Minutes from 21 May 2008

2008/05/21 CSEP meeting

Attendees: Yan, Phil, Danijel, Jeremy, John, Masha

1. Introducing Western Pacific testing region into the CSEP:
   • Jeremy send out rules for Western Pacific testing region to Dave and Yan:
     
     1. Yan expressed a concert about combining SW and NW Pacific regions into one
        • Overlapping region near the equator for SW and NW regions
        • New set of parameters for the models
        • Would not be comparable to experiment that is done at UCLA
   • Amit applied changes he's done to the NW Pacific Fortran code to the SW Pacific Fortran code
   • Masha automated Yan's models for NW and SW Pacific regions within CSEP:
     
     1. Automatic CMT download
     2. Creates input catalog
     3. Invokes one-year and one-day models for each region
     4. For validation purposes only uses zero days of delay to be able to compare against UCLA generated forecasts
   • Masha validated CSEP binaries for Yan's models using Yan's input catalog:
     
     1. Copied Yan's catalog to CSEP
     2. Set up crontab job for NW Pacific at 23:50pm (time when forecast gets generated at UCLA)
     3. Generated forecast matched 100% to the forecast generated at UCLA
   • Masha changed CMT download functionality:
     

     1. Final events for 2008/01 were placed in file 2008 as opposed to standard location (for example, 2007/jan07.ndk) at http://www.ldeo.columbia.edu/~gcmt/projects/CMT/catalog/NEW_MONTHLY

   • Danijel will respond to Warner with testing region rules:
     
     1. Will ask to split testing region in SW and NW Pacific regions
   • Danijel is working on ForecastML master template for Yan's model:
     
     1. Masha will write conversion utility to populate template with Yan's forecast rates
     2. Yan confirmed that rates are in earthquakes per day per square kilometer units:
        • Jeremy will work with Masha on calculating areas for each bin - to be used to calculate bin rate in earthquakes per day units
     3. Can evaluate models with N and L RELM tests once master ForecastML is populated with forecasts rates
   • Danijel will write GMT script for testing regions to generate forecasts maps
   • Masha will work on:
     
     1. Acceptance tests for Yan's models:
        • Models are time-dependent
        • Have to modify Fortran code to read time from the parameter file instead of making a system call to acquire the time
     2. Utility to split Yan's file that contains long-term and short-term forecasts into two files - for forecast evaluation purposes
   • John will work on web-site changes to accomodate new testing regions
     
     1. California, SW Pacific, NW Pacific
2. Software development:

   • There were "Maximum recursion limit of 100 reached. Use set(0,'RecursionLimit',N)" Matlab failures in STEP code:

     1. Masha changed the limit to 800, got failed days to process on csep-devel
     2. Jeremy said that STEP code has a scientific error, might need to withdraw the model from the testing center:
        • Danijel will contact Matt
        • Phil suggested to ask for replacement model
3. Hardware:

   • John has set up csep-cert-new server:

     1. John is working with Masha on getting the CSEP to build