Moving towards the operational seismogeodesy component of earthquake and tsunami early warning

A robust technique for rapid estimation of extended earthquake source properties from seismogeodetic (combined seismic acceleration and GPS displacement) time series has been demonstrated for recent large events. By simulating the recording and analysis of GPS and seismic data in real time using Precise Point Positioning with Ambiguity Resolution (PPP-AR) we have determined the latency in creating a fast-Centroid Moment Tensor (fast-CMT) solution to be less than 3 minutes following the initiation of rupture for the 2011 M9.0 Tohoku-Oki tsunamigenic earthquake. This would have made an estimate of the ~290 km source length available in time to update an early warning of the impending tsunami with a better estimate of the wave height. Combining this information with wave gauge data, which could also be implemented in real time, provides improved estimates of tsunami run-up heights at coastal sites (see Melgar et al., 'Near-field tsunami models with rapid earthquake source inversions', this meeting). The fast-CMT method produces continuous updates of a static source inversion using many possible line source fault geometries to determine a rough estimate of the fault slip and area that does not rely on a priori assumptions about fault zone geometry. The key to this new capability is robust, fast, and accurate analysis of real-time 1Hz GPS data that incorporates the following advances: 1) a real time GPS service that produces satellite clocks and fractional cycle biases available for positioning individual sites within the San Andreas fault system and Cascadia subduction zone. This means cpu requirements increase linearly with the number of sites in PPP mode, increasing the feasibility of analyzing data from large dense networks in real time; 2) the tightly-coupled Kalman filter combination of GPS observations with local accelerometer data. This provides ambiguity resolution comparable in quality to a network solution and real-time precision of < 15mm horizontal standard deviation; 3) simultaneous solution for ground motion biases to mitigate errors due to accelerometer tilt; 4) real time integration of accelerometer data to velocity and displacement without baseline corrections, providing the fundamental input for rapid finite fault source inversion; 5) low frequency estimates of P-wave arrival displacement to support single station earth quake early warning. The operational real-time GPS analysis was implemented in time to provide waveforms from the August 2012 Brawley, CA, seismic swarm. Now the full real-time seismogeodetic analysis is operational for GPS sites we have upgraded with low-cost MEMS accelerometers, meteorological sensors and an in-house geodetic modules for efficient real-time data transmission. The analysis system does not yet incorporate an alert system but is currently available to serve as a complement to seismic-based early warning systems to increase redundancy and robustness. It is anticipated to be especially useful for large earthquakes (> M7) where rapid determination of the fault parameters is critical for early assessment of the extent of damage in affected areas, or for rapid tsunami modeling.