CORe used 13 streams from 1950 to 202401 on Gaea. Each stream had varying start dates
due to the loss of the Gaea file system, and there're year to long overlaps over
most of the streams.  The stream boundaries have caused viable transitions in earlier
reanalyses.  However, there are less obvious in CORe because

 (1) used inial conditions from a CORe phase 1 run that had to be rerun because
       of a file system crash
 (2) The crashed CORe phase 1 used initial conditions from CORe phase 0
 (3) many streams had one year of overlap.

Initialization for 1950-01 analysis

Started in 1949-09-30 
using intial conditions 1951-01-01 from CORe version 0: GFS global spectral version
run to 1951-9-30
Used results from 1951-09-30 as initial conditions for 1949-09-30
Run from 1949-09-30 to 1959-05
Used results from 1951-09-30 as initial conditions for 1949-09-30
Rerun from 1949-09-30 to 1951-12

Using this procedure, the 1950-1951 analyses were run 3 times. The idea behind
this procedure is to try to get an initial conditions for the 1950-01 analysis with

  1) reasonable QBO state
  2) The 1950's has many regions with little obsevations, so the soil moisture
     would "model climatological" state rather than a more modern climatological
     state. By running the system multiple times over 1950-1951, we would
     get a state more resembling the model climatology.  Of course, there are
     droughts and flood in the 1950-1951 period so this method has limitations.


Here are the periods during which each stream covered: 
195001 to 195112
195201 to 195905
195906 to 196403
196404 to 197009
197010 to 197604 
197605 to 198204 
198205 to 199207 
199208 to 200009 
200010 to 200812 
200901 to 201603 
  later hardware glitch was detected, reran 201401 to 201412
  minor discontinuity between 201412 and 201501 
201604 to 202401

Starting in February 2024, CORe transitioned to running on WCOSS2. 


LETKF was turned off when there was not enough observations data 
1950010100, 1950010112, 1952010100, 1953010100 and 1999091206.