Résumé
Restoring the M9-class earthquake cycle in subduction zones is challenging in particular when dealing with the vertical deformation through the different phases of the cycle. Field investigations using geologic, geomorphologic, geodetic and seismological approaches including independent (and significant) uncertainties to estimate the Pleistocene, Holocene, coseismic and postseismic active deformation come to an independent estimation of the return period for M9-class earthquakes. The 2011 Tohoku-Oki earthquake (Mw 9.0) caused significant land-level changes in NE Honshu where GPS and leveling data reveal a 500-km-long section with >0.10 m coseismic subsidence reaching 1.15 m close to the epicenter. A similar trend is observed for the 120 years preceding the earthquake. Conversely, CGPS data during the 880 days (2.4 years) following the 2011 main shock suggest a logarithmic uplift trend with up to 0.25 m of postseismic upheaval. Uplift is similarly recorded on the long term by late Quaternary marine terraces that exhibit 0.1-0.4 mm/yr uplift rate. The drainage network that incises terraces also documents long-term uplift as shown by the pattern of incision depth, river profile convexity and knickpoint migration. Since offshore M8 earthquakes have negligible permanent effect on the coast and no major active fault occurs along the coastline north of the epicentre, the long-term vertical deformation can be used as a proxy to reveal the M9-class earthquake cycle. On the Pleistocene timescale, the coseismic M9-class coastal subsidence appears to be concealed by the post- and interseismic deformation. The deformation budget across the whole cycle reflects temporal variations in plate boundary coupling and suggests a concealing time of 500-1000 years for the 2011 earthquake, which we propose as the minimum recurrence interval of M9-class earthquakes. This matches well the suggested 3 to 4 giant earthquakes identified in tsunami deposits for the past 3 kyr, and the 350-700 yr interval derived from the slip deficit.