Crosswalking refactored hydrofabric

[awlostowski-noaa]

Very short channel segments create a performance bottleneck for the diffusive wave model. The diffusive wave model adjusts timestep duration to maintain numerical stability on the shortest segment in the entire domain. For example, if there is a single short segment (say, 4 meters) in a domain composed of hundreds of segments, then the routing timestep is made as small as it needs to be to generate stable solutions on that segment. That same short timestep is applied to ALL segments in the domain, too. So, we need to refactor diffusive domain hydrofabrics so as to eliminate the presence of very short segments.

Refactoring may include merging together short segments to create longer ones and/or moving junctions to facilitate the creation of longer segments. Either way, the refactored hydrofabric will have a different number of channel segments than the native hydrofabric coming from RouteLink. This fact is problematic because channel forcing's (lateral inflows) are generated for each segment in the RouteLink and we also need to write flow results out for each segment in the RouteLink.

We need a scheme to crosswalk forcings to the newly refactored hydrofabric, and crosswalk flow results back.
@BrianAvant-NOAA

key considerations

• Diffusive network graph must be contiguous
• Channel parameters may need to be averaged (slope, alt) or summed (length)
• mass balance must be maintained
• need to write results to all (non-waterbody) segments listed in the RouteLink.
• Can we include the crosswalking data into the natural cross section data product? That is, everything is in a single .nc file.

[BrianAvant]

Steps to resolve this issue (this assumes I have a serially complete network of surveys):

• Update natural xs netcdf with proper attributes (issue created; PR submitted)
• ensure no duplicates between comid and refactor_id (issue created; PR submitted)
• check nnu.build_connections for diffusive wave
• update MC stream boundaries to attribute with refactor_id
• convert q_lat timeseries for diffusive domain
• convert flows/depth from refactored network to NWM v2.1