The use of an outlier detecting method in time series of continuous daily measurements of underground water level and temperature in earthquake prediction investigation

Continuous daily measurements of underground water level and temperature ha ve been monitored in four selected shallow wells in a seismically active ar ea close to Thessaloniki Greece. The aim was to study the changes in the un derground water level and temperature with respect to earthquakes occurring in the same area during 1 January 1988 to 31 December 1989. A graphical me thod has been applied to analyze the time series in order to detect outlier s. Our results reveal that a strong correlation exists the earthquakes occu rred in the area, and the water level time series. The correlation was foun d to be weak in the case of the water temperature time series. Most of the outliers found correspond to a time inter-Val varying between 1 and 7 days before or after the earthquakes. The magnitude of the underground water lev el change varies from 3 to 10 cm, and that of temperature from 0.3 to 0.8 d egreesC. The strongest outliers correlate to earthquakes of magnitude great er than 3.0. The software developed to detect outliers, use immediate prece ding observations in the time series of continuous daily measurements, at t he time of their occurrence. This is very important in studying precursory phenomena of impending earthquakes. (C) 2001 Elsevier Science B.V. All righ ts reserved.