Referring to the post here on the limitation of D8 algorithm, or watershed algorithm in general.
From the post I have the impression that even though the urban areas have a lot of flat surfaces ( which contradicts or at least render the D8 assumption inconvenient), and the flow paths don't always follow the naturally shaped areas, and thus might make watershed algorithms in general less accurate and useful, but I think one can alleviate these concerns by
- For flat areas, we just apply artificial gradient so that they are no longer flat ( the general rule for the artificial gradient is that the flat areas should be the highest at the middle, and the height gradually disperses as the water goes to the edge)
- We artificially burn known flow paths ( eg: streamlines or drains) into the topology before we apply the watershed algorithm.
But am I right on these? Or are they other constraints that are unfixable by the above remedies?
