EARTHQUAKE EARLY WARNING (EEW) SYSTEMS PROVIDE SECONDS TO MINUTES OF WARNING TO BOTH PEOPLE AND AUTOMATED SYSTEMS BEFORE STRONG SHAKING OCCURS AT THEIR LOCATION (GIVEN ET AL. 2014). RAPID AND ACCURATE MAGNITUDE ESTIMATION IS CRITICAL TO THE SUCCESS OF EEW SYSTEMS HOWEVER FAST MAGNITUDE DETERMINATION FOR LARGE (MW8+) EARTHQUAKES IS VERY CHALLENGING FOR MODERN SYSTEMS (HOSHIBA&OZAKI 2014). FOR EXAMPLE THE 2004 MW9.2 SUMATRA EARTHQUAKE WAS INITIALLY CLASSIFIED AS A MW8.0 EARTHQUAKE UNTIL FEW DAYS LATER WHEN OBSERVATIONS OF EARTH'S FREE OSCILLATIONS REVISED THE MAGNITUDE TO MW9.2 (PARK ET AL. 2005; STEIN&OKAL 2005). ADDITIONALLY THE 2011 MW9.0 TOHOKU-OKI EARTHQUAKE A CATASTROPHIC TSUMANIGENIC EARTHQUAKE WAS FIRST ESTIMATED TO BE A MW8.1 WHICH RESULTED IN UNDERESTIMATING OF BOTH THE SHAKING AND TSUNAMI INTENSITY. MAGNITUDE UNDERESTIMATION OR SATURATION (GELLER 1976) IS A RESULT OF TWO FACTORS. THE FIRST IS OBSERVATIONAL; IT IS PRODUCED BY LIMITATIONS OF SEISMIC MONITORING EQUIPMENT (E.G. CLIPPING OR BASELINE DRIFTING). THE SECOND REASON IS PHYSICAL. LARGE EARTHQUAKES TAKE A FINITE AMOUNT OF TIME TO RELEASE THEIR MOMENT AND IT IS A MATTER OF SOME DEBATE HOW FAR INTO THIS SOURCE PROCESS EARTHQUAKES OF DIFFERING MAGNITUDES SEPARATE AND HAVE DIFFERENT OBSERVABLE BEHAVIOR (MORI&KANAMORI 1996; RYDELEK ET AL. 2006; MEIER ET AL. 2016 2017; MELGAR&HAYES 2017). IN THE FIRST SECONDS TO MINUTES WITH CURRENTLY EXISTING OPERATIONAL ALGORITHMS BOTH OF THESE LIMITATIONS MAKE AN MW 8.0 EARTHQUAKE AND AN MW 9.0 EARTHQUAKE INDISTINGUISHABLE (HOSHIBA&OZAKI 2014). THIS SECOND ISSUE MERITS FURTHER DISCUSSION. WHETHER LARGE RUPTURES ARE DETERMINISTIC OR NOT LIES AT THE HEART OF THIS PROPOSAL. STRONG DETERMINISM WAS POSTULATED EARLY ON (OLSON&ALLEN 2005) AND IN THIS MODEL THE FIRST FEW SECONDS OF RUPTURE CONTAIN INFORMATION ON THE FINAL FATE OF AN EARTHQUAKE. THE OPPOSING VIEW THAT THERE IS NO DETERMINISM AT ALL (RYDELEK&HORIUCHI 2006; MEIER ET AL. 2016 2017) SUGGESTS THAT THE RUPTURE INITIATION SHOWS NO DIFFERENCES BETWEEN SMALL AND LARGE EARTHQUAKES BEFORE THE RUPTURE IS COMPLETE. HOWEVER THOSE ARGUMENTS WERE BASED ON THE BAND-LIMITED SEISMIC-BASED OBSERVATIONS. ANOTHER RECENT STUDY CONSIDERING A LARGE DATABASE OF FINITE FAULT MODELS AND GPS OBSERVATIONS (MELGAR&HAYES 2017) PROPOSED A COMPROMISE MODEL OF WEAK DETERMINISM. IN THIS VIEW LARGE EARTHQUAKE AND VERY LARGE EARTHQUAKE ARE NOT DISTINGUISHABLE AT THEIR NUCLEATION (A FEW SECONDS) BUT IN GPS WAVEFORMS HAVE OBSERVABLY DIFFERENT BEHAVIOR WITHIN THE FIRST TENS OF SECONDS AND WELL BEFORE THE RUPTURE HAS FINISHED. GOLDBERG ET AL. (2018) RECENTLY SHOWED THAT THIS TIME EVOLUTION OF PGD RELIABLY IMAGES THE UNFOLDING RUPTURE PROCESS BUT STOPPED SHORT OF PROPOSING AN ALGORITHM TO LEVERAGE THESE FINDINGS. IN THIS RESEARCH WE WILL EXPAND THE PGD CONCEPT TO INDUCE NOT JUST ITS FINAL VALUE BUT ITS EVOLUTION AS A FUNCTION OF TIME. WE PROPOSE TO TEST THE MINIMUM RESOLVABLE TIME NEEDED TO RAPIDLY DETERMINE A RELIABLE MW. IN ORDER TO ACCOMPLISH THIS WE WILL APPLY MACHINE LEARNING (ML) TECHNIQUES TO RECOGNIZE AND CLASSIFY THE PGD-MW PATTERNS IN THE EARLY STAGES OF A LARGE EARTHQUAKE. IN THIS PROPOSAL WE WILL USE A MIXTURE OF SYNTHETIC EARTHQUAKE DATA COMBINED WITH A LARGE CATALOG OF OBSERVED EVENTS TO TRAIN AND THEN TEST THE ML ALGORITHM. THE WHOLE PROCESS WILL FIRST FOCUS ON LARGE CASCADIA SUBDUCTION ZONE RUPTURES (M8+) WHERE THERE IS A PRESSING NEED FOR SUCH AN ALGORITHM AND THEN EXPAND TO A WORLDWIDE FOCUS. THE PROJECT'S OBJECTIVE RELATES TO NASA'S GOAL: "HOW CAN EARTH SCIENCE PROVIDE SOCIETAL BENEFIT?" AND THE STRATEGIC OBJECTIVE "CHARACTERIZE THE DYNAMICS OF EARTH IMPROVING THE CAPABILITY TO ASSESS AND RESPOND TO NATURAL HAZARDS AND EXTREME EVENTS". NASA SUPPORTED GPS PROVIDES THE KEY SOLUTION TO EARTHQUAKE HAZARD BY WARNING THE PUBLIC AT THE EARLIEST STAGE AND REDUCING THE DISASTER.