Relation between earthquakes, weather, and soil tilt

Two years of local earthquake, temperature, and rainfall data taken near a tiltmeter site were used in a study of the numerical relation between these phenomena and the recorded tilt response. A least-squares shaping and predictive error filter approach was used. The relations were ranked in part according to the root mean square (r.m.s.) error of fit across the entire sample space. The tilt data with an annual range of tilt of approximately 10 microradians were fitted to the combined weather data of temperature and rainfall with a 0.75-microradian r.m.s. error. The best fit of earthquakes to these same tilt data is the subclass of events with magnitude (M) > 2.5 within 30 kilometers of the tilt site. The filter that mapped earthquakes to tilt yielded a 1.03-microradian r.m.s. error. The most unusual tilt anomaly over the entire 2-year period has the best fit of rainfall to the data for any single month of the entire data set. This unusual anomaly was the basis of an erroneously predicted earthquake (M ??? 5). These data indicate that if there are premonitory earthquake signals, they are buried in local meteorlogical noise. Separating an earthquake anomaly from the response to surface phenomena becomes more difficult as the earthquake anomaly lead time approaches the rise time of the soil to weather and seasonal variations.