SST update delayed by one history interval in wrfout_d02

[elkro21]

Hi, I am preparing a multi-year WRF simulation and need to update SST. I am not using an external SST dataset but want to update SST from ERA5 data, which is used as model input at the same time (I therefore did not run ungrib.exe twice to create separate ‘SST’ files for metgrid). ERA5 driving data is injected into the simulation every 3 hours, while SST only needs to be updated every 24 hours since SST provided by ERA5 is daily. In my namelist, I have set the following:

&time_control
…
interval_seconds = 10800,
input_from_file = .true.,.true.,
history_interval = 1440, 180,
frames_per_outfile = 124, 8,
adjust_output_times = .true.,
restart = .true.,
restart_interval = 175680,
override_restart_timers = .true.,
write_hist_at_0h_rst = .true.,
auxinput4_inname = "wrflowinp_d<domain>",
auxinput4_interval = 1440, 1440,
io_form_auxinput4 = 2,
…
/

&physics
…
sst_skin = 1,
sst_update = 1,
…
/

I did a 4 month test run with the above namelist options. WRF completed successfully, and the output looks basically reasonable. SST is updated every 24 hours in both domains. However, in D2, where the history interval is 3-hourly (opposed to daily in D1), SST is updated at 03:00:00 instead of 00:00:00. This is not a huge issue, but I wonder why and where in the model simulation/code his happens or if I did something wrong. In the original ERA5 data as well as in the met_em.d02* files and in the wrflowinp_d02 file created by real.exe, SST changes at 00:00:00. The “delay” by one history interval is only present in the wrfout_d02* files and must therefore be produced during the actual WRF simulation. Does anyone have an idea why SST is not updated at 00:00:00 in the nest?

Here is my rsl.error.0000 file as well as the SST variable from my wrflowinp_d02 and wrfout_d02* files.

Thanks a lot!

 

[Ming Chen]

Did you run real.exe using the namelist you attached? I wonder why "Times" in wrflowinp is 3-hr interval, i.e.,
...
"2011-08-03_09:00:00",

"2011-08-03_12:00:00",

"2011-08-03_15:00:00",

"2011-08-03_18:00:00",

"2011-08-03_21:00:00",

"2011-08-04_00:00:00",
...

I suppose they should be at 24-hr intervals.

 

[elkro21]

Hi Ming,

Please excuse the long response time, I was out of the office.

Yes, I have run real.exe with exactly the above namelist options. Although I specified auxinput4_interval = 1440, 1440, the time interval is 3-hourly in both wrflowinp_d01 and wrflowinp_d02. This is indeed strange, and I also wonder why the input is written in 3-hour time steps. Unfortunately, I ran real without debugging, so I cannot see in detail what it did. I could run again with debugging enabled.