Steve Kopp 5 years, 2 months ago

Emily - the primary cause for differences in your analysis is from using the 10m NED. The Ready to Use watershed delineation service in Texas is based upon the NHDPlus 30m DEM, with additional hydroconditioning work by Esri at stated in the documentation.
https://pro.arcgis.com/en/pro-app/tool-reference/ready-to-use/watershed.htm
https://developers.arcgis.com/rest/elevation/api-reference/source-data-for-hydrology-analysis-tasks.htm
Using a different DEM source which was not hydroconditioned is guaranteed to give a difference answer. I am assuming you were not using the NHDPlus High Resolution 10m beta, although that would still differ from the Esri result.
As stated in the doc link above, our hydroconditioning process differs from that of the NHDPlus team. This was done in order to provide continuous flow in places we knew a stream flowed even though the DEM said it did not. This is clearly illustrated in figure 4 of your paper and you attempt to explain it but need to dig a little deeper. The Brazos River runs through Lubbock, so your green polygon is off by roughly 50 miles. Also, the Esri watershed is accounting for the noncontributing area you are asking for, illustrated by the big hole in the blue polygon north of Lubbock.
Hydroconditioning a DEM followed by a quality control step to ensure streams flow when and where they should is a time consuming process. The team spent months on this process in communication with the NHDPlus team, and investigating ancillary data sources. The last section of the documentation referenced above says to email the team for more information on this data processing. And the online help for the ArcGIS Pro tool has a link where to provide feedback, which would include asking how something works that you couldn't find in the documentation, or reporting what you believe are errors in the output.
The team would have been happy to explain all this and help you design a more apples to apples comparison.