topic DEM Artefacts in ArcGIS Spatial Analyst Questions

DEM Artefacts

PaulHeersink — Thu, 05 Mar 2015 20:49:27 GMT

I have a collection of data points from which I am trying to generate a DEM and, ultimately, a hillshade for use as a backdrop in a map. Whichever way I try to process it (using Topo to Raster) I get artefacts where the points are located. See the attached example:

Any ideas what might be causing this?

Re: DEM Artefacts

ChrisDonohue__GISP — Thu, 05 Mar 2015 21:22:45 GMT

You may want to use a different surface creation method than "Topo to Raster".

Off the top of my head, Inverse Distance-Weighted (IDW) may work out better as a surface creation for then creating a hillshade.

Also, if I remember correctly, Natural Neighbors is the one that makes high points more extreme than in real life (much like the result you posted), so avoid that one.

Additionally, note that the number and distribution of your input points can factor into your results too, so you may have to experiment with several methods to see which works best considering your available data.

Surface creation and analysis

ArcGIS Help (10.2, 10.2.1, and 10.2.2)

Chris Donohue, GISP

Re: DEM Artefacts

PaulHeersink — Fri, 06 Mar 2015 12:27:55 GMT

Well, I tried your suggestion and the result is worse than the Topo to Raster result. See below.

I'll be trying some of the other interpolation methods and getting back to you on my success (or lack thereof).

Re: DEM Artefacts

PaulHeersink — Fri, 06 Mar 2015 13:08:38 GMT

Kriging (top), Natural Neighbour (middle) and Spline (bottom) are not much better.

Re: DEM Artefacts

OskarKarlsson1 — Mon, 09 Mar 2015 12:35:44 GMT

Have you tired creating a TIN or Terrain with those points and then transforming it into a DEM? Are your points dense enough? Maybe you dont have enough to come with a proper result

Re: DEM Artefacts

PaulHeersink — Mon, 09 Mar 2015 13:02:43 GMT

Oskar - I have tried creating a TIN. The TIN looks fine but converting that to a DEM results in an entirely different set of artefacts, one that appears to have many of the single cells outlined.

Most of the elevation points are about 195 m apart with other points added in along streamlines. I am contacting the original creator of the data to find out what process they use and what issues they've come across.

Re: DEM Artefacts

OskarKarlsson1 — Mon, 09 Mar 2015 13:07:47 GMT

Its gotta be something on the data then...Are the points for the elevation and the streamlines in the same units?

Re: DEM Artefacts

BenLeslie1 — Mon, 09 Mar 2015 13:54:40 GMT

I too get artefacts in my DEMs when using natual neighbours. I have two solutions:

1. Create the DEM at a lower resolution and then resample to a higher res.

2. Use LINEAR interpolation instead.

Neither is ideal, 1 can take a few goes to find the low res that eliminates artefacts so I use 2 - fortunately for me this is adequate.

Re: DEM Artefacts

RajinderNagi — Mon, 22 Jun 2015 22:56:46 GMT

You may want to use Bilinear or cubic re-sampling method.

Re: DEM Artefacts

DanPatterson_Retired — Mon, 22 Jun 2015 23:03:15 GMT

Rajinder... as an addendum to your post could you check your resources and see if this article has been updated for ArcMap 10.x, the reference only goes to 9.3 but I would assume that the situation is the same

29127 - Why are there raster artifacts visible in hillshade results when something is performed on a projected DEM?

Re: DEM Artefacts

RajinderNagi — Mon, 22 Jun 2015 23:14:27 GMT

Dan, It seems the article you referred is from tech support and I added comment for them to update this for 10.x.

For any continuous surface, bilinear or cubic re-sampling should be used instaed of default nearest neighbor to avoid artifacts (such as stair case).

Re: DEM Artefacts

DanPatterson_Retired — Wed, 16 Sep 2015 02:10:47 GMT

thanks for updating the thread with the information... still no update but the link will be useful because of the image documentation

Re: DEM Artefacts

DEAF_PROBERT_68 — Sat, 12 Dec 2015 02:22:27 GMT

Does the DEM you have are in known projection rather than geographic coordinate systems ? I'd check on that ..