This is all assuming falcooat is trying to render a micro polygon export in Arnold.
If it is, then this is what is probably going on:
Looks like an import of an object into Arnold that has the normals turned off, hence the faceted appearance. This is probably compounded by the camera position in the render that is looking at very small section of the terrain and there is just not enough resolution in the mesh.
Turning on normals in the micro polygon exporter in Terragen can smooth the triangulation look / faceted appearance but the render can then look a bit "soft" . Also in some instances the file size can double with normals turned on (more on file sizes later) so I usually leave it off and smooth the mesh in Maya if I need to.
Main problem is, if you export your entire terrain as a mesh and then zoom into it,. you're going to see those individual polygons. You then come to appreciate how Terragen handles it's geometry - always rendering a proceedural version no matter how close you are to the terrain.
Ajcgi has a good approach, and it's something I've tried during my vector displacement experiments, but there is a limit to how much detail you can get.
I tried rendering 8k and even 16k displacement maps from Terragen and using them on a simple plane in Maya/Arnold, but in order to realise the detail in the map in the render a very high subdivision level is needed. This can take quite a while to calculate and uses lots of memory. I have 64gb in my machine and it ran out of memory before it could render a surface subdivision high enough for a 16k map!
There are ways around it, and my approach is to use tiling. Many smaller files (meshes or displacement maps) can be handled much better by Maya / Arnold than one gargantuan file :-)
This is how the texture UDIMs work, rather than create a huge 32k texture for a character, it is split into tens or even hundreds of smaller 4k tiles that are easier to deal with.
Matt mentioned you can render the micro from the camera perspective - for a still shot this will work fine, but if the camera moves then the still mesh capture will be redundant and a top down projection will be needed.
In my experiments, I've found the best approach is to export the entire terrain to the micro exporter, previs to find where the camera is going to be animated, then replace just the close up section with a much higher resolution mesh (or vector displacement map)
I'm going to publish a new version of my Terraman script soon. In it, you'll be able to use an orthographic camera in Maya to Boolean out a chunk of terrain close to a render camera, then export that same orthographic camera to Terragen to create a micro polygon mesh of the area in high resolution (this can even be tiled further for even more detail)
You will then be able to import the higher resolution terrain (or terrain tiles), and it will slot right into the hole created by the Boolean .The Maya scene will only have detail where it needs it and low poly detail in the background.
I'm working on an animation using this at the moment and will post when completed!
Anyway, I hope this was what falcooat was asking :-)
If not, hopefully someone else might find it useful!
