The great Hokkaido Toho-Oki earthquake (M_JMA=8.1) occurred at 13h 23m on October 4, 1994 (GMT), in the region off the Shikotan Island. Since all STS-1 seismographs in Eastern Japan were saturated by the earthquake, broad-band strong-motion seismographs (velocity response is flat from 0.015s to 600s) installed at two stations in the region of the northern end of the main land of Japan, where the seismic intensity of 4 was reported by JMA, were used to make a long-periods (≥70s) moment tensor inversion. Our solution in terms of best double couple is (strike, dip, slip)=(69°, 67°, 119°), M_c=1.8×10²¹ Nm (M_w=8.0), and centroid depth=42km. This depth of 42km and aftershock distribution determined by the Research Center for Earthquake Prediction of Hokkaido University may suggest that the earthquake occurred within the Pacific plate.
Since the data of this great Hokkaido Toho-Oki earthquake are recorded at epicentral distances of about 5 to 6°, the spatiotemporal source finiteness must affect our estimate of moment tensor even at long-periods (≥70s). Our numerical experiments using finite sources (e. g., unilateral moving line source) show that the source finiteness has a small influence on the source mechanism estimate, but causes the seismic moment to be underestimated. The moment reduction rate depends on the mode of moment release (moving source) and is about 50% in the case of the October 4 earthquake. This result may suggest that our estimation of the moment of the October 4 earthquake should be multiplied by a factor of about 2. The experiments also show that the source finiteness effect can be reduced by using data of longer periods when the station coverage is good.