Running mHM in large areas

[diegodinamarca]

Hello mHM team,

we are planning to run mHM in all the continental territory of Chile, which is a pretty large area. The precipitation alone can be up to 40gb at 3km resolution, so the processing of the model is highly demanding. We are planning to simulate a long period from 1960-2021

What insights could you share about the best ways to run mHM at large areas?
Should we split the area in smaller chunks?
Would it be best to run it without routing and then do a postprocessing with routing?

Thank you for your help
Regards
Diego

[rakovec]

dear @diegodinamarca
for large jobs, I surely recommend to switch for meteo chunk reading, using time_step_model_inputs(1), e.g. as yearly or monthly intervals, see this in mhm.nml

  !> switch to control read input frequency of the gridded meteo input,
  !> i.e. precipitation, potential evapotransiration, and temperature
  !>      >0: after each <timeStep_model_inputs> days
  !>       0: only at beginning of the run
  !>      -1: daily
  !>      -2: monthly
  !>      -3: yearly
  !> if timestep_model_inputs is non-zero, than it has to be non-zero
  !> for all domains
  time_step_model_inputs(1) = 0

If that does not help enough, yes, you can further split Chile into several sub-domains, which follow hydrologic boundaries, if you will use mRM.

[diegodinamarca]

ty @rakovec we finally had to split the domain into sub domains because ultimately we want to run it in hourly timesteps.

ty for your help