|
|
 |
|||||||||||||||||
|
|||||||||||||||||
 |
|||||||||||||||||
| JOURNAL HOME | HELP | CONTACT PUBLISHER | SUBSCRIBE | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Article |
Department of Civil Engineering
Imperial College of Science, Technology, and Medicine
Exhibition Rd., London SW7 2BU, U.K.
n.ambraseys{at}ic.ac.uk
Manuscript received 10 December 2000.
 |
Abstract |
|---|
 Top Abstract IntroductionThe Twentieth-Century SeismicityEvaluation of Long-Term...DiscussionConclusionsAcknowledgmentsReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionThe Twentieth-Century SeismicityEvaluation of Long-Term...DiscussionConclusionsAcknowledgmentsReferences |
|---|
|
|
To answer this question, Ambraseys and Jackson (2000) examined the long-term seismicity of the region over the last 500 yr. In this article the investigation is extended to the past 2000 yr for which both instrumental and macroseismic events are appraised and historical data relating to earthquakes in previous centuries are critically studied. These studies of past earthquakes, which are still in progress, provide a practical context within which early and modern events in the region can be calibrated on a uniform basis and permit creation of yardsticks in terms of qualitative information against which early earthquakes can be classified according to their size (Ambraseys, 2002). In the process of acquiring and classifying this information, new earthquakes were found and known ones were discarded as spurious.
Background
Although scattered indications of earthquakes in the Marmara Sea region go back as far as the third century B.C., adequate documentary coverage of individual events for location and magnitude does not begin until the first century A.D. The coastal area of the Marmara Sea region was always a major trade route, and for more than 20 centuries it was the crossroads between the West and the East, a region of great importance. The Roman land route called the Via Egnatia, the main trade route from Rome to the East, used by the Byzantines and later by the Ottomans (
ol Qol), was built ca. 130 B.C. It crossed the Balkans to Byzantium and followed the Thracian coast of the Sea of Marmara via Thessaloniki, Enez (Enos), and Tekirda
(Raedestos) to Istanbul. From Istanbul, the trade route ran along the north coast of the Gulf of 
zmit to zmit (Nicomedia), the main route crossing over the Gulf of zmit at Hersek (Helenopolis) to Iznik (Nicaea). From there, one branch turned to the west, passing from Bursa to Çanakkale and the Dardanelles, thus encircling the Sea of Marmara, with the sea routes connecting most of the ports around the Sea and thus supplementing the communication of information with the interior (e.g., Taeschner, 1926).
At the center of the region, Byzantium, later Constantinople, or Istanbul, was the thriving, populous capital of the Lower Roman, Byzantine, and later, Ottoman Empires, with a sustained tradition of historiography. The historical record of this region over the past 20 centuries, as provided by literary sources, is remarkably full and relatively continuous. The Roman, and in particular, the Byzantine, tradition, was necessarily waned with the Ottoman conquest about six centuries ago but was superseded by the fairly rich, but not so accessible, Ottoman historiography, supplemented by occidental sources. The historical record of the Marmara Sea region regarding information about earthquakes is rivaled only by that of a small part of northeast China. The benefit of being able to use observations over a period 20 times longer than the 100 yr of instrumental seismology is obvious.
Tectonics
The active tectonics of the Marmara Sea region in northeast Turkey is dominated by the right-lateral North Anatolian fault zone, running from Karliova in the east (41° E) to the Gulf of zmit (30° E). In the eastern part of the Marmara Sea region, the westward motion of Turkey relative to Europe occurs mostly along the North Anatolian Fault zone. East of about longitude 30° E, the North Anatolian Fault system, which accommodates most of the westward motion of Turkey, has a narrow and localized character, clearly defined by the predominantly strike-slip surface along its entire 1000-km length, which is associated with a series of major earthquakes.
West of about 31° E, the North Anatolian Fault system branches into three subparallel strands; a northern strand passes by Sapanca, enters the Gulf of zmit, and forms the Marmara Sea Basin, and a southern strand runs toward Lake Iznik and Bursa. Near 30° E, the southern strand splits up again into a middle and southern branch, the former passing south of Iznik and the latter south of Bursa, by the Lakes of Apolyont and Manyas into the North Aegean extensional province (Fig. 2).
The Marmara Sea Basin is about 230 km long and 70 km wide with a shallow shelf to the south and a series of subbasins to the north, namely, the Tekirda, Central, Çinarcik, Karamürsel, and zmit basins. Active faulting on land in the region is relatively well documented. However, the pattern of active faulting in the Basin is much less well established. Originally the Basin was considered to be a graben or a structure of right-lateral faults exhibiting an overall normal motion (Barka and Kadinsky-Cade, 1988). Recently it was proposed that the Marmara Sea Basin was controlled by a strike-slip fault that extended between the Gulf of zmit and the Galipoli Peninsula (Le Pichon et al., 2000). However, its bathymetry, supplemented by faults identified by conventional seismic reflection surveys and focal mechanisms of a few earthquakes (Ambraseys and Jackson, 2000), shows a series of pull-apart basins bounded by a system of relatively short strike-slip and normal faults, which clearly imply significant regional extension responsible for the formation of the Marmara Sea Basin. The depths and steep bathymetric gradients of the subbasins of Tekirda in the west and Çinarcik in the east parts of the Basin suggest high local seismic activity (Fig. 3) (Smith et al., 1995; Wong et al., 1995; Parke et al., 2000; Okay et al., 2000).
|
West of the Marmara Basin (west of 27.5° E), the motion is taken up by the strike-slip Saros zone and by the extension of the crust in and around the North Aegean Sea. In the Aegean the present-day deformation is distributed over a broad region, but in the seismogenic upper crust most of the motion is apparently accomodated by a few subparallel major fault systems including the Saros zone (Taymaz et al., 1991).
|
The Twentieth-Century Seismicity |
|---|
|
TopAbstractIntroductionThe Twentieth-Century SeismicityEvaluation of Long-Term...DiscussionConclusionsAcknowledgmentsReferences |
|---|
5.0 in the period 1900-1999, for which the locations have been reappraised and their magnitudes recalculated from amplitude/period readings of long phases and the Prague formula (Ambraseys and Douglas, 2000). The most active parts of the Marmara Basin are at the terminus of the North Anatolian Fault system in the east and at the junction of the Basin with the North Aegean in the west, regions controlled by strike-slip faults.
Compared with other parts of the North Anatolian Fault, the Marmara Sea region has been one of rather high seismicity in the twentieth century, releasing a total moment of 6.4 x 1027 dyne cm, more than half of which comes from the earthquakes of August 1912 in Ganos and August 1999 in zmit, with the central Marmara Basin in between contributing only 8% of the total seismic moment released.
|
Evaluation of Long-Term Seismicity |
|---|
|
TopAbstractIntroductionThe Twentieth-Century SeismicityEvaluation of Long-Term...DiscussionConclusionsAcknowledgmentsReferences |
|---|
The present study concerns chiefly the long-term seismicity of the Marmara Basin, but my group's investigation extends over the larger area of the Marmara Sea region.
Approach
The macroseismic data retrieved for the historical period have been interpreted following an improved version of the method described by Ambraseys and Melville (1982). This method includes assessment of the observed intensities, epicentral locations, and a calculation of the magnitude of the earthquakes as defined by the intensities assessed at different sites and by their associated source distances.
The source material used is far too long to present in a journal article. The full database gives the various sources that have been found to contain information and relates the record of events to specific earthquakes. For events up to 1500 and after 1800, all the source material, which is in various languages dead and alive, has been translated into English, whereas for events between 1500 and 1800, the full database includes a revised and updated version of the article by Ambraseys and Finkel (1995). The full database of information, which is still under development, is then used to assess epicentral areas, intensities, and magnitudes (Ambraseys, 2001a). The intricacies of the interpretation of the literary information in the sources have been presented elsewhere (Ambraseys, 2002) and some may be mentioned here.
Dating
With historical earthquakes, dating is not always easy. Several calendars are used to date earthquakes over the last twenty centuries, and details of the different dating systems in use may be found, for instance, in Grumel (1958). Dating errors in primary and in particular secondary sources are common, and whether each event reported was really distinct or whether multiple reports of the same event were interpreted as separate events should be examined. Failure to identify and eliminate doublets or amalgamated events, a common error in parametric catalogs, renders assessment of the epicentral area and magnitude particularly problematic.
Simultaneity
Another problem in locating historical events and assessing their size is the simultaneity of the earthquake mentioned in various places. This is not always specifically stated by the early authors, who refer to several places being shaken or damaged in a particular week, month, or year. This confusion occurs chiefly with early chroniclers who mention all the events of a particular period in a single description. Failure of earthquake compilers to recognize lack of simultaneity often results in spurious doublets or in the amalgamation of different events, which leads to catalog entries of great earthquakes, the sizes of which are grossly exaggerated (Ambraseys and White, 1997).
Earthquake Effects
Assessment of the strength of ground shaking from earthquake effects reported in historical and modern accounts depends largely on the vulnerability of the building stock exposed to the shaking and on the way this information is transmitted in historical documents.
Until relatively recently, houses in Istanbul and in towns and villages around the Sea of Marmara were almost entirely made of wood, and as a consequence, although damaged beyond repair in a severe earthquake, they rarely collapsed and did not cause great loss of life (Refik, 1935). A general observation about the typical timber house is that its inherent strength was considerable but variable and that its vulnerability to earthquakes was rather low. In contrast, stone and brick constructions, which were relatively few in the region, can collapse with great loss of life. Further inland, houses in villages in Thrace, to a great extent of made of mud bricks and covered with flat heavy roofs, were also vulnerable. It is interesting that for Istanbul, damage and loss of life resulting from fires were always higher than from earthquakes (Cezar, 1963).
Quite often, the fact that a few historical monuments made of stone or brick are still to be found in the region in a state of tolerable preservation is interpreted as an indication that either the vulnerability of this class of structures is low or that they have been free from destructive earthquakes. However, our evidence shows that the historical structures still extant have in fact, during their lifetime, been subjected to a number of destructive earthquakes, and they have partly or wholly survived through a process of natural selection. They represent a small sample of the best final design and construction, achieved through the ages by trial-and-error techniques or by chance, and their outcome cannot and should not be used to assess seismic hazard.
Another modern class of man-made structures that seem to have little extra resistance to earthquakes is that of houses built in the last few decades with nontraditional materials, such as reinforced concrete. As the recent earthquakes have shown, in the absence of proper building codes and enforcable regulations, the introduction of new materials and methods of construction has produced a class of highly vulnerable structures, the excessive damage of which tends to inflate intensity grading.
Intensity
As a consequence of these differences in vulnerability of the building stock and its changes with time, it is difficult to assign intensity. We find that damage or collapse of aging houses and public buildings or of city walls is often an indication of the weakness of these structures, which frequently collapse without the help of earthquakes, rather than of the severity of ground shaking, which in the absence of additional corraborating evidence makes it difficult to assess intensity.
Natural exaggeration in the sources, historical and modern, is also a problem; however, it is not difficult to solve. The authenticity of the source, the style of its narrative, internal evidence in the account, and the experience gained by the assessor from processing historical information all combine to permit a realistic, albeit subjective, estimate of the degree of impact or the extent of the area over which earthquake effects were serious. For instance, statements such that after a "destructive" earthquake in Istanbul people took refuge in churches or mosques, or that they organized religious processions in the city, suggest that the shock was not in fact all that destructive. There are also cases in which a historian unintentionally exaggerated damage by reporting collectively effects from more than one earthquake or from fire and thunderstorms. Intentional exaggeration of damage, or syncretization of an earthquake with political or religious happenings, is also known. The substance of such accounts is the same, but the impressions they leave about the seriousness of the event with those who use historical data without scrutiny or at second hand are very different. These problems are discussed elsewhere in connection with the assessment of intensities in the region (Vogt, 1996; White, 2000; Ambraseys, 2002).
Statements referring to fatalities among leading citizens, large-scale reconstruction undertaken by the state, remission of taxes, new coinage, adverse effects on trade, or emigration after an earthquake all suggest a large destructive event. Some of these effects are not always recorded together with or attributed to an earthquake but can be inferred from internal evidence and from the historical context of the event.
Casualty figures are rather difficult to check and are not necessarily indicative of the magnitude or the intensity of an earthquake. Because of the emphasis on effects in major centers, reported figures depend to a large extent on population distribution and density. Even today, with fuller coverage of events, the same bias exists, although this may reflect the genuine situation: quite often the largest proportion of people killed is found in small towns and villages away from urban centers.
The assessment of intensity in this study is made on the Medvedev-Sponheuer-Karnik (MSK) scale, somewhat simplified by disregarding critieria meant for modern structures, which allows for consistent subjective observations and correlation with intensities estimated from twentieth century earthquakes to be made. This simplification is also dictated by the fact that local conditions in the Marmara Sea area, such as siting of settlements and towns, building materials, and techniques, until the early 1900s, had changed little over much of the period under review. Intensity estimates for historical earthquakes, made by the writer and by various independent assessors at different times during the last decade, are accurate only to ±0.5 intensity unit, in the sense that two assessors assigned intensities differing by one unit and occasionally for some early events are only accurate to ±1 unit.
Seismic Sea Waves
Little is known in any detail about the 14 seismic sea waves in the Sea of Marmara, six of which were damaging, associated with earthquakes of MS 6.8. It is not possible to say much about the source area of these waves except perhaps that most of them were reported from the northeast coast of the Sea of Marmara and from the Gulf of zmit, and that the source mechanism of some of them seems to be coastal mass failures and sliding of submarine sediments (Ambraseys, 2001b). In Figure 3, we have assumed sea wave sources to be close to the inferred offshore epicentral area of the associated event.
Epicentral Region
Data provided by historical sources for earthquakes on land are adequate to permit the general location of the general epicentral area, particularly for the larger events, which show some clustering toward large centers of habitation. Early earthquakes are less well located, and it is often difficult to ascertain their true epicentral region, although for the larger events, aided by local tectonics, there is rarely great ambiguity about their general position within the Marmara Sea region. Figure 4 shows the general location of the epicentral regions of historical and modern events.
|
|
Study of twentieth-century earthquakes is additionally valuable because it permits correlations to be made between the effects of historical earthquakes and those of the twentieth century. This is possible despite shortcomings in the documentary evidence, for one can, in many cases, determine from internal or contextual evidence such details as the size of the area affected, athe extent and type of damage caused, the duration of aftershocks, and the association of the event with ground deformations, such as surface faulting, landslides, and proximity to known active faults. All such details, ideally found in or inferred from historical sources, when calibrated against similar information derived from the study of modern events, permit assessment of the different intensities experienced throughout the affected area and estimation of the magnitude of the historical earthquake.
For earthquakes that occurred before the 1920s, locations are the centers of the macroseismic regions as defined by the higher intensities. For events with epicenters in the Sea of Marmara, we have attempted to associate earthquakes with active faults. This is often determined based on knowledge of the general location and magnitude of the event and on the nearest known fault in the Marmara Basin that could accomodate the associated seismic moment release.
Classification of Events
All events were classified according to the location of their epicentral region, that is, on land, offshore, or not known, and their quality, that is, according to whether they were well located, less well defined, adopted, or instrumental. They are also classified according to the number of sites used to estimate their magnitude. Depending on maximum effects, earthquakes with epicentral region on land or offshore were classed into those that caused considerable damage, heavy damage, or destruction and social or economic effects resulting in extensive reconstruction. These events are ranked again in terms of the amount of casualties they caused, that is, small, significant, or great. Further classification was made according to whether an event exhibited a small or large area of perceptibility and caused localized or widespread damage. Events accompanied by a seismic sea wave, surface faulting, and ground failures are also flagged.
Table 1 shows the classification of events in 10 fields, the reason for which has at least three purposes: (1) to inspect all events regardless of when and where they happened; (2) to compare the location and size of the events; and (3) to test the completeness of the information. Only in this way might we succeed in getting an overall qualitative evaluation of the historical data before we use them in a final quantitative analysis.
Magnitude
The surface-wave magnitude MS of pre-1898 earthquakes was estimated from the intensity and epicentral distance using the regional equation
 | (1) |
where MSi is the uniformly recalculated surface-wave magnitude from the Prague formula, in which Ii (
VIII MSK) is the intensity at a distance ri (in km) from the seismic source, and p is 0 for mean value and 1 for 84 percentile. Equation (1) was derived from earthquakes in Greece and Western Turkey and is valid for far-field conditions (Ambraseys, 1992). In the near field, source distance is defined as the closest distance to the surface fault rupture, and in the far field, it is defined by the epicentral distance. Site magnitudes MSi were calculated from equation (1), the average value of which is the event magnitude MS, which, together with the number of station magnitudes used to assess it, is listed in Table 1.
Intensity estimates for historical earthquakes in this study are accurate only to ±0.5 intensity units at best, and from equation (1) to an uncertainty in MS of ±0.35 magnitude units, but for events reported from very few places this uncertainty can be larger.
Seismic Moment and Moment Magnitude
Relations between surface-wave magnitude MS and seismic moment M0, and vice versa, provide suitable functions for the correlation between one source size indicator and the other. Current scaling log(M0) - MS laws for shallow earthquakes have been derived from global or large subglobal datasets for active regions and for stable continental regions. Ekström and Dziewonski (1988) derived global average scaling log(M0) - MS laws in which the independent variable is log M0, which quite often are assumed to be orthogonal although they are not. These are
 | (2a) |
 | (2b) |
 | (2c) |
These authors then rewrite equations (2) in the form
 | (3a) |
 | (3b) |
 | (3c) |
However, since equations (2) and (3) are rewritten, formally, they are not the correct relationships for estimating log M0 from MS. Regional bias in M0 does exist, and global average scaling log(M0) - MS laws, such from equation (3), may be inappropriate. To overcome the problem with equations (3) that have been derived by fitting the data with log(M0) as the independent variable, and at the same time to take into account regional bias, we derived the following set of bilinear relationships:
 | (4a) |
and
 | (4b) |
for the Eastern Mediterranean and the Middle East region up to 70° E, using Harvard centroid moment tensor (CMT) or P/SH moments from special studies and the corresponding reassessed MS values from the Prague formula of 577 shallow (h < 40 km) earthquakes in the log M0 range 22.4-27.3; in the regression MS is the independent variable (Ambraseys and Douglas, 2000). Seismic moments estimated from equation (4) are by a factor 0.7 and 0.85 smaller than from the global relations (3) for M < 6.0 and M > 6 respectively.
Seismic energy magnitude MW, moment magnitude M, and surface-wave magnitude MS are defined as a linear transformation of the logarithm of the seismic moment M0 given by
 | (5) |
in which M0 is in dyne cm (10-7 N m) (Kanamori, 1977), M being nothing more than a definition or a transformation of M0 through equation (5). What is often overlooked is that equation (5) is valid only for events that rupture the entire thickness of the seismogenic zone and that the validity of equation (5) holds only for MS values larger than about 6.5 (Ekström and Dziewonski, 1988). For smaller magnitudes M 
MS. Seismic moments in Table 1 are either Harvard CMT estimates for events after 1977 or values calculated from equation (1) and from Ekström and Dziewonski's (1988) global scaling log(M0) - MS law (??)(3c) for events before 1977.
|
Discussion |
|---|
|
TopAbstractIntroductionThe Twentieth-Century SeismicityEvaluation of Long-Term...DiscussionConclusionsAcknowledgmentsReferences |
|---|
The main factors seen to influence the survival of historical data are the quality of the contemporary literary record, the prevailing historical circumstances, the geographical location of events and their magnitude. The total number of earthquakes identified in the Marmara Sea area during the last 20 centuries is 581, of which 408 occurred in the preinstrumental period, before 1900, and 173 in the twentieth century. Of the 408 earthquakes in the preinstrumental period, 301 are small events that caused no damage or great concern, the magnitude of which was not assessed. For the remaing 107 larger earthquakes, magnitudes were assessed from equation (1) with values ranging between 5.0 and 7.4. The magnitude of all 173 earthquakes of the instrumental period, after 1899, were calculated from the Prague formula; these ranged between 4.0 and 7.4. Table 1 lists 54 events from the dataset of MS 6.8, together with their seismic moments, which for earthquakes before 1977 were calculated from equation (3). Table 1 shows also the classification of these events.
For historical earthquakes at sea, locations are necessarily approximate and do not warrant a sophisticated procedure for their determination. Bearing in mind that earthquakes of MS between 6.8 and 7.4 will have ruptured faults 30 km to more than 100 km in length, locations at sea are necessarily approximate but adequate indicating nothing more than the general location of an event. Indeed, a too sophisticated location procedure carries with it the danger that its weaknesses and assumptions may not be appreciated. Conversely, a too simple method may be discredited just because it exposes the underlying assumptions too clearly.
Macroseismic Magnitude
For shallow earthquakes of about MS < 6.4, which have source dimensions sufficiently small, there is little difference between source distance and epicentral distance of a site, certainly not larger than the uncertainty in the determination of a twentieth-century epicenter. As the size of an earthquake increases, epicentral distance in the near field loses its meaning as a measure of the proximity of a site to the source, and distance r in equation (1), which has been derived for regional conditions, must be measured from the fault trace.
The use of magnitude calibration relations derived for other regions, using epicentral distance, has an important effect on the assessment of hazard in the Marmara Sea area. As an example magnitudes for some of the historical earthquakes in the Marmara Sea region have been calculated by Parsons et al. (2000) from
 | (6) |
a calibration formula derived by Bakun and Wentworth (1997) from 22 Californian earthquakes in the magnitude range 4.4 M 6.9, in which d is the epicentral distance in km. For M 6.0, Parsons et al. (2000) assumed that MI = M, and that for MH < 6.0, MH is the local magnitude ML in Hubert-Ferrari et al. (2000); d is the median epicentral distance for a particular intensity (I). Equation (6) cannot be compared directly with equation (1) because these two relationships differ in the definitions of distance, magnitude, and regional bias. The only observation that can be made from Figure 5, which shows a plot of equations (1) and (6), is that the latter gives larger magnitudes.
|
Table 2 lists the values of MI calculated by Parsons et al. (2000) from equation (6) for 13 historical earthquakes in the Marmara Sea region, together with MH estimates made by Hubert-Ferrari et al. (2000) from assumed fault lengths that are judgmental in nature, and the scaling relation of Kanamori and Anderson (1975). In the absence of other relevant information, the length of an active fault is the best guide to the maximum earthquake that might occur along it, although any such guide is a gross approximation at best, particularly with offshore faults for which any assumption regarding their location and length is little more than an arbitrary judgment. For comparison, Table 2 also shows MS calculated in this study from equation (1). These three datasets are plotted in Figure 6 from which, considering the diverse assumptions made in their derivation in terms of MS and the small number of observations, the large differences, particularly for MS < 7.0, are to be expected.
|
|
Completeness
Much statistical ingenuity has been spent on devising techniques for tackling this problem, but there are doubts about how useful and how well-founded some of these techniques really are, and the statistician here should take a background role. He or she can point to features in the data that look anomalous because they depart from some standard model, particularly when large events are clustered in time. Whether the anomalies are to be ascribed to peculiarities of the tectonic process, or to peculiarities of the process by which the events were recorded in history, is not a question the statistician should be asked to answer. It should be referred back to the historian or the geophysicist. With historical data, if an effect is really present it should not take a statistician to bring it out (Vere-Jones, 1987).
Turning to Table 1, it is unlikely that all small and perhaps even moderate historical earthquakes in the interior, away from the coast of the Sea of Marmara, before the middle of the eighteenth century, would have been recorded. It is very probable, however, that many moderate and any large earthquake have been noted, although not necessarily fully identified. After the middle of the eighteenth century, documentary information improved rapidly, further supplemented by occidental sources (consular correspondence, travelers, and the press), so that after the mid-eighteenth century, data about moderate and large magnitude events should be complete. It is reasonable to suppose, therefore, that the available data for the whole region is almost complete only for moderate earthquakes or greater.
We can devise no formal method to test completeness of the data other than by testing their implications. Formal statistical tests are as valid as the distributional assumptions on which they are based. Since these assumptions are rarely likely to be satisfied by historical data, such tests may best be regarded as indicators to the extent to which a particular conclusion would be supported, or not, by the data, if in fact the assumptions were justified, and hence, of the extent to which that conclusion is likely to remain valid despite departure from those assumptions (Lomnitz, 1994).
Contribution of Small Events to the Total Moment
Relatively complete datasets of long-term seismicity in active regions are necessarily restricted to magnitudes greater than 6.8 or 7.0. The contribution of smaller magnitudes to the total moment depends on the piecewise linear frequency distribution of the data and also on the choice of the scaling log(M0) - MS law (e.g., Molnar, 1979; Ambraseys and Sarma, 1999). It is obvious, therefore, that the total moment release and slip rates calculated from such datasets, without consideration of the contribution of smaller magnitudes, will be underestimated by a factor of 1.5 to 2.0.
The contribution of smaller events to the total seismic moment alters the view about regions of high aseismic creep and fits better with direct measurements of velocities derived from geodetic methods. The possibility of aseismic slip in the Marmara Basin is an important question, and solutions to the problem of seismic versus aseismic slip might well come from geodetic observations.
Seismic Slip Rates
There are several features in the time distribution of earthquakes in the Marmara Sea region that would seem to require discussion. These features show up clearly when presented as a plot of cumulative moment versus time, shown in Figure 7. The most obvious is perhaps the concentration of large events during the fifth century and to a lesser extent in the tenth, fourteenth, and twentieth centuries, as well as the relative seismic quiescence in between these periods that may be thought to be due to lack of information rather than a genuine lack of earthquakes. But from Figure 8, which shows the volume of historical sources, we notice that there is no lack of information about earthquakes in the quiescent periods, and it is rather unlikely that large earthquakes over these relatively well-documented periods have been missed. It may seem unlikely that such features could result from a sudden excess or loss of enthusiasm among local historians during these periods to record earthquakes.
|
|
In order to investigate the effect of the size of the study area on clustering, we reduced the size of the study region by taking a 4 times smaller area (e.g., region B in Fig. 1). We find that this removes clustering, and as Figure 9 shows, the rate of moment release over the whole of the last 2000 yr becomes almost constant. Next we estimated the average right-lateral shear velocity through regions A and B by summing the scalar seismic moments of the earthquakes of MS 6.8, using a seismogenic thickness of 12.5 km (10-15 km) and, assuming a value for the rigidity of l = 3.0 x 1011 dyne cm-2, each event contributing to this motion. Some events in the Marmara Basin do have normal faulting mechanism, in keeping with the generally short normal fault segments, but we do not think that in these calculations the assumption that all events are strike slip is an important source of error. The main source of uncertainty in the shear velocity is likely to come from the uncertainty in M0, which is caused by its assessment from global (equation 3c) or from the regional MS - M0 formulae (equation 4b), and also by the uncertainties in MS. We therefore estimate the uncertainty in velocity from the uncertainty in observed surface-wave magnitude. Because of the possibility of systematic bias, we do not calculate standard errors, but simply extreme maximum and minimum possible values allowing for all the magnitudes to be underestimated or overrestimated by 0.3 magnitude units and converted into M0 through equations (3c) or (4b) (see Table 3).
|
|
We find that for the Marmara Sea region (A in Fig. 1), which is 420 km long, allowing for missing earthquakes of MS < 6.8 that increase the total moment by a factor of 1.7, earthquakes of the last 2000 yr account for right-lateral velocity that varies with time between 1.5 and 2.5 cm/yr with a mean value of 2.0 ± 0.25 cm/yr, which is shown in Figures 10 and 11. These values are very close to those estimated from Global Positioning System (GPS) measurements by Straub et al. (1996) and by Reilinger et al. (1997), which are between 2.2 ± 0.3 and 2.6 ± 0.3 cm/yr and correspond to the elastic strain to be accounted for by future earthquakes and aseismic creep (Walcott, 1984).
|
|
For the smaller Marmara Basin (B in Fig. 1), which is half as long and half as wide, we find almost identical velocities of 1.9 ± 0.20 cm/yr on average. The faster rates in the first centuries shown in these figures are partly due to the short time intervals used to calculate the ratio at the beginning of the plot. Very similar rates are obtained from the global and regional scaling log(M0) - MS laws (equations 3 and 4), which differ little for Ms < 6.5.
Given the uncertainties in (a) the original Ms values, (b) the unaccounted for seismic moment from subevents of large multiple earthquakes, (c) depth, and (d) the log(M0) - Ms scaling law, it is difficult to estimate more realistic velocities obtained from seismicity. We consider it likely that the global relation, equation (3), yields an upper bound to the seismic moment release and that the regional relation, equation (4), gives a more realistic estimate. We may then conclude that the major portion, perhaps effectively all of the motion in this region during the last 2000 yr, is probably achieved by seismic slip on faults, and that aseismic creep, is relatively unimportant.
Frequency-Magnitude Distribution
It is usually agreed that regional seismicity is well described by the Gutenberg cumulative frequency-magnitude relation
 | (7) |
in which (N/ya) is the annual number of earthquakes of magnitude equal to or greater than Ms per unit area (a). However, whether the same distribution is the proper description of seismicity over a long period of time along a fault zone seems to remain a question.
Using our 100-yr-long dataset, which is more complete for Ms > 4.5 than the 2000-yr-long dataset, we examined the frequency-magnitude distributions of earthquakes in the Marmara Sea region (A) and separately in the Marmara Basin (B), which are shown in Figure 12. For Ms < 6.5, the results are very similar and follow equation (7) with a b-value of about 0.6; but for larger magnitudes they diverge from equation (7) and show a bilinear distribution, a change in slope that is not well defined because the period of observation is too short to provide adequate statistics for earthquakes of Ms > 6.5.
|
If we extend the period of observations from 100 to 2000 yr we notice that, for both regions A and B, the distribution of events at large magnitudes becomes smoothly asymptotic. Also from Figure 13, which is a detail of Figure 12 for Ms 6.8, we notice that in the time between large earthquakes, the smaller Marmara Basin (B) is generally quiescent but for the occurrence of a low-level background activity and that the distribution follows what one might call a "characteristic earthquake" model (Wesnousky et al., 1983). Figures 12 and 13 suggest a genuine departure of the observed data from Gutenberg's equation (7) for large magnitudes and an asymptotic behavior of b at the upper end of the recurrence curve where it steepens. This implies that the true nature of the distribution is hampered not only because the twentieth-century record is too short but also because the test area is too small to define the repeat time of large earthquakes and that the characteristic model could be an artifact of incompleteness of data in space and time. This also implies that the shape of the frequency-magnitude distribution for short-term observations cannot be defined at larger magnitudes. In our case the implication of this is that large earthquakes in the Marmara region are less frequent when predicted from the long-term dataset than from the usual 100-yr instrumental period, making the notion of a recurrence time, in its usual definition, questionable. Preliminary results from the study of a region larger than A in Figure 1, which includes a considerable part of the Anatolian Fault Zone, confirm this and suggest that a greater east-west extent of the Zone needs to be considered (Ambraseys, 2001a).
|
Figures 12 and 13 show that, in hazard modeling from short-term, 100-yr-long datasets, it is simply not reasonable to ignore the chance that much or all of our records may be from a quiescent or energetic period in the seismic activity. This is one of the possibilities that must be considered when making assumptions with incomplete datasets, and this is the principal reason why statistics alone cannot quickly and simply answer the question of seismic hazard evaluation (Vere-Jones, 1987).
The North-Central Marmara Basin
The question now arises of whether the long-term dataset suggests the existence of a throughgoing fault along the shelf break bordering the Tekirda, Central, and Çinarcik subbasins, which could rupture along a length of 150 km or more and produce a major earthquake.
The answer to this question depends much on whether such events have left some significant trace in the bathymetry (Parke et al., 2000). As we have seen, the Sea of Marmara is located where deformation on the North Anatolian Fault changes from being localized strike-slip on a single narrow fault zone striking roughly east-west, to being distributed over a broad region, bearing more to the south, with a substantial normal component of slip in the Basin that is taken up by short fault segments of a few tens of kilometers in length. Faults are generally less continuous offshore than onshore (Wong et al., 1995), and the implication is that rupture of these short faults should be associated with moderate to small earthquakes, which is compatible with both short-period (100 yr) and long-period (2000 yr) evidence shown in Figures 3 and 7. One would expect, therefore, motion transferred from the Anatolian Fault Zone to the Marmara Basin east of 30° E to be taken up by the southern strands of the Anatolian zone, and in particular, with considerable components of normal slip, by the segmented floor of the Basin. This then would have caused the subbasins in the west extremity of the Tekirda, and in the east of Çinarcik where the faulting style changes, to show higher activity, which is exactly what we see in Figures 2 and 4.
This apparent deficiency of large events in the Basin, and consequently of the total long-term seismic moment released is shown in Figure 14, a deficiency that is not all that significant, chiefly caused by the contribution in the total seismic moment of events of Ms < 6.8 that are missing from our historical dataset.
|
The relatively low seismic activity in the Marmara Basin during the last 100 yr is unequivocal (Fig. 2). Whether we can say with confidence the same for the much longer period of 2000 yr depends critically on the correct interpretation of the macroseismic information of the two largest earthquakes of 10 September 1509 and 22 May 1766. We may discuss briefly these two cases for which the interpretation of the supporting macroseismic data have led to debatable conclusions regarding the size and location of some of these events.
The earthquake of 10 September 1509 is of special interest as it has been considered by modern writers (e.g., Parsons et al., 2000) to be a major event of magnitude 7.6 caused by an alleged rupture of a 380-km-long fault along the Marmara Basin extending from near the Dardanelles to zmit. However, a careful reinterpretation of macroseismic effects in new sources show that this earthquake was not an exceptionally large event (Ambraseys, 2002). It excited widespread contemporary interest because of its location, close to the capital of the Ottoman empire, rather than because of its destructive effects. Its reappraisal shows that, with an epicentral area in the Sea of Marmara, the destruction it caused did not extend very far from Istanbul, certainly not beyond Silivri in the west and Gebze in the east, and that damage at Tekirda, Demitoka, and Gelibolu was not more serious than what one would have expected at these distances from an earthquake of MS 7.2.
Hubert-Ferrari et al. (2000), using the data in Ambraseys and Finkel (1991), associated the earthquake of 22 May 1766 with a 100-km-long fault rupture offshore the north-central part of the Marmara Basin, which extends from Tekirda to near Istanbul, and for which they assume a magnitude of 7.4. However, the epicentral region of this earthquake of magnitude 7.2 is clearly east of Istanbul, in the Çinarcik Basin, a location correctly identified by Parsons et al. (2000).
With an historical record 20 times longer than that of the twentieth century, no truly large earthquake of MS > 7.5 in the Marmara Basin has been found extending from Ganos to the Çinarcik subbasins and beyond, of a size comparable to those in the North Anatolian Fault zone, any one of which would account for 700 yr worth of moment release in the Basin. Whether 2000 yr is long enough to obtain a reliable answer to this question is less clear.
Thus, although one cannot rule out the possibility of large strain release in the Marmara Basin, it would not be adequately justifiable to assume at the present time that this can result from rupture of a strike-slip fault that extends from the Gulf of zmit to the Galipoli Peninsula.
Seismicity of the Iznik-Bursa Area
The south branch of the fault system, in the interior of the mainland south of the Marmara Basin, is less active than the north branch through zmit and the Marmara Basin. The morphology of the south branch through Geyve, Bursa, and Gönen to the Aegean Sea suggests late Quaternary activity (Barka, 1996) and a long-term seismicity that accounts for an average shear velocity of about 0.3 cm/yr, which is compatible with GPS measurements (Straub, 1996). However, we must admit that the long-term activity of the inland part of this region is not perfectly known.
Seismicity of the Thrace Basin
It is perhaps interesting that in spite of efforts to investigate the long-term seismicity of Thrace, we could find virtually no significant events in the region to the north of the Marmara Basin, except a moderate earthquake, southeast of Edirne in 1752.
We find that, throughout its 17-century-long history, Istanbul has suffered serious damage from near moderate and more distant large offshore earthquakes, originating chiefly from the Çinarcik subbasin. However, the potential high risk today is in great part due to the very rapid growth of Istanbul in the last decades and also due to the increased vulnerability of its modern building stock. Considering that in 1480 the population of Istanbul was just over 100,000 and today has exceeded 10,000,000, with greater Istanbul occupying vulnerable sites that extend along the coast for tens of kilometers, it is obvious that the repetition of one of the damaging historical earthquake today would be the küçük kiyamat (Little Apocalypse) of the Ottoman historiographers and cause great losses.
|
Conclusions |
|---|
|
TopAbstractIntroductionThe Twentieth-Century SeismicityEvaluation of Long-Term...DiscussionConclusionsAcknowledgmentsReferences |
|---|
zmit. |
Acknowledgments |
|---|
|
TopAbstractIntroductionThe Twentieth-Century SeismicityEvaluation of Long-Term...DiscussionConclusionsAcknowledgmentsReferences |
|---|
|
References |
|---|
|
TopAbstractIntroductionThe Twentieth-Century SeismicityEvaluation of Long-Term...DiscussionConclusionsAcknowledgmentsReferences |
|---|
Ambraseys, N. (1992). Soil mechanics and engineering seismology, Proc. 2nd Natl. Conf. on Geotech. Eng. (invited paper), 21-23 October 1992, Thessaloniki, Greece.
Ambraseys, N. (2001a). Material for the investigation of the seismicity of the Marmara Sea region, NW Turkey, 1-2000, Eng. Seism. and Earth. Eng. Report no. 01.5, Dept. Civil Eng., Imperial College of Science, London.
Ambraseys, N. (2001b). Seismic sea-waves in the Marmara Sea region during the last 20 centuries, in Proc. NATO Advanced Research Workshop on Underwater Ground Failures on Tsunami Generation, Modeling, Risk and Mitigation, Istanbul, May 2001.
Ambraseys, N. (2002). The earthquake of 1509 in the Sea of Marmara, Turkey, revisited, Bull. Seism. Soc. Am. 91, no. 6, 1397-1416.[CrossRef]
Ambraseys, N., and J. Douglas (2000). Reappraisal of surface wave magnitudes in the Eastern Mediterranean region and the Middle East, Geophys. J. Int. 141, 357-373.[CrossRef]
Ambraseys, N., and C. Finkel (1991). Long-term seismicity of Istanbul and of the Marmara Sea region, Terra 3, 527-539.
Ambraseys, N., and C. Finkel (1995). The Seismicity of Turkey and Adjacent Areas: A Historical Review, 1500-1800, Eren, Istanbul, 240 pp.
Ambraseys, N., and J. Jackson (2000). Seismicity of the Sea of Marmara (Turkey) since 1500, Geophys. J. Int. 141, F1-F6.[CrossRef]
Ambraseys, N., and C. Melville (1982). A history of Persian earthquakes, Cambridge Univ. Press, New York.
Ambraseys, N., and S. Sarma (1999). The assessment of total seismic moment, J. Earthquake Eng. 3, 439-461.[CrossRef]
Ambraseys, N., and D. White (1997). The seismicity of the eastern Mediterranean region 550 to 1 BC: a re-appraisal, J. Earthquake Eng. 1, 603-632.[CrossRef]
Bakun, W., and C. Wentworth (1997). Estimating earthquake location and magnitude from seismic intensity data, Bull. Seism. Soc. Am. 87, 1502-1521.
Barka, A. (1996). Slip distribution along the North Anatolian Fault associated with the large earthquakes of the period 1939 to 1967, Bull. Seism. Soc. Am. 86, 1238-1254.
Barka A., and K. Kadinsky-Cade (1988). Strike-slip fault geometry in Turkey and its influence on earthquake activity, Tectonics 7, 663-684.[Web of Science][GeoRef]
Castenetto, S., and F. Galadini (2001). 13 Gennaio 1915 il terremoto nella Marsica, Publ. Servizio Sismico Nazionale, Rome, 788 pp.
Cezar, M. (1963). Osmanli devrinde stanbul yapilarinda tahribat yapan yanginlar ve tabii afetler, Türk Sanati Tarihi Ara
tirma ve ncelemeleri, Istanbul 1, 380-392.
Ekström, G., and A. Dziewonski (1988). Evidence of bias in estimation of earthquake size, Nature 332, 319-323.
Finkel, C. (2000). Earthquakes of the Marmara Sea Basin: a historian's perspective, in Proc. NATO Adv. Res. Seminar, Istanbul, May 2000.
Grumel, V. (1958). Traité d'études byzantines. I: La chronologie, Paris.
Hubert-Ferrari, A., A. Barka, E. Jacques, S. Nalbant, B. Meyer, R. Armijo, P. Tapponnier, and G. King (2000). Seismic hazard in the Marmara Sea region following the 17 August 1999 zmit earthquake, Nature 404, 269-273.[CrossRef][Medline]
Kanamori, H. (1977). The energy release in great earthquakes, J. Geophys. Res. 82, 2981-2987.[CrossRef]
Kanamori, H., and D. Anderson (1975). Theoretical basis of some empirical relations in seismology, Bull. Seism. Soc. Am. 65, 1073-1095.
Le Pichon, X., T. Taymaz, and C. Sengör (2000). Important problems to be solved in the Sea of Marmara, in Abstracts NATO Adv. Res. Seminar on Integration of Earth Sciences Research, Istanbul, 14-17 May 2000, 66-67.
Lomnitz, C. (1994). Fundamentals of Earthquake Prediction, Wiley, New York.
Molnar, P. (1979). Earthquake recurrence intervals and plate tectonics, Bull. Seism. Soc. Am. 69, 115-133.
Okay, A. I., A. Kalitar-Özcan, C. Imran, A. Boztepe-Güney, E. Demirba, and I. Kuçu (2000). Active faults and evolving strike-slip basins in the Marmara Sea, north-west Turkey: a multichannel seismic reflection study, Tectonophysics 321, 189-218.[CrossRef][Web of Science][GeoRef]
Parke, J., T. Minshull, G. Anderson, R. White, D. McKenzie, I. Kuscu, M. Bull, N. Gorur, and C. Songer (1999). Active faults in the Sea of Marmara, western Turkey, imaged by seismic reflection profiles, Terra Nova 11, 223-227.[CrossRef][Web of Science][GeoRef]
Parsons, T., S. Toda, R. Stein, A. Barka, and J. Dieterich (2000). Heightened odds of large earthquakes near Istanbul: an interaction-based probability calculation, Science 288, 661-665.
Refik, A. (1935). On altinci asirda Istanbul hayati, Devlet Basimevi, Istanbul.
Reilinger, R., S. McClusky, and M. Oral, et al. (1997). Global Positioning System measurements of present-day crustal movements in the Arabia-Africa-Eurasia plate collision zone, J. Geophys. Res. 102, 9983-9999.
Smith, A., F. Oktay, T. Taymaz, J. Jackson, H. Basaran, B. Alpar, M. Simsek, and S. Kara (1995). High resolution seismic profiling in the Sea of Marmara-NW Turkey: Late Quaternary sedimentation and sea-level changes, Geol. Soc. Am. Bull. 197, 923-936.
Straub, C. (1996). Recent crustal deformation and strain accumulation in the Marmara Sea region, NW Anatolia, inferred from GPS measurements, Mitteil. Inst. Geod. and Photogramm. ETH, no. 58, Zürich.
Stucchi, M. (1993). Historical investigation of European Earthquakes, Vol. 1, Pub. Istituto di Ricerca sul Rischio Sismico, Milan, 258 pp.
Taeschner, F. (1926). Die Verkehrslage und das Wegentz Anatoliens im Wandel der Zeiten, Peterm. Mitteil. 72, 202-6.
Taymaz, T., J. Jackson, and D. McKenzie (1991). Active tectonics of the north and central Aegean Sea, Geophys. J. Int. 106, 433-490.[CrossRef]
Vere-Jones, D. (1987). Statistical aspects of the analysis of historical earthquake catalogues, in Historica Seismicity of the Central-Eastern Mediterranean Region, C. Margottini and L. Serva (Editors), Rome, pp. 271-295.
Vogt, J. (1996). The weight of pseudo-objectivity, Ann. Geofisica 39, 1005-1011.
Walcott, R. I. (1984). The kinematics of plate boundaries through New Zealand: a comparison of the short and long term deformation, Geophys. Journ. 79, 613-663.
Wesnousky, S., C. Scholtz, K. Shimazaki, and T. Matsuda (1983). Earthquake frequency distribution and mechanics of faulting, J. Geophys. Res. 88, 9331-9340.[Web of Science]
White, D. P. (2000). Quality evaluation of sources for historical earthquakes in the East Mediterranean, Ph.D. Thesis, University of London.
Wong, H., E. Lüdmann, T. Ulug, and E. Özel (1995). The Sea of Marmara: a boundary sea in an escape tectonic regime, Tectonophysics 244, 231-250.[CrossRef][Web of Science][GeoRef]
This article has been cited by other articles:
|
|
 |
|
  E. Kalkan, P. Gulkan, N. Yilmaz, and M. Celebi Reassessment of Probabilistic Seismic Hazard in the Marmara Region Bulletin of the Seismological Society of America, August 1, 2009; 99(4): 2127 - 2146. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Rockwell, D. Ragona, G. Seitz, R. Langridge, M. E. Aksoy, G. Ucarkus, M. Ferry, A. J. Meltzner, Y. Klinger, M. Meghraoui, et al. Palaeoseismology of the North Anatolian Fault near the Marmara Sea: implications for fault segmentation and seismic hazard Geological Society, London, Special Publications, January 1, 2009; 316(1): 31 - 54. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. Palyvos, D. Pantosti, C. Zabci, and G. D'Addezio Paleoseismological Evidence of Recent Earthquakes on the 1967 Mudurnu Valley Earthquake Segment of the North Anatolian Fault Zone Bulletin of the Seismological Society of America, October 1, 2007; 97(5): 1646 - 1661. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. M. Scharer, R. J. Weldon II, T. E. Fumal, and G. P. Biasi Paleoearthquakes on the Southern San Andreas Fault, Wrightwood, California, 3000 to 1500 B.C.: A New Method for Evaluating Paleoseismic Evidence and Earthquake Horizons Bulletin of the Seismological Society of America, August 1, 2007; 97(4): 1054 - 1093. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Yolsal, T. Taymaz, and A. C. Yalciner Understanding tsunamis, potential source regions and tsunami-prone mechanisms in the Eastern Mediterranean Geological Society, London, Special Publications, January 1, 2007; 291(1): 201 - 230. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. Seeber, M.-H. Cormier, C. McHugh, O. Emre, A. Polonia, and C. Sorlien Rapid subsidence and sedimentation from oblique slip near a bend on the North Anatolian transform fault in the Marmara Sea, Turkey Geology, November 1, 2006; 34(11): 933 - 936. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. B. Pavlides, A. Chatzipetros, Z. S. Tutkun, V. Ozaksoy, and B. Dogan Evidence for late Holocene activity along the seismogenic fault of the 1999 Izmit earthquake, NW Turkey Geological Society, London, Special Publications, January 1, 2006; 260(1): 635 - 647. [Abstract] [PDF]
|
|
|
|
|
|
H. Kondo, Y. Awata, O. Emre, A. Dogan, S. Ozalp, F. Tokay, C. Yildirim, T. Yoshioka, and K. Okumura Slip Distribution, Fault Geometry, and Fault Segmentation of the 1944 Bolu-Gerede Earthquake Rupture, North Anatolian Fault, Turkey Bulletin of the Seismological Society of America, August 1, 2005; 95(4): 1234 - 1249. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. S. Martin Intensity Distribution from the 2004 M 9.0 Sumatra-Andaman Earthquake Seismological Research Letters, May 1, 2005; 76(3): 321 - 330. [Full Text] [PDF]
|
|
|
|
|
|
R. B. Stothers Ancient and Modern Earthquake Lights in Northwestern Turkey Seismological Research Letters, March 1, 2004; 75(2): 199 - 204. [Full Text] [PDF]
|
|
|
|
|
|
Paleoseismic Evidence of Characteristic Slip on the Western Segment of the North Anatolian Fault, Turkey Bulletin of the Seismological Society of America, December 1, 2003; 93(6): 2317 - 2332.
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| JOURNAL HOME | HELP | CONTACT PUBLISHER | SUBSCRIBE | ARCHIVE | SEARCH | TABLE OF CONTENTS |
