Infrequent megathrust earthquakes, with their complex cycles and rupture modes, require a high-resolution spatiotemporal record of tsunami inundations over thousands of years to provide more accurate long-term forecasts. The geological record suggests that Mw>8 earthquakes in the Kuril Trench occurred at intervals of several hundred years. However, uncertainties remain regarding the rupture zone, owing to the limited survey areas and chronological data. Therefore, we investigated the tsunami deposits in a coastal wetland of southeastern Hokkaido, Japan, to characterize the tsunamis that have originated from the Kuril Trench over the last 4000 years. On the Erimo coast, more than seven sand layers exhibited the common features of tsunami deposits, such as sheet distributions of several hundred meters, normal grading structures, and sharp basal contacts. According to numerical tsunami simulations, the 17th-century sand layer could be reproduced by using a multiple rupture zone model (Mw~8.8). We used high-resolution radiocarbon dating and tephras to correlate the tsunami deposits from the last 4000 years with those reported from regions ~100 km away. The tsunami history revealed here shows good agreement with histories of adjacent regions. However, the paleotsunamis reported to have occurred in regions > 200 km away include some events that differ from those in this study, which suggests a diversity of Mw>8 earthquakes in the Kuril Trench. We clarified the history and extents of earthquake-generated tsunamis along the southwestern end of the Kuril Trench, which were previously unknown. Our results provide a framework for magnitude estimations and long-term forecast of earthquakes.

Sediment transport modeling (STM) is a potentially effective tool for estimating the magnitude of tsunamis and earthquakes without historical records. However, applying STM to prehistorical tsunamis is challenging because of many uncertainties in topography and roughness. In the coast of Hidaka, Hokkaido, Japan, there is potential to conduct STM even in the absence of historical records because of the comprehensive geological data that reveal the coastal evolution during the Holocene in addition to tsunami sediment surveys. The tsunami deposits in Hokkaido suggest the presence of events on a larger scale than historical tsunamis; particularly the 17th-century tsunami had multiple potential wave sources other than a Kuril Trench earthquake, inhibiting its magnitude estimation. In this study, we applied STM to paleotsunamis in the coast of Hidaka, where the wave source is unknown and there are comprehensive geological data. The modeling parameters—paleotopography, roughness, grain size, initial sand source, sea level, and beach ridge height—were estimated using data obtained from geological surveys and sensitivity tests. The modeling of a tsunami induced by a Kuril Trench earthquake reproduced the sediment distributions and sedimentary structures of the observed sand layers better than that of the extreme storm and volcanic tsunami. The paring down wave sources of the sand layer implies that a wider rupture zone in the Kuril Trench is less likely. This case study provides information on the parameters that geologists and modelers should consider when applying STM to paleotsunamis.