• from Altunel et al. (1999)

• from Altunel et al. (1999:297)

The NW–SE-trending Dinar fault is an active normal fault upon which the 1 October 1995 earthquake (M = 6.1) occurred. The 1995 earthquake resulted in a c. 10-km-long surface rupture with the south side down-thrown by ≤ 50 cm. Investigations of two trench sites perpendicular to the 1995 rupture suggest at least two prior large earthquakes in historical times. Radiocarbon dates and historical records constrain the age of events between 1500 BC and AD 53, event 2 possibly coinciding with the earthquake that damaged Dinar (the ancient city of Apamea Kibotos) in c. 80 BC and event 1 around 1500 BC. Surface displacements determined for events 1 and 2, compared to the 1995 surface faulting, indicate that M > 6.8 earthquakes were associated with each rupture. Using the total displacement in trenches, a slip rate of about 1 mm yr^-1 can be estimated for the Dinar fault. Observations suggest that the return period for large earthquakes in the Dinar area is about 1500–2000 years.

• from Altunel et al. (1999:297-298)

The major tectonic features of southwestern Turkey are E–W-trending grabens (e.g. the Büyük Menderes graben, Gökova graben) and NE–SW-trending strike-slip faults (e.g. the Fethiye–Burdur fault zone) (Fig. 1a). Several NW–SE-trending active normal faults cut across these E–W and NE–SW-trending major structures. One of these NW–SE-trending normal faults is the Dinar fault named after a nearby town by Ozturk (1981). The Dinar fault delimits the NE–SW-trending Baklan, Acıgöl and Burdur basins in the north (Fig. 1b). These actively forming basins have developed near the northeastern end of the NE–SW-trending Fethiye–Burdur fault zone, which is the northeastern continuation of the Pliny–Strabo Fault zone (Barka et al., 1995). Although not clearly supported by structural and seismic data, Sengör et al. (1985) and Price and Scott (1994) have suggested that the Dinar fault, which reactivated during the 1 October 1995 Dinar earthquake (M = 6.1), might be one of the major break-away faults in the north based on the fault geometry and kinematics. Price and Scott (1994) further discussed the width of the Dinar basin and speculated that the Dinar fault might not be presently active. However, the 1 October 1995 earthquake contradicts this idea. In the present study, we investigate palaeoseismic events along the Dinar fault within the 1995 rupture zone to understand the long-term behaviour of the Dinar fault and its importance and role within the neotectonics of southwestern Turkey.

The 1995 Dinar earthquake occurred on the Dinar fault, which is about 60-km long and dips c. 60° SW. It seriously damaged the nearby town of Dinar and resulted in 96 deaths and hundreds of injuries. Furthermore, the earthquake destroyed 4400 houses and damaged a further 10,000 houses. The total cost of damage was estimated at US$500,000. The 1995 event was associated with faulting in the hanging wall of the NW-trending Dinar fault with the southwestern side down-thrown. A c. 10-km-long surface rupture parallel to the pre-existing Dinar fault was mapped by Eyidoğan and Barka (1996). The trace of the rupture zone is sinuous in plan but its general trend is NW–SE. The maximum vertical offset is about 50 cm and the rupture zone has a slight strike-slip component in places where the fault changes in strike from its general trend (Eyidoğan and Barka, 1996).

The earthquake record in the Dinar area extends back at least 2800 years. Dinar is the modern name of the ancient city of Apamea Kibotos which was founded in the 8th century BC (Akurgal, 1995); historical sources in the Dinar area provide reliable information about macroseismic events before the instrumental period. Historical sources interpreted by, e.g., Ergin et al. (1967), Soysal et al. (1981), Strabo and Guidoboni et al. (1994), report that the ancient city of Apamea Kibotos was damaged by earthquakes in 400 BC, 88 BC and AD 53. More recent earthquakes also damaged the modern town of Dinar in 1875, 1914, 1925 and 1971. Historical sources show no records of large earthquakes between AD 53 and 1875. Although the 1875 (I_O = IX–X) and 1925 (I_O = VII–IX) earthquakes damaged Dinar, Ambraseys (1975, 1988) reported that they occurred on the NE–SW-trending Baklan fault (Fig. 1). Similarly, the 3 October 1914 (I_O = IX) and 12 May 1971 (M = 6.2) earthquakes, that caused damage in Dinar town, clearly occurred on the NE–SW-trending Fethiye–Burdur fault zone (Ergin et al., 1967; Ering et al., 1971; Ambraseys and Finkel, 1987).

Historical records report that earthquakes destroyed the ancient city of Apamea Kibotos. However, it is not clear whether these events occurred on the Dinar fault; in the present study, we use field evidence and historical information to elucidate this.

• from Altunel et al. (1999:298-299)

Figure 2 shows the locations of trench sites. Trench 1, which is about 10-m long and 3-m high, was first dug perpendicular to the 1995 earthquake break by Demirtaş et al. (1995) to investigate whether the 1995 break followed a pre-existing fault. The trench wall clearly shows six different units which are nearly horizontal (Fig. 3). Units B, C, D, E and F can be correlated with the hanging wall of the 1995 break but units B and C disappear in the footwall where only D, E and F can be correlated on the footwall. Unit A exists in the footwall of the 1995 break but it suddenly disappears in the hanging wall (Fig. 3). There is an angular unconformity between unit A and units B, C and D. Units A and C are blanketed by unit D, which extends horizontally from one end of the trench wall to the other. All of these units are colluviums comprising angular clasts in various sizes derived from the basement rock. Only unit D is rich in clay and organic materials and it has fewer clasts. A charcoal sample from unit D yielded an age of 590 ± 50 yr BP. As shown in Fig. 3, the 1995 break offsets units A, D, E and F by up to 30 cm. A nearly vertical fissure that developed during the 1995 earthquake follows the unconformity between unit A and units B and C (Fig. 3). This S-facing unconformity probably reflects a degraded fault scarp.

Trench 2 is a quarry site located on the hanging wall of the Dinar fault (Fig. 4a). This trench is about 50-m long and 15-m high and it cuts the 1995 earthquake break perpendicularly. The trench site displays well-stratified Holocene colluvium deposits and complex normal faulting (Figs 4b and 5). Unit a, exposed at the base of the cliff, includes two key stratums that provide reliable information about the amount of vertical displacement. The unconsolidated coarse gravels and a gently S-dipping tuffite layer within unit a are clearly offset by conjugate normal faults (see Figs 4(b) and 5). Fault F1, dipping 80° S, cuts through unit a and offsets the tuffite layer vertically by about 78 cm. The F1 fault is truncated by unit b. Fault F1-1, offsetting the tuffite layer 20 cm vertically, is parallel to the F1 fault and disappears within unit a. Fault F2, dipping 75° S, cuts both units a and b, with vertical displacements of c. 215 cm and 140 cm, respectively. Faults F2-1 and F2-2, cutting both units a and b, dip north with vertical displacements of 72 cm and 45 cm, respectively. Fault F2-3 dips S with just a few centimetres vertical displacement. Faults F2, F2-1 and F2-2 are blanketed by unit c but the F2-3 fault disappears within unit b. The F3 fault represents the 1995 earthquake break, offsetting all units exposed on this cliff by about 30 cm.

• from Altunel et al. (1999:299-301)

The 1 October 1995 Dinar earthquake (M = 6.1) showed that the Dinar fault is active and capable of producing large earthquakes. Surface rupture, source characteristics and aftershocks of the 1995 earthquake were studied in detail (e.g. Demirtaş et al., 1995; Eyidoğan and Barka, 1996; Kalafat, 1996; Kara et al., 1996; Aktar et al., 1997; Koral et al., 1997) but the long-term behaviour of the Dinar fault is unknown. Sengör et al. (1985) and Price and Scott (1994) suggested that the NW–SE-trending Dinar fault might be a major break-away fault from the NE–SW-trending faults. Thus, estimation of the rates, style of present-day deformation and the assessment of seismic hazard in the region have become extremely important.

The 1995 rupture zone is located in the hanging wall of the main Dinar fault. The rupture zone is parallel to the main Dinar fault-line but does not follow it on the surface. The sense of motion on the 1995 earthquake break is normal with a slight strike-slip component (Eyidoğan and Barka, 1996).

Does the 1995 earthquake break follow a pre-existing break or is it a new rupture? As Figs 3 and 5 show, the 1995 fault does not follow exactly pre-existing faults (e.g. F1 and F2 faults) in trench walls, but they are close to each other. The horizontal distance between the 1995 break and pre-existing faults varies near the surface; for example, it is about 1 m in trench 1 (Fig. 3) but the nearest fault is located about 8 m away from the 1995 break in trench 2 (Fig. 5). This variation in distance probably indicates that faults developed at different times are anastomosing in the hanging wall of the main fault. As Fig. 6(a) illustrates, it is possible that faults developing in the hanging wall of the main fault are branching and rejoining (e.g. Dresen et al., 1991). In a cross-section from the branching part, faults appear as different segments and the total displacement is partitioned unequally (Fig. 6b). However, where faults rejoin they appear as a single fault in cross-section and the displacement is cumulative (Fig. 6c). On the basis of this observation, it can be concluded that the 1995 earthquake break is branching and interconnecting with the pre-existing faults.

The trench sites across the 1995 break showed evidence of at least two past events (called event 1 and event 2). The F1 and F1-1 faults cut unit a but they are truncated by unit b (Fig. 5). Thus, F1 and F1-1 are simultaneous, produced during the same event (event 1). However, although the F2, F2-1, F2-2 and F2-3 faults cut both units a and b they are covered by unit c (Fig. 5). The F2-1 and F2-2 faults are antithetic and F2-3 is synthetic to the F2 fault. Thus, the F2 fault is the main fault and F2-1, F2-2 and F2-3 are accommodation structures developed at the same time during the same event (event 2). The amount of displacement caused by F2 varies between stratigraphic units. For example, the coarse gravel level and unit b are offset by F2 but the displacements are about 215 cm and 140 cm, respectively (Fig. 5). Different amounts of displacement on F2 indicate that it reactivated before event 2. The c. 140 cm offset of unit b indicates that the previous displacement on F2 fault is about 75 cm, which is similar to that on F1 fault. Thus, it is possible that the F2 fault moved during event 1 with F1 fault. On the basis of these observations, it can be concluded that there were at least two different faulting events before the 1995 earthquake and that event 1 is relatively older than event 2.

Samples were collected from stratified units in trench walls in order to determine precise ages of events but only two samples yielded reliable ages. Samples collected from unit D in trench 1 (see Fig. 3 for location) and from the tuffite layer yielded ages of 590 ± 50 yr BP and about 3500 yr BP, respectively. Certain dates of events cannot be assessed unambiguously with these obtained ages but they do bracket the ages of events 1 and 2. The age of 590 ± 50 yr BP shows that except for the 1995 Dinar earthquake there has not been any large earthquake for at least the last c. 600 years. Although the age of 590 ± 50 yr BP is the upper boundary for event 2 according to historical sources, there is no record of any large earthquake in the Dinar area between AD 53 and 1875. Thus, it can be concluded that the Dinar fault reactivated at two different times between 3500 BC and AD 53. It is significant that when an important city was destroyed by an earthquake in historical times in western Turkey either the city centre was shifted or the city was rebuilt (e.g. the ancient cities of Hierapolis, Priene, Tralles; Hancock and Altunel, 1997; Altunel, 1998, 1999). It is also noteworthy that Strabo (a travel writer and geographer who lived from 63 BC to AD 21) reports that Apamea Kibotos, one of the most important cities of Phrygia, was damaged by an earthquake around 80 BC, and when King Mithridates (king of Pontus 120–63 BC) won a war against Rome (88–84 BC) he gave financial help for reconstruction of the city. In view of the historical records and of the field evidence it is suggested that event 2 probably occurred in 88 BC and that the last reactivation of the F2 fault and offset in unit b were associated with the 88 BC earthquake.

The history of the Dinar area before the 8th century BC is unknown because the ancient city of Apamea Kibotos was founded around the 8th century BC (Akurgal, 1995). There is a historical record of an earthquake in 400 BC in the Dinar area, but there is no account of reconstruction of the ancient city. This information indicates that event 1 probably occurred before the 8th century BC. The age of about 3500 yr BP is the lower boundary for event 1 which corresponds with 1500 BC. As outlined above, although the 1875, 1914, 1925 and the 1971 earthquakes damaged Dinar, they did not occur on the Dinar fault (Ergin et al., 1967; Ambraseys, 1975, 1988; Ering et al., 1971; Ambraseys and Finkel, 1987). Similarly, as pointed out by several historical records, the 400 BC and the 53 AD earthquakes damaged the ancient city of Apamea Kibotos but they probably did not occur on the Dinar fault. Thus, on the basis of historical records and of the field observations, a return period of about 1500–2000 years for large earthquakes can be estimated in the Dinar area. However, the differing amounts of displacement on faults associated with earthquakes indicate that their magnitudes were not the same.

Fault motion at depth does not all reach the surface but this depends on the length of straight fault (Vita-Finzi and King, 1985). Because the Dinar fault is straight (Fig. 1b), as the 1995 event has shown, this fault can produce surface scarp. Assuming that the rupture length and surface offset are directly proportional to magnitude, further interpretations can be made. The magnitude of the 1995 earthquake was 6.1 and it produced a c. 10-km-long surface break with a maximum 50 cm vertical displacement. The vertical displacement is about 78 cm on fault F1 and 140 cm on fault F2. Considering that 50 cm vertical displacement occurred during the 1995 earthquake, the 140 cm vertical displacement seems exaggerated. However, normal faults in southwestern Turkey are capable of producing such offsets during one event. For example, the 20 September 1899 Menderes earthquake (I₀ = IX) was associated with normal faulting giving as much as 2 m vertical displacement (Ergin et al., 1967; İlhan, 1971; Allen, 1975; Sipahioğlu, 1979; Ambraseys and Finkel, 1987; Altunel, 1999). The maximum vertical displacements of 78 cm and 140 cm on F1 and F2 faults suggest that the magnitudes of events 1 and 2 were larger than 6.8 and that a large portion of the Dinar fault must have been ruptured.

The total displacement measured in the trenches for the last 3500 years is about 3.50 m. This gives rise to about 1 mm yr^-1 slip rate on the Dinar fault. Using this value, about 0.4 mm yr^-1 extension rate can be estimated for the Dinar fault. Assuming that the initiation of the neotectonic regime in southwestern Anatolia began with the anticlockwise rotation of the Anatolian block (Barka and Reilinger, 1997) about 4–5 Ma ago, and that the Dinar fault was formed in this new regime, the total extension across the Dinar fault should be about 1.6–2 km. About 2 km width of the basin in front of the Dinar fault (Fig. 2) is consistent with this value.

In summary, our palaeoseismological investigation into the 1995 earthquake rupture illustrates that two large events have occurred on the Dinar fault during the last 3500 years and that these events were larger than the 1995 earthquake. The approximate recurrence interval of large earthquakes is estimated to be 1500–2000 years and there is a slip rate of c. 1 mm yr^-1.

• from Altunel et al. (1999)

Trench 1 – The trench wall clearly shows six different units which are nearly horizontal (Fig. 3). Units B, C, D, E and F can be correlated with the hanging wall of the 1995 break but units B and C disappear in the footwall where only D, E and F can be correlated. Unit A exists in the footwall of the 1995 break but it suddenly disappears in the hanging wall (Fig. 3). There is an angular unconformity between unit A and units B, C and D. Units A and C are blanketed by unit D, which extends horizontally from one end of the trench wall to the other. All of these units are colluviums comprising angular clasts in various sizes derived from the basement rock. Only unit D is rich in clay and organic materials and it has fewer clasts. A charcoal sample from unit D yielded an age of 590 ± 50 yr BP. As shown in Fig. 3, the 1995 break offsets units A, D, E and F by up to 30 cm. A nearly vertical fissure that developed during the 1995 earthquake follows the unconformity between unit A and units B and C (Fig. 3). This south-facing unconformity probably reflects a degraded fault scarp.

Trench 2 – a quarry site located on the hanging wall of the Dinar fault (Fig. 4a). This trench is about 50 m long and 15 m high and it cuts the 1995 earthquake break perpendicularly. The trench site displays well-stratified Holocene colluvium deposits and complex normal faulting (Figs 4b and 5). Unit a, exposed at the base of the cliff, includes two key strata that provide reliable information about the amount of vertical displacement. The unconsolidated coarse gravels and a gently south-dipping tuffite layer within unit a are clearly offset by conjugate normal faults (see Figs 4b and 5). Fault F1, dipping 80° S, cuts through unit a and offsets the tuffite layer vertically by about 78 cm. The F1 fault is truncated by unit b. Fault F1-1, offsetting the tuffite layer 20 cm vertically, is parallel to the F1 fault and disappears within unit a. Fault F2, dipping 75° S, cuts both units a and b, with vertical displacements of c. 215 cm and 140 cm, respectively. Faults F2-1 and F2-2, cutting both units a and b, dip north with vertical displacements of 72 cm and 45 cm, respectively. Fault F2-3 dips south with just a few centimetres vertical displacement. Faults F2, F2-1 and F2-2 are blanketed by unit c but the F2-3 fault disappears within unit b. The F3 fault represents the 1995 earthquake break, offsetting all units exposed on this cliff by about 30 cm.

The 1995 rupture zone is located in the hanging wall of the main Dinar fault. The rupture zone is parallel to the main Dinar fault line but does not follow it on the surface. The sense of motion on the 1995 earthquake break is normal with a slight strike-slip component (Eyidogan and Barka, 1996). Does the 1995 earthquake break follow a pre-existing break or is it a new rupture? As Figs 3 and 5 show, the 1995 fault does not follow exactly pre-existing faults (e.g. F1 and F2 faults) in trench walls, but they are close to each other. The horizontal distance between the 1995 break and pre-existing faults varies near the surface; for example, it is about 1 m in trench 1 (Fig. 3) but the nearest fault is located about 8 m away from the 1995 break in trench 2 (Fig. 5). This variation in distance probably indicates that faults developed at different times are anastomosing in the hanging wall of the main fault. As Fig. 6a illustrates, it is possible that faults developing in the hanging wall of the main fault are branching and rejoining (e.g. Dresen et al., 1991). In cross-section from the branching part, faults appear as different segments and the total displacement is partitioned unequally (Fig. 6b). However, where faults rejoin they appear as a single fault in cross-section and the displacement is cumulative (Fig. 6c). On the basis of this observation, it can be concluded that the 1995 earthquake break is branching and interconnecting with the pre-existing faults.

The trench sites across the 1995 break showed evidence of at least two past events (called event 1 and event 2). The F1 and F1-1 faults cut unit a but they are truncated by unit b (Fig. 5). Thus, F1 and F1-1 are simultaneous, produced during the same event (event 1). However, although the F2, F2-1, F2-2 and F2-3 faults cut both units a and b they are covered by unit c (Fig. 5). The F2-1 and F2-2 faults are antithetic and F2-3 is synthetic to the F2 fault. Thus, the F2 fault is the main fault and F2-1, F2-2 and F2-3 are accommodation structures developed at the same time during the same event (event 2). The amount of displacement caused by F2 varies between stratigraphic units. For example, the coarse gravel level and unit b are offset by F2 but the displacements are about 215 cm and 140 cm, respectively (Fig. 5). Different amounts of displacement on F2 indicate that it reactivated before event 2. The c. 140 cm offset of unit b indicates that the previous displacement on F2 fault is about 75 cm, which is similar to that on F1 fault. Thus, it is possible that the F2 fault moved during event 1 with F1 fault. On the basis of these observations, it can be concluded that there were at least two different faulting events before the 1995 earthquake and that event 1 is relatively older than event 2.

Samples were collected from stratified units in trench walls in order to determine precise ages of events but only two samples yielded reliable ages. Samples collected from unit D in trench 1 (see Fig. 3 for location) and from the tuffite layer yielded ages of 590 ± 50 yr BP and about 3500 yr BP, respectively. Certain dates of events cannot be assessed unambiguously with these obtained ages but they do bracket the ages of events 1 and 2. The age of 590 ± 50 yr BP shows that except for the 1995 Dinar earthquake there has not been any large earthquake for at least the last c. 600 years. Although the age of 590 ± 50 yr BP is the upper boundary for event 2 according to historical sources, there is no record of any large earthquake in the Dinar area between AD 53 and 1875. Thus, it can be concluded that the Dinar fault reactivated at two different times between 3500 BC and AD 53.

Speculation on the date of Event 2 – It is significant that when an important city was destroyed by an earthquake in historical times in western Turkey either the city centre was shifted or the city was rebuilt (e.g. the ancient cities of Hierapolis, Priene, Tralles; Hancock and Altunel, 1997; Altunel, 1998, 1999). It is also noteworthy that Strabo reports that Apamea Kibotos, one of the most important cities of Phrygia, was damaged by an earthquake around 80 BC, and when King Mithridates (king of Pontus 120–63 BC) won a war against Rome (88–84 BC) he gave financial help for reconstruction of the city. In view of the historical records and of the field evidence it is suggested that event 2 probably occurred in 88 BC and that the last reactivation of the F2 fault and offset in unit b were associated with the 88 BC earthquake.

Speculation on the date of Event 1 – The history of the Dinar area before the 8th century BC is unknown because the ancient city of Apamea Kibotos was founded around the 8th century BC (Akurgal, 1995). There is a historical record of an earthquake in 400 BC in the Dinar area, but there is no account of reconstruction of the ancient city. This information indicates that event 1 probably occurred before the 8th century BC. The age of about 3500 yr BP is the lower boundary for event 1 which corresponds with 1500 BC.

• from Altunel et al. (1999:297)

The NW–SE-trending Dinar fault is an active normal fault upon which the 1 October 1995 earthquake (M = 6.1) occurred. The 1995 earthquake resulted in a c. 10-km-long surface rupture with the south side down-thrown by ≤ 50 cm. Investigations of two trench sites perpendicular to the 1995 rupture suggest at least two prior large earthquakes in historical times. Radiocarbon dates and historical records constrain the age of events between 1500 BC and AD 53, event 2 possibly coinciding with the earthquake that damaged Dinar (the ancient city of Apamea Kibotos) in c. 80 BC and event 1 around 1500 BC. Surface displacements determined for events 1 and 2, compared to the 1995 surface faulting, indicate that M > 6.8 earthquakes were associated with each rupture. Using the total displacement in trenches, a slip rate of about 1 mm yr^-1 can be estimated for the Dinar fault. Observations suggest that the return period for large earthquakes in the Dinar area is about 1500–2000 years.

• from Altunel et al. (1999:297-298)

The major tectonic features of southwestern Turkey are E–W-trending grabens (e.g. the Büyük Menderes graben, Gökova graben) and NE–SW-trending strike-slip faults (e.g. the Fethiye–Burdur fault zone) (Fig. 1a). Several NW–SE-trending active normal faults cut across these E–W and NE–SW-trending major structures. One of these NW–SE-trending normal faults is the Dinar fault named after a nearby town by Ozturk (1981). The Dinar fault delimits the NE–SW-trending Baklan, Acıgöl and Burdur basins in the north (Fig. 1b). These actively forming basins have developed near the northeastern end of the NE–SW-trending Fethiye–Burdur fault zone, which is the northeastern continuation of the Pliny–Strabo Fault zone (Barka et al., 1995). Although not clearly supported by structural and seismic data, Sengör et al. (1985) and Price and Scott (1994) have suggested that the Dinar fault, which reactivated during the 1 October 1995 Dinar earthquake (M = 6.1), might be one of the major break-away faults in the north based on the fault geometry and kinematics. Price and Scott (1994) further discussed the width of the Dinar basin and speculated that the Dinar fault might not be presently active. However, the 1 October 1995 earthquake contradicts this idea. In the present study, we investigate palaeoseismic events along the Dinar fault within the 1995 rupture zone to understand the long-term behaviour of the Dinar fault and its importance and role within the neotectonics of southwestern Turkey.

The 1995 Dinar earthquake occurred on the Dinar fault, which is about 60-km long and dips c. 60° SW. It seriously damaged the nearby town of Dinar and resulted in 96 deaths and hundreds of injuries. Furthermore, the earthquake destroyed 4400 houses and damaged a further 10,000 houses. The total cost of damage was estimated at US$500,000. The 1995 event was associated with faulting in the hanging wall of the NW-trending Dinar fault with the southwestern side down-thrown. A c. 10-km-long surface rupture parallel to the pre-existing Dinar fault was mapped by Eyidoğan and Barka (1996). The trace of the rupture zone is sinuous in plan but its general trend is NW–SE. The maximum vertical offset is about 50 cm and the rupture zone has a slight strike-slip component in places where the fault changes in strike from its general trend (Eyidoğan and Barka, 1996).

The earthquake record in the Dinar area extends back at least 2800 years. Dinar is the modern name of the ancient city of Apamea Kibotos which was founded in the 8th century BC (Akurgal, 1995); historical sources in the Dinar area provide reliable information about macroseismic events before the instrumental period. Historical sources interpreted by, e.g., Ergin et al. (1967), Soysal et al. (1981), Strabo and Guidoboni et al. (1994), report that the ancient city of Apamea Kibotos was damaged by earthquakes in 400 BC, 88 BC and AD 53. More recent earthquakes also damaged the modern town of Dinar in 1875, 1914, 1925 and 1971. Historical sources show no records of large earthquakes between AD 53 and 1875. Although the 1875 (I_O = IX–X) and 1925 (I_O = VII–IX) earthquakes damaged Dinar, Ambraseys (1975, 1988) reported that they occurred on the NE–SW-trending Baklan fault (Fig. 1). Similarly, the 3 October 1914 (I_O = IX) and 12 May 1971 (M = 6.2) earthquakes, that caused damage in Dinar town, clearly occurred on the NE–SW-trending Fethiye–Burdur fault zone (Ergin et al., 1967; Ering et al., 1971; Ambraseys and Finkel, 1987).

Historical records report that earthquakes destroyed the ancient city of Apamea Kibotos. However, it is not clear whether these events occurred on the Dinar fault; in the present study, we use field evidence and historical information to elucidate this.

• from Altunel et al. (1999:298-299)

Figure 2 shows the locations of trench sites. Trench 1, which is about 10-m long and 3-m high, was first dug perpendicular to the 1995 earthquake break by Demirtaş et al. (1995) to investigate whether the 1995 break followed a pre-existing fault. The trench wall clearly shows six different units which are nearly horizontal (Fig. 3). Units B, C, D, E and F can be correlated with the hanging wall of the 1995 break but units B and C disappear in the footwall where only D, E and F can be correlated on the footwall. Unit A exists in the footwall of the 1995 break but it suddenly disappears in the hanging wall (Fig. 3). There is an angular unconformity between unit A and units B, C and D. Units A and C are blanketed by unit D, which extends horizontally from one end of the trench wall to the other. All of these units are colluviums comprising angular clasts in various sizes derived from the basement rock. Only unit D is rich in clay and organic materials and it has fewer clasts. A charcoal sample from unit D yielded an age of 590 ± 50 yr BP. As shown in Fig. 3, the 1995 break offsets units A, D, E and F by up to 30 cm. A nearly vertical fissure that developed during the 1995 earthquake follows the unconformity between unit A and units B and C (Fig. 3). This S-facing unconformity probably reflects a degraded fault scarp.

Trench 2 is a quarry site located on the hanging wall of the Dinar fault (Fig. 4a). This trench is about 50-m long and 15-m high and it cuts the 1995 earthquake break perpendicularly. The trench site displays well-stratified Holocene colluvium deposits and complex normal faulting (Figs 4b and 5). Unit a, exposed at the base of the cliff, includes two key stratums that provide reliable information about the amount of vertical displacement. The unconsolidated coarse gravels and a gently S-dipping tuffite layer within unit a are clearly offset by conjugate normal faults (see Figs 4(b) and 5). Fault F1, dipping 80° S, cuts through unit a and offsets the tuffite layer vertically by about 78 cm. The F1 fault is truncated by unit b. Fault F1-1, offsetting the tuffite layer 20 cm vertically, is parallel to the F1 fault and disappears within unit a. Fault F2, dipping 75° S, cuts both units a and b, with vertical displacements of c. 215 cm and 140 cm, respectively. Faults F2-1 and F2-2, cutting both units a and b, dip north with vertical displacements of 72 cm and 45 cm, respectively. Fault F2-3 dips S with just a few centimetres vertical displacement. Faults F2, F2-1 and F2-2 are blanketed by unit c but the F2-3 fault disappears within unit b. The F3 fault represents the 1995 earthquake break, offsetting all units exposed on this cliff by about 30 cm.

• from Altunel et al. (1999:299-301)

The 1 October 1995 Dinar earthquake (M = 6.1) showed that the Dinar fault is active and capable of producing large earthquakes. Surface rupture, source characteristics and aftershocks of the 1995 earthquake were studied in detail (e.g. Demirtaş et al., 1995; Eyidoğan and Barka, 1996; Kalafat, 1996; Kara et al., 1996; Aktar et al., 1997; Koral et al., 1997) but the long-term behaviour of the Dinar fault is unknown. Sengör et al. (1985) and Price and Scott (1994) suggested that the NW–SE-trending Dinar fault might be a major break-away fault from the NE–SW-trending faults. Thus, estimation of the rates, style of present-day deformation and the assessment of seismic hazard in the region have become extremely important.

The 1995 rupture zone is located in the hanging wall of the main Dinar fault. The rupture zone is parallel to the main Dinar fault-line but does not follow it on the surface. The sense of motion on the 1995 earthquake break is normal with a slight strike-slip component (Eyidoğan and Barka, 1996).

Does the 1995 earthquake break follow a pre-existing break or is it a new rupture? As Figs 3 and 5 show, the 1995 fault does not follow exactly pre-existing faults (e.g. F1 and F2 faults) in trench walls, but they are close to each other. The horizontal distance between the 1995 break and pre-existing faults varies near the surface; for example, it is about 1 m in trench 1 (Fig. 3) but the nearest fault is located about 8 m away from the 1995 break in trench 2 (Fig. 5). This variation in distance probably indicates that faults developed at different times are anastomosing in the hanging wall of the main fault. As Fig. 6(a) illustrates, it is possible that faults developing in the hanging wall of the main fault are branching and rejoining (e.g. Dresen et al., 1991). In a cross-section from the branching part, faults appear as different segments and the total displacement is partitioned unequally (Fig. 6b). However, where faults rejoin they appear as a single fault in cross-section and the displacement is cumulative (Fig. 6c). On the basis of this observation, it can be concluded that the 1995 earthquake break is branching and interconnecting with the pre-existing faults.

The trench sites across the 1995 break showed evidence of at least two past events (called event 1 and event 2). The F1 and F1-1 faults cut unit a but they are truncated by unit b (Fig. 5). Thus, F1 and F1-1 are simultaneous, produced during the same event (event 1). However, although the F2, F2-1, F2-2 and F2-3 faults cut both units a and b they are covered by unit c (Fig. 5). The F2-1 and F2-2 faults are antithetic and F2-3 is synthetic to the F2 fault. Thus, the F2 fault is the main fault and F2-1, F2-2 and F2-3 are accommodation structures developed at the same time during the same event (event 2). The amount of displacement caused by F2 varies between stratigraphic units. For example, the coarse gravel level and unit b are offset by F2 but the displacements are about 215 cm and 140 cm, respectively (Fig. 5). Different amounts of displacement on F2 indicate that it reactivated before event 2. The c. 140 cm offset of unit b indicates that the previous displacement on F2 fault is about 75 cm, which is similar to that on F1 fault. Thus, it is possible that the F2 fault moved during event 1 with F1 fault. On the basis of these observations, it can be concluded that there were at least two different faulting events before the 1995 earthquake and that event 1 is relatively older than event 2.

Samples were collected from stratified units in trench walls in order to determine precise ages of events but only two samples yielded reliable ages. Samples collected from unit D in trench 1 (see Fig. 3 for location) and from the tuffite layer yielded ages of 590 ± 50 yr BP and about 3500 yr BP, respectively. Certain dates of events cannot be assessed unambiguously with these obtained ages but they do bracket the ages of events 1 and 2. The age of 590 ± 50 yr BP shows that except for the 1995 Dinar earthquake there has not been any large earthquake for at least the last c. 600 years. Although the age of 590 ± 50 yr BP is the upper boundary for event 2 according to historical sources, there is no record of any large earthquake in the Dinar area between AD 53 and 1875. Thus, it can be concluded that the Dinar fault reactivated at two different times between 3500 BC and AD 53. It is significant that when an important city was destroyed by an earthquake in historical times in western Turkey either the city centre was shifted or the city was rebuilt (e.g. the ancient cities of Hierapolis, Priene, Tralles; Hancock and Altunel, 1997; Altunel, 1998, 1999). It is also noteworthy that Strabo (a travel writer and geographer who lived from 63 BC to AD 21) reports that Apamea Kibotos, one of the most important cities of Phrygia, was damaged by an earthquake around 80 BC, and when King Mithridates (king of Pontus 120–63 BC) won a war against Rome (88–84 BC) he gave financial help for reconstruction of the city. In view of the historical records and of the field evidence it is suggested that event 2 probably occurred in 88 BC and that the last reactivation of the F2 fault and offset in unit b were associated with the 88 BC earthquake.

The history of the Dinar area before the 8th century BC is unknown because the ancient city of Apamea Kibotos was founded around the 8th century BC (Akurgal, 1995). There is a historical record of an earthquake in 400 BC in the Dinar area, but there is no account of reconstruction of the ancient city. This information indicates that event 1 probably occurred before the 8th century BC. The age of about 3500 yr BP is the lower boundary for event 1 which corresponds with 1500 BC. As outlined above, although the 1875, 1914, 1925 and the 1971 earthquakes damaged Dinar, they did not occur on the Dinar fault (Ergin et al., 1967; Ambraseys, 1975, 1988; Ering et al., 1971; Ambraseys and Finkel, 1987). Similarly, as pointed out by several historical records, the 400 BC and the 53 AD earthquakes damaged the ancient city of Apamea Kibotos but they probably did not occur on the Dinar fault. Thus, on the basis of historical records and of the field observations, a return period of about 1500–2000 years for large earthquakes can be estimated in the Dinar area. However, the differing amounts of displacement on faults associated with earthquakes indicate that their magnitudes were not the same.

Fault motion at depth does not all reach the surface but this depends on the length of straight fault (Vita-Finzi and King, 1985). Because the Dinar fault is straight (Fig. 1b), as the 1995 event has shown, this fault can produce surface scarp. Assuming that the rupture length and surface offset are directly proportional to magnitude, further interpretations can be made. The magnitude of the 1995 earthquake was 6.1 and it produced a c. 10-km-long surface break with a maximum 50 cm vertical displacement. The vertical displacement is about 78 cm on fault F1 and 140 cm on fault F2. Considering that 50 cm vertical displacement occurred during the 1995 earthquake, the 140 cm vertical displacement seems exaggerated. However, normal faults in southwestern Turkey are capable of producing such offsets during one event. For example, the 20 September 1899 Menderes earthquake (I₀ = IX) was associated with normal faulting giving as much as 2 m vertical displacement (Ergin et al., 1967; İlhan, 1971; Allen, 1975; Sipahioğlu, 1979; Ambraseys and Finkel, 1987; Altunel, 1999). The maximum vertical displacements of 78 cm and 140 cm on F1 and F2 faults suggest that the magnitudes of events 1 and 2 were larger than 6.8 and that a large portion of the Dinar fault must have been ruptured.

The total displacement measured in the trenches for the last 3500 years is about 3.50 m. This gives rise to about 1 mm yr^-1 slip rate on the Dinar fault. Using this value, about 0.4 mm yr^-1 extension rate can be estimated for the Dinar fault. Assuming that the initiation of the neotectonic regime in southwestern Anatolia began with the anticlockwise rotation of the Anatolian block (Barka and Reilinger, 1997) about 4–5 Ma ago, and that the Dinar fault was formed in this new regime, the total extension across the Dinar fault should be about 1.6–2 km. About 2 km width of the basin in front of the Dinar fault (Fig. 2) is consistent with this value.

In summary, our palaeoseismological investigation into the 1995 earthquake rupture illustrates that two large events have occurred on the Dinar fault during the last 3500 years and that these events were larger than the 1995 earthquake. The approximate recurrence interval of large earthquakes is estimated to be 1500–2000 years and there is a slip rate of c. 1 mm yr^-1.

• from Altunel et al. (1999)

Trench 1 – The trench wall clearly shows six different units which are nearly horizontal (Fig. 3). Units B, C, D, E and F can be correlated with the hanging wall of the 1995 break but units B and C disappear in the footwall where only D, E and F can be correlated. Unit A exists in the footwall of the 1995 break but it suddenly disappears in the hanging wall (Fig. 3). There is an angular unconformity between unit A and units B, C and D. Units A and C are blanketed by unit D, which extends horizontally from one end of the trench wall to the other. All of these units are colluviums comprising angular clasts in various sizes derived from the basement rock. Only unit D is rich in clay and organic materials and it has fewer clasts. A charcoal sample from unit D yielded an age of 590 ± 50 yr BP. As shown in Fig. 3, the 1995 break offsets units A, D, E and F by up to 30 cm. A nearly vertical fissure that developed during the 1995 earthquake follows the unconformity between unit A and units B and C (Fig. 3). This south-facing unconformity probably reflects a degraded fault scarp.

Trench 2 – a quarry site located on the hanging wall of the Dinar fault (Fig. 4a). This trench is about 50 m long and 15 m high and it cuts the 1995 earthquake break perpendicularly. The trench site displays well-stratified Holocene colluvium deposits and complex normal faulting (Figs 4b and 5). Unit a, exposed at the base of the cliff, includes two key strata that provide reliable information about the amount of vertical displacement. The unconsolidated coarse gravels and a gently south-dipping tuffite layer within unit a are clearly offset by conjugate normal faults (see Figs 4b and 5). Fault F1, dipping 80° S, cuts through unit a and offsets the tuffite layer vertically by about 78 cm. The F1 fault is truncated by unit b. Fault F1-1, offsetting the tuffite layer 20 cm vertically, is parallel to the F1 fault and disappears within unit a. Fault F2, dipping 75° S, cuts both units a and b, with vertical displacements of c. 215 cm and 140 cm, respectively. Faults F2-1 and F2-2, cutting both units a and b, dip north with vertical displacements of 72 cm and 45 cm, respectively. Fault F2-3 dips south with just a few centimetres vertical displacement. Faults F2, F2-1 and F2-2 are blanketed by unit c but the F2-3 fault disappears within unit b. The F3 fault represents the 1995 earthquake break, offsetting all units exposed on this cliff by about 30 cm.

The 1995 rupture zone is located in the hanging wall of the main Dinar fault. The rupture zone is parallel to the main Dinar fault line but does not follow it on the surface. The sense of motion on the 1995 earthquake break is normal with a slight strike-slip component (Eyidogan and Barka, 1996). Does the 1995 earthquake break follow a pre-existing break or is it a new rupture? As Figs 3 and 5 show, the 1995 fault does not follow exactly pre-existing faults (e.g. F1 and F2 faults) in trench walls, but they are close to each other. The horizontal distance between the 1995 break and pre-existing faults varies near the surface; for example, it is about 1 m in trench 1 (Fig. 3) but the nearest fault is located about 8 m away from the 1995 break in trench 2 (Fig. 5). This variation in distance probably indicates that faults developed at different times are anastomosing in the hanging wall of the main fault. As Fig. 6a illustrates, it is possible that faults developing in the hanging wall of the main fault are branching and rejoining (e.g. Dresen et al., 1991). In cross-section from the branching part, faults appear as different segments and the total displacement is partitioned unequally (Fig. 6b). However, where faults rejoin they appear as a single fault in cross-section and the displacement is cumulative (Fig. 6c). On the basis of this observation, it can be concluded that the 1995 earthquake break is branching and interconnecting with the pre-existing faults.

The trench sites across the 1995 break showed evidence of at least two past events (called event 1 and event 2). The F1 and F1-1 faults cut unit a but they are truncated by unit b (Fig. 5). Thus, F1 and F1-1 are simultaneous, produced during the same event (event 1). However, although the F2, F2-1, F2-2 and F2-3 faults cut both units a and b they are covered by unit c (Fig. 5). The F2-1 and F2-2 faults are antithetic and F2-3 is synthetic to the F2 fault. Thus, the F2 fault is the main fault and F2-1, F2-2 and F2-3 are accommodation structures developed at the same time during the same event (event 2). The amount of displacement caused by F2 varies between stratigraphic units. For example, the coarse gravel level and unit b are offset by F2 but the displacements are about 215 cm and 140 cm, respectively (Fig. 5). Different amounts of displacement on F2 indicate that it reactivated before event 2. The c. 140 cm offset of unit b indicates that the previous displacement on F2 fault is about 75 cm, which is similar to that on F1 fault. Thus, it is possible that the F2 fault moved during event 1 with F1 fault. On the basis of these observations, it can be concluded that there were at least two different faulting events before the 1995 earthquake and that event 1 is relatively older than event 2.

Samples were collected from stratified units in trench walls in order to determine precise ages of events but only two samples yielded reliable ages. Samples collected from unit D in trench 1 (see Fig. 3 for location) and from the tuffite layer yielded ages of 590 ± 50 yr BP and about 3500 yr BP, respectively. Certain dates of events cannot be assessed unambiguously with these obtained ages but they do bracket the ages of events 1 and 2. The age of 590 ± 50 yr BP shows that except for the 1995 Dinar earthquake there has not been any large earthquake for at least the last c. 600 years. Although the age of 590 ± 50 yr BP is the upper boundary for event 2 according to historical sources, there is no record of any large earthquake in the Dinar area between AD 53 and 1875. Thus, it can be concluded that the Dinar fault reactivated at two different times between 3500 BC and AD 53.

Speculation on the date of Event 2 – It is significant that when an important city was destroyed by an earthquake in historical times in western Turkey either the city centre was shifted or the city was rebuilt (e.g. the ancient cities of Hierapolis, Priene, Tralles; Hancock and Altunel, 1997; Altunel, 1998, 1999). It is also noteworthy that Strabo reports that Apamea Kibotos, one of the most important cities of Phrygia, was damaged by an earthquake around 80 BC, and when King Mithridates (king of Pontus 120–63 BC) won a war against Rome (88–84 BC) he gave financial help for reconstruction of the city. In view of the historical records and of the field evidence it is suggested that event 2 probably occurred in 88 BC and that the last reactivation of the F2 fault and offset in unit b were associated with the 88 BC earthquake.

Speculation on the date of Event 1 – The history of the Dinar area before the 8th century BC is unknown because the ancient city of Apamea Kibotos was founded around the 8th century BC (Akurgal, 1995). There is a historical record of an earthquake in 400 BC in the Dinar area, but there is no account of reconstruction of the ancient city. This information indicates that event 1 probably occurred before the 8th century BC. The age of about 3500 yr BP is the lower boundary for event 1 which corresponds with 1500 BC.