ArcHydro drainage mismatch

216
5
02-07-2018 03:45 PM
GusBonorino
New Contributor

Hello!

Working with ArcHydro in ArcGIS 10, the confluences of river segments (str, strlnk, drainage line) do not exactly coincide with catchment boundaries. The confluence is 1/2 pixel farther downstream from the boundary in all cases. See picture attached.

I use ASTER GDEM. 

If possible I would appreciate learning how to correct this defect.

Thank you

Gustavo Bonorino

Argentina

Reply
0 Kudos
5 Replies
DavidBetts
New Contributor II

Gustavo,

The disagreement appears to be in the method used to draw the catchment boundaries.  It looks like the catchment boundaries are drawn to include only WHOLE pixels, so the boundaries stop after the last WHOLE pixel.  This leaves the last half pixel before the confluence outside of each catchment.

This may not be a problem.  Your estimates may not be exact, but they might be good enough.  There is always a loss of precision compared to reality when you are modeling a watershed.  Your picture displayed one example of the loss of precision.  The catchment boundaries are drawn with 90 degree angles.  In the real world, you wouldn't expect to see the catchment boundaries to look like a flight of stairs.  Even the confluence may not be as precise as is drawn on the map.  Instead, the boundaries might be much "fuzzier" due to differences in subsurface and ground water return flows, or changing over time because of changes in stream-bed morphology.

Perhaps, before dedicating your time and effort to the 1/2 pixel differences, the more important question to ask is, ¿What kind of error are you willing to accept?

  • ¿How much difference does that half pixel make to your model?
  • ¿Would your model of the catchments be better represented by expanding the original boundaries, keeping the boundaries where they are currently drawn, or finding a boundary somewhere in between?
    • ¿Should you be looking for a way to draw the catchments with a smooth boundary, rather than along the 90 degree angles of pixel boundaries?
  • ¿How much would your model improve if you had more precision in catchment boundaries?
    • ¿Is that "precision" truly a better representation of the landscape?
    • ¿Is that improvement in "precision" worth the effort?

Y al final, di me por favor si no me entiendas.  Te podría responder en castellano también.  Aprendí el castellano al vivir en la zona de cuyo hace un montón, pero siempre trendré un cariño para Argentina.

Suerte

Reply
0 Kudos
GusBonorino
New Contributor

Hola David,

Cuyo, la tierra del sol y del vino.

Thanks for your message.

My concern is not with precision, however. When I go ahead with GeoHMS to construct the subbasins the misalignment resolves into spurious subbasins, as shown in the picture attached.

It must be something related to the configuration of ArcGIS or ArcHydro in my PC, but what?

All the best.

Gustavo

Gustavo González-Bonorino Centro Austral de Investigaciones Cientificas CADIC Av. Bernardo Houssay 200 Ushuaia, Tierra del Fuego (9410) Argentina

Reply
0 Kudos
GusBonorino
New Contributor

Hola David,

Cuyo, la tierra del sol y del vino.

Thanks for your message.

My concern is not with precision, however. When I go ahead with GeoHMS to construct the subbasins the misalignment resolves into spurious subbasins, as shown in the picture attached.

It must be something related to the configuration of ArcGIS or ArcHydro in my PC, but what?

All the best.

Gustavo

Gustavo González-Bonorino Centro Austral de Investigaciones Cientificas CADIC Av. Bernardo Houssay 200 Ushuaia, Tierra del Fuego (9410) Argentina

Reply
0 Kudos
MarkBoucher
Regular Contributor II

I think what you see happening is normal. If you click on the pixels to see the flow accumulation value, you will see those at the bottom of the watershed (along the strlnk line and the last pixel before leaving the subbasin) are related to the area of the watershed (multiply fac by pixel area).  The confluence would NEED to be inside the next downstream catchment/subbasin/watershed.

 

I also agree that what we are dealing with is a digitized representation of something that is analog. So we have to except a more “grainy ” result in our analysis. Hydrology is a subset of the overall Civil Engineering discipline and C.E. can also stand for "close enough".  

Mark Boucher, PE

Senior Hydrologist

Contra Costa County Flood Control

& Water Conservation District

DavidBetts
New Contributor II

Are the extra subbasins always at the lowest point of your analysis?

With the reply from Mark Boucher, I wonder if the "spurious subbasins" are just the product of GeoHMS needing to put the confluence in the next subbasin.

For these kinds of questions I've used TauDEM, so unfortunately I'm not familiar with the limitations and assumptions of GeoHMS.  No quick answers from me today.

Reply
0 Kudos