You are on page 1of 6

RESEARCH ARTICLES

Mechanism of triggered seismicity at Koyna:


An evaluation based on relocated earthquakes
Kusala Rajendran* and C. M. Harish
Centre for Earth Science Studies, P.B. No. 7250, Akkulam, Thiruvanathapuram 695 031, India

stimulated by reservoirs. We will not discuss the physical


Seismicity associated with the Koyna Reservoir, west- basis of these attributes, but refer the reader to McGarr
ern Maharashtra, is unique because it is one of the few
and Simpson6 who prefer to use the adjective triggered,
sites in the world where earthquakes of M 5.0 con-
tinue to occur three decades after the initial spurt of to describe earthquakes caused by artificial reservoirs.
activity. This contrasts with most other seismogenic The recent report of the USCOLD (US Committee on
reservoirs where temporal patterns of seismicity are Large Dams)7 also recommends the same. We also favour
characterized by one or two peaks that taper-off with the adjective triggered and use it in this paper to des-
time. Persistence of moderate seismicity at Koyna is cribe the seismicity at Koyna.
enigmatic, considering also the low seismic producti- The ongoing seismicity at Koyna suggests that annual
vity of the peninsular shield. Temporal association of loading cycles in the reservoir that was impounded in
major earthquakes with annual reservoir loading has 1962, continue to influence the overall stability of the
been observed for long. In this paper, we report a dis- region. Filling of the Warna Reservoir in 1985 is another
cernible difference between the spatial pattern of important factor that must have affected the hydro-
earthquakes associated with the reservoir loading and geological environment. Sustained increase in fluid pres-
unloading. A conceptual model is proposed to explain
sure, complemented by repeated failure in the fault zone
the continued occurrence of earthquakes.
may have led to many complex processes that are not dir-
ectly observable. However, we expect that the spatial and
SEISMICITY associated with the Koyna Reservoir is unique temporal patterns of earthquakes may reflect some of
in many ways. Koyna is among the four reservoirs that these processes. The quality of data, especially those prior
have triggered earthquakes of magnitude 6.0 (ref. 1). to 1993, was not good enough for detailed interpretations.
Earthquakes of M 5.0 continue to occur here nearly This limitation has been recognized for long and there
three decades after the major burst of activity in 1967 have been several efforts to improve the quality of
(ref. 2). Peaks of seismic activity show remarkable tempo- locations811.
ral association with the increased rate of filling3. In the
backdrop of low strain rate and low seismic productivity,
characteristic of stable continental regions4, seismicity at Relocation of earthquakes during 19831993
Koyna, located in the Archean craton is rather unusual.
Over 100,000 earthquakes of M 0.0; nearly 100 events Earthquakes of M 3.0 were relocated after reexamining
of M 4.0 and 10 earthquakes of M 5 have been rep- the P and S arrival times and fixing the station coordinates
orted from this region5. Such frequency of earthquakes, using a hand-held GPS receiver. The velocity model by
especially those exceeding M 5.0, in a short span of time, Kaila12 which gives smaller standard errors (RMS, ERH
is a rarity in the stable shield regions. As reported by and ERZ) was used. Several precautions were taken to
McGarr and Simpson6, an event of M 6.3 and a sequence improve the quality of data, although corrections for intrin-
continuing for more than 30 years do not fit the usual sic errors could not be made (see ref. 13 for more details).
mainshockaftershock decay pattern observed for tectonic Out of the 207 relocated epicentres, 125 met the B and
earthquakes. They note that, the reservoir and annual C quality criteria prescribed in HYPO71 and only these
cycles in water level continue to interact with, and aug- are used in this study (Figure 1). Most earthquakes
ment the tectonic stress field in a complex manner that has seemed to originate from depths shallower than 10 km, as
led to a persistent sequence of anomalous seismicity. In indicated also by some recent studies using broadband
this paper we explore some important factors that may be data14,15.
conducive to the continued seismicity at Koyna.
The adjectives induced and triggered have often been Inference of a N-S trending seismogenic structure
used liberally and interchangeably, to describe earthquakes
While the earlier studies16,17 defined a general zone of
activity around the Koyna Reservoir, more specific pat-
*For correspondence. (e-mail: kusala@vsnl.com) terns were identified later. Based on 39 relocated events
358 CURRENT SCIENCE, VOL. 79, NO. 3, 10 AUGUST 2000
RESEARCH ARTICLES

of ML 4.0, Rastogi and Talwani9 suggested two trends, earthquakes relocated by Rastogi and Talwani9 occurred
in the NNE and NW directions. Dube10 defined a NNE- in this region. Filling of the Koyna Reservoir 23 years
oriented fault, based on the distribution of 380 relocated prior to Warna must have weakened much of these
earthquakes. Using 71 relocated earthquakes during Oct- regions and addition of a new reservoir must have pro-
ober 1973 to December 1976, Gupta et al.8 suggested moted instability and widened the areas of influence.
the existence of a N-S trending fault close to 7345 E Existence of a roughly N-S trending regional structure
longitude. has been inferred from geomorphologic, tectonic and
Earthquakes relocated in this study define a nearly N-S geophysical data1922. The gravity data19 suggest a deep
zone, extending south of the Koyna Reservoir (Figure 1). fault oriented roughly in the N-S direction. Two promi-
It is important to note that most of the activity during the nent sets of joints in N-S (N 10E to N 10W) and E-W
last three decades has generally been confined to an area directions and shatter or shear zones running for consi-
of about 25 10 km2, extending south of the Koyna Res- derable distances have been mapped in the Koyna
ervoir. Regions to the west and east and below the deepest region21. Many N-S fractures in this region are reported to
parts of both Koyna and Warna Reservoirs have generally be water-bearing22. Based on a variety of data, two recent
been less active all through these years. The pattern studies23,24 have also suggested the prominence of the N-S
observed during the period 1983 to 1993 is not different structure. Drilling at Koyna25 has confirmed that the fault
(Figure 1). dips ~ 60 towards WNW. Although there may be other
Recent analyses using broadband data suggested the potential structures in the region, we believe that the N-S
existence of two parallel, near vertical NNE-SSW trend- to NNE-oriented structure plays an important role in the
ing faults, the Koyna and Bhogiv faults, about 5 to 10 km earthquake process in the KoynaWarna region.
apart5,14. Strike-slip fault zones often display en chelon
character18, and we believe that Koyna and Bhogiv faults
are not two independent faults, but strands of a major Seismicity during filling and drawdown phases
fault system. It should be noted that although the Bhogiv
fault is closer to the Warna Reservoir, this region was not Temporal association of seismicity with filling cycles in
totally free of earthquakes in the past. For example, a few the Koyna Reservoir is quite striking1. Earthquakes of
M 4.0 are reported to occur 2 to 3 months after the
increased rate of filling ( 40 ft/week) in the Koyna Res-
ervoir3. Post-monsoon refilling generally starts by July
and the major peaks in seismic activity occur during or
after September. Based on the six-year delay between the
initial filling and the largest earthquake, seismicity at
Koyna was classified as delayed response26, where pore
pressure diffusion plays a significant role. The two/three
months lag between peak refilling and the major burst of
activity consistently observed here is a probable indica-
tion of diffusion of pore pressure to earthquake source
zones. If so, can it in some way be observed from the
spatial distribution of earthquakes? Spatial correlation
of earthquakes with advancing pressure fronts has been
observed in field experiments at Rangely27. We made an
effort to explore the spatial pattern of earthquakes, in rela-
tion to loading and unloading in the reservoir.
In this study, we have considered the extended effect of
diffusion to define the refilling and drawdown phases,
which are different from the actual refilling/drawdown
periods. In this context, it is important to understand the
role of pore pressure diffusion in triggering earthquakes
near seismogenic reservoirs. In explaining the spatial and
temporal pattern of induced seismicity using the mechani-
cal effect of pore pressure diffusion, Talwani and Acree28
used a term seismic hydraulic diffusivity denoted as as.
Using a simple relation as = L2/t, where L is the distance
between the source of the pore pressure front (reservoir)
Figure 1. Relocated earthquakes of M 3 during 19831993 near
Koyna and Warna Reservoirs. Seismic stations are shown by filled and the location of earthquakes and t is the delay between
squares. the time of filling and the onset of seismicity, they com-
CURRENT SCIENCE, VOL. 79, NO. 3, 10 AUGUST 2000 359
RESEARCH ARTICLES

puted a range of values for as. They further reported that a refilling phase. By extending the refilling phase for two
characteristic value of about 5 104 cm2/s was applicable months after the peak reservoir levels are reached, we are
to most regions. Thus, filling and drawdown processes do incorporating the delayed effect of diffusion. The period
have prolonged effects, justifying our definition of refill- from March to June is treated as the drawdown phase.
ing and drawdown phases. Refilling and drawdown phases for 1983, a representative
During a typical year, refilling in the lake starts in year, are shown in Figure 2.
June/July and the maximum lake levels are reached by We realize the difficulty in prescribing time limits for
August/September. Significant changes in the lake level processes such as the pore pressure diffusion, but this
may occur only after the third week of June. The level demarcation is justified by the consistent delay between
remains close to maximum till December and then starts the onset of refilling and major spurt of activity. It should
to drop. This pattern has been consistently observed for be remembered that during the days immediately follow-
the ten-year period discussed in this study13. Assuming a ing the refilling, pore pressure field is complicated by the
hydraulic diffusivity of 5 104 cm2/s, and distance of undrained response of the reservoir26. Since most initial
more than 5 km from the deepest part of the lake, delay of days of refilling are devoid of significant activity, we have
about two months between filling time and onset of activ- not considered the effect of undrained response in this
ity is justifiable. Thus, in this study, we treat the period study.
from the first week of July to the end of February as the Distribution of earthquakes during the refilling and
drawdown phases of 1983 provides a good illustration of
these processes during one cycle (Figures 2 and 3). The
activity during the refilling phase occurs in a wider area,
but it shrinks during the drawdown phase. Peaks in earth-
quake activity generally occur during September to Dec-
ember, two to three months after the start of refilling. The
westward extension of the active zone during the refilling
phase is a probable manifestation of pore pressure growth
to farther areas and the lateral shrinkage may indicate
retreat of the pressure front.
Composites of all relocated earthquakes during refilling
and drawdown phases for the period 1983 to 1993 are
shown in Figure 4 a and b. A wider zone of activity with a
westward extension is noted during the refilling phase,
whereas there is shrinkage in the zone of activity during
Figure 2. Average lake levels at Koyna during 1983, considered as a the drawdown phase. Thus, what we observe at Koyna
representative year (scale on right). Refilling and drawdown phases
are indicated. Frequency of earthquakes of M 3.0 also shown (scale may be considered analogous to breathing of the fault,
on left). guided by the time-dependent growth of pore pressure

a b

Figure 3. Relocated earthquakes of M 3.0 during (a) refilling phase, and (b) drawdown phase during 1983.

360 CURRENT SCIENCE, VOL. 79, NO. 3, 10 AUGUST 2000


RESEARCH ARTICLES

front. The spatial pattern, with westward growth of acti- directly below the reservoir in a vertical strike-slip fault.
vity is a probable indication of a west dipping fault zone. However, the preferred regions of activity are away from
The narrow zone of activity noted during the drawdown what is predicted by the model and we attribute this
phase occurs close to 7345 E, the inferred location of deviation to the geological characteristics of the region.
the fault (Figures 3 b and 4 b). Thus, none of the existing models satisfactorily explains
the mechanism of triggered seismicity at Koyna, particu-
Mechanism of triggered seismicity at Koyna larly its continuing nature.
We propose a conceptual model based on our observa-
Although several papers have been published on Koyna tions and also using some well-known characteristics of
seismicity, the mechanism of triggering seismicity has not mature faults3335. A continuously degenerating fault zone
been discussed in detail. In one of the earlier papers, that may respond to smaller stress changes is fundamental
Balakrishna and Gowd29 used a theoretical model to gen- to the mechanism we propose. The fault zone at Koyna
erate failure conditions at Koyna. Predominant horizontal has been experiencing hundreds of earthquakes each year
stress and excess fluid pressure in the fault zone provide and due to its longevity, we treat this as a mature fault.
the boundary conditions in this model. Athavale30 used The annual filling cycles continue to weaken this fault and
various lines of evidence to establish the existence of a we hypothesize that the fault degenerates with the passage
N-S striking fault, west of the Koyna Reservoir. Seepage of time, promoting failure at lower values of stress changes.
of large quantities of water into this fault zone was sug- Response to lake level fluctuations of the order of
gested as a possible mechanism for earthquakes near 1.5 m/day near Lake Jocassee36, suggest that some seis-
Koyna. mogenic reservoirs are sensitive to stability changes intro-
In two recent studies, Talwani23,31 suggested that earth- duced by small changes. The temporal and spatial pattern
quakes at Koyna occur near the intersections of major of seismicity observed at Koyna suggests that the region
faults oriented in N-S, NW-SE and NE-SW directions. is quite sensitive to changes in stresses introduced by res-
His model assumes that intersections of these faults, aided ervoir fluctuations.
by annual loading in the reservoir provide means of As we discuss the behaviour of fault zones under high
stress build-up. However, it may be noted that all the fluid pressure, it is convenient to recall the following con-
larger earthquakes in the region have occurred along the ditions6 for frictional failure on planar surfaces:
~ 20-km-long N-S zone8, and not in clusters, as one would
expect from localization of stresses near intersections. We = 0 + (n P). (1)
examined the possibility of applying the poroelastic mod-
els by Roeloffs32 which predict maximum instability In the above equation, is the strength of a fault or the

a b

Figure 4. Composite of relocated earthquakes of M 3.0 during (a) refilling phases of 18931993;
and (b) drawdown phases of the same period.

CURRENT SCIENCE, VOL. 79, NO. 3, 10 AUGUST 2000 361


RESEARCH ARTICLES

it fail in response to small changes in shear stresses, as


suggested by eq. (1)? With the passage of time, will the
fault zone become further unstable, progressively lower-
ing the threshold changes required to initiate failure? The
recurring temporal and spatial patterns of seismicity at
Koyna suggest this to be a possibility. These ideas need to
be explored further, but if they are applicable, we have a
unique site to study the critical failure conditions in a hyd-
romechanically weakened fault zone.

Discussion and conclusions


Figure 5. Schematic diagram illustrating the geological and hydro-
geological conditions in the Koyna region. The top layer of basalt is Relocation has generally improved the quality of loca-
assumed to permit flow only through vertical fractures. Lateral flow is
restricted, due to the lower permeability of the country rock and occur- tions, enabling more detailed studies than what was possi-
rence of an impervious seal between the fault and the country rock. The ble previously. The relocated seismicity in the Koyna
section is assumed to be in the WNWSSE direction, perpendicular to region during 1983 to 1993 defines a N-S trending zone
the fault.
extending south of the Koyna dam, reinforcing earlier
observations. Earthquakes during the refilling phase are
noted to occur in a wider region, whereas those during the
shear strength required for failure; 0 is the cohesion; is
drawdown phase are restricted to a narrower zone. We
the coefficient of friction; n is the normal stress across
consider this to be the indirect evidence for an advancing
the fault and P is the pore pressure within the fault zone.
pore pressure front, guided by reservoir load. With more
An increase in shear stress ( ); a decrease in the normal
detailed examination of higher quality data, the spatial
stress (n) or an increase in pore pressure (P) may cause
association of earthquakes with filling cycles can be better
slip in the fault zone. It may become weak also if the
understood.
coefficient of friction () is transiently decreased. A suffi-
The crucial question that is always being asked is about
ciently low intrinsic coefficient of friction seems impro-
the continuing seismicity at Koyna. With a setting where
bable because such behaviours have not been found in
the effect of fluid pressure on failure process is dem-
earth materials37. Thus, one probable explanation for low
onstrated by recurring sequences of earthquakes, Koyna
fault strength would be the simple hydromechanical reduc-
offers an excellent location to understand the conditions
tion in strength, resulting from abnormally high fluid
of criticality necessary for failure. Continuous weakening
pressure within the fault zone, as suggested by the above
of the fault zone enabling failure at lower threshold levels
equation.
of stress changes is proposed here as a likely mechanism.
Increase in fluid pressure caused by compaction due to
Quantifying the threshold values of pore pressure and stress
small amounts of ductile creep is a favoured mechanism
changes required for the initiation of failure is a major
of failure in many mature fault zones like the San
step in understanding the mechanism of earthquakes stimu-
Andreas fault36. Unlike these fault zones which derive
lated by increased pore fluid pressure and in that context,
fluids from deeper sources38, growth in pore pressure in a
studies at Koyna have far-reaching implications.
reservoir environment is guided by other mechanisms.
The conceptual model presented in this paper needs to
Elastic compression of the rock matrix due to the load-
be supported by quantitative modelling. Study of three-
induced compaction as well as pore pressure diffusion due
dimensional pattern of earthquakes, physical properties of
to the increased hydraulic head are two important pro-
the rocks and pore pressure changes in the area will help
cesses that lead to increase in pore pressure26. Chambers
to constrain some of these observations. Ongoing experi-
of high fluid pressure may be developed and sustained, if
ments at Koyna, including measurements of pore pressure
geological conditions are favourable.
changes in deep observation wells, together with higher
A perpetually weakening fault zone that is hydraulically
quality broadband data will be useful in understanding
connected to the reservoir is the most important feature of
these processes in future.
our model (Figure 5). The assumption of a near vertical
fault is supported by a variety of data discussed earlier. It
is assumed that the N-S vertical fractures are more con- 1. Gupta, H. K., Reservoir Induced Earthquakes, Elsevier, 1992,
ductive22 and that they help to recharge the fault zone. p. 364.
Assumption of relatively higher permeability within the 2. Gupta, H. K., J. Geol. Soc. India, 1993, 42, 413415.
3. Gupta, H. K., Bull. Seismol. Soc. Am., 1983, 73, 679682.
fault zone and the existence of a low permeability seal
4. Johnston, A. C. and Kanter, L. R., Sci. Am., 1990, 262, 6875.
that separates the fault zone from the country rock are 5. Rastogi, B. K., Chadha, R. K., Sarma, C. S. P., Mandal, P., Satay-
based on observations in fault zones elsewhere35. If the narayana, H. V., Narendrakumar, Raju, I. P., Satyamurthy, C. and
fault zone is assumed to be under high fluid pressure, will Nageshwar Rao, A., Bull. Seismol. Soc. Am., 1997, 87, 14841494.

362 CURRENT SCIENCE, VOL. 79, NO. 3, 10 AUGUST 2000


RESEARCH ARTICLES
6. McGarr, A. and Simpson, D. W., in Rockbursts and Seismicity in 25. Gupta, H. K., Rao, R. U. M., Srinivasan, R., Rao, G. V. and
Mines (eds Gibowicz, S. J. and Lasocki, S.), Balkema, Rotterdam, Reddy, G. K., Geophys. Res. Lett., 1999, 26, 19851988.
1997, pp. 385396. 26. Simpson, D. W., Leith, W. S. and Scholz, C. H., Bull. Seismol.
7. The United States Committee on Large Dams (USCOLD) report, Soc. Am., 1988, 78, 20252040.
Denver, 1997, p. 20. 27. Raleigh, C. B., Healy, J. H. and Bredehoft, J. D., Science, 1976,
8. Gupta, H. K., Rao, C. V. R. K. and Rastogi, B. K., Bull. Seismol. 191, 12301237.
Soc. Am., 1980, 70, 18331847. 28. Talwani, P. and Acree, S., PAGEOPH, 1984, 78, 20252040.
9. Rastogi, B. K. and Talwani, P., ibid, 18491868. 29. Balakrishna, S. and Gowd, T. N., Tectonophysics, 1970, 9, 291
10. Dube, R. K., ibid, 1986, 76, 395407. 300.
11. Talwani, P., Kumaraswamy, S. V. and Sawalwade, G. B., Techni- 30. Athavale, R. N., Bull. Seismol. Soc. Am., 1975, 65, 183191.
cal report, University of South Carolina, 1996, p. 343. 31. Talwani, P., PAGEOPH, 1995, 145, 167174.
12. Kaila, K. L., 1983 (pers. commun. to MERI). 32. Roeloffs, E. A., J. Geophys. Res., 1988, 93, 21072124.
13. Rajendran, K., Harish, C. M. and Kumaraswamy, S. V., Project 33. Byerlee, J. D., Geophys Res. Lett., 1990, 17, 21092112.
report to the DST, 1996, p. 94. 34. Rice, J. R., in Fault Mechanics and Transport Properties of Rocks
14. Gupta, H. K., Rastogi, B. K., Chadha, R. K., Mandal, P. and (eds Evans, B. and Wong, T. F.), Academic Press, 1992, pp. 475
Sarma, C. S. P., J. Seismol., 1997, 1, 4753. 503.
15. Mandal, P., Rastogi, B. K. and Sarma, C. S. P., Bull. Seismol. Soc. 35. Sleep, N. H. and Blanpied, M. L., Nature, 1992, 359, 687692.
Am., 1998, 88, 833842. 36. Rajendran, K., PAGEOPH, 1995, 145, 8795.
16. Guha, S. K., Gosavi, P. D., Nand, K., Padale, J. G. and Marwadi, 37. Morrow, C., Randey, B. and Byerlee, J., in Fault Mechanics and
S. C., Report Central Power and Power Research Station, Khadak- Transport Properties of Rocks (eds Evans, B. and Wong, T. F.),
wasala, Poona, 1974, p. 24. Academic Press, 1992, pp. 6988.
17. Guha, S. K., Gosavi, P. D., Agarwal, B. N. P. and Marwadi, S. C., 38. Sibson, R. H., Tectonophysics, 1992, 211, 283293.
Eng. Geol., 1974, 8, 5977.
18. Yeats, R. S., Sieh, K. and Allen, C. R., The Geology of Earth-
quakes, Oxford University Press, 1997, p. 568.
ACKNOWLEDGEMENTS. This work is part of a project funded by
19. Kailasam, L. N. and Murthy, B. G. K., Indian J. Power River Val.
the Department of Science and Technology, New Delhi. We thank Dr
Dev., 1971, Special No. 17, 2730.
R. K. Midha and Dr G. D. Gupta for their encouragement and support
20. Balasundaram, M. S. and Srinivasan, P. B., ibid, 1114.
and the Director, CESS, for facilities. Earthquake data was provided by
21. Sahasrabudhe, Y. S., Rane, V. V. and Deshmukh, S. S., ibid, 47
Maharashtra Engineering Research Institute (MERI). Comments by Dr
54.
H. K. Gupta and Dr S. K. Arora on earlier versions of this manuscript
22. Snow, D. T., Geol. Soc. Am., Spec. Pap., 1980, 189, 317360.
have been useful.
23. Talwani, P., PAGEOPH, 1997, 150, 511550.
24. Seeber, L., Ekstrom, G., Jain, S. K., Murty, C. V. R., Chandak, N.
and Armbruster, J. G., J. Geophys. Res., 1996, 101, 85438560. Received 4 March 1999, revised accepted 5 June 2000

CURRENT SCIENCE, VOL. 79, NO. 3, 10 AUGUST 2000 363

You might also like