Ask A Cartographer

Generalizing Hillshades for small scale display

October 12 2009 | 3 comments
Categories: Map Data

I have downloaded 30 m elevation data for all of the Middle East. How do I generalize the data so that it looks nice at smaller scales? I would like just major landforms highlighted at 1:1,000,000 and more detail appear as the user zooms in. This would be used for web applications.


Mapping Center Answer:

It may be easier to use the SRTM 90 meter data for your area.  I'm not sure what you mean by elevation data or "downgraded".  We often get people calling a hillshade elevation, which is not quite true--a hillshade is a relief depiction that is derived from an elevation dataset. There's not much to be done for generalizing a hillshade--the elevation dataset, i.e. a DEM should instead be genearlized and then a new hillshade would ideally be produced from that.

So, let's assume you've got a downgraded DEM, maybe it's integer values instead of floating point values for elevation.  Convert the DEM to floating point and then resample it, which will give you a more refined set of values in the result, even though the cell size will be larger--the idea is not to lose any more information than is already lost. Then you can produce a new hillshade.

That said, for your scale, you may have an easier time just downloading the SRTM 90 meter elevation data.  It is always better to use data that has been captured at a given resolution than to generalized more refined data to that same resolution (generalizing will never give you the specificity or accuracy, often arbitrarily removing details). You can download that data from The National Map Seamless Server.



Hillshaded Service posted by Bruce Burwell on Jan 18 2010 11:10PM
What we ideally would like to do is produce a Hillshaded service for our web applications. When a user logs in, they would see a hillshade of the entire Arabian Penninsula. As the user zooms into an area, better hillshading would appear. I assumed I would start with the best resolution data we currently have, which is 30 m resolution, and then resample the data so that it looks more generalized at different scales. Is that right?
Before you go to the trouble... posted by Charlie Frye on Jan 19 2010 11:11AM
Before resampling and making additional hillshades, try changing the resampling method for your pyramids. If you haven't tried it, make a copy of the hillshade you have now (Use the Copy Raster tool) and in the Geoprocessing Environment Settings for Raster Storage Setting, set the pyramid resampling technique to Bilinear.

Let's say, for argument's sake, that you don't totally like the result of changing the pyramid resampling. If you end up resampling the DEM to produce new hillshades, consider setting the Z-factor for those hillshades to introduce a little more vertical exaggeration for each progressively smaller scale you produce a hillshade for. That will help enhance the terrain, and potentially let you get more out of each dataset.
Or try smoothing it instead posted by Aileen Buckley on Jan 19 2010 1:53PM
You could also try smoothing the higher resolution hillshade instead of resampling it. We used the Focal Statistics tool along with the vertical exaggeration that Charlie mentioned to get seamless looking hillshades across multiple scales for a world terrain data set. We will be publishing those tools in a toolbox in the next mornth of so here on Mapping Center on the ArcGIS Resources page.

If you would like to post a comment, please login.

Contact Us | Legal | Privacy |