4:00 PM - 4:15 PM
[SCG53-03] Status of REGARD: Japanese Real-time GNSS Analysis System
Keywords:REGARD, GNSS, real-time analysis, PPP, MCMC
The displacement data produced by GNSS observations never saturate for large earthquakes in contrast to seismometer data that has a limitation of saturation due to its intrinsic frequency. Geospatial Information Authority of Japan (GSI) and Tohoku University have developed a real-time GNSS analysis system named REGARD (Real-time GEONET Analysis System for Rapid Deformation monitoring), which estimates finite fault models using observation data of the Japanese nationwide GNSS CORS network named GEONET.
REGARD has three subsystems: (1) the real-time positioning subsystem, (2) the event detection subsystem and (3) the fault model inversion subsystem. The real-time positioning subsystem always processes real-time kinematic analysis of ~1300 stations belonging the GEONET. When the event detection subsystem receives an earthquake information from earthquake early warning system operated by the Japan Meteorological Agency or detects large displacement according to RAPiD algorithm (Ohta et al., 2012), the subsystem starts to calculate displacement of the GEONET stations to be used and send the displacement data to the fault inversion subsystem. The fault inversion subsystem estimates two types of fault model. The One is a single rectangular fault model, which is available for earthquake event which occurs at inland and subduction zone. The other is slip distribution model, which is suitable for earthquake events which occurs at great subduction zone. Estimated Mw from each fault model are sent to the relevant people such as Japanese Agency which is in charge of response to national disaster and railway company so that they can decide what they should do next.
In addition to above features, GSI has improved REGARD aiming more stable operation and high reliability of estimated fault model. GSI introduced Precise Point Positioning (PPP) as method of 1 Hz real-time positioning. It allows us to be free from analysis problem due to base station. RTK method may include positioning error of rover station due to it of base station. It sometimes causes difficulty of calculation of station displacement and noises of estimated fault model. Another improvement is introduction of Markov-Chain Monte-Carlo (MCMC) for estimation of single rectangular fault according to Ohno et al. 2021. It provides more detail information regarding uncertainty of each estimated parameter such as geometry of fault, slip, etc. and more reliable estimated result avoiding local minimum problem. These improvements make REGARD more stable and more reliable.
In this presentation, we will explain overview of former REGARD and detail of improvements of REGARD in these days. Then we show long-term evaluation of REGARD such as artificial noises of real-time positioning, over fitting of fault inversion.
REGARD has three subsystems: (1) the real-time positioning subsystem, (2) the event detection subsystem and (3) the fault model inversion subsystem. The real-time positioning subsystem always processes real-time kinematic analysis of ~1300 stations belonging the GEONET. When the event detection subsystem receives an earthquake information from earthquake early warning system operated by the Japan Meteorological Agency or detects large displacement according to RAPiD algorithm (Ohta et al., 2012), the subsystem starts to calculate displacement of the GEONET stations to be used and send the displacement data to the fault inversion subsystem. The fault inversion subsystem estimates two types of fault model. The One is a single rectangular fault model, which is available for earthquake event which occurs at inland and subduction zone. The other is slip distribution model, which is suitable for earthquake events which occurs at great subduction zone. Estimated Mw from each fault model are sent to the relevant people such as Japanese Agency which is in charge of response to national disaster and railway company so that they can decide what they should do next.
In addition to above features, GSI has improved REGARD aiming more stable operation and high reliability of estimated fault model. GSI introduced Precise Point Positioning (PPP) as method of 1 Hz real-time positioning. It allows us to be free from analysis problem due to base station. RTK method may include positioning error of rover station due to it of base station. It sometimes causes difficulty of calculation of station displacement and noises of estimated fault model. Another improvement is introduction of Markov-Chain Monte-Carlo (MCMC) for estimation of single rectangular fault according to Ohno et al. 2021. It provides more detail information regarding uncertainty of each estimated parameter such as geometry of fault, slip, etc. and more reliable estimated result avoiding local minimum problem. These improvements make REGARD more stable and more reliable.
In this presentation, we will explain overview of former REGARD and detail of improvements of REGARD in these days. Then we show long-term evaluation of REGARD such as artificial noises of real-time positioning, over fitting of fault inversion.