Estimate river width and depth if no river geometry data available

[TBovenschen]

When no river geometry data is available (hydroAtlas or Lin 2019 do not have small rivers included) HydroMT does not create the riverwidth and -depth layers and the wflow model can not be run.
Maybe the old deprecated method of setup_riverwidth could be used as backup to the new methods, to estimate the riverwidth based on discharge/precip/upstream area when no data is available. I think it would be most logical if this would be included as an option in the setup_river method.
Another option would be to estimate the river width based on a raster (like merit_hydro). Merit_hydro has the river width data included, and it covers some area that are not covered by for example Lin 2019, so might be useful sometimes. I am however not sure whether this is possible and would give reliable results.

[hboisgon]

Good suggestions! I would also be in favor of being able to still use the 'old' setup_riverwidth method.

For river depth, another suggestion could be to use directly discharge from wflow outputs based on kinematic wave approach (under the Lin database a return period of 2 years is used). Note that for the kinematic wave, a default bankfull_width of 1 meter an be used and has almost no impact on the results for water level and very little to no impact on the discharge.

[hboisgon]

Suggestion: split setup_rivers into setup_rivers (mask only and dem adjust) and setup_river_cross_sections (width, depth, roughness)

setup_rivers: derive mask and data from high res dem and width if available in this dataset

• mask
• length
• slope
• optional: width_raw (as this may not be complete)

setup_river_cross_sections_from_vector and setup_river_cross_sections(_from_raster) should allow to update only width or only depth or only roughness. (later like d3d setup_river_cross_sections_from_point xy crosssection)

roughness methods:

• mapping table based on streamorder --> current setup_rivers maybe improve with table per resolution or upstream area

width methods:

• vector: from width in gdf (new: fill locally with powlaw(uparea) instead of smoothing?) --> current setup_rivers
• raster: based on power law (discharge, precip, uparea) --> old setup_riverwidth method - discharge is best - uparea can be okay locally
• raster: add fill gap and smooth from width_raw using a power law(uparea)

depth methods:

• vector: from depth in gdf --> current setup_rivers
• vector: from width + qbankfull in gdf using powlaw/manning/gvf --> current setup_rivers
• raster: from width + qbankfull derived from gridded qbankfull (not discharge timeseries 2yr RP postprocessing script on wflow output or postprocessing method) (e.g. wflow kin wave run where depth has a very small influence) using powlaw/manning/gvf --> new method
• both: maybe not make it mandatory to create it especially for kinematic wave

Also possible: create new river database merging lins and merit hydro data
Add new river dataset GRIT: https://zenodo.org/records/7629908
Interesting abstract (data not yet available?): https://meetingorganizer.copernicus.org/EGU24/EGU24-19997.html