Manage runs¶
A run is a search and its run directory. This guide shows how to restart, continue and re-evaluate runs.
Start again¶
qseek search does not overwrite an existing run directory. To start the search again with the same name:
qseek search my-search.json --force # keep the old run as a backup
qseek search my-search.json --force --no-backup # overwrite the old run
With --force, Qseek renames the old run directory with a .bak-<time> suffix.
Continue a search¶
If a search stops, e.g. on a machine restart, continue it where it stopped:
Qseek reads the configuration of the run from its search.json and the progress from progress.json. Detections after the last processed window are removed and searched again.
Recalculate magnitudes and features¶
To change the magnitudes or features of a finished run, edit the magnitudes and features in the run's search.json and recalculate them for all detections:
--nparallel sets how many detections are processed in parallel.
Search again with station corrections¶
After a first run, extract station corrections from its travel time residuals and run a second search with them. The corrections refine the locations and can reveal more events.
Export the detections¶
Export the detections of a run to other formats, e.g. a HypoDD project for double-difference relocation:
| Format | Description |
|---|---|
hypodd |
A HypoDD project folder for double-difference relocation, see relocate with HypoDD. |
simple |
Travel times of the picks in CSV format. |
velest |
A VELEST project folder for 1D velocity model inversion, see exporters. |
--force overwrites an existing export directory. --config reads the settings of the export module from a JSON file. The exporters page describes the formats.