Step 1 (took 30 minutes to run)
  python process_ghcnd_by_stn.py stnlist_all_2021Jan29.txt 1940 2020
  
  This works fine, even with the ISD stations in the station list. It quickly
  identifies that the station is unavailable and goes to the next one. No
  error/warning message is output, which I think is fine.
  
  Note that processing time should be somewhat similar regardless of the 
  specified start/end years. While there will be some differences in the
  processing time, the files being downloaded from NCEI contain the entire 
  period of record for each station. So specifying a smaller subset of years
  has no effect on download time. The ftp server load and other similar 
  factors will also be relevant, making the actual processing time variable.
  
  One thing to note... When this script finishes, check the output to be sure 
  that you were able to download ALL relevant files. The script is configured
  to gracefully handle an issue with any particular station being missing
  or unavailable. But for some reason, once it fails it seems to fail every 
  subsequent one.
  

Step 2 (just under 8 hours to run)
  python get_all_ish.py stnlist_isd_2021Jan29.txt 1940 2020
  
  This will work even with the GHCND stations in the list. But it wastes a
  *LOT* (several hours) of time trying to download each of those missing 
  GHCND files. Thus, I strongly recommend editing the stnlist file to only 
  contain the ISD stations.  The listed time of 8 hours was when I ran it
  with only the isd stations in the list. My first attempt had the ghcn
  stations at the top of the list, and I noted after a couple hours that 
  it was only partway through that section of the list. I don't remember
  exactly how far, but it was still in the Canadian ones and hadn't even 
  reached the U.S. yet. So I'm guessing it would have taken 6-12 additional
  hours had I not stopped it and restarted with an edited list.

Step 3 (just over 32 hours to run)
  run GLSHFS to add all of the data
     1. initial setup
     2. Edit the GLSHFS config file:
        AddStationData = YES
        BuildSubbasinMet = YES
        UpdateHistorical = NO
     3. run the procedure
     
Step 4 (took 1h 12m)
  Run GLSHFS (without adding new data) for the modeling phase
     1. Edit the GLSHFS config file:
        AddStationData = NO
        BuildSubbasinMet = NO
        UpdateHistorical = YES
    
Note that steps 3 & 4 can be combined by simply setting all 3 of those 
switches to YES. I recommend doing the big historical set in separate steps
only to avoid wasted time if there is an error in the station data update.
  
Step 5  
  Build station met for the "operational" update, to simulate USACE process
     1. Delete all of the met_*.csv files in the STN directory
     2.   python process_ghcnd_by_stn.py stnlist_all_2021Jan29.txt 2020 2021
          (took 18 minutes)
     3.   python get_all_ish.py stnlist_isd_2021Jan29.txt 2020 2021
          (took under an hour, but not exactly sure)
     4. Edit the GLSHFS config file:
        AddStationData = YES
        BuildSubbasinMet = YES
        UpdateHistorical = YES
     5. run GLSHFS (took 2h 50m)
     
Step 5 (take 2)
  Hmm... That step 5 above didn't make forecasts. USACE wants to know how long
  it will take to do EVERYTHING. So let's do step 5 again, but this time we
  will actually do the forecast part, too.
     1. Delete all of the met_*.csv files in the STN directory
     2.   python process_ghcnd_by_stn.py stnlist_all_2021Jan29.txt 2020 2021
          (took 26 minutes)
     3.   python get_all_ish.py stnlist_isd_2021Jan29.txt 2020 2021
          (took 14 minutes)
     4. Edit the GLSHFS config file:
        AddStationData = YES
        BuildSubbasinMet = YES
        UpdateHistorical = YES
        RunForecasts = YES
        MakeSummaryFiles = YES
        OutputMetDetails = []
     5. run GLSHFS (took 3h 17m)