Questioning the Precursory Tilt Change Before the 1944 Tonankai Earthquake

It is well known that a precursory tilt change was detected by leveling survey just before the 1944 Tonankai earthquake (M7.9), which occurred at 13:30 on December 7, 1944, along the Nankai Trough subduction zone in southwestern Japan (Mogi, 1984). This precursor was considered to be caused by a pre-slip on the plate boundary megathrust (e.g. Linde and Sacks, 2002), and the Tokai seismic gap experiment of Japan is based on such an interpretation. However, this observation was not authorized as a reliable earthquake precursor by the IASPEI subcommittee for earthquake prediction (Wyss and Dmowska, 1997), partly because the description of the observation was not enough. In order to investigate this precursory in more detail, I reviewed the original leveling logs. The leveling survey started in late November near Kakegawa City, and continued until late December. The same leveling route was measured twice, that is, before and after the Tonankai earthquake. As Mogi (1984) pointed out, relative height for two 700m-long subsections measured after 10:30 of December 7 had large (>4mm) discrepancy from the measurement of the same subsection in the opposite direction 1-3 days before. This may look significant, but these were not the only example of unusually large two-way misfits. There are two other large misfits which are not related to precursory phenomena. Thus the large two-way misfits cannot be a sufficient evidence of precursory deformation. Mogi (1984) also pointed out that there was also accelerating tilt change starting from about 3 days before the main shock. However, in the morning of December 7, multiple subsections did not show significant changes from 3 days before. Therefore, even if the precursory tilt change existed, the observation do not necessarily indicate an acceleration. Finally, repeated leveling before and after the main shock yields vertical displacement profile, which reflects the coseismic as well as the possible preseismic deformation. A finite fault modeling of this displacement profile shows that the responsible fault must be as shallow as 10-15km. If this is correct, the precursory deformation cannot be attributed to a pre-slip around the source region of the main shock on the plate boundary. In summary, the precursory tilt change might occur before the 1944 Tonankai earthquake. But it is not impossible to prove by the existing dataset. It will be more important to seek for a new evidence with state-of-the art observations such as continuous GPS and borehole measurements.