Difference between revisions of "Staff Priorities - Masha"

From SCECpedia
Jump to navigationJump to search
Line 2: Line 2:
  
 
=== Yesterday ===
 
=== Yesterday ===
* Posted CSEP meeting minutes, sent next meeting announcement
 
 
* Trac ticket #372: Support Bayesian ensemble models for one-day forecasts:
 
* Trac ticket #372: Support Bayesian ensemble models for one-day forecasts:
** Committed code changes to SVN
+
** Debugging of NaN values in BMA forecasts since 2014/3/15 on certification server
** Set up forecast group on csep-cert to create Bayesian ensemble forecasts with BMA and SequencialBMA skill weighting
+
** Sent debugging summary to Max
** Generated 2012/10/1 - 2016/2/15 BMA and SequencialBMA daily forecasts
 
  
 
=== Today ===
 
=== Today ===
 
* Trac ticket #374: Remove temporary input catalog generated by CA ETASV12_10 model
 
* Trac ticket #374: Remove temporary input catalog generated by CA ETASV12_10 model
 
* Trac ticket #372: Support Bayesian ensemble models for one-day forecasts:
 
* Trac ticket #372: Support Bayesian ensemble models for one-day forecasts:
** N test evaluation failed due to NaN in latest BMA forecast, debugging the problem, emailed to Max with status update
+
** Meeting with Max to discuss the problem and possible solution
 +
** Implement possible solution
  
 
'''Blocked''': No
 
'''Blocked''': No

Revision as of 22:21, 21 March 2016

Current Activities

Yesterday

  • Trac ticket #372: Support Bayesian ensemble models for one-day forecasts:
    • Debugging of NaN values in BMA forecasts since 2014/3/15 on certification server
    • Sent debugging summary to Max

Today

  • Trac ticket #374: Remove temporary input catalog generated by CA ETASV12_10 model
  • Trac ticket #372: Support Bayesian ensemble models for one-day forecasts:
    • Meeting with Max to discuss the problem and possible solution
    • Implement possible solution

Blocked: No

Followups: No

TODO

  • csep-op failure: evaluation was issued on the very first day of new one-year Global forecasts - should not be the case
  • Trac ticket #353: Optimize RELM N-test by avoiding event to bin look up when masking bit is set for the whole forecast

Responsiblities