Hello,
I am generating a stream raster using RouteDEM. If it aligns fine with real stream then I will use it for SWY and SDR models. However, In my case, the results seems weird and I do not know where did get wrong and how to correct it.
As you can see by comparing DEM and the output of stream mask, there is an area in the middle especially problematic (I circled with blue line in flow direction map).
I have played around from FTA 500 to 5000 (both D8 and MFD algorithm) but none of them change the middle part close to the reality. I appreciate if anyone can give me some ideas or suggestions.
The reason this is important is that, at least from the screenshot, the elevation does does not allow for water flow through that bridge. From the model’s perspective, that causeway is forming a dam, which means that just upstream of the dam is a hydrological sink, which must be filled for the model to route water correctly.
Your solution here is simple: break the dam. You will need to modify your source DEM so that there isn’t a hydrological sink there at the dam. Once you do that, you should not see these strange flow patterns.
Let us know how it goes! I’d be very interested to see a screenshot of the flow after you break the dam too
James
Thank you very much for checking on it. Yes, it is a road across the middle of this area and a small bridge above the river.
Regarding editing on DEM, I am considering creating a small area (rasterized polygon) around that river bridge cross to adjust the elevation. Still, I have no experience in editing DEM maps (such as dealing with very small pixels). Do you have any suggestions?
I’m not as advanced a GIS user as some other folks at NatCap, but rasterizing a small polygon at an appropriate elevation sounds like a good solution here.
I haven’t had to do something like this, and you’ve probably made something work by now, but it is an interesting DEM artifact. I might try something like this:
Note the elevation of the pixels around the bridge, to determine the elevation that the bridge pixels should be.
Create a small polygon that completely covers the bridge. Assign that polygon the elevation value that you want to change the pixels to.
Convert the polygon to raster. The raster should be the same extent as the DEM. All raster values that are not part of the bridge polygon should have a value of 0. The bridge polygon area should have the target elevation value.
This is where it gets tricky, and you might need to experiment.
I’m not sure which GIS you’re using, but in Arc you can use the Con tool to say “in the places where the bridge elevation raster is [target elevation], assign the pixels in the output raster to that value; otherwise, assign the other pixels to the same value as in the original DEM”. Last I checked, there wasn’t a direct equivalent of this in QGIS, which is unfortunate, it’s an incredibly useful tool.
Now run the results through InVEST and see what the streams look like. If they still look funny in that area, you could try smoothing out the area you just edited, by doing an interpolation of the DEM just within the bridge area (again, Con is probably useful, so you can do this just within the bridge area). You might need to slightly expand the bridge area to make this produce a good result.
That’s all I have off the top of my head. If you’ve found a way to fix this, please let us know!
Thank you so much for helping me! I was able to get rid of two bridges and replace them with some elevation values as close as to their surroundings. The stream map looks normal and the main river is captured and continues.
I have a further question about the TFA. The guidance says the stream map should be “as close as possible to the real-world stream network in the watershed”. How do we know it is close enough if we only have the main river maps as a reference? The main river map does not help us to know how many details are too much. Taking my map as an example, we can see the main river from both stream maps, FTA=1000 gives more flows than 3000. How can I know FTA at 1000 is not exaggerated or is 3000 too simplified?
@urgendalai here are some notes on helping to find the threshold flow accumulation…
First, keep in mind that the selection is somewhat subjective, not everyone will choose the same value and the one you use doesn’t have to be exact. What you are trying to do is approximate where the first order streams begin.
You can overlay your flow accumulation raster onto various AGO basemaps (such as National Geographic or Topographic) and other stream datasets, such as those you can download from HydroRivers (HydroRIVERS). Here’s a link to high resolution Google satellite basemap that might help, just add it to Pro using Add Data from path.
Not every stream in the FAC will match perfectly with your basemap or other streams datasets, so just use your own judgment. Consider the shape of the channel, does it look natural? Consider the DEM, can you see the cutting into it? It’s hard for me to tell from your maps here, but my initial guess is that the 3000 value is what you should start with. You should make everything below that value transparent so you can see other layers beneath it and play with your symbology until you get a decent match.
