

3

THE MECHANISM OF SETTLEMENT

The main parts of Jatiluhur dam are clay core and rockfill. Both of these, by virtue of their nature, behave differently under various loading conditions.

The rockfills are made up of solid stones with many voids and shapes. The dumped rockfills with high lift thickness, as at Jatiluhur dam, have a very loose structure. The contacts between stones consist of points and edges. On application of load, crushing of points and edges occurs which permits movement of the overlying and adjacent mass in the general direction of applied forces. The response of the rockfill to load application is generally ‘immediate’. There may be some creep action in some rockfills due to ageing effects, but it is generally at negligible rates. That is the reason why, in concrete face rockfill dams, in which the main body of the dam is only rockfill, about 85 % of the settlement occurs in the first year of filling.

On the other hand in earth cores, especially constructed of clay, compacted wet of the optimum moisture content, a major portion of the applied load is transferred as pore pressures. Due to low permeability of the material these pressures dissipate at a very slow rate and result in long term consolidation or creep settlements. These two distinct behaviors of clay and rockfill materials dictate the pattern of strain and stresses within the dam body and its long term behavior during the operational life of the dam.

The forces which can induce movements in the dam body are embankment body loads, forces on first reservoir filling, the effect of reservoir level variations and any external load such as shaking by earthquake.

Under the embankment body loads, the process of settlement is rapid in rockfill and is completed to large extent during construction. But in clay core material the applied body loads are partially absorbed in strains and partially in development of pore pressures. If, for any reason, the construction is halted for a considerable period the core will settle on dissipation of pore pressures, resulting in differential settlement between the core and rockfill. The development of cracks and differential settlement observed at Jatiluhur Dam in January 1965 when the construction of dam was stopped for a period of

6 months is the manifestation of this phenomenon.

The effect of reservoir filling on an earth core rockfill dam are multi directional :

The water load on the upstream foundation causes upstream settlements.

The buoyant uplift forces in the upstream shell cause upward movements within the dam.

The upstream rockfill gets a ‘lubrication’ effect on wetting and rock pieces adjust themselves, causing settlement.

The movements observed on Jatiluhur Dam at the dam crest on first reservoir filling are reported to be of the following order :

Downstream edge: Horizontal towards d/s: 580mm (64% of the

total to date)Vertical settlement: 177mm (51% of the

total to date)

Upstream edge : Vertical settlement: 430mm (23% of the

total to date)

It may be noted that the vertical settlement of the downstream edge, which depends solely on rockfill behavior, after first reservoir filling was about 51% of that recorded to date (after 32 years) while for the upstream edge which rests directly on the core, the amount of settlement was about 23% of that observed to date.

As already discussed, the rockfill portion of an embankment generally adjusts itself within the first few years of loading and normally does not indicate any movement due to reservoir fluctuations at later stages. On the other hand, the clay core portion will settle with its controlling rate of dissipation of pore pressures. If a reservoir is kept at a constant level for a long duration, the settlement pattern of the core will resemble the time-settlement curve of an oedometer test carried out on clay samples in the laboratory. This curve represents a typical hyperbola with diminishing rate of settlement with time. When the reservoir level drops below the level at which the core has attained a state of

ACE/SEJD/Tech. Rep.

4

steady seepage, the following conditions will develop :

· The effective stresses within the core will increase.· The dissipation of pore pressures will accelerate.

· The weight of rockfill overlying the core will increase from its submerged weight to its bulk weight for the portion above water level.

These conditions will accelerate the rate of settlement. The amount of settlement will depend on the amount of reservoir drawdown and the time while the reservoir remained at lower levels. As the general behavior of settlement in clay suggests, when the reservoir rises again to the previous steady state level, the effective

stresses will reduce to the original values, but the rate of settlement will follow a flatter curve.

The observed settlement of the upstream edge of Jatiluhur dam crest, which reflects the deformations of the clay core, represents the above explained behavior.

PROCEDURE OF PREDICTING SETTLEMENT DUE TO RESERVOIR VARIATION

If time-settlement curves are developed for an embankment dam under loading conditions at various reservoir levels, the settlement behavior due to reservoir fluctuation can be predicted by the procedure given in Figure 4 and explained below :

Figure 4. Method of estimating settlement with reservoir variations

ACE/SEJD/Tech. Rep.

5

- Suppose at any given time to the settlement already attained is So when the reservoir is at elevation Ro. If the reservoir level changes to R1 in time increment t1, the settlement during this period can be calculated by extending a horizontal line from point 0 to meet the time-settlement curve for reservoir level at average of Ro and R1 at point 0’.

- The amount of settlement will be the vertical intercept traversed on this line in time t1.

- Similarly when, in the next step, the reservoir level changes from R1 to R2, between time t1 and t2 ( interval t2 = t2

– t1 ), the settlement is calculated by extending a horizontal line from point 1 to 1” on the time-settlement curve for reservoir level equal to average of R1

and R2. The amount of settlement will be the vertical intercept on this curve in t2 time. In this way a settlement curve can be developed for successive reservoir changes.

From the curve, developed by the above procedure, it shall be noted that, at any particular time, the settlement attained should be on the time-settlement curve of a particular reservoir loading. If the actual reservoir level remains above that particular reservoir level then the rate of settlement with time will be slower. But when the reservoir level drops below that particular level, the settlement rate will be accelerated.

The above explained method was applied to estimate settlements at Jatiluhur Dam and to compare them with the observed settlements. The method was also applied to predict the quantum and rate of future settlements.

DEVELOPMENT OF TIME-SETTLEMENT CURVES FOR VARIOUS RESERVOIR LEVELS

The accurate quantification and prediction of the settlement of a complex structure, in which variation in materials can be expected in vertical as well as in horizontal directions, is rather a difficult task.

However, attempts were made to estimate settlements in the core at the highest section of the dam, at St. 75L, where the settlement is recorded at marker C7b.

When the levels recorded during 32 years operation of Jatiluhur reservoir are averaged the value comes to El. 99.7m, which is around 100m. Thus if the sudden drops of major drawdown episodes of 1972, 1982 and 1997-98 are omitted the remaining settlement curve should represent the time-settlement curve for loading with the reservoir level at El. 100m. The adjusted time settlement curve was developed from the actual observed readings of marker C7b. The curve is a typical hyperbola with horizontal asymptote. A regression was done to determine constants ‘a’ and ‘b’ to represent the curve by the standard hyperbolic equation :

S = t / (a + b x t)

Where S = settlement, in mmt = time in years

The vertical intercept of the asymptote, which is the reciprocal of constant ‘b’, was found to be 1996 mm. The value represents the ultimate settlement which could be attained if the reservoir level remained constantly at El. 100m. An estimate of total settlement was also made by the Terzaghi formula :

S = Cc / (1 + eo ) x H x log (p100 / pc)

Where Cc= compression indexeo = initial void ratioH = height of clay core

p100 = effective stress at mid of core when reservoir is in steady state condition at reservoir level 100m

pc = pre-consolidation pressure

Adopting average values of Cc = 0.18, eo = 0.8 from the laboratory tests done on clay core samples and pc = 400kPa, from the co-relationship given by De Mello (1980) for 95% compaction, the above equation gives a total settlement of 2100mm. This amount is in close agreement with the value of 1996mm found from the adjusted time-settlement curve of the observed data.

ACE/SEJD/Tech. Rep.

6

The comparison of the hyperbolic curve with the theoretical time-settlement curve developed from Terzaghi's consolidation theory with Cv = 6 x 10-8

m2/sec and an average length of seepage path equal to 10 m, indicates a very close agreement between the curves. The value of Cv was adopted from the consolidation tests on core samples.

The ultimate settlements for other reservoir level loading conditions were determined by applying the principle of e-log p curve. For any loading at reservoir level other than El. 100m, the ultimate value can be calculated from the equation :

SultR = S ult100 + {Cc / (1+e0)} x log (pR/p100)

Where,SultR,Sult100 = ultimate settlement for loading conditions of reservoir at Elevation R and 100 respectively

p100 and pR= effective pressure at mid of core when reservoir is at El.100m and R respectively.

After determination of ultimate settlements for various reservoir levels, the time-settlement curves could be developed on the assumption that the degree of consolidation at any time is the same for any or all loading conditions. Curves developed for reservoir levels (between El. 107m and El. 75m) at 2.5m intervals are shown in Figure 5.

ANALYTICAL PROCEDURE TO PREDICT SETTLEMENT FROM RESERVOIR RECORDS

Using reservoir level records, a field time-settlement curve could be estimated by, the graphical procedure explained earlier.

The processing of huge amount of data of reservoir levels to predict core settlements by the graphical method is rather tedious. Therefore an analytical version of the above procedure was developed to calculate settlement increments with reservoir variations. An outline of the

analytical procedure is given below:

- Collect initial data time = to, reservoir level = Ro and settlement already attained = So

- Next reservoir data at time = t1 and reservoir level R1

- Calculate average reservoir level Rav = (Ro + R1)/2 and time increment, t = t1 - to

- Calculate the extended time on the time- settlement curve for the reservoir at Rav by the formula :

to' = So x a / [(S ult Rav/S ult 100) – So x b]

Where a and b are constants of the hyperbolic settlement curve for the reservoir at El. 100m and SultR and Sult100 are ultimate settlements for loading condition at reservoir level at Rav and El. 100m.

- Calculate corresponding time on time- settlement curve for reservoir at Rav by:

t1’= to’ + t

- Calculate new value of settlement by formula :

St1 = t1' x (Sult Rav/Sult100) / ( a + b x t1’)

- Continue calculations for the subsequent reservoir data assuming previously calculated settlement as initial data.

COMPARISON OF CALCULATED AND OBSERVED SETTLEMENTS AND FUTURE PREDICTION

The values calculated by the above procedure are shown in Figure 5. It is interesting to note that, except for minor variations at few places, the calculated curve matches very closely the observed field data. The curve even explains the minor as well as major kinks in settlement when the reservoir level drops.

The analysis shows that, when the reservoir remains at higher level for a long time, the rate of settlement is slow and, when any major drawdown occurs, very high settlement rates are observed during drawdown.

ACE/SEJD/Tech. Rep.

7

1965 1975 1985 1995 2005

-3500

-3000

-2500

-2000

-1500

-1000

-500

0

40

50

60

70

80

90

100

110

YEAR

Set

tlem

ent (

mm

)

Res

ervo

ir Le

vel (

m )

Observ ed reserv oir lev el

Observ edsettlement

Estimated settlement f or

observ ed reserv oir lev els

RL. 105.0 m

RL. 107.0 m

RL. 102.5

RL. 100.0 m

RL. 97.5 m

RL. 95.0 m

RL. 92.5 m

RL. 90.0

RL. 87.5mRL. 85.0

RL. 82.5

RL. 80.0

RL. 77.5

Estimated settlement f or

Reserv oir at El. 75 m

Figure 5. Observed and estimated settlement at the highest section of the embankment

This close agreement between the two curves provided enough confidence to utilize the procedure to estimate expected settlement, if the reservoir is utilized to maximum capacity and emptied in a period of half the year and filled in the next half. The prediction has been extend to the year 2065, a century of Jatiluhur reservoir operation. Total settlement after 100 years of dam operation, i.e. in the year 2065, will be about 2.5 m.

SUMMARY AND CONCLUSIONS

The review and analysis of settlement data for Jatiluhur dam shows that, in the case of a rockfill dam with clay core inclined upstream, the major portion of settlement of the downstream edge takes place during the first few years after reservoir filling.

On the other hand, the settlement of upstream edge, which depends on settlement of the core, will rely on the controlling rate of dissipation of

pore pressures. Due to operational restrictions the Jatiluhur reservoir was kept all the time high which did not allow quick dissipation of pore pressures and resulted in a delayed rate of settlement. However, events of major drawdown in years 1972, 1982 and 1997-98 allowed quicker dissipation of pore pressures and in turn accelerated settlement.

A method of analysis was developed using basic theory of consolidation to estimate time-settlement behaviour of the embankment core. The comparison of calculated values with the observed field data shows a very close match. The method even explains the minor as well as major kinks in settlement when reservoir level drops.

The proposed procedure can be utilised to predict the time-settlement history of an existing or a new earth-core rockfill dam for the planned reservoir operation.

ACE/SEJD/Tech. Rep.

8

ACKNOWLEDGEMENT

[ Please add here a note of thanks for assistance from POJ and PU]

REFERENCES

Bohn, M., and Hamon (1967), “The Djatiluhur Project”, Water Power, Aug. 1967.

De Mello, V.F.B (1980), Comparative Behavior of Similar Compacted Earth rock Dam in Basalt Geology in Brazil, Symposium on Problems and Practice of Dam Engineering, Bangkok, December 1980.

Sherard, J. L (1973), “Embankment Dam Cracking”, Embankment Dam Engineering, Casagrande Volume, John Wiley & Sons, New York.

Mis/kh\msw\kh/’JATIL’17.5.99

ACE/SEJD/Tech. Rep.

9