Segmentation of Multi-Temporal Earthquake Density Imagery for the Detection of Geophysically Related Geothermal Activity

The objective of this research is to present a methodology in order to characterize the spatial and temporal evolution of seismicity in Greece by studying the occurrence of earthquakes as it is recorded in international catalogues. Such effort will be useful in defining regions that would be further studied on the basis of the thermal response in specific time intervals. The U.S. Advanced National Seismic System composite earthquake catalogue was used. Only the locations of the epicentres are considered for the calculation of the density. Density is calculated by assuming the annual density of point features around each grid cell by the use of circular neighbourhood. The grid size is 5 km while the circle radius is 100 km. The grid size selection was affected by the positional accuracy of the epicentres (<5 km) while the search radius was defined according to the size of the thermal anomalies (10,000 square km). So a series of seismic density images was calculated. Finally unsupervised classification searched for natural groupings or clusters of the seismic density time properties of pixels, and assigned each pixel to a time class based on initial clustering parameters defined. After the classification was completed, each of the 12 initial density time clusters was assigned to a thematic information class on the basis of the interpretation of the corresponding cluster centroid. The 4 thematic information classes allowed the definition of regions with distinct seismic density behaviour within a 35 years period of study.