Why difference in delineation of watershed by two online watershed modeling tools?

Question

I am using two online watershed delineation tools- eRAMs and Model My Watershed (MMW)- to auto delineate watersheds using water quality monitoring sondes as my outlet points. I have three pieces of equipment on different streams so I have three different outlet locations. Two out of three outlets/sondes have similar watershed delineations, but one sonde location has dramatically different watershed delineations. Here is pic of sonde location, the green marker is our coordinates and the red circles are different outlet locations I've manually added in eRAMs tool:

Even after trying multiple outlet locations between the bridge and the streams confluence with another stream, as well as even trying locations upstream of the bridge, the tools result in the same watershed delineations shown below:

Note: eRAMS result is green boundary, MMW result is orange boundary, Creek where sonde is located is highlighted in red.

I lean toward the MMW delineation, as the delineation encompasses the creek on which the sonde is located. Note that these tools move the manually placed outlet points to nearest downstream NHD cell, but even the moved outlets are above the confluence.

What might be the reason these tools have such different delineations?

I believe that both tools delineate based on 30 x 30 m DEMs and the NHDPlusV2 Medium Resolution stream network. eRAMs uses TauDEM tools/algorithms found here: https://erams.com/documentation/watershed-delineation-tools/ It's difficult for me to interpret the subprocesses of MMW's Rapid Watershed Delineation algorithms listed on GitHub (limited development background), but I think they are modified versions of TauDEM algorithms, similar to eRAMs.

Answer 1

There are lots of reasons why automated watershed delineation can be wrong. Sometimes it has to do with the underlying source data, which does not accurately represent the flow directions of the terrain. Errors are frequently due to the step of "snapping the pour point," where the software relocates your watershed outlet to correspond with a digital stream, or a grid cell with a high "flow accumulation." The algorithms for doing this often fail around confluences.

In this case, the watershed in green looks wrong, as it fails to include the drainage area of the "East Branch."

If your monitoring equipment downstream of the "Middle Branch," you should try nudging the outlet downstream to capture that drainage area.

Or you could merge the two polygons in GIS to get a better result.

All in all, knowing how to do "manual" watershed delineation remains a useful skill. If you can interpret a topographic map to figure out the flow pathways and watershed divides, it will help you determine if the software's output is correct.