Minutes from 10 December 2008
2008/12/10 CSEP meeting Attendees: Yan K., Phil M., Jeremy Z., Fabian E., Danijel S., John Y., Masha L.
- Upgrade of New Zealand testing center to CSEP V8.10:
- Danijel provided GMT script for forecasts maps
- Nightly crontab jobs run automated CSEP V8.10, results and maps are published to the nz.cseptesting.org/results
- Wish list from Jeremy:
- Masha re-processed test date of 2008-07-01 for all RELM forecasts groups on csep-cert:
- Jeremy identified an extra event for 2008/04/26 - Masha traced it to the differences in raw ANSS catalog data
- Includes fixes to cutting ANSS catalog to geographical region
- Masha re-processed test date of 2008-07-01 for all RELM forecasts groups on csep-cert:
- Masha is helping Kurt Rucker, UCLA student of Dave Jackson, to run some evaluation tests for his and Yan's models
- Global testing:
- Data source: CMT
- No collection or test area
- Danijel provided master ForecastML template:
- 1.0 x 1.0 degree cells
- M5.95 (or M6.0) and above
- Danijel provided GMT script to generate forecasts maps
- Jeremy provided a fix for already submitted Triple-S one-year rate-based and alarm-based model
- Code and reference data are updated in SVN
- Forecasts with start date of 2008-11-09 were created and evaluated on csep-cert
- Alarm-based: ROC test
- Rate-based: N, L, ROC tests
- John has updated web viewer with new testing region
- Danijel will submit a model for V9.4
- Anna Maria has not responded to email - will probably submit the model for V9.4
- Masha has sent email to Yan summarizing rules for one-day and five-year models within California:
- Yan recommended to change MCUT value from 400 to 395 to be consistent with existing forecasts groups
- Masha will make the change and will update reference data for an acceptance test
- CSEP V9.1 remaining tasks:
- Masha evaluating forecasts differences generated with Yan's catalog vs. CSEP generated input catalog
- Trac ticket #117: Switch from wget to Python's urllib for ANSSDataSource.py data retrieval
- Should decide later which data sources and corresponding acceptance tests should be part of generic CSEP distribution
- Code freeze is on Monday, 2008-12-15
- CSEP V9.1 installation on csep-op is on Monday, 01-05-2009
- CSEP V9.4:
- Catalog archive of CSEP available raw catalog data
- Code distributions for testing centers
- Integration of Japanese codes
- Staging of existing raw catalog data
- Global models from Yan, Danijel, Warner and Anna Maria
- AGU - next week
- Next week CSEP meeting should be a Skype conference to give an update on release candidate