Gridded (coarse resolution ) SNOWPACK simulations using RESAMPLING_STRATEGY = GRID_EXTRACT_PTS
Within the following repository the team of the avalanche warning service Tyrol is developing a model chain to run gridded SNOWPACK simulations. 5 (flat field and four aspects) SNP simulations shall be conducted for each grid point of the available NWP model data.
https://gitlab.com/avalanche-warning/models/snowpack/forecasts
The goal would be to provide the NWP data in NETCDF format. However, SNOWPACK/METEOIO is really slow with processing the NETCDF data compared to simulations driven by SMET files. The NWP data is formatted according to the CF1.6 standard, variables are recognized and simulations are running, but though RESAMPLING_STRATEGY = GRID_EXTRACT_PTS simulations are really slow. Even with the hack of Simon Horton in the GridsManager.cc for selecting grid indices i,j directly instead of easting, northing (see attached pdf file), the processing is not faster in the newest Meteoio Gitlab version. I had to add a few more lines compared to Simon's hack, because otherwise it misses lat,lon coordinates for calculating sun angle stuff. Maybe this has potential for improvement. I am hoping the slow processing is due to a wrong SNP setup (ini-file). A complete ini-file besides multiple VSTATIONS and sno-files would be:
AC_Horton_Custom_MeteoIO_code.pdf
Since you have way more experience with SNP, maybe one of you has an idea for a quick fix by just looking at the ini-file. Otherwise it could also be an option to adapt the NETCDF plugin.
@bavay @reisecker @Christoph.Mitterer @Adrien_Michel
Side note: Since we are interested in running SNP on quite coarse nwp model data, we are running flat field and 4 slopes for each grid point. Actually, we don't need the DEM for these simulations and setting slope angle and aspect based on the DEM (as done in GridsManager.cc is not needed). Maybe it is possible at some point to allow gridded NETCDF input without providing a DEM?