Possible Error Sources

The possible error sources in this study can be summarized as follows:

• measurement errors in backazimuth and travel times
• errors in the backazimuth due to the complex three-dimensional structure around Garm caused by arrivals off-azimuth
• large coda levels that mask small amplitude phases
• one-dimensional velocity model
• possible variations in the velocity model, e.g. velocity variations or variations in the thickness of the crust

An additional study showed, however, that a change in the velocities does not affect the result very much. A decrease in the crustal thickness of 20 km results in a change of $\pm 10$ km in the event depth. This would still be a reasonable result. Furthermore, it is unlikely that the thickness varies as much as 20 km. Therefore, the error caused by possible variations in the crustal model should not be too large.