Merged Real Time GNSS Solutions for the READI System

Thursday, 18 December 2014
Victor M Santillan, Central Washington Univ, Ellensburg, WA, United States and Jianghui Geng, University of California San Diego, La Jolla, CA, United States
Real-time measurements from increasingly dense Global Navigational Satellite Systems (GNSS) networks located throughout the western US offer a substantial, albeit largely untapped, contribution towards the mitigation of seismic and other natural hazards. Analyzed continuously in real-time, currently over 600 instruments blanket the San Andreas and Cascadia fault systems of the North American plate boundary and can provide on-the-fly characterization of transient ground displacements highly complementary to traditional seismic strong-motion monitoring. However, the utility of GNSS systems depends on their resolution, and merged solutions of two or more independent estimation strategies have been shown to offer lower scatter and higher resolution. Towards this end, independent real time GNSS solutions produced by Scripps Inst. of Oceanography and Central Washington University (PANGA) are now being formally combined in pursuit of NASA’s Real-Time Earthquake Analysis for Disaster Mitigation (READI) positioning goals. CWU produces precise point positioning (PPP) solutions while SIO produces ambiguity resolved PPP solutions (PPP-AR). The PPP-AR solutions have a ~5 mm RMS scatter in the horizontal and ~10mm in the vertical, however PPP-AR solutions can take tens of minutes to re-converge in case of data gaps. The PPP solutions produced by CWU use pre-cleaned data in which biases are estimated as non-integer ambiguities prior to formal positioning with GIPSY 6.2 using a real time stream editor developed at CWU. These solutions show ~20mm RMS scatter in the horizontal and ~50mm RMS scatter in the vertical but re-converge within 2 min. or less following cycle-slips or data outages. We have implemented the formal combination of the CWU and SCRIPPS ENU displacements using the independent solutions as input measurements to a simple 3-element state Kalman filter plus white noise. We are now merging solutions from 90 stations, including 30 in Cascadia, 39 in the Bay Area, and 21 from S. California. Six months of merged time series demonstrate that the combined solution is more reliable and can take advantage of the strengths of the individual solutions while mitigating their weaknesses. The merging can be easily extended to three or more independent analysis strategies, which may be considered in the future