GEOPHYSICAL RESEARCH LETTERS, VOL. 25, NO. 1, PAGES129-130, JANUARY 1, 1998 Comment on"Geodetic investigation of the 13 May Kozani-Grevena (Greece) earthquake" by Clarke et al. B. Meyer 1,R. Armijo 1,D. Massonnet 2,J.B. deChabalier 1,C. Delacourt 1, J.C. Ruegg 1,j. Achache 1, and D. Papanastassiou 3 In a recent paper, Clarke et al. [1997] try to describethe coseismic displacement field of the 1995, Grevena (Northern Greece) earthquake by comparing triangulation surveys of a geodetic network made in 1984-1986 and a post-earthquake survey of the same network with GPS made in 1995. The composite triangulation-GPS horizontaldisplacement obtained and the model proposed are inconsistent with important tectonic, seismological and geodetic observations that Clarke et al. [1997] have ignored in their inversion procedure. Concerning the tectonicsthe Clarke et al. [1997] paper is internally inconsistent. On the one hand Clarke et al. [1997] postulate that "no obvioussurface break was found so it was not possible to locatethe fault plane a priori" (p. 707), but on the other hand they incorporate in their map (their Fig. 1), without appropriate citation, the approximate location of the surfacebreak along the Palaeochori Fault described by Meyer et al. [1996]. This break and the other active faults in the area have been identified with modern morphological techniques using the SPOT satellite imagery and the field evidence for Holocene slip [Meyer et al., 1996]. The tectonic results ignored by Clarke et al. [1997] are as follows. The 1995 surfacebreak was mappedin detail for 8 km along the pre- existing Palaeochori normal fault, a feature with topographic relief of about 50 m clearly seenin the SPOT images(see Fig. la in Meyer et al. [1996]). The fault strikes N70øE with near- surface dip of 70 ø to the northwest and the 1995 break was a continuous strand formed by open fissuresand small scarps with 2-4 cm down-to-the-northwest normal slip (see Figs. 1, 2a and 2b in Meyer et al. [1996]). There is no evidence in the SPOT satellite imagery nor in the field, of significant surface faulting southwestward beyond the mapped southwest tip of the Palaeochori Fault. Several smaller segments, however, are foundstepping NE and splayingSE of the northeast tip of the Palaeochori fault. The presenceof numerous small slumps over the surfacetraces of these fault segments suggests that they also rupturedduring the main event. One important result was to establish that the much more prominent Servia and Paliuria faults did not break, in spite of the clear evidencefor Holocene activity (see Figs. 1, 2c and 2d in Meyer et al. [1996]). Selectively, Clarke et al. [1997] use this tectonic lInstitut dePhysique duGlobe, Pads, France 2CNES, Toulouse, France 3National Observatory, Athens, Greece Copyright 1998 by the American Geophysical Union. Paper number 97GL03447. 0094-8534/98/97GL-03447505.00 information to fix the position of the fault plane, not the upward nor the lateralextentof the rupture. The model by Clarke et al. [1997] is consistent with the fault plane solution of the main shock [Hatzfeld et al., 1997] but it proposes that the ruptureextended 27 km along strike, far beyondthe observed tips of the Palaeochori Fault, WNW across the Vourinos Range and especially ESE into the Mesohellenic Trough. This gives a seismic scalar momentof 16.3 x 1018 Nm, twice that ofthe Harvard's CMT solution (7.6 x 1018 Nm); improbable though not impossible. Finally, the compositetriangulation-GPSvectors and the model by Clarke et al. [1997] are in serious conflict with geodetic information provided by the SAR interferometry, a technique that has now proved to be of prime importancein resolving the coseismic [e.g., Massonnet et al., 1993] and even the more subtle postseismic displacement fields [Peltzer et al., 1996]. For the 1995 Grevena event the basic SAR information is given by Meyer et al. [1996]. To summarize, the S AR interferograms obtained with the ERS-1 satellite scenes covering two-years intervals including the date of the earthquake reveal 11 well-definedconcentric fringesoutlining a kidney-shaped area of subsidence of400 km 2 with maximum of about 30 cm (Fig. la). The model by Clarke et al. [1997] impliesmany more fringes (20), corresponding to about 53 cm subsidence (Fig. l c), and a longer along-strike subsidence zone of 50 km compared to the about 25 km observed in the interferograms. The surface displacement field seen in the interferograms can be modelledusing a dislocation embedded in an elastic half space,incorporating the tectonicconstraints and increasing progressively the complexity of the modelto fit quantitatively, by trial and error, the details of the SAR information(Fig. lb). These detailsallow for modellingwith some precisionthe faulting complexities of the NE end of the rupture zone [Meyer et al., 1996]. The overall model is consistentwith the tectonic constraints,the hypocenter, the fault plane solutionand the CMT scalarmomentin addition to the inference that the surface deformation seen in the interferograms is mostly coseismic. The horizontal displacement field derived from this model (Fig. lb) can be compared with the composite triangulation-GPS vectors (Fig. l c). The corresponding vectors are roughlycomparable in the small area close to the Palaeochori Fault, but they are completely inconsistent in the far field. The horizontal vectors predicted from the model by Clarke et al. [1997] fit somewhat better the triangulation-GPS vectors in the near field, but important discrepancies persist in the far field. Overall the fit is, in our opinion, unsatisfactory. The large composite displacements (10-20 cm) seen with the triangulation-GPS technique at significant distance from the Palaeochori fault cannot be explained by any reasonable coseismic model (Fig 129