One of the biggest challenges was handling non-drainage depressions — they can be both real hydrological features and digital artefacts.
Instead of “filling sinks”, I used a mixed approach:
– identifying real depressions manually (tailing ponds, collapse funnels, drainage portals),
– and letting the model naturally route flow into them.
It turned out to be an effective way to simulate realistic runoff in a heavily disturbed terrain.