How to write out multiple forcing files?

[Martin20494]

Hi team,

I am trying to write out multiple forcing files to avoid one heavy forcing files. I can read multiple original forcing files but still cannot write out multiple forcing files. I tried to use the *, but did not succeed. Could you please help me with how to perform this task? I am using hydromt_wflow version 0.8.0.

Martin Nguyen

• Read multiple original forcing files in data_catalog.yml

era5_hourly:
  crs: 4326
  data_type: RasterDataset
  driver: netcdf
  meta:
    category: meteo
    history: Extracted from Copernicus Climate Data Store
    paper_doi: 10.1002/qj.3803
    paper_ref: Hersbach et al. (2019)
    source_license: https://cds.climate.copernicus.eu/cdsapp/#!/terms/licence-to-use-copernicus-products
    source_url: https://doi.org/10.24381/cds.bd0915c6
  path: era5_hourly_*.nc
  unit_add:
    temp: -273.15
  unit_mult:
    # kin: 0.000277778
    # kout: 0.000277778
    precip: 3600
    press_msl: 0.01

• Write out multiple forcing files. I tried to use *, but it just wrote out one big file era5_hourly_new_.nc.

setup_config:      
  starttime: 2000-01-01T00:00:00
  endtime: 2000-03-31T23:00:00
  timestepsecs: 3600
  input.path_forcing: era5_hourly_new_*.nc

[hboisgon]

It's in the docs: https://deltares.github.io/hydromt_wflow/stable/_examples/update_model_forcing.html

[Martin20494]

Hi @hboisgon,

Thank you very much for your instructions. I can use freq_out to write multiple forcing files at some areas, but in other areas I still encountered an ArrayMemoryError (see attached hydromt.log). This error happened when I included precip_clim_fn. I have tried the following:

• different freq_out settings (H, D, M, and Y)
• different chunksize values (from 2 to 48)
• reducing the simulation period from a 3-year range (from 1997-01-01T00:00:00 to
  1999-03-31T23:00:00) to single-year runs (year 1997, 1998, and 1999 separately), which does work with freq_out: M

Is there another way to keep the full 3-year period and write the forcing (hourly, monthly, or yearly) while still using precip_clim_fn?

Thanks again for your help.

Martin Nguyen

[shartgring]

Dear Martin,

This is related to something else. precip_clim_fn should contain a climatology grid that can be used to correct the precipitation generated in setup_precip. This is typically a grid with monthly timesteps. However, I would suggest that such corrections can better be done outside of HydroMT Wflow as this feature is not often used anymore so not rigurous

Note that your log file states:

2025-11-30 23:25:46,649 - build - main - ERROR - Unable to allocate 35.5 GiB for an array with shape (19680, 518, 935) and data type float32

This often happens when arrays are multiplied with each other by xarray. It is very well possible that the multiplication of the precipitation dataset and the climatology dataset is resulting in 19680 coordinates being included in the time axis (as I assume that x and y dimensions are 518 and 935)

It is better to leave precip_clim_fn empty and only work with precip_fn and apply correction factors to the precipitation before reading it with HydroMT. That should work

[Martin20494]

Thanks @shartgring very much for your really helpful explanation and suggestion!