Minutes from 8 September 2009

2009/09/08 CSEP Meeting

Attendees: Phil M., Danijel S., James H., Jeremy Z., John Y., Masha L.

1. csep-op:
   1. V9.10.0 release:
      • Need to upgrade SciPy (V0.7.0) and NumPy (V1.3.0) packages
   2. V9.7.0 release:
      • Re-processing of California's one-day and RELM forecasts groups due to the ANSS catalog fixes
        1. RELM-mainshock-corrected forecast group failed for 2009-08-03
           • Exception is raised when directory of random numbers is renamed
           • Started happenning once we switched to NFS server
           • Masha will re-process the date
        2. Some test dates failed to rsync to the cseptesting.org web server due to the "Connection unexpectedly closed" error
           • Masha manually issued failed rsync command to publish results for failed data
2. CSEP meeting at USC on September 24-25, 2009
3. Software development:
   1. Code freeze for V9.10 release candidate scheduled for some time this week
   2. Trac tickets to be addressed:
      • FIXED: Trac ticket #170: Replace use of xml.dom.minidom by cElementTree module in plotting functions
      • FIXED: Trac ticket #176: Spawned simultaneously Dispatcher's use the same run-time directory
      • Trac ticket #171: Cumulative results still use '<' value instead of '<=' in probability function
        1. No re-processing required, can recompute cumulative based on daily information within it
        2. Will implement the fix as ResultsCumulativeSummary class method to be invoked in standalone mode
        3. Will implement after release candidate will be installed on csep-cert
   3. • FIXED: Trac #173: Check for downloaded raw catalog file not being empty
        1. Danijel suggested to add another sanity check: Trac #178 Add a sanity check for number of observed events in cumulative result
      • Trac #174: Re-process 2009/05/14-16 test dates that use CMT catalog due to the ticket #173
      • FIXED: Trac #175: Change configuration files for SW and NW Pacific forecasts groups to use local copies of observation catalogs
      • Trac #177: XML format of forecast should still be passed through master XML template
      • FIXED: Trac #172 Install Alvaro Gonzalez models for California
        1. Two one-day alarm-based models for California: UNIFORM and NEAREST
        2. UNIFORM model for testing in three-month and five-year forecasts groups
        3. Alvaro validated forecasts maps and tests results as pubished by csep-cert to intensity
4. mini_CSEP distribution
   • Distribute generic package instead:
     • Some scientists responded that they would be interested in series of experiments, might as well provide them with all generic CSEP codes
   • Jeremy distributed requirements to modelers community of RELM experiment and Italian testing regon
     • Some scientists are interested in contributing to the CSEP
     • Jeremy and Danijel will set up priorities for suggested by community requirements
     • Masha is working on acceptance tests that exercise each of the requirements:
       * convert a ForecastML forecast to MATLAB format
       * convert a MATLAB forecast to ForecastML format
       * generate a plot of the spatial distribution of the forecast (i.e.,
       map-view of the forecast)
       * generate a plot of the magnitude distribution of the forecast (i.e.,
       magnitude versus normalized probability--I have attached an example )
       * execute an L-test (including appropriate plot)
       * execute an M-test (including appropriate plot)
       * execute an N-test (including appropriate plot)
       * execute an R-test (including appropriate plot)
       * execute an S-test (including appropriate plot)
       * Optional: declustering
   • We should provide list of external package dependencies and links to these package
   • Jeremy prioritized plot of magnitude vs. normalized probability as low priority for V9.10 release
5. James gave a work overview for his group:
   • One-day to a few years forecasts models using ANSS catalog (M>=1.5-2)
   • James has evaluation tests he would like to contribute to the CSEP
   • James is interested in work on results presentation
6. Keilis-Borok model:
   • Danijel reported that UCLA group will prepare forecast according to the forecast format used by the CSEP
   • Will send forecast by email
   • New forecast class within CSEP
   • Will decide on how to evaluate the forecast
7. Phil talked about EEW:
   • Masha to spend 4 hours per week
   • Design review early next year
   • EEW Testing document describes evaluation tests EEW is using
   • Phil would like to abandon current EEW web page and create EEW testing center link on CSEP web page
8. Danijel mentioned that current results presentation is not enough to make scientific conclusions:
   • Jeremy and Danijel will propose what data and how should be displayed by the results viewer on the web server
9. SCEC Annual Meeting:
   1. Masha submitted CSEP software/system abstract
   2. Jeremy submitted CSEP science abstracts:
      • SCEC Testing Center results
      • RELM results
      • M and S tests results
   3. Danijel asked Masha to give a short talk about testing center implementation on Sunday during "Source Inversion Validation" workshop
10. CSEP proposal:
    • Danijel reported that we're waiting for USGS feedback for acceptance
11. Hardware:
    1. csep-cert
       • CSEPWebTest.py is not part of nightly build - some issues connecting to the web server
         1. John is looking at the problem, but most likely won't make it into V9.10.0
         2. Masha suggested to invoke the tests manually once V9.10 is installed on csep-op