Various location procedures and velocity models are compared for the ML 3.3 Kara Sea event of 16 August 1997. This event has been the subject of considerable discussion because of its geographical position and the difficulty in obtaining a reliable focal-depth estimate. A comprehensive data set was extracted by (re-)reading the records from all available stations. These readings have then been used in a sequence of location experiments to examine the effect of using different velocity models to describe the travel times of the phases, and also to compare the use of a fully nonlinear scheme (shakeNA; Sambridge and Kennett, 2001) and a linearized location algorithm (hyposat; Schweitzer, 2001, 2002). A standard least-squares misfit criterion has been used for direct comparisons between the two methods.
The results confirm both the importance of S-wave information in assessing the depth of regional events, and the need to apply a reliable velocity model to place the strongest constraints on the location of the event. Even with only a limited data set, but an adequate velocity model, it is possible to find the position of the Kara Sea event close to the most probable locations; however, there is then no depth resolution. Reported error ellipses from standard data centers tend to have relatively small error ellipses. With the commonly made assumption that the reading errors and the a posteriori residuals have an unbiased normal distribution, such inversion results may indicate an unreasonably high resolution and accuracy of the solution.
The epicenter estimates for the whole data set using the range of different techniques agree quite well, with some overlap of the estimated confidence regions. The observed seismic source was most likely an earthquake in the middle or lower crust at about 10–30 km depth.