Hello, I am looking to perform high-frequency DA cycling (e.g., 30-min cycling) within the regional configuration of UFS/FV3-LAM. I have a few questions about whether this type of cycling is feasible in the current code. Ignoring the DA aspect, I should note that I am currently using the RRFS_baseline branch of ufs-srweather-app which allows for subhourly output by setting nhours_fcst and restart_interval to a real in the model_configure file (in my case, 0.5).
My questions primarily relate to the LBC updating for this subhourly restarting. My understanding from Black et al. (2021) is that for FV3-LAM, LBCs are updated at each model timestep after being linearly interpolated onto that timestep using the available LBC files. However, it seems to me that the LBCs read in for interpolation are based on relative forecast time. So for example if I start a forecast at 00:30 UTC with bc_update_interval = 1, then the model will expect LBC files at 00:30 UTC and 01:30 UTC. This presents a problem since GFS forecasts are only available on the hour, and chgres_cube does not allow for time interpolation as far as I can tell. Second, I see that bc_update_interval is defined as an integer in fv_regional_bc.f90. This means I can't get around the subhourly LBC problem by setting bc_update_interval = 0.5 to try and read the 01:00 UTC LBC file instead of 01:30 UTC.
Just wondering if anyone had come up with a solution to this problem, or if perhaps I might be misunderstanding something. Perhaps I could do my own offline linearly interpolation to get LBC files available at off-hour times, but maybe someone has come up with something better. Any feedback would be greatly appreciated.