Giters

nansencenter / DAPPER

Data Assimilation with Python: a Package for Experimental Research

Home Page:https://nansencenter.github.io/DAPPER

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Fix and simplify Chronology

patnr opened this issue · comments

commented

Contrary to the original ambition, the Chronology is among the ugliest elements of DAPPER.

  • The ticker should be oriented around obs-analysis-cycles. The intermediate forecast times should be provided by a sub-loop/ticker.
  • HMM.Dyn and HMM.Obs (operators) should take k (or ko) as parameter, not t.
  • Should change the meaning of K (and Ko) from "last index" (which we get with -1 anyway) to "length of array". Fixes the current heavy use of K+1 and Ko+1. This issue is a hangover from when I started I was used to Matlab/Fortran (1-based indexing).
  • Should have observation time 0 ? This would be nice coz then len(tto) == len(tt) when dko = 1. But the 0th cycle would then have two obs-times instead of one...
  • BurnIn complicates everything. Simplify? Provide automatic averaging for custom-length time series?
  • Last I checked I could change the dko of a Chronology without it raising an exception nor having the desired effect. BUG!