topic Problems rendering hillshade with USGS NED data in ArcGIS Spatial Analyst Questions

Problems rendering hillshade with USGS NED data

BradleySmith2 — Tue, 06 Nov 2012 04:02:28 GMT

I used to download the digital elevation models from the USGS seamless server to render elevation contours and hill shades (shaded relief) for my work. The DEMs from the seamless server worked flawlessly for me, rendering beautiful hill shades for my projects, such as this example.

[ATTACH=CONFIG]19042[/ATTACH]

Ever since the USGS switched to the National Elevation Dataset, I've been having all kinds of problems rendering accurate and useable hill shades. Has anyone else had this problem? I even tried using a z factor of 0.0003, which was suggested on various websites. I still get poor quality hill shades, such as this, which is the same location as the example above.

[ATTACH=CONFIG]19043[/ATTACH]

Re: Problems rendering hillshade with USGS NED data

Anonymous User — Tue, 06 Nov 2012 05:15:20 GMT

Original User: m.gasior

It seems that your z-factor is still to high.

This article explains how to use z-factor, how it depends on latitude and provides z-values for elevation data in meters and feet:
http://blogs.esri.com/esri/arcgis/2007/06/12/setting-the-z-factor-parameter-correctly/

Despite the factors presented in this article, for cartographic purposes a bit higher z-factor in Hillshade tool gives more vivid terrain picture.

Other solution is to project a DEM which is in geographical coordinate system to projected coordinate system which xy units are meters/feet not degrees.

Re: Problems rendering hillshade with USGS NED data

BradleySmith2 — Wed, 07 Nov 2012 04:50:49 GMT

Thanks for the post. I projected the NED data set I have to NAD 83 UTM Zone 11N. It looks like things are heading in the right direction as you can see from the example below. Unfortunately, now I'm getting these weird lines across the raster that are at right angles, almost like the texture of a course fabric. Do you think its a problem with my data set or is there something more I should be doing in addition to projecting the data set?

[ATTACH=CONFIG]19082[/ATTACH]

Re: Problems rendering hillshade with USGS NED data

Anonymous User — Wed, 07 Nov 2012 05:11:19 GMT

Original User: m.gasior

I experienced this same behavior when hillshade cell size was the same as DEM cell size.
Setting a bit larger cell size of hillshade may help in this situation.

Re: Problems rendering hillshade with USGS NED data

MarcoBoeringa — Wed, 07 Nov 2012 06:55:58 GMT

You may also be able to solve this last issue by setting the "Resample during display using" option under the Layer properties "Display" TAB to bilinear interpolation or cubic convolution instead of the default nearest neighbour resampling. This will interpolate a value during display and zooming, giving smoother results.

Re: Problems rendering hillshade with USGS NED data

Anonymous User — Wed, 07 Nov 2012 15:58:32 GMT

Original User: erictrice

Bradley,

The default resampling in the Project Raster tool is Nearest. You can't use that otherwise you get the result you did. Here is the help for this parameter.

"The NEAREST and MAJORITY options are used for categorical data, such as a land-use classification. The NEAREST option is the default since it is the quickest and also because it will not change the cell values. Do not use NEAREST or MAJORITY for continuous data, such as elevation surfaces."

Go back to your original data and project it again using bilinear, then run the hillshade with z factor 1 (assuming X,Y,& Z all have same units). Do not change the cell size of the output hillshade.

You should also build pyramids with Bilinear, which can be set in the environment settings. If you resample with bilinear when projecting but build pyramids with nearest (the default) it could still look weird. Also, as mentioned above, set the display resampling to Bilinear for a better display.

Best,
Eric

Re: Problems rendering hillshade with USGS NED data

BradleySmith2 — Sat, 10 Nov 2012 16:30:16 GMT

Thanks for everyone's feedback. Projecting the data set from a geographic coordinate system to a projected coordinate system and using bilinear sampling solved the problem.