Introduction

The 2011.03.11 Mw9.1 Tohoku (Japan) earthquake and the 2017.11.12, Mw7.3 IraneIraq earthquake are two recent events which highlight the problem of underestimation of the maximum magnitude of earthquakes expected in various regions. A solution of this problem requires expanding the time interval of noninstrumental seismological observations which are necessary to provide reliable estimates of the maximum magnitudes of expected earthquakes and of the seismic hazard and risk in various regions. Analysis of historical records is the simplest approach for that in many parts of the world [1,2]. However, the efforts at expanding the length of records of pre-instrumental earthquakes face two limitations: First, many events may be missing from earthquake catalogs; for example earthquakes not producing surface faulting and important ground deformation are not usually be recorded by typical, geology-oriented palaeoseismic studies [3], while strong earthquakes can be missed for various reasons, even in regions with a rich historic record. These may give the wrong impression of broad aseismic areas [4]. Second, literary evidence of earthquakes is usually recorded in major towns and in cultural, military or commercial centers, giving the impression that historical earthquakes selectively occur around such areas [1]. In addition, the destructive effects of earthquakes are dominated by the strong motions generated by the fault rupture and which are subsequently modified by local soil dynamics effects (soil amplification and topographic local aggravation or on the contrary attenuation of seismic ground accelerations [5,6]. Reports of damage and of the feeling of the strong motion in antiquity, roughly expressed by seismic intensities, are usually limited and may have a fragmentary character producing biased perceptions of the characteristics of the ancient earthquakes. This problem is highlighted in the following two recent examples. The 2017.08.21 M4.0 earthquake in Ischia Island (Naples, Italy) produced death toll and localized major destruction [7], while the shallow, 2014 M6.9 earthquake in the North Aegean, at the extension of the North Anatolian Fault, produced very limited ground shaking and practically no damage close to its epicenter, but increased intensities about 50 km away from the fault [8]. Hence, if the Ischia earthquake had occurred some centuries ago, its damage reports would have probably been interpreted as evidence of a much stronger earthquake (biased result). On the contrary, the North Aegean earthquake would have been practically ignored (missed event), and if recorded, it would have been interpreted either as a smaller magnitude event, or as a rather remote event (biased result). These cases are definitely not unusual, because some of the major seismic destructions in the last 50e60 years, for example the 1960 Agadir (Morocco), the 1963 Skopje (FYR of Macedonia) and the 1999 Athens (Greece) earthquakes were associated with moderate earthquakes (M
۶٫۰) and with no significant signs of seismic surface faulting [6,9]. Hence, a requirement for a reliable analysis of the seismic risk and hazard is to know whether and how it is possible to combine the seismological, geological and engineering characteristics of earthquakes of the pre-instrumental period. This article examines this problem focusing on the 1750 century earthquake in Croatia. Based on a combination of coastal palaeoseismic and tectonic data and Finite Fault Modeling on one hand, and on the other hand, on detailed historical and archaeological information for seismic damage, two alternative scenarios for this event are examined: a small (~M5) earthquake with intensities amplified due to soil dynamic effects, and a strong (M > 6.0) earthquake (preferred scenario).