The Headquarters for Earthquake Research Promotion (HERP) commonly uses two methods to estimate a crustal earthquake magnitude, one of which is the rupture area-based model based on its rupture area and the other is based on its fault length. Each model incorporates uncertainties related to fault characterization (HERP, 2017a). The rupture area-based model is used on a combination of scaling relations, called “3 stage scaling model” (e.g., Irikura and Miyake, 2011), in which an accurate magnitude can be estimated. However, there could be so much uncertainty about fault width obtained from fault dipping angle and seismogenic layer thickness, which makes it difficult to calculate an appropriate rupture area. On the other hand, fault length is the sole independent variable needed for the length-based model. This model does not take its fault width into account, thus it is thought that this model could lead to overestimation or underestimation of the magnitude.

　 In this study, we performed multiple regression analysis to determine the relationships between rupture parameters and earthquake magnitude with an intermediate model between the above two HERP models using a qualitative parameter of fault type. We hypothesized that the relationship between the rupture area and the magnitude could be modeled easily by its fault length and its fault type, which can be both acquired from tectonic geomorphological surveys. The obtained multiple regression coefficients were consistent with the general understanding of fault mechanics and geometry. We then compared the outcomes of our multiple regression analysis with the results indicated by Irikura and Miyake (2001) and Matsuda (1975), which are adopted in HERP (2017a) for the second stage of earthquake scaling. All results were generally consistent if the fault type in their regression data and newly obtained empirical relationship between Mw and M_J are considered. The multiple regression equation proposed in this study is valuable as a simple intermediate model that bridges the two models used in HERP (2017a) by incorporating the difference of fault types into earthquake magnitude estimation.

　 The Iyo-nada Sea is the inland sea in southwestern Japan and is located at the western end of the Median Tectonic Line (MTL) that extends east-west through southwestern Japan. We conducted seismic reflection and refraction surveys to understand geological structures and to evaluate activity of faults at the Iyo-nada Sea. We recognized three geological structures accompanied with the MTL in this survey. Those are the boundary between the Ryoke granites and the Sambagawa metamorphic rocks (R/S boundary), the boundary between the Izumi Group and the Sambagawa metamorphic rocks (Iz/S boundary) and the MTL active fault system (MTLAFS). The R/S boundary in the Iyo-nada Sea strikes southwest, coinciding with the linear trace of the MTLAFS. On the other hand, the Iz/S boundary extends southwest along a curved coastline of the Sadamisaki Peninsula and is not concordant with the trace of the MTLAFS. Additionally, the Iz/S boundary in the central Iyo-nada Sea is overlaid with sedimentary layers (Neogene to Quaternary) based on shallow seismic survey results. These results indicate that the Iz/S boundary is not an active fault in the central Iyo-nada Sea.

　 Marine terraces (probably Holocene terraces) divided into three levels of L1 to L3, representing the activity of off-shore active faults, have been developed along the northern coast of the Noto Peninsula, central Japan. To evaluate the mean vertical displacement rate and recurrence interval of off-shore active faults, this study tries to estimate the emergence age of terraces using an idea of time predictable model with the rate of relative sea-level fall. In the Saruyama-oki segment off the northwestern coast, assuming the ages of the L1 terrace to be 6,000 or 3,500 years ago based on previously reported peak age of post-glacial transgression, it is inferred that the mean vertical displacement rate is 0.87 mm/year or 1.49 mm/year, and recurrence interval is 2,000 or 1,200 years. However, the emerged sessile assemblages indicate the shortest interval of 300 years because the height and age suggest that two uplift events of 0.7 m and 0.8 m occurred in the 9th century and 12th to 13th centuries, respectively. In the Wajima-oki and Suzu-oki segments off the north-central to northeastern coast, the height distribution of the lower terraces indicates undulating deformation with two peaks, which is consistent with the height distribution of the late Pleistocene terrace. This suggests the characteristic displacement and its accumulation due to the fault activity of these segments. Under the same assumptions for estimating the age of terraces, the mean vertical displacement rate and recurrence intervals are 0.67-0.72 mm/year or 1.14-1.23 mm/year and 900-1,400 years or 500-800 years respectively. The age of emerged sessile assemblage indicates that the latest event of the Wajima-oki segment can be correlated to the historical earthquake of AD 1729.