Hello, I am thinking about using the DelineateIt tool to create watershed boundaries in an urban environment. Has anyone used it in this way and can offer advice? Will the tool produce sub watersheds or just one big watershed? Should I try to use the approximate scale that I will ultimately be using for my research or use a larger area in order to produce more accurate results? What are the pros and cons of using the “detect pour points” option vs. inputting watershed outlet features? Thank you!
Hey @kahorl -
I’ve not delineated watersheds in a specifically urban environment, but I have used DelineateIt a lot. Can you describe what you’d be using the watershed for? Is it input to a specific model? Is it likely to be entirely urban, or mostly natural with some urban included? Depending on your use, I’d think that the main consideration is that the DEM that DelineateIt uses probably only represents natural flow, not gutters, sewers, etc (unless you have a fancy DEM that does take these into consideration?)
DI can produce one large watershed, or multiple sub-watersheds, depending on how you use it. If you provide a single outlet point on a river, then it will produce a single, possibly big, watershed. If you provide multiple outlet points, it will create the watershed that drains to each one.
Regarding pour points, as the User Guide says, “There is a pour point located at the center of each pixel that flows off the edge of the raster or into a nodata pixel.” So if you choose this option, you’ll get watersheds that flow to the edge of the DEM, or where there’s a NoData pixel in the DEM. I haven’t found this option to be very useful, since it doesn’t necessarily create sub-watersheds that cover the whole larger watershed (like you might use in the Annual Water Yield model). I generally provide my own pour point layer, that way I know that it will create sub-watersheds that are useful for my needs, and I can adjust their location as needed (for example, if one is too far away from a stream to create a good corresponding watershed).
~ Stacie
2 Likes
Hi @swolny!
Thank you for your reply. I am intending on using it in a purely urban context, the densest part of a city, with the eventual use as an input to the urban InVEST model for stormwater. I’m using a DEM I downloaded from the USGS National Map, so fairly standard, nothing fancy. Ideally I would like to produce multiple sub watersheds, or at least one smallish watershed that is closer to my intended study area. I’m using Cincinnati, OH as a case study, which sits on the Ohio River, so I could create an outlet point layer using that.
In a preliminary attempt to use DelineateIt, I downloaded two adjacent DEMs and created a new mosaic raster combing the two together using the same characteristics as the original (.img) files, so floating 32 bit and 1 band. I have tried a few different file formats, but when I try to use them in DelineateIt (using pour points for now just to see how this all works because I haven’t created an outlet layer yet), it won’t recognize the rasters as a valid input. Any suggestions on how the DEM/raster needs to be formatted to work?
Thank you for your help! There is definitely a slight learning curve to this.
Katie
Hi @kahorl,
I’ll let Stacie speak to using this on a dense urban landscape, but I can help with the model failing to accept the raster input. InVEST uses GDAL for handling raster operations so should except a wide variety of formats. Normally we see people use the GeoTIFF format .tif. Some information that will help us debug what’s going on:
- What tool did you use for the mosaic operation and what was the output format?
- What formats have you tried that failed?
- Are you seeing a red validation error from the user interface or is the model failing after you’re able to run it?
Thanks,
Doug
Hi @dcdenu4,
I looked up the GDAL formats just to make sure I wasn’t missing something, and from what I can tell, I have used valid file formats. The input DEMs/rasters were .img files downloaded from the USGS National Map. I used the Mosaic to New Raster tool in ArcMap with no problems. At first I created a new raster as another .img using 32 bit floating point and 1 band, to maintain the same properties as the input DEMs. After I unsuccessfully tried using this file as an input for DelineateIt, I made another mosaic as a .tif, again using 32 bit float and 1 band. This also failed to be recognized as a valid input. I am getting the red x validation error in the user interface. It will not allow me to run the tool.
Any trouble shooting ideas you have would be most welcome. Thank you so much for your help!
Katie
Hi @kahorl -
Can you upload a screenshot of your model input window so we can see the input file name and error message?
I haven’t used the Urban Stormwater model, but looking at its chapter in the User Guide, it looks like watersheds are only used for aggregating results. Since the model doesn’t use the DEM to route water, and says that any type of area of interest may be used for aggregation, I’m less concerned about the watersheds being hydrologically correct. So I guess that it’s up to you to decide what are the meaningful units in which to aggregate results, which may be watersheds/sewersheds, or perhaps neighborhoods, etc.
What I do know is that DelineateIt will create a watershed based on the DEM, and since the DEM is only based on elevation, you’ll get a result that is mapped to something like a natural stream network (which is an output of the tool, so you can check it out). But in an entirely urban setting, it’s rather unlikely that water is moving naturally, so the watershed result may not accurately represent the area that’s actually flowing to your defined outlet point on the Ohio river.
Perhaps the model authors have more advice about this?
~ Stacie
1 Like
Hi @kahorl ,
Could you also share what version of InVEST you’re running?
If you’d like to share the DEMs that are failing to load I can try to replicate the issue and debug further. If the DEMs are small enough you can upload them here, otherwise over a shared file system like Google Drive usually works pretty well.
Thanks,
Doug