In the period of Ocean Drilling Program Legs 186, 191, and 195, we successfully deployed a set of seismo-geodetic195, we successfully deployed a set of seismo-geodetic sensors in deep seafloor boreholes in four locations (JT1, JT2, WP1, and WP2). The JT1 and JT2 borehole observa-tories are located in the landward slope of the Japan Trench,the Japan Trench,Japan Trench, where we expect seismic activity beneath the observatories. These borehole observatories gave us good platforms to look to look at the dynamic process on the plate boundary in thethe dynamic process on the plate boundary in thedynamic process on the plate boundary in the seafloor.The WP1 and WP2 are considered a part of a globala globalglobal seismic network of observatories currently distributed mostly on land and ocean islands. For permanent seismic observation in the seafloor setting, installation of broadband seismic sensors in the seafloor borehole was recommended, and candidate locations were chosen (Suyehiro et al., 2006).., 2006)., 2006). The WP1 and WP2 sites are one of the first borehole obser-vatories to implement such a permanent seafloor obser-permanent seafloor obser-seafloor obser-vatory.We called these seafloor borehole observatories �NEREID� (�Neath Seafloor Equipment for Recording Earth�s Internal Deformation). The NEREID observatories are characterized by several key features that are thought tothat are thought to to acquire optimum performance from the borehole sensors. optimum performance from the borehole sensors.Firstly, sensors are cemented at the bottom of the borehole. This is very important because long-period seismic and tilt/strain sensors are sensitive to fluid flow around them, so it isthem, so it is necessarytocouplethemtothegroundverywell.Cementing to couplethemtothegroundverywell.Cementing themtothegroundverywell.Cementing to thegroundverywell.Cementingthe groundverywell.Cementingground very well. Cementing sensors at the bottom of the borehole was able to fix sensors rigidly in place, and eliminated fluid around the sensors.in place, and eliminated fluid around the sensors. place, and eliminated fluid around the sensors.place, and eliminated fluid around the sensors., and eliminated fluid around the sensors.eliminated fluid around the sensors. fluid around the sensors.the sensors.sensors. Cementing is necessary especially for strainmeters, becauses, because, because the strain of the ground should be transmitted to the sensing cylinder of the strainmeter by some means. On land, it hasmeans. On land, it has. On land, it has been widely practiced for that purpose.at purpose. purpose.Secondly, separate cables connect borehole sensors uphole for power feeding and seafloor data recording. This is to deal with possible failure of the sensors, cables, and and connectors in the borehole during and after the installation in the borehole. It worked very well, and we were able to and we were able to replace failed sensors after the installation without affecting failed sensors after the installation without affecting operation of other sensors.Sensors deployed in these observatories were configured to target specifications needs in each location of the observa-tories. For JT1 and JT2 observatories, a tiltmeter (AG510), a a strainmeter (dilatometer and three component types), and and broadband seismometers (CMG1T, PMD) are chosen to mainly target the earthquakes and slow deformation processes below the observatory. The WP1 and WP2 obser-The WP1 and WP2 obser- WP1 and WP2 obser-vatories are important for monitoring global seismic activity,for monitoring global seismic activity, monitoring global seismic activity,ing global seismic activity, global seismic activity, so two identical broadband seismometers (Guralp CMG1T) were installed for redundancy.Data from borehole sensors are recorded in the seafloor. Electric power necessary for these sensors is provided by ais provided by a provided by a stack of seawater batteries or lithium batteries stored in aithium batteries stored in a titanium sphere. Data areitanium sphere. Data are recovered by ROV by retrievingtrievinging the data recorder because the amount of the recorded seismic data was as large as 10GB per year.We conducted visits to each observatory for data recovery and maintenance once a year or more. From most of the observatories,