Empirical criteria for the accuracy of earthquake locations on the Croatian territory

This paper presents the empirically based ground truth criteria, or shorter GT criteria, for the estimation of the epicentral location accuracy of the seismic events recorded at network stations within 400 km around the city of Zagreb. The criteria are based only on the network coverage metrics and the GT5 level represents an absolute location error lower than 5 km. They have been developed using a bootstrap resampling method: same earthquakes have been relocated many times but with different, randomly selected seismic stations. We used 330 reference events taken from the pages of ISC (ISC Reference Event Bulletin, 2008) and showed that the location accuracy is most affected by the distance to the farthest station in the seismic network, while not at all influenced by the distance to the nearest. The developed GT criteria for GT595% level of accuracy require 10 or more network stations, all within 125 km from the epicentre, and the secondary azimuthal gap (the largest gap when any given station is removed from the network) less than 200°, or the network quality metric (the deviation between the optimal uniformly distributed network and the actual network) less than 0.41. The obtained results revealed that the global criteria are too restrictive and unsuitable for the studied area since they require more regular networks. With our criteria, it is possible to achieve higher accuracy for the networks with a bigger secondary azimuthal gap or greater network quality metric. In addition, our criteria limitations are shown for the areas with simpler geological structure.