Can We Predict Earthquakes?Earthquake PredictionMethods Employed In Earthquake PredictionReal Data or Simulated?Popular media statementsStatistical ModelsPoisson Model Weibull ModelParkfield and WrightwoodThe ExperimentSo how regular are the recurrence times of these earthquakes?Probability PlotsCan we rule out the possibility that even EQs at Parkfield are random in time?WrightwoodThe Recurrence Times of the EQs at WrightwoodSlide 15Simulation for the Wrightwood areaSummaryPotential Future WorkAcknowledgments1 Can We Predict Earthquakes?Can We Predict Earthquakes? Andrea NemethAdvisor: Dr. Mark Schilling2Earthquake PredictionEarthquake Predictionlocationtimemagnitudeprobability of occurrencereliableaccurate The collapse of part of Jefferson Junior High School in Long Beach in 1933.(Photo: Portland Cement Association)3Methods Employed In Methods Employed In Earthquake PredictionEarthquake Predictionstatistical probabilityphysical measurements geochemical observationsobservations of animal behavior Seismicity of California (USGS)4Real Data or Simulated?Real Data or Simulated?Significant CA Earthquakes1 (1800-2003)(Magnitude M>=6)66.26.46.66.877.27.47.67.880 10 20 30 40 50 60 70 80Significant CA Earthquakes 2 (1800-2003)(Magnitude M>=6)66.26.46.66.877.27.47.67.880 10 20 30 40 50 60 70 805Popular media statementsPopular media statements“the Big One is overdue” “the longer it waits, the bigger it will be”(USGS)6Statistical ModelsStatistical Modelstime-independent Poisson (exponential)modeltime-dependent Gaussian gamma log-normal Weibull distributionsBrownian Passage Time7 Poisson ModelPoisson Model Weibull Model Weibull ModelMagnitudes of EQs and the time intervals between EQs are each assumed to be independently distributed. memorylessThe probability of rupture is a function of the accumulated strain.F t et( ) 1F t etn( ) 18Parkfield and WrightwoodParkfield and Wrightwood Parkfield area medium-sized EQs occur here fairly regularly Wrightwood arealong term data is availableLA(USGS)91857, 1881, 1901, 1922, 1934, 1966USGS prediction:an earthquake of ~M6 would occur in Parkfield between 1983 and 1993The The ExperimentExperiment10So how regular are the recurrence So how regular are the recurrence times of these earthquakes?times of these earthquakes?Mean: 24.5 years Standard deviation: 9.25 years.The Recurrence Times of the EQs in the Parkfield Experiment05101520253035400 1 2 3 4 5 6 7Time (years)The intervals between these EQs: 24, 20, 21, 12, 32, 3811Probability PlotsProbability Plots0 50 100 150 200 2501030506070809095979899DataPercentExponential Probability Plot for T in the PEML EstimatesMean: 24.510 100 1 2 3 51020304050607080909599DataPercentWeibull Probability Plot for T in the PEML EstimatesShape:Scale:3.2075427.421012Can we rule out the possibility that even Can we rule out the possibility that even EQs at Parkfield are random in time?EQs at Parkfield are random in time?Result: 8.8% of all simulated interval sequences had standard deviation less than 9.25.Conclusion: This sequence is somewhat regular, but not extremely unusual. T R  11l n ( )24 7 3 10 25 3120 20 25 29 7 5321 4 5 27 20 512 13 9 7 47 1832 17 16 32 2 2238 11 41 27 39 129.259.255.795.7914.5914.5910.6310.6317.7517.7516.9316.9313WrightwoodWrightwood534, 634, 697, 722, 781, 850, 1016, 1116, 1263, 1360, 1470, 1536, 1610, 1690, 1812, 185714The Recurrence Times of the The Recurrence Times of the EQs at WrightwoodEQs at WrightwoodThe time intervals between successive EQs:100, 63, 25, 59, 69, 166, 100, 147, 97, 110, 66, 74, 80, 122, and 45 years.mean: 88.2 years standard deviation:37.8 years.0204060801001201401601800 2 4 6 8 10 12 14 16EQ Interval Index (1-15)Time (years)15Probability PlotsProbability Plots0 100 200 300 400 500 6001030506070809095979899DataPercentbetween EQs at WrightwoodExponential Probability Plot for the Time IntervalsML EstimatesMean: 88.200010 100 1 2 3 51020304050607080909599DataPercentIntervals between EQs at WrightwoodWeibull Probability Plot for the TimeML EstimatesShape:Scale:2.5975499.428816Simulation for the Wrightwood Simulation for the Wrightwood areaareaResult: Only 1.5% of all simulated interval sequences had standard deviation less than 37.8 years.Conclusion: This sequence of 16 EQs at Wrightwood is more regular than the Parkfield sequence.17SummarySummarySeveral factors make EQ prediction difficult:the cycle of EQs is long the fundamental physics of EQ faulting is not yet understoodno clearly recognizable precursor has been observedEQ history is short for most faults18Potential Future WorkPotential Future WorkFurther investigation of the Wrightwood dataAnalysis of other data sets from the San Andreas FaultStudy of other statistical models with our data19AcknowledgmentsAcknowledgmentsThis project was sponsored by the NASA/JPL PAIR program.I thank Dr. Carol Shubin for her continuous support, interest and encouragement.I’m very grateful to Dr. Mark Schilling, my advisor, for his comments on the data analysis and preparation, for his valuable insights