While playing around with Fractal Terrains, I like to start with a rough sea floor and no land and then raise up sections to create continents. Although the parameters are set to tell the program not to generate any land, a number of extremely high spikes are created almost every time. It took a while to figure out how to flatten these out without generating artifacts.

First off, here are the settings that I have been using:

  • Highest Peak: 32000 feet*
  • Lowest Depth: -16000 feet
  • Circumference: 7000 miles
  • Roughness: 0.75
  • Percent Sea: 100
  • Land Size: 10.00
  • Fractal Function: RMF with Perlin’s Improved Noise
  • Editing Size: 4096 (Custom)
  • Allow Prescale Offset Editing: Checked

*Setting highest peak to 0 for world generation seems to do away with the spikes, but later raising it back up restores the spikes. Percent sea=100% should make this irrelevant anyway.

The Problem

The settings above should generate a world with no landmasses and a fairly uniform ocean floor. For the most part this works. But almost every map will have one or more extreme spikes forming small islands well in excess of the upper limit for landmasses (around 100k when the limit is set at 32k), even though no landmasses should be created at all.

Here is an example:

The Fix

Simply changing the altitude doesn’t work as it creates significant artifacts. Feathering the selection before lowering altitude doesn’t work either because of the extreme differences in altitude involved.

Here’s what I’ve found that works:

  1. Select by altitude everything above 0
  2. Feather the selection by 1
  3. Tools > Global Lower > Prescale Land Offset by 0.1

Generally, that improves the situation but doesn’t entirely correct it. In that case, do all three steps again. Twice usually corrects the problem down to a number of spots just slightly above sea level. To get rid of these, I perform the steps once more but this time I select everything above -500 feet and lower by 0.01 instead.

Why Is This Happening

Sadly, I’m still new to this program so I really don’t know. I assume it is an inherent issue with the fractal routines. I’ve observed this behavior in both fractal methods that use improved noise. I haven’t experimented with others yet to see if they all have this problem. I am using these two methods because they are the ones that generate the most detail at the closest zoom.

That completely takes care of the problem for me with absolutely zero edge rippling around the selection. Different parameters will most likely require slightly different settings but the adjustment should be minimal.