Development easyQuake

[JLSanchezRoldan]

I added some modifications to easyQuake.py to include new functions for computing local and mblg magnitudes. Also, I created a function that merges all the picks into a single file to study the evolution of the detection.

There are some minor modifications in mseed_predictor.py for allowing to change between the "original" and the "conservative" versions of the EQTransformer model. Both .h5 files were taken from the EQTransformer GitHub repository, and are supposed to be the latest version.

Finally, the readme.md file includes some indications for installing this software on workstations with different versions of CUDA.

[jakewalter]

Why did you change the for both models here? Is that based on what worked best for you or based on original EQT documentation for original model?

[JLSanchezRoldan]

Why did you change the for both models here? Is that based on what worked best for you or based on original EQT documentation for original model?

It's based on comments written by Mousavi on EQT repository:

Original model (the one in the paper) has been optimized to minimize the false negative rate (maximize the number of detected events). This is a good choice when high false positive rates is not an issue in your application. For instance, in event location workfolow most of the false positives will automatically be removed during the association, location, and - especially - relocation procedure. You can use the original model with higher threshold values (~ 0.3 for P and S and 0.5 for detection).

Conservative Model, on the other hand, has been optimized to minimize the false positive rate (maximum number of valid detections). This model is useful when your application can not tolerate a high false positive rate. For example, in travel time tomography or using the detected events by EqT as the templates for template matching. You should use this model with a much lower threshold levels (0.03 for P and S).

For the conservative model, I set slightly different values for the thresholds (0.05 instead of 0.03 for P, S, and 0.2 instead of 0.3 for detection), but they work well with my data. I also tested what results could I get by changing the overlapping to 0.6 (Mousavi recommends 0.9 here: smousavi05/EQTransformer#128, although it seemed too much for me), but the original model showed a poor performance. There were too many false positives when I ran the association with 3 stations*.

*I cared only for the earthquake detection and epicentral location, so 3 stations were OK to test every model. Later, I run a new association with 4 stations with fewer detections for every model.