Eric, I have some USGS data of Death Valley, California. The map encompasses a valley floor with an elevation 282 feet below sea level and an escarpment rising over 11,000 feet in elevation. It is a MUCH better set of data for testing heightmap algorithms. St. Louis is not exactly known for its mountainous terrain.

Your problem likely results from over scaling of the data and a limited resolution in the dataset. Even my Death Vallye data has this, but is virtually not noticable because the terraces are miniscule compared to the mountainous relief.

If you would like, I can email you the data, which I have reorganized into a very simple file of integers. I can provide you the source code to read the data too: about 10 lines of code.

I would like to see how your algorithm displays the data. Also, I may have a solution to completely eliminate the terraces.

Cool!

I wish I could show you my screenshots of the same data, but I have since thrashed the code with modifications and that was a long time ago so I can''t easily get it back to working correctly.

If you want to see something kind of cool, I would turn the fog off, and pull the camera back enough to get the whole height map in the frame. And you might want to tone the scale of the altitude down a little. Try about 1/3 of what you are currently using.

Eagerly awaiting more screenshots if you have them...

Eric,
I have a couple comments. What you are doing is cool, and I would imagine you are having fun with it. I notice that you have created an algorithm that selects whether to render sand or dirt. Although you argued against me (only a little though) in my Snow on terrain? thread, I think that thread (at least what I was proposing) was in the same spirit as your sand vs. dirt algorithm. Essentially, it proposes a preprocess of the terrain to identify those areas which would have snow cover, and contrary to many of the people who posted in that thread, the whole concept was not that complex. In my last few posts in that thread, which are very recent, I provided an equation to figure out the snow coverage. Bottomline: even Death Valley gets snow in it''s higher elevations! If you are interested, I could help you implement the snow calculations and we could see how they would look on your data sets.

I have some more data sets for you in the same format I gave you Death Valley in. I just have to find them. One is Owens Valley, which is about 100 miles northwest of from Death Valley and is actually considered a deeper valley (one of the deepest) because it has 14,000 mountain ranges rising in escarpments 10,000 feet above the valley floor on both sides of the valley. I think I have the Grand Canyon floating around on my hard drive also.

Fall Canyon! Is that the Fall Canyon in Death Valley? I hiked up that canyon about a year and a half ago. It becomes a slot canyon the further in you go.

Another useful calculation to run on the terrain as a preprocess is the contributing area function. It is a really simple function (about 5 lines) but it is really powerful. It will tell you the amount of water which flows through a cell, and you can use this data to determine the locations of rivers, creeks, streams, and areas of increased vegetation.