GPS data processing

The data of the 1439 GEONET stations have been processed in double difference using GAMIT/GLOBK software⁶². For each day, the GEONET data were split into sub-networks of about 40 sites chosen to minimize the baseline between stations and improve the resolution of phase ambiguity. Sub-networks share 2 common sites with nearby other sub-networks to “tie” the solutions together, following a similar approach as that presented in the framework of the PBO project⁶³. A reference tie network containing at least 1 station from every other sub-network is also constructed to provide additional stability to the entire network combination. For each sub-network, we reduce 24-hr measurement sessions to daily estimates of station position, choosing the ionosphere-free combination and fixing the ambiguities to integer values. We use IGS final products for the satellite orbits, satellites clocks and Earth orientation parameters (https://www.igs.org/products). Following Herring et al.⁶³, the orbit parameters are fixed to the IGS values. Ocean loading corrections are applied at each stations, using FES2004 (Finite Element Solution)⁶⁴ ocean tidal loading. The Vienna Mapping Function (VMF1)⁶⁵ is used to map the tropospheric delay in zenithal direction. The zenith delay is estimated every 2-hours and 1 gradient parameter is estimated per day. We apply atmospheric tidal and non-tidal loading correction, following Tregoning and van Dam⁶⁶ recommendation. The different sub-networks are then combined together into a single daily solution in a regional stabilization approach using a Kalman filter with GLOBK software to obtain loosely constrained daily solutions. Then daily solutions are combined into a multiyear solution to derive the time series, and to express the solutions in the ITRF2014⁶⁷ with a 7-parameter transformation using regional IGS sites. The daily GNSS position time series⁶⁸ are available at https://doi.osug.fr/staging/GNSS_products/GNSS.products.Japan.html. These GNSS position time series are then corrected from any jumps associated with documented material changes. The change of velocity before and after the Tohoku-oki earthquake is simply calculated as the difference of slope in each time series, between a linear trend before the earthquake from 2008/01/01 to 2011/03/08 and a linear trend after the earthquake from 2011/03/13 to 2013/03/10.