1.  Tell DART what to extract from the truth run

  a. Run create_obs_sequence: defines an observation network
  b. Run create_fixed_network: spreads it in tim

2.  Create my synthetic obs

  a. Select a random seed in the wrf1d namelist

  b. Copy input.nml.perfect_model_obs_default to input.nml

  c. Run perfect_model_obs: create obs_seq.out

3.  Create an "ensemble mean" about which we will perturb

  a. Run gen_init: creates file member1_ics

  b. Copy member1_ics to filter_ics

4.  Assimilation expt

  a. Change the random seed in the wrf1d namelist (first member != truth)

  b. Copy input.nml.filter to input.nml
    start_seconds = 43200
    cutoff = 1000.0
    start_from_restart = .false.

  c. Run filter

NOTES:

1.  All obs are located by model vertical coordinate right now (type=1).  Surface obs are at coordinate 0.
  
2.  To add another variable to the state vector:

  a.  Select a type name and index
  b.  add one to the total_num* 
  c.  Make sure the locations are defined (init)
  d.  Put in roll and unroll
  e.  To output, add in write_atts and write_vars
