Sudden Jump in PREC_ACC_NC After Restart in Idealized TC Simulation

[khlau03]

Hello,

I'm using WRF version 4.2.2, configured for automatic vortex following nests.

I’m running an idealized tropical cyclone simulation and using prec_acc_dt = 1 to output 1-minute accumulated precipitation (PREC_ACC_NC) at each history output time. I then multiply this value by 60 to estimate the instantaneous rain rate in mm/hr.

However, I’ve observed a significant spike in PREC_ACC_NC at the first history time step following a restart. For example, the figure below compares PREC_ACC_NC (units: mm) at a particular output time step in the innermost domain between the original and restart runs. The restart was initiated at 2007-09-06 00:00:00 using wrfrst files from the original run. Both simulations use a history output interval of 60 minutes.

This spike only occurs at the first time step after the restart, after which the PREC_ACC_NC remains significantly greater than that in the original run for the remainder of the run.

I’ve attached my namelist.input file of the restart run for reference. I’d greatly appreciate any insights into why this jump is occurring.

Many thanks,
KH

 

[Ming Chen]

Hi KH,

In your namelist.input, I notice that you set

override_restart_timers = .true.,
write_hist_at_0h_rst = .true.,

I suppose you have changed wrfrst output intervals for the restart run, is this correct? If so, can you keep it same as your original run and see whether you still have the spike? I ask this because we are aware of some issues related to changed wrfrst intervals when using the restart capability. This is why the option "override_restart_timers" is added. Back then this option is treated as a temporary fix, but we never really further explored those issues.

Did you look at prec_acc_nc in both your wrfrst files and wrfout files at the restarting time and over all your domains? Do they look consistent?

Also, if you set write_hist_at_0h_rst = .false., how does the model behave?

 

[khlau03]

Hello Ming,

Many thanks for the adivces. I can confirm that at the restarting time, the prec_acc_nc in the restart run is the same as that in the original run.

I have tried:

override_restart_timers = .false.,
write_hist_at_0h_rst = .false.,

However, the problem still persist (i.e., prec_acc_nc is still much higher than that in the original run).

- KH

 

[khlau03]

I found that this problem only occurs when I run the simulation on a beta plane. Gradients in F and XLAT in the wrfinput are added to create a beta plane with 20N at the centre of the domain.

On f-plane (i.e., without modification of the wrfinput), the PREC_ACC_NC in a restart run does not jump.

 

[Ming Chen]

Hi KH,
Thanks for the update. The prec_acc option is originally added for real-data case studies, although we also expect it should work for ideal cases.
Sorry that I don't have an immediate answer to explain why it shows this weird behavior only because you change wrfinput.