CISM Evaluation Metrics: Difference between revisions

From SCECpedia
Jump to navigationJump to search
Kmilner (talk | contribs)
Created page with "List of metrics/regression tests for all catalogs as a way to compare them, ensure that RSQSim is behaving the same on different machines, and to compare/validate results from..."
 
Kmilner (talk | contribs)
No edit summary
Line 3: Line 3:
* Magnitude-Frequency distribution
* Magnitude-Frequency distribution
** Whole-system
** Whole-system
*** [http://opensha.usc.edu/ftp/kmilner/ucerf3/2013_05_10-ucerf3p3-production-10runs/fss_mfd_plots/2013_05_10-ucerf3p3-production-10runs_COMPOUND_SOL_MFD_RELM_TESTING_Region.pdf UCERF3 example]
*** [http://opensha.usc.edu/ftp/kmilner/ucerf3/2013_05_10-ucerf3p3-production-10runs/fss_mfd_plots/2013_05_10-ucerf3p3-production-10runs_COMPOUND_SOL_MFD_RELM_TESTING_Region.pdf UCERF3 example]: Gray is off fault and on fault sub-seismogenic, Blue is on fault supra-seismogenic, Red is total
** Region- and fault-specific (for more complex models)
** Region- and fault-specific (for more complex models)
• Inter-event time and distance distributions
*** [http://opensha.usc.edu/ftp/kmilner/ucerf3/2013_05_10-ucerf3p3-production-10runs/parent_sect_mfds/participation_cumulative/ UCERF3 fault participation MFD examples]
• Scaling relations
* Inter-event time and distance distributions
o Magnitude-area
* Scaling relations
o Slip-area
** Magnitude-area
o Stress drop - magnitude  
** Slip-area
• Rupture speed
** Stress drop - magnitude  
• Recurrence time distributions for event sets of interest
* Rupture speed
* Recurrence time distributions for event sets of interest


 Most of these are already in plotEqs(), one of the RSQSim post-processing functions that Keith includes with RSQSim, but we probably want to do regional tests that focus on particular sites or faults as well.
 Most of these are already in plotEqs(), one of the RSQSim post-processing functions that Keith includes with RSQSim, but we probably want to do regional tests that focus on particular sites or faults as well.

Revision as of 16:56, 13 April 2016

List of metrics/regression tests for all catalogs as a way to compare them, ensure that RSQSim is behaving the same on different machines, and to compare/validate results from updated versions of RSQSim:

  • Magnitude-Frequency distribution
  • Inter-event time and distance distributions
  • Scaling relations
    • Magnitude-area
    • Slip-area
    • Stress drop - magnitude
  • Rupture speed
  • Recurrence time distributions for event sets of interest

 Most of these are already in plotEqs(), one of the RSQSim post-processing functions that Keith includes with RSQSim, but we probably want to do regional tests that focus on particular sites or faults as well.  We should create a new function or package of them, similar to plotEqs() that does all the tests and reports/plots the results.


Additional tests and ways of comparing RSQSim catalogs, to both simulated and real data:

 We also want some measure of the divergence of these catalogs when small changes are made to the parameters, as well as the number of cores the simulations are run on. o Lyapunov exponent  Set up a reference rupture set or catalog.  Compare regional RSQSim MFD’s to UCERF3’s on-fault MFD’s.