Geophys. J . Int. (1991) 106, 293-300

RESEARCH NOTE

Microseismicity and structure of the Germencik area, western Turkey

Malcolm Jones and Rob Westaway Department of Geological Sciences, University of Durham, South Road, Durham DH1 3LE, UK

Accepted 1991 March 7. Received 1991 March 7; in original form 1990 October 1

SUMMARY Microseismicity at the boundary between two segments of the Buyuk Menderes active normal fault zone, western Turkey, was monitored for two weeks during 1990 April and May, using a dense network of six portable seismographs with spacing -1-2 km. Extension rate across this fault zone is investigated by three independ- ent methods; our preferred estimate is 1 .2f0 .4mmyr- ' . The area contains a geothermal field, but microseismicity appears unrelated to geothermal well positions and was thus lower than expected; six local events were recorded, none larger than magnitude 2. Microearthquakes in this size range contribute negligibly to local tectonic deformation. However, frequency of occurrence of these and local magnitude 7 events both satisfy the standard Gutenberg-Richter relationship. The local geomorphology includes an example of river capture associated with elevation changes accompanying changing patterns of slip on individual fault segments, which appears to have occurred less than 1 Myr ago.

Key words: extension, microseismicity, normal faulting, Turkey.

INTRODUCTION

Western Turkey is a region of active continental extension, which is taken up predominantly on E-W striking normal faults. Extension began in late Miocene time, most likely in the Tortonian stage between 7 and 11 Myr ago (e.g. Kaya 1981). Evidence from the Greek island of Samos (Fig. l) , a few kilometres offshore of western Turkey, confirms that local extension began less than 11 Myr ago (Sen & Valet 1986). During the period 1989 August to 1990 May a temporary network of over 30 seismometer stations was installed in western Turkey to monitor the seismicity associated with this extension, supplementing the sparse distribution of permanent stations (Fig. 1). The local microseismicity of a smaller area near the town of Germencik in the Buyuk Menderes valley (Figs 1 and 2) was monitored in detail as a part of this wider project, and is the subject of this article. The northern flank of this valley follows one of the principal active normal fault zones in western Turkey, the Buyuk Menderes fault zone (BMFZ). It trends roughly ENE from the Aegean Sea coast, south of Izmir, for over 100 km (longitudes 27"-29"E at latitude 38"N), and for most of its length comprises two fault branches offset en 6cheZon by -4 km (e.g. Roberts 1988). As with many other normal faults, this BMFZ is segmented on a scale of 10-20km. The Buyuk Menderes valley is an actively forming sedimentary basin in the hanging wall of

the southern branch of this fault zone. Two documented major earthquakes have occurred on this southern branch in recent centuries. One in 1653, appears to have ruptured several fault segments from Omerbeyli eastward for 70 km (Ambraseys 1988), including the -20 km long Omerbeyli segment (Roberts 1988). The second, in 1899 (M,=6.9), again ruptured the eastern 40 km of this 70 km length of fault, with -1 m of slip (Ambraseys 1988).

An uplifted Neogene basin, known locally in the Germencik area as the Arzular basin, in the footwall of the southern branch of the BMFZ, is situated in the hanging wall of the northern branch of the BMFZ that bounds it to the north, and Palaeozoic schist of the Menderes massif is exposed in the uplifted footwall of the northern BMFZ branch. Recent uplift of this basin has presumably occurred as a result of the greater slip in the recent geologic past on the normal fault that bounds this basin to the south compared with the other normal fault to the north. These faults and basins are shown on the map in Fig. 2 and in cross-section in Fig. 3. Exposures of fault planes on major normal faults in and around the area shown in Fig. 2 typically dip southward -45" (Roberts 1988). Samos island further west shows a similar structural style, although with the opposite polarity of major faults, which locally dip northward. The north coast of Samos is a major active normal fault, with a -3 km deep Neogene sedimentary basin in its hanging wall to the north (Turgut 1988). On

293

294 M . Jones and R. Westaway

KKB pLD+ \

4 KAP Permanent seismometer stations

A Turkey 0 Greece

+ Bulgaria

- 100 krn

Figure 1. Map of the Aegean Sea region, showing the locations of Germencik, Samos island, and the permanent seismograph stations in Greece, western Turkey, and southern Bulgaria. The sparse distribution of stations in southwestern Turkey prompted the installation of the temporary regional network mentioned in the text.

Samos, a Neogene sedimentary basin has been uplifted in the footwall of this fault. Sen & Valet (1986) reported ages of these sediments from 11-5Myr, with most extensive deposition around 7 Myr ago. Presumably sedimentation in this basin ceased when activity shifted from other active normal faults to the major normal fault north of Samos, causing Samos to begin to be uplifted. Bearing in mind this evidence, we adopt 7Myr ago as the nominal time when extension began, both on Samos and along the nearby BMFZ.

MICROSEISMICITY OF THE GERMENCIK AREA

During the period from 1990 April 21 to 1990 May 3, a temporary network of portable seismographs was operated near the southern branch of the BMFZ northeast of Germencik. This area was thought in advance to most likely exhibit substantial local microseismicity for the following reasons. First, it lies close to overlapping segments of the southern branch of the BMFZ. Previous work elsewhere in the Aegean region (e.g. Vita-Finzi & King 1985) has shown that basement between segments of major faults may be broken up by many smaller faults. One would expect these small faults to take up deformation in frequent microearthquakes. Second, it is an active geothermal field, and a major objective was to attempt to determine the extent of microseismicity associated with this geothermal field, compared with the microseismicity associated with tectonic extension.

The network consisted of six Sprengnether MEQ-800 seismographs with Geospace HS10-2 short-period vertical

component seismometers. These were positioned - 1-2 km apart in order to best record the relatively shallow geothermal microearthquakes that were expected (Fig. 2). Recording was on smoked paper on a drum; records of vertical ground velocity were scretched by a stylus. A single seismometer station had operated nearby from 1989 December, but was dismantled in 1990 April before the others were installed (A28, Fig. 2).

Absolute timing was obtained by synchronizing the MEQ internal clocks with a 9996 kHz radio time signal from Moscow, so that the start of the minute markers on the smoked paper records coincided with the start of the minute time signal. By synchronizing clocks every 2 days as the paper on the drum was changed, a timing accuracy of -0.1 s was maintained. Station sites were carefully chosen in localities where noise levels were low, enabling most seismographs to be set up to amplify ground velocity by 120 dB, the maximum gain possible. High frequency cut-off filters at 5 or 1OHz reduced high-frequency noise, but no low-cut filter was employed.

During the 13 days of operation, six local earthquakes were recorded (characterized by a <I s interval between P- and S-wave arrivals), along with other more distant events that will be studied separately. Only one of these local events (event 6) was large enough to enable confident identification of the P and S phases at all stations operating at the time (Fig. 4). The similarity of the waveforms from this event at stations GEOl and GE02 in Fig. 4 suggests a similar ray path was followed. The time interval between P- and S-wave arrivals is smallest for station GE03, indicating that this station lies closest to the hypocentre, which is thus west or southwest of the three stations.

Microseismicity of the Germencik area 295

57-

56'

55'

54'

53'

37O 52' N

0 0

0 0 0 0 0 0

A r z u l a r / b a s i n

27'34' E 35' 36' 37- 38' 39- 40'

(__I Palaeozotc basement A Geostore network station

Neogene sediments A Drum seismograph Earthquake eptcentre - 7 Principal normal fault

Holocene alluvial deposits '6W13 event number depth(km)

Built up area (ticks on downthrow side)

Erosional contact ---

Figure 2. Simplified geological map of the Germencik area showing locations of temporary seismograph stations and the six local earthquakes recorded by them. Error bars indicating formal uncertainty in epicentral coordinates are omitted to avoid clutter, but formal error estimates are given in Table 1. NB and SB indicate positions of the northern and southern branches of the Biiyiik Menderes fault zone. Most geothermal wells are situated between lat. 37" 53-54" and Ion. 27" 36-37'E.

Arrival times of the P and S phases for this and the other local events relative to the nearest minute markers on the records were measured using an engraved glass ruler, graduated in 0.1mm intervals, that was viewed through a magnifier. Although the absolute station positions were not known to high precision, their relative positions were measured to -30 m accuracy. This ensured that the limiting factors controlling uncertainties in hypocentral location relative to the local seismograph network were the -0.05 s uncertainty in picking arrival times, the <-0.1 s uncertainty in absolute timing, and the uncertainty in the local velocity structure.

The program HYP071PC was used to locate the earthquakes. This is a version of the standard HYPO71 program (Lee & Lahr 1975), adapted to run on MS-DOS workstations. Velocity structure of this region was approximated using an upper layer of low velocity (representing unconsolidated Neogene and Quaternary sediments) underlain by a half-space of higher velocity (basement; Palaeozoic schist). A first approximation to the sediment thickness was made using results from a gravity survey by Akqig (1988), in which a gravity anomaly of -30 mgal was measured in a north-south traverse across the Biiyiik Menderes valley to the east of the study area. The

296 M. Jones and R. Westaway

S.25"W. South branch of

N.25"E. North branch of

Figure 3. Schematic SSW-NNE cross-section across the Biiyuk Menderes fault zone (BMFZ) along the line XY in Fig. 2. Horizontal and vertical scales are equal, with depths measured relative to sea level (i.e. -10 m below the Biiyuk Menderes valley floor). Sediment thickness and bedding dips within the Arzular basin (projected downdip from the observed dip of beds at the surface) are from Roberts (1988). The northern branch of the BMFZ and the Omerbeyli segment of the southern BMFZ branch are projected downdip at 45", the dip of their surface traces. The same dip was assumed on the Hidirikoylu segment of the southern BMFZ branch. The position of the block in the hangingwall of the Hidirikoylu segment is not well constrained (hence its broken outline), but its upper surface must lie below the earths' surface since no basement crops out locally in this hangingwall. The location of event 6 which presumably occurred within the basement suggests that top basement is unlikely to be deeper than -2 km.

infinite slab formula gives an underestimate for the mean sediment thickness of 2.4 km, assuming a density contrast between basement and sediment of 300 kg m-'. Given that the seismograph network was positioned near the flank of the valley, and given that the actual density contrast between basement and sediments is likely to exceed this estimate, the local sediment thickness is probably less than this prediction. Starting with a sediment thickness of 1 km and a basement velocity of 6kms-' (a typical value for schist), the sediment velocity was varied in order to minimise root mean square residual time differences between observed and predicted arrival times (RMS). A minimum RMS of 0.1s was obtained with a sediment velocity of 2.4 km s-'. Using this velocity, the sediment thickness and basement velocity were systematically varied in an attempt to reduce RMS values further. However, no such adjustment of the basement velocity or sediment thickness reduced the RMS further. Locations for the six local events obtained using this velocity model are listed in Table 1, and plotted in Fig. 2. Using this procedure, the most reliable location (event 6) was WSW of the network at 2.1 km depth [labelled 6(a) in Table 1 and Fig. 21.

To estimate the extent of non-uniqueness of hypocentral locations, the P- and S-wave arrival times for event 6 were altered by f 0 . 0 5 ~ from their initial values (i.e. the estimated timing uncertainty). By increasing the P- and S-wave arrival times at station GE03 and reducing the basement velocity to 5.5 kms-', the RMS of event 6 could be reduced to 0.06s. This fit [labelled 6(b) in Table 1 and Fig. 21 had a hypocentre on or near the downdip continuation of the Hidirikoylu segment of the southern branch of the BMFZ (Fig. 3). This revised hypocentre is -1 km away from its original position. It is likely that making similar adjustments to the P and S arrival times of other events would alter their locations by similar amounts.

Formal standard errors in the hypocentral coordinates are

listed in Table 1. For events where few S-wave arrival times could be picked, these are mostly larger than the -1-2 km station spacings. In these cases the order in which each seismic phase arrived at stations provides a stronger constraint than the formal standard errors on hypocentral position. In the case of event 6, the relative timings of P- and S-wave arrivals are consistent with a hypocentre a few km WSW of the stations. The limiting factor controlling the reliability of this hypocentre is probably the limitation of the HYP071PC program that allows it to include only horizontal boundaries in velocity structure. Given that the unconformity at the base of the Neogene basin in the Biiyiik Menderes valley probably dips northward at a substantial angle, those refracted ray paths travelling through this structure probably differ substantially from those assumed by this program. Allowing for this effect would displace the hypocentre of event 6 east, probably to a point near lat. 37'53'N, lon. 27"36'E: also close to a downdip projection of the Hidirikoylu segment of the southern BMFZ branch. The duration of the seismogram for this event at GE03 is substantially shorter than that at GEOl and GE02. Assuming the main factor affecting the duration of the seismograms for event 6 is scattering on passing through Neogene sediments, this indicates the path length through Neogene sediments to GE03 is substantially less than to the other stations. This is consistent with the cross-section in Fig. 3.

There appears to be no correlation between microearthquake hypocentres and positions of geothermal wells. In particular, focal depths are typically deeper than the -1 km depths of these wells. Consequently, it is likely that the observed microearthquakes are tectonic rather than geothermal.

Microearthquake magnitudes Md were calculated from seismogram durations and hypocentral distances using a standard empirical formula derived originally by Lee & Lahr

Figure 4. Seismograms of the largest microearthquake recorded (event 6; Md = 2.0; at 00:29:30.0 on 1990 May l ) , at stations GEOl, CEO2 and GE03. Minute marker ticks were 60mm apart on the original records. P-wave arrival times were picked at 00:29:31.20 (GEOl), :31.10 (GE02), and :30.85 (GE03). S-wave arrival times were picked at 00:29:32.10 (GEOl), :31.90 (GE02), and :31.45 (GE03). All times are expressed using Greenwich Mean Time.

Microseismicity of the Germencik area 297

Table 1. Hypocentral coordinates of the six local earthquakes recorded during the period 1990 April 22-May 3. The magnitude Md given is calculated from an empirical relationship between earthquake size, source-station distance and seismogram duration, used by Lee & Lahr (1975) for Californian earthquakes. ERH and ERZ are the semimajor axes of the hypocentral error ellipsoid and approximate to the formal standard errors in the epicentral position and hypocentral depth. RMS is the root mean square residual between observed arrival time and arrival time expected from the hypocentral coordinates and the assumed velocity structure. Velocity models used to obtain these locations are described in the text.

Event Date GMT Latitude Longitude Depth(km) Md RMS(s) ERH (km) ERZ (km) Origin time Hypocentre Magnitude

1 22-Apr 21:30:43.87 37'52.35' 27'38.72 2.0 0.6+/-0.1 0.07 1 .I 1

2 28-Apr 00:33:09.10 37'53.55' 27'36.26 5.3 0.4+/-0.3 0.07 1.8 1.3

3 28-npr 01:00:31.11 37'52.25' 27'37.87' 4.0 0.4+/-0.3 0.05 2.1 1.4

4 28-Apr 03:21:51.85 37'53.31' 27'38.67' 5.0 1.8+/-0.4 0.03 2.2 0.4

5 28-Apr 0558:35.92 37'53.19' 27'37.24' 3.4 1.4+/-0.3 0.03 2.3 0.5

6(a) 1-May 00:29:29.88 37'53.12 27'34.63 2.1 2.0+/-0.2 0.10 0.3 0.3

6(b) I-May 00:29:29.97 37'52.89' 27'35.07' 1.3 2.0+/-0.2 0.06 0.2 0.6

(1975) for events in California. The unavailability of necessary seismometer calibration information prevented seismogram amplitudes from being used to calculate magnitudes instead.

The contribution to tectonic extension from the observed microseismicity can be estimated from Kostrov's (1974) equation relating corresponding elements of the strain rate tensor Eij in a volume of crust V, to the summed seismic moment tensor Mij of earthquakes within it:

Mij ( i , j = 1 to 3), 1 E. . = -

" 2pvt where I is the time over which seismic moments are summed and p is the shear modulus within the crust. The seismic moment released by a magnitude 2 earthquake is -lo1' N m [extrapolating the moment-magnitude relation of Hanks & Kanamori (1979)]. Suppose E l is the extensional eigenvalue of the strain rate tensor and Ml is the corresponding eigenvalue of the summed seismic moment tensor, which corresponds approximately to the T axes of individual earth- quakes. If the volume V is that of a parallelepiped with height h equal to the -10km upper crustal brittle layer thickness, width w and length L, then V = wLh. If w is measured parallel to the regional extension direction and L parallel to fault strike, then the rate of separation of the faces of the parallelepiped in the direction of w, is u,, where

Ml 2ptLh '

v,= E,w =-

If the observed activity is approximated by one magnitude 2 earthquake per week, and the seismicity occurs within a volume of crust between the earth's surface and depth h = 10 km, and with length L = 4 km along strike, then with shear modulus p = 3 X 10" Pa, the extension rate predicted by equation (2) is -0.02mmyr-' or -10-2mmyr-1 as an approximate order of magnitude estimate.

GEOMORPHOLOGY OF THE GERMENCIK AREA

Figure 5 shows the topography and present-day drainage of an area near the western edge of the seismograph network,

where the Omerbeyli and Hidirikoylu segments of the southern branch of the BMFZ overlap. The Hidirikoylu and Alangiillii rivers flow southward, approximately perpendicu- lar to the local fault strike, and have presumably incised the footwalls of these faults segments as they have uplifted. Between the two rivers a dry valley at -110m elevation runs sub-parallel to fault strike. It thus seems likely that uplift of the part of the footwall nearest its cut-off at one time constrained river flow to be sub-parallel to strike. The direction of flow was most likely eastward (see broken line, Fig. 5), causing flow to enter the Biiyiik Menderes valley where the Hidirikoylu fault segment dies out at its eastern end. This drainage pattern has most likely subsequently been altered by headward erosion of south-flowing streams through the uplifted footwall of this Hidirkoylu segment, which eventually captured the former river at the western end of this present-day dry valley.

The steepest bed dips in the footwall of the Omerbeyli segment of the BMFZ are -25". In contrast, the steepest bed dips in the footwall of the Hidirikoylu segment are 7". The steepest beds in the footwall of the Omerbeyli segment are near the base of the Neogene sedimentary sequence (Fig. 3), and therefore were presumably deposited shortly after extension began. Assuming uniform rate of tilting during extension, tilting has occurred at -25" per 7 Myr or 3.5" Myr-'. The 7" dips of the beds exposed in the footwall of the Hidirikoylu segment thus imply that these beds are -2 Myr old.

An estimate of the timing of river capture can be made by projecting the 7" bed dip southwards from the northern edge of the footwall of the Hidirikoylu segment to the position of the dry valley. With 150 m elevation at the northern edge of the footwall and 350m elevation at this southern footwall cut-off, the projected elevation of these 2 Myr old beds at the dry valley is -210111. The present day elevation of the valley is -110m, so the former river can be estimated to have incised its valley by -100m before capture. The present Hidirikoylu river valley has an elevation of -60x11 where it passes the dry valley, so it has been incised by -50m relative to the dry valley since capture. If it is assumed that the erosion rates for both rivers have been the same, the former river existed for roughly twice as long as

298 M . Jones and R. Westaway

Figure 5. Topographic map, with elevations in metres, of the boundary zone between the Hidirikoylu and Omerbeyli segments of the southern branch of the Biiyiik Menderes fault zone, showing present-day drainage and the postulated former course of the Hidirikoylu river. Observed bed dips are from Roberts (1988). See text for discussion.

the present Hidirikoylu river. This predicts that capture took place -0.7Myr ago. This reasoning implies that -180m thickness of sediments has been eroded from the footwall of the Hidirikoylu fault segment since uplift began. If less erosion than this has occurred, the date of capture would be correspondingly older, up to a maximum of -2 Myr.

SLIP A N D EXTENSION RATES ACROSS THE B ~ ~ Y U K MENDERES FAULT ZONE

The slip rate, us, on and extension rate, u,, across the active faults in the BMFZ can be estimated in several ways. One method is to divide the estimated total displacement across the fault zone by the time since extension began. North of the northern BMFZ branch, the Menderes massif rises to -2km elevation, whereas south of the southern BMFZ branch, the basement appears buried under -2km of sediments. Total throw of at least 4 km has thus occurred since the BMFZ became active, say 7Myr ago. Given present day fault dips of 45", 4 km of throw is equivalent to a slip of -6km, or us -0.9mmyr-' over the last 7Myr, corresponding to u, -0.6 mm yr-'. This estimate excludes erosion of the uplifted footwall of the northern BMFZ branch in the Menderes massif, so is likely to be a lower limit.

Second, the last documented major earthquake that caused slip on the Omerbeyli segment of the BMFZ occurred in 1653. If -1 m of slip occurred in this event, then over the last 340 yr average us is -3 mm yr-', corresponding to u, -2mmyr-'. This is likely to be an upper limit since this time interval excludes the time between the present and

that time in the future when the next major earthquake will occur.

A third, more indirect method to estimate us is to compare the relative footwall and hanging wall displace- ments with those predicted from isostatic equilibrium constraints. By considering the isostatic equilibrium of each side of a fault with dip 6, cutting through crust of density pc and thickness tc, overlain by sediment of density p, and underlain by a mantle of density p, that can be regarded as a fluid over long time-scales, Jackson & McKenzie (1983) show that the ratio of footwall motion to hanging wall motion, f, satisfies

where

(3)

(4)

1, the elastic thickness of the upper crust, has typical value 3-10 km. Substitution of p, = 3300 kg mP3, p, = 2800 kg r K 3 , ps = 2400 kg mP3, and t, = 30 km gives, for example, f = -0.80 for 1 = 0 and f = -0.85 for 1 = 10 krn.

Table 2 shows how this ratio varies with slip rates across the southern fault branch in this area, using structural information from the area. The footwall uplift on the Hiridikoylu segment of the southern BMFZ branch over the past 2Myr can be estimated, as before, from a southward continuation of the 7" bed dip from -150 m at the northern edge of its footwall, giving elevation -350m at the projected footwall cut-off. Helice, -350 m of footwall uplift

Microseismicity of the Germencik area 299

M d - 2 events per week or -250 per year. Gutenberg & Richter (1954) showed that frequency of occurrence of earthquakes of different magnitudes M in any region is related as

log,, N ( M ) = a - bM, (5 )

where a and b are constants for any given region with b - 1, and N ( M ) is the number of earthquakes with magnitude M or greater. Given this value of b, around one event with M - 7 would be expected every -400yr if - 2 5 0 M - 2 events occur per year. This is in broad agreement with the documented local historical seismicity (Ambraseys 1988), and suggests that equation (5) may well be followed on this fault segment over the magnitude range from 2 to 7.

Table 2. The variation with slip rate of the ratio f of footwall uplift to hanging wall subsidence of the southern Biiyiik Menderes fault branch (see text for explanation). Simple isostatic equilibrium considerations constrain the ratio f to the range - -0.6 to -0.9, inferring a slip rate of between 1.0-1.5 mm yr-' on the Hidirikoylu segment.

Sllp rate

0.5 0.7 0.8 0.9 1

1.1 1.2 1.5

(mm/yr)

Hldlrlkdylu Throw

(m) 710 990 1130 1270 1410 1550 1700 2120

segment Footwall upllft (m) 350 350 350 350 350 350 350 350

Hanglng wall subsldence (m)

360 640 780 920 1060 1200 1350 1770

Total lootwall upllll (m) 1000 1000 1000 1000 1000 1000 1000 1000

Ratlo f

-2.8 -1.6 -1.3 -1.1 -0.9 -0.8 -0.7 -0.6

has taken place since these beds were horizontal -2 Myr ago (column 3, Table 2). Using a similar argument for the footwall uplift of the Omerbeyli segment of the BMFZ gives -650 m of uplift over the last 2 Myr, or a total uplift due to slip on the two fault segments of -lOOOm (column 5). Estimated throw on this part of the Omerbeyli segment over the past 2 Myr is thus -500 m (i.e. 650-150 m), indicating us -0.4 mm yr-'. Assuming -350 m of footwall uplift on the Hidirikoylu segment, hanging wall subsidence was deduced by subtracting this uplift from the throw produced by the various slip rates acting over 2 Myr (i.e. in Table 2, column 4 = column 2 - column 3).

By comparing the expected value of the ratio f, - -0.8 to -0.85, calculated from equation (3) above, to the range of possible values given in Table 2 , v, on the Hidirikoylu segment can be constrained to -1.0-1.1 mm yr-'. Howe- ver, equation (3) assumes sediment is loaded on both sides of a single fault, whereas in this case the ratio is calculated from displacements either side of two fault segments with basement outcropping in part of the northern footwall, and so must be considered as an approximation to the real situation in this particular fault zone (although it is likely that the block between the two fault segments is sufficiently narrow to not significantly affect isostatic equilibrium). In addition, the ratio f is a sensitive function of the assumed densities. For example, if the assumed sediment density were increased by 100 kg m-' to p, = 2500 kg m-', the ratio f becomes -0.60 to 0.68, inferring v, - 1.4-1.5 mm yr-'. Given the -45" dips of both fault segments, and allowing for the -0.4 mm yr-' slip rate on the Omerbeyli segment, f- -0.8 to -0.85 in Table 2 is consistent with ve-l.O-l.lmmyrF1, and f- -0.6 to 0.68 with ve- 1.3mmyrF' across both fault zones. These estimates are consistent with the earlier reasoning that -180 m thickness of sediment has been eroded from the footwall of the Hidirikoylu fault segment. If less erosion than this has occurred, estimated extension rate would be less across the Hidirikoylu segment and correspondingly greater across the Omerbeyli segment, but overall extension rate would be unaffected.

As already mentioned, the microseismicity around Germencik observed with Md < -2 contributes only -lo-* mm yr-' of extension, which is negligible compared with the overall extension rates across the BMFZ estimated from structural studies. This microseismicity comprises around one Md - 2 event per week on a -4 km length of this fault zone. The overall -20 km length of the Omerbeyli segment would thus be expected to experience around five

CONCLUSIONS

The microseismicity within the study area near the boundary zone between the Hidirikoylu and Omerbeyli segments of the Biiyiik Menderes fault zone was lower than expected during the period of fieldwork, with an average of less than one local microearthquake recorded per day. The largest local event recorded (event 6, Md = 2.0) occurred near the Hidirikoylu segment. Of the three methods used to constrain slip rate, us, on and extension rate, v,, across the Biiyiik Menderes fault zone, that based on isostatic equilibrium considerations appears the most reliable, and predicts u s - 1.7 f 0.6 mm yr-l and v, - 1.2 f 0.4 mm yr-'. The observed seismicity during the study period contributes -lo-* mm yr-' to this extension rate, indicating that most of the extension across this fault segment is associated with large, infrequent earthquakes such as the event in 1653. However, the local occurrence rate of magnitude 2 and 7 events approximately satisfies the standard relationship of Gutenberg & Richter (1954). An example of river capture has been identified, due to rapid headward erosion through unconsolidated sediments in an uplifted fault footwall. Capture is estimated to have taken place -0.7 Myr ago.

ACKNOWLEDGMENTS

We thank the Natural Environmental Research Council (NERC) for financial support through a research student- ship (MJ) and research grant GR3/6967 (RW). We also thank the NERC Geophysical Equipment Pool for loan of seismological equipment (loan no. 171-0988). John Knipe (Newcastle University) kindly made available an instrument for precisely measuring arrival times of seismic phases. Kuwet Atakan (Durham University), Alkan Tetik, Mustafa Ergiin (Dokuz Eyliil University, Izmir), and numerous Turkish Geological Survey personnel at the Omerbeyli geothermal research station provided logistical support. We are also grateful to two anonymous referees for their helpful comments.

REFERENCES

Aksig, Z., 1988. Analysis of the tectonic problems of Western Anatolia with gravity data, Geol. Bull. Turkey, 31, 63-70 (in Turkish with English abstract).

Ambraseys, N. N., 1988. Engineering seismology, Earthquake Eng. struct. Dyn., 17, 1-105.

300 M. Jones and R. Westaway

Gutenberg, B. & Richter, C. F., 1954. Seismicify of the Earth and Associated Phenomena, Princeton University Press, Princeton, NJ.

Hanks, T. C. & Kanamori, H., 1979. A moment-magnitude scale, J. geophys. Res., 84, 2348-2350.

Jackson, J. A. & McKenzie, D. P., 1983. The geometrical evolution of normal fault systems, J. muct. Geol., 5 , 471-482.

Kaya, O., 1981. Miocene reference section for the coastal parts of west Anatolia, Newsl. Stratigr., 10, 164-191.

Kostrov, V. V., 1974. Seismic moment and energy of earthquakes, and seismic flow of rock, Izv. Acad. Sci. USSR, Phys. Solid Earth, 10, 23-44 (English translation).

Lee, W. H. K. & Lahr, J. C., 1975. HYPO71 (Revised): A computer program for determining hypocentre, magnitude, and

first motion pattern of local earthquakes, Open-File Reporr 75-311, US Geological Survey, Menlo Park, CA.

Roberts, S., 1988. Active normal faulting in central Greece and western Turkey, PhD thesis, University of Cambridge.

Sen, S. & Valet, J. P., 1986. Magnetostratigraphy of late Miocene continental deposits in Samos, Greece, Earth planet Sci. Lett.,

Turgut, S., 1988. Observations of the Aegean Sea from the point of view of hydrocarbon exploration, Turk Assoc. Petrol. Geol. Bull., 1, 27-38 (in Turkish with English abstract).

Vita-Finzi, C. & King, G. C. P., 1985. The seismicity, geomorphology and structural evolution of the Corinth area of Greece, Phil. Trans. R. SOC. Lond., A, 314,379-407.

80, 167-174.