.I. Geodynnmics Vol. 26, No. 24, pp. 461486, 1998 0 1998 Published by Elsevier Science Ltd Pergamon All rights reserved. Printed in Great Britain PII: SO264-3707(!W)ooo66-5 0X&3707/98 $19.00+0.00 SEGMENT BOUNDARIES, THE 1894 RUPTURES AND STRAIN PATTERNS ALONG THE ATALANTI FAULT, CENTRAL GREECE ATHANASSIOS GANAS,‘* GERALD P. ROBERTS2 and TZETTA MEMOU3 ‘Department of Geography, University of Reading, Whiteknights, Reading RG6 6AB, U.K. *The Research School of Geological and Geophysical Sciences, Birkbeck and University College London, Gower Street, London WClE 6BT, U.K. ‘Division of Geophysics, IGME, 70 Messoghion Str., 115 27 Athens, Greece (Received 27 April 1997; revised 2 June 1997; accepted 5 June 1997) Abstract-The Atalanti Fault is a large active normal fault segment inside the Gulf of Evia Rift system (Central Greece), that last ruptured during the April 1894 earthquake sequence. Using structural and geomorphological interpretations of digitally processed Landsat TM satellite imagery, two regions of i) low topography, ii) minimum hinterland development and iii) transverse bedrock ridge development, 34 kilometres apart were identified; these regions are suggested to be segment boundaries constraining the length of the fault. From throw profiles and displaced syn-rift strata, we estimate a minimum slip of 810m at the central region of the fault (Tragana), increasing to a value of 1200 meters within the Asprorema embayment area. These figures averaged over a time span of 3 million years (age of oldest offset syn-rift), yield mean slip rates of at least 0.27 to 0.4 mm/year. Field studies were also conducted along the length of the Atalanti Fault Segment to re- examine and map the 1894 ruptures. The surface break is only preserved locally where the footwall comprises a resistant bedrock lithology (limestone), whilst the rest of the rupture noted in historical records propagated along the contact with the volcanic pre-rift, as well as within the syn-rift, and has since been eradicated due to man-made changes in surface morphology. The surface breaks appear not to have crossed over the segment boundaries that we propose, but seem to have ruptured the full length of the Atalanti Fault Segment, that is, 34 km. These observations suggest that the 1894 rupture is the longest mapped within Central Greece. However, it remains unclear whether the ruptures were produced solely by the 271411894 earthquake, or by two events, one week apart. We discuss the implications for fault-behavioural models and seismic hazards for the Atalanti area. 0 1998 Published by Elsevier Science Ltd. All rights reserved *Author to whom all correspondence should be addressed: E-mail: [email protected]. 461 I. INTRODUCTION In Central Greece (Fig. I 1. cruhtal cxtcnhion 14 ;lcc<mmvdated by slip along large normal fault segments of planar geometry. These faults ~ODIIC! it series of E-W to NW-SE half- grabens exhibiting major. but systematic variations in topography and bathymetry along their axes (e.g. Roberts and Jackson. 199 I ; Roberts and Gawthorpe, 1995). Syn-rift deposits of Neogene and Quaternary age have been mapped within the continental basins of Eastern Central Greece (Evia. Phiotis. Viotia: IGME. 1965. 197X; Mettos el al.. 1992). The basins are arranged in a ‘domino-style‘ of fault-controlled continental extension (e.g. Westaway. 1991 ). More than 60”/0 of the extensional deformation is accommodated by large earth- quakes ( > MS 5.8; Ambraseys and Jackson. 199~):Billiris r/ ul.. 1991). Published earthquake catalogues (e.g. Papazachos and Papazachou. 19X9: NEIC-PDEs) show that most historical and instrumental seismicity is localised within the major marine basins on either side of the Gulf of Evia. These are the Gulf of Corinth (Roberts and Koukouvelas. 1996; Armijo CI N/.. 1996) lying 60 km to the south, and the Gulf of Pagasai region in Southern Thessaly (Pavlides. 1993) lying SO km to the north. The Gulf of Evia rift has been a seismically-quiet area. with the last surface rupturing events on 77 /4; I X94 along the Lokris coast (Skouphos. 1X94). Fig. 1. C;ener,dgwlogd map01’ Easlern c cn~ral~~reecc. ~ho~~~n~ geological formations (adapted from IGME. 1989) and reglonal neotectomc deformation (fault segments and marine basins). Pre-rift rocks are mainly Mesozoic carbonates and ophiolites. Syn-rift is mostly composed of Pliocene lacustrme formations and Quaternary alluviA deposits. Fault segments shown have been defined In Ganas (‘I N/. (1996). Symbol RB is Renginion basm. Box at lower left show5 a map of Greece with the region of study. Segment boundaries, the 1894 ruptures along the Atalanti Fault, Central Greece 463 A remarkable feature of the 1894 earthquakes is that the ruptures are purported to have been 60 km in length (e.g. Richter, 1958 pages 617618; Karnik, 1971 page 104; Lemeille et al., 1977; Rondoyianni-Tsiambaou, 1984; Stiros and Rondoyianni, 1985, 1988; Papazachos and Papazachou, 1989 page 100; IGME, 1989). In contrast, earthquake ruptures mapped along fault segments in central Greece, such as those on the southern shore of the Gulf of Corinth from the 1981 and 1995 earthquakes, seem to grow up to a length of < 15-20 km (Jackson et al., 1982; Ambraseys and Jackson, 1990; Roberts and Koukouvelas, 1996; Hubert et al., 1996). Also, further north, the 13/5/95 Grevena earthquake (MS 6.6) produced surface ruptures along 8-12 km (Pavlides et al., 1995; Meyer et al., 1996) of a much larger normal fault (27 km; Clarke et al., 1997). In this paper we re-examine the extent of the 1894 surface breaks, and the length of the host fault segment using field and satellite observations. Our conclusions show that although the entire length of the fault segment we define was ruptured, a length of 34 km, it is unclear whether the ruptures were produced solely by the 271411894 earthquake, or by two events, one week apart. Sections 24 of this work present new information concerning decametre-and-km scale structures extracted from digital processing of low-sun angle satellite images (Landsat Thematic Mapper; see Salomonson and Stuart, 1989). These structures include hinterland geomorphology, transverse bedrock ridges and cross faults, all of which can be used to infer the positions of persistent segment boundaries along the strike of seismogenic normal faults (see Crone and Haller, 1991 for methodology). In addition, data (field and strati- graphic) on the throw distribution across the Atalanti Fault are presented together with the results of electrical surveying across the Atalanti plain conducted by TzM by use of the resistivity method (see also Memou, 1986). Structural data concerning fault displacement, fault-slip directions and bedrock scarp morphology are presented to support observations of fault-segment length. Section 5 includes our re-examination report of the rupture path of the 27/4/1894 M7 earthquake (see also Skouphos, 1894, pages 442448). 2. THE GEOMETRY AND KINEMATICS OF THE ATALANTI FAULT SEGMENT 2.1. Remote sensing and geomorphic analysis Remote sensing, a relief-sensitive mapping technique can be used to infer fault segment positions in rift systems because in such areas the configuration of relief may be linked with systematic variations of fault displacement (Anders and Schlische, 1994). For example, in the Gulf of Corinth rift system (Roberts and Koukouvelas, 1996; Armijo et al., 1996) as well as within the Basin and Range extensional province (Wallace, 1978; Crone and Haller, 1991) the surface topography is controlled by long-term growth of normal faults. The footwalls of such faults are marked by high mountains, where pre-rift rocks are exposed; the elevations of these mountains change along fault strike. Areas of relatively low elevation lie at the ends of the fault segments. These areas of low topography can be either recognised from space by the construction of high resolution digital elevation models (e.g. Ganas et al., 1996), or from a simple examination of I:100 000 or other suitable topographic maps of rift systems (e.g. Agar and Klitgord, 1995). These low-relief positions are characterised by a large cumulative deficit of seismic slip along the main rift-bounding fault, together with an increased ‘distributed deformation by smaller faults (Zhang et al., 1991), and, therefore, are primary candidates for recognition as persistent segment boundaries (PSBs). To map the length of the Atalanti Fault Segment and segment boundaries from space, we used a winter Thematic Mapper (TM) scene (Figs 2 and 4; scene ID = 4202208284, Path/Row: 183/33, 28.5 x 28.5 m pixel size), provided by IGME (Athens), The image was processed by AG using a Sun Sparc20rM workstation, using the Erdas ImagineTM image processing software. This scene is particularly suited for structural interpretations because of the low sun-angle (25 degrees) during its acquisition date (28 January, 1988) and its geometric rectification into the UTM projection, that conforms to the Greek national cartographic grid. The low sun-angle combined with a southeastern sun azimuth in the image selected, provide an almost ideal imaging geometry for E-W ( + 20 ) trending normal faults downthrowing to the North, such as the N290-striking Atalanti fault segment (Fig. 2). This is because relief created by the dome-shaped footwall uplift we describe below (see Fig. 5a,b; compare with Fig. 4) casts long shadows to the north of the fault segment. Shadow widths diminish towards segment boundaries because of the low values of relief. In addition, UTM-geocoded imagery can be used to measure geometrical properties of structures and landforms within acceptable accuracies (_+ 1OOm) for regional-scale mapping. To provide an insight into the digital manipulation of imagery, it was found that footwall topography along the Atalanti fault is better imaged using principal component trans- formations and linear stretching of the near-infrared spectral bands of the sensor (TM4.