[21pt] Missing inundation due to missing flows in BLE benchmark data

[mluck]

Even with the stream routing fixes addressing #916, the evaluation metrics such as CSI and MCC still decrease in some HUCs, for example 12090301. Stream routing was fixed so that the DEM-derived branch stream (red dotted line) follows the NWM stream levelpath (thicker black line) as seen in the images below. However, with the corrected stream routing there are two noticeable areas where inundation has erroneously decreased according to the agreement raster.

The problem is that some feature_ids do not exist in the BLE flow file. This seems to originate in the generation of the benchmark dataset where stream segments are assigned flows depending on the BLE cross-sections that they intersect. However, there are a number of stream segments that don't intersect BLE cross-sections and are not assigned flow (i.e., their feature_id doesn't exist in the BLE flows file). The NWM streams are crosswalked to the DEM-derived streams and assigned flow from the BLE table for inundation, so the reaches without flow values are not inundated. In the images below, the yellow segments are missing flow values.

The NWM stream segments with missing flow values (yellow) are crosswalked to the DEM-derived reaches (bold red dotted lines) and no inundation is produced for those reaches.

[mluck]

Inundation using 100 year recurrence flows (dark blue: overinundation, white: underinundation):