Minutes from 27 August 2008

2008/08/27 CSEP meeting

Attendees: David J., Yan K., Jeremy Z., Phil M., Danijel S., Masha L.

1. Dave and Yan reported that short-term (1-day) and longterm (1-year, 5-year) forecasts models for California and Nevada are ready to be installed within SCEC Testing Center:
   • Short-term forecast is dependent on long-term forecast
   • Uses ANSS catalog
   • 0.1 degree grid points:
     
     1. Danijel asked to shift grid corners by 0.05 degrees to get the same cells as in RELM forecasts
   • Jeremy raised a concern that result forecast does not have magnitude binning
     

     1. Forecasts are given for M >= 4.65

     2. Yan will provide constant table for magnitude binning:
        • CSEP to apply the table to each cell rate to get bin rates
        • Yan will validate the bin rates for CSEP generated forecast with magnitude binning
   • Phil scheduled installation of the model within SCEC Tesing Center for Version 9.1 release (January 2009)
2. Jeremy posted Java code for:
   • Alarm-based evaluation tests: ROC and Molchan/ASS
     
     1. To be applied to all installed 5-year and 1-year models
   • Forecasts for:
     
     1. 1-yr rate forecast and an alarm-based forecast for the California natural lab
     2. Alarm-based forecasts for the NW and SW Pacific natural labs
   • Jeremy will provide reference data for acceptance tests
3. Double Branching one-year model from Anna Maria for Western Pacific testing regions:
   • SW and NW Pacific 1-year forecasts are fully automated within CSEP:
     
     1. Acceptance tests are part on nightly CSEP build
     2. Crontab nigtly jobs evaluate forecasts

     3. Masha added new one-year-aug2008 forecast group with later entry date into the center:

        • TripleS, KaganJacksonSmoothSeismisity, and DoubleBranching models

        • Invokes N, L and R tests for the models
4. Fixed Trac ticket # #89: Automate TODAY to UTC adjustment in Yan's forecasts:
   • CSEP calculates current offset in hours and passes the value to the model through input parameter file
5. Yan suggested to have more detailed forecasts maps:
   • Existing Trac ticket #99: will fix for Version 8.10 release
6. Danijel reported on CSEP Testing Center in Japan:
   • TripleS, ALM (Wiemer/Schorlemmer), RI one-year models
     
     1. Uses JM catalog
     2. 0.1 degree binning
     3. No declustering
     4. M4-9
   • Danijel will send out specifications
   • Dave and Yan will consider submitting their model to the center
   • Phil expressed concern about supporting different CSEP software branches
7. Phil pointed out the CSEP report for Keck is due September 15, 2008
8. SCEC 2008 Annual Meeting:
   • Danijel will have a few slides on CSEP in RFP discussion
   • CSEP results poster from Jeremy
   • CSEP roadmap poster from Masha
9. Code freeze for CSEP Version 8.10 is scheduled for September 15, 2008