Seismic Observation Survey
■ April 1973 – A study of seismic activity in and around the southern Hidaka Mountains began using film- and ink-records from four stations: the observations were composed of the Urakawa (KMU), Erimo (ERI) and Sapporo (SAP) stations of Hokkaido University, and a commissioned private observatory in Hiroo (MYR). Using these records, a survey of seismic activity in and around the southern Hidaka Mountains, Hokkaido, was initiated.　In addition, we read the smoked-paper records from the JMA high-sensitivity volcanic observations in Hokkaido and ink-records from the high-density, high-sensitivity seismic network of Kitakami Observatory, Tohoku University, as necessary, to improve the accuracy of the earthquake locations.

■ July 1976- A seismic network with a radio telemeter transmission system consisting of 9 stations (6 stations in the Hidaka area, 3 stations in Sapporo, Esan and Akkeshi) was constructed to routinely monitor seismic activity in a wider area. Using the digital seismic waveform data transmitted and accumulated from the network to the Sapporo campus of Hokkaido University, we have been involved in the development of analysis software and seismic wave recordings for the investigation of seismic activity of microearthquakes (M3 or less) in Hokkaido and its vicinity.

■ Furthermore, we were involved in the introduction of a satellite telemetry transmission system for the whole of Hokkaido, data exchange with the Sapporo District Meteorological Observatory, centralization of seismic waveform data by introducing a real-time data distribution system and a common seismic waveform data processing system (WIN system developed by the Earthquake Research Institute of the University of Tokyo). Through the upgrading of the seismic observation system and the centralization of vast seismic wave data as described above, we have achieved data distribution with universities, the Japan Meteorological Agency (JMA), the National Research Institute for Earth Science and Disaster Resilience (NIED), and others. Currently, seismic data from almost the whole country is available to anyone via the Internet.

■ Participated in joint observational studies of subsurface structure exploration by the Research Group for Explosion Seismology: the subsurface structure survey of the Matsushiro seismogenic area in 1967, the Shakotan Peninsula to Cape Erimo subsurface structure survey in southern Hokkaido in 1968 and 1969, the seabed structure survey off Tohoku in 1974 and 1978, and the Chugoku area reflection wave observation survey in 1978.

■ Participation in surveys of aftershock activity based on emergency aftershock observations after large earthquakes: aftershock activity surveys after large earthquakes such as the 1973 Nemuro Peninsula earthquake, 1982 Urakawa earthquake, 1993 Hokkaido southwest offshore earthquake, 1995 Hyogo-ken Nanbu earthquake, 2003 Tokachi-oki earthquake, etc.

■ Production of a total of about 40 solar panel and battery-powered long-duration unmanned cassette recorder seismic observation systems with digital clocks, and investigation of microearthquake activities and crustal velocity structure surveys in various districts of Hokkaido using these systems: for example, the southern Hokkaido subsurface structure survey using the Usu Volcano eruption earthquake, the Hidaka and Tokachi regions, the southern Oshima Peninsula and, eastern Hokkaido, etc.

■ Involved in geophysical surveys at the hotspot Erebus volcano (3794 m), Ross Island, Antarctica: the world’s first unmanned telemetric observation in collaboration with Hokkaido University, the Polar Research Institute, the Earthquake Research Institute, the University of Tokyo, the University of Alaska, and the University of Wellington, New Zealand. The world’s first unmanned telemetered seismic observation system was built and maintained for about four years after its installation. A tripartite seismic network composed of handmade, long-term unmanned cassette recorders was also deployed on the summit where we stayed for a month to observe seismic data during the eruption.

■ In order to respond to the request from the Government of the Republic of Indonesia to establish a wireless unmanned seismic observation network in West Java in collaboration with the Indonesian Museum of Geology (Bandung Museum of Natural History) and to support its observation and analysis, I has been involved in four support activities based on Japanese Official Development Assistance.　In addition, on the request of the Japan International Cooperation Agency (JICA), I accepted one staff member of the Indonesian Museum as a short-term field trainee.

■ Involved in the exploration of seafloor seismic activity and oceanic crustal structure by the self-floating ocean bottom seismic observation systems (OBSs) in the offshore areas of Kagoshima, Niigata, Shakotan Peninsula and the intersection of the Kuril Trench and the Japan Trench, which were jointly conducted by Hokkaido University, Tohoku University, Kagoshima University, JAMSTEC, and the Earthquake Research Institute of the University of Tokyo. Hokkaido University, as its own, has conducted annual self-floating ocean bottom seismic observations to monitor subsequent seismic activity in the source area of the 1952 Tokachi-oki earthquake (M8.2), the largest ever recorded in the seas surrounding Hokkaido.

■ Involved in global oceanic crustal structure exploration in the North Atlantic Ocean off Norway to Iceland and in the Caribbean Sea near Guadeloupe, in collaboration with the University of Bergen, Norway, and the University of Reykjavik, Iceland.

■ International collaboration with the Carnegie Institution for Science, USA, on crustal strain observations using remotely operated borehole high-sensitivity, high-sampling volumetric strain meters and long-period seismometers (both developed by the Carnegie Institution for Science, USA) at the Urakawa observatory (KMU) and three sites (TES, NIT, KUT) in Teshikaga town, Hokkaido. Each observation system was connected to Hokkaido University and the Carnegie Institution for Science via the Internet.

Mathematical Investigation of Seismic Waveforms
■ Involved in the development of an automatic real-time determination method of seismic wave arrival times based on the representation of seismic waves by autoregressive state-space modeling, in collaboration with the Institute of Statistical Mathematics. This work has attracted widespread interest worldwide, and I gave an invited lecture at a meeting of young researchers, the International School-Symposium on “Earthquake Prediction and Earthquake Engineering” organized by the European Geophysical Union in Plovdiv, Bulgaria, from 16 to 23 April 1990. The automatic P- and S-wave determination software introduced there was provided to earthquake-related institutions (e.g. University of Memphis, Geological Survey of Edinburgh, Earthquake Research Institute, University of Tokyo, Department of Oceanography, Tokai University.).

■ Involved in the development of algorithms for high-precision signal separation and extraction of crustal deformation records (here, volumetric strain) using statistical state-space modeling: the development work was carried out with the Institute of Statistical Mathematics, and its performance was verified in collaboration with the Earthquake Research Institute, University of Tokyo, Carnegie Institution for Science, Japan Meteorological Agency, and others. The true strain record was extracted from data recorded in the JMA’s volumetric strain observation network, which has been set up to predict earthquakes in the Tokyo metropolitan area, by removing crustal response components due to atmospheric pressure, rainfall, tide and artificial noise. Furthermore, the Kalman filter was applied to the period of missing data due to system troubles such as sudden power failures during earthquakes, to estimate the continuous crustal strain as much as possible.

■ As a fundamental study of seismic wave theory, I was involved in numerical experiments of the exact ray theory of the Cagniard-de Hoop method using the supercomputer at the Large Computing Centre of Hokkaido University. As a case study using the Cagniard-de Hoop method, we attempted to estimate the crustal structure beneath the survey line by comparing theoretical reflected and refracted waves calculated using seismic wave velocity parameters obtained from a seabed crustal structure survey conducted in the Japan Sea.

■ Similarly, using the supercomputer at the Large Computing Centre of Hokkaido University, numerical experiments were carried out to investigate how seismic energy incident on a seismic velocity boundary is divided into reflected and transmitted waves at the boundary and refracted waves propagating along the boundary using the exact ray theory of the Cagniard-de Hoop method.

■ When seismic waves propagate inside the earth, the observed seismic waveform is generally expressed as a convolution of the source response, the propagation path response, the geometric decay coefficient, the subsurface structural response at the receiving point and the seismometer response used for the observation. By introducing the simple assumption that the source response, the propagation path response, and the geometric decay coefficient are almost the same among the components of the seismic wave, the response function of the velocity structure at the observation point can be obtained following Phinney’s method based on Haskell theory. Seismographs with a 5-second eigen-period were installed at several JMA seismic stations in Hokkaido, Japan, to observe earthquakes that occurred in various parts of the world. In my master’s thesis, I calculated the amplitude spectra of the vertical and horizontal components of P-waves of deep far-field seismic waves arriving from various directions recorded at the JMA stations of Sapporo and Urakawa and estimated the P-wave velocity structure under both stations from the mean distribution of the amplitude spectral ratios of the vertical and horizontal components.

Major Findings
■ The stationary-state microearthquake observation network established for the first time by Hokkaido University made it possible to investigate the degree of stationary-state microearthquake activity in southern Hokkaido, which was previously unknown. For example, the foreshock and aftershock activities of the 1982 Urakawa-Oki earthquake and the 2003 Tokachi-oki earthquake, which occurred in the area, could be investigated in detail. The seismic activity in the Hidaka Mountains and surrounding land areas is also high, and its spatio-temporal distribution can now be accurately determined. Furthermore, the self-floating seafloor seismic observation network that Hokkaido University temporarily deployed in the hypocenter area of the 1952 Tokachi-oki earthquake (M8.2) confirmed many clusters of microearthquakes immediately before the 2003 Tokachi-oki earthquake (M8.0). The microearthquakes were small in scale and were rarely detected by the stationary seismic network on land.　A few years later, we deployed a similar temporary seismic network again, but found that this type of clusters of microearthquakes could hardly be detected in the cluster areas identified immediately before the 2003 Tokachi-oki earthquake. Recently, the permanent submarine cable-type stationary seismic observation network called “S-net” has been deployed on the seafloor in the southern Kuril trench and northern Japan trench, and data from this permanent network are added to the JMA’s usual hypocenter location calculations. If there is a sudden increase in small clusters of earthquakes, as described above, they will be included in the JMA’s earthquake catalogue.

■ From seismic mechanism analysis of earthquakes occurred in the shallower southern Hidaka Mountains, we inferred an east-west compressive stress field, and suggested a stress field different from the seismic mechanism solution inferred from the direction of subduction of the Pacific plate directly under the Hidaka Mountains. This newly discovered stress field is the basis for today’s well-known theory of the westward motion of the Kuril forearc (sliver) carrying the Habomai Islands, Shikotan Island, and Nemuro Peninsula.

■ The frequency distribution of earthquakes by magnitude and the pattern of seismic distribution in and around the hypocenter areas of major earthquakes such as the 1982 Urakawa-Oki earthquake, the 1993 Hokkaido southwest offshore earthquake, the 1994 Hokkaido east offshore earthquake and the 1995 Hyogo-ken Nanbu earthquake were found to change before the major earthquakes.

■ The new three-dimensional seismic tomography method by Aki and Lee (1973) was applied for the first time to natural earthquake travel-time data obtained from long-term seismic observations in the Hidaka district of Hokkaido, and Imaged distribution of P-wave velocity anomalies to a depth of about 100 km below the Hidaka Mountains. From this tomographic analysis, a large low-velocity zone in the deep western part of the Hidaka Mountains and, conversely, a high-velocity zone in the eastern part are inferred.　 This asymmetric structure beneath the Hidaka Range cannot be explained by the geosynclinal orogenesis, which was once accepted as the orogenic movement of the Hidaka Range. Furthermore, the low-velocity zone estimated by this tomography analysis was a hypocenter area of high seismic activity, suggesting a peculiar relationship between seismic wave velocity and seismic activity under the Hidaka Mountains. This led to the continuation of similar studies in the region, and the tomographic image of the results has been updated with each study. Further analysis with the addition of temporary ocean bottom seismic network data and national centralized seismic data has produced a broader subsurface structure model.

■ A state-space model of microearthquake waveforms was developed, enabling the identification and extraction of wave phases such as P-waves, S-waves and background noise from complex seismic waveforms using the Akaike Information Criterion (AIC). Furthermore, a real-time high-precision method for determining the arrival times of P- and S-waves using an auto-regressive state-space modeling was developed, enabling fast and automated calculation of hypocenter locations.

■ Flat amplitude spectrum up to the vicinity of DC is required for highly accurate estimation of the attenuation structure in the vicinity of the observation point. For this purpose, it is necessary to remove the background component and the P-wave coda component superimposed on the S-wave part from the raw broadband seismic waveform in advance. Here, the state-space modelling was applied to the raw broadband seismic waveforms to estimate the optimal autoregressive models for the background, the P- and the S-wave parts, respectively. We selected earthquakes that occurred in the slab (Wadati-Benioff zone) beneath the Sawauti station of Tohoku university and developed the state-space modeling method to estimate P- and S-wave forms from their observed waveforms recorded by the wide-band, high-dynamic-range seismometer made by the Carnegie Institution for Science and installed at Sawauti station. And the amplitude spectral distribution ratio method for the P- and S-wave forms extracted by the state-space modelling, could provide the distribution attenuation values (Q values) in the vicinity of Sawauchi. As a result, the remarkable low Q values zone near Sawauchi was already found to be the same region as the clear low-velocity region just below the volcanic front, which had already been estimated from the seismic tomographic analysis conducted by Tohoku University. The existence of partial melting of dry mantle material was inferred from high-temperature and high-pressure rock experiments.

■ State-space modelling was applied to strain data before and after the 2003 Tokachi-oki earthquake (M8.0) recorded by the highly sensitive and high-sampling Sacks-Evertson borehole volumetric strainmeter developed by the Carnegie Institution of Japan and installed at the Urakawa Seismological Station (KMU), Hokkaido University, in order to separate atmospheric pressure, precipitation and tides, We attempted to separate crustal deformation caused by atmospheric pressure, precipitation, tides and anthropogenic noise. We attempted to estimate the rectangular fault motion calculated from the strain records extracted by removing these effects from the raw records and the displacement records of GPS stations deployed in the Tokachi and Hidaka regions. A volumetric strainmeter located 105 km from the hypocenter of the main shock recorded a clear and gradual strain event after the 2003 Tokachi-oki earthquake (Mw8.0, 19:50:06 UTC 25 September 2003). This consisted of a 4-day contraction and a 23-day expansion period. Displacements were also recorded at a GPS site in southeastern Hokkaido during the same time period. Quasi-static calculations were used to produce a composite waveform of the measured quantities. All data were satisfied by a propagating source two-stage model with slow reverse slip, a uniform amplitude of 50 cm and a constant 9 cm/s (first stage) and exponentially decreasing rupture propagation velocity from 3 to 0.7 cm/s (second stage). The post seismic slip event was considered to be in the same plane as the mainshock rupture on the upper surface of the Wadachi-Benioff double seismic zone (DSZ), approximately coinciding with the seismic rupture. Normal earthquakes release only about 30% of the plate motion in this section of the subduction zone.

■ I was able to realize from the world’s first long-term unmanned telemetry continuous seismic observation at the hotspot Ross Island Erebus volcano (3,794 m) in Antarctica that it is one of the most active volcanoes today, with plumes of smoke rising from the lava lake at the summit, sometimes accompanied by ground rumbling and earthquakes. Around the lava lake at the summit, all possible observations were carried out, including seismic observations, geodetic surveys, surface temperature measurements, geoelectric current measurements and rumbling observations. This continuous observation demonstrated that long-term unmanned telemetry seismic observation is possible in the Antarctic circle, where the temperature at the summit in midwinter can be as low as minus 50-60°C, using solar cells and batteries buried underground where the ground is always warm.