search

yetinam / pyocto

PyOcto: A high-throughput seismic phase associator
MIT License
42 stars 2 forks source link

The associated results exhibit a striped pattern in the depth #4

Closed onebota closed 10 months ago

onebota commented 10 months ago

Hello, when using the pyocto program, I always observe a striped distribution in the depth of the associated results. I have tried changing the velocity model and increasing location_split_depth, but there hasn't been any noticeable improvement. What could be the possible issue causing this? 截图 2024-01-05 15-22-49

yetinam commented 10 months ago

Hi @onebota , thanks for your interest in PyOcto! I guess you've also tried increasing the location_split_return parameter? I'd expect that increasing both location_split_depth and location_split_return together should reduce stripes (at the cost of runtime). Otherwise, I think you've explored basically all options inside PyOcto. I'd recommend using a dedicated location tool like NonLinLoc to get better locations. As an association tool, PyOcto runs a very high number of location procedures. For this reason, the location procedure needs to be tuned to low run times at the cost of some inaccuracies from local minima. That's why locations won't be as good as those from specialized tools. Especially for depth, there will always be local minima because the differential travel times at the surface are small. From my experience so far, this does not have larger implications for association, only for the locations, but I'd be happy to see more results on this.