The terrain is created and rendered, but due to limitations with World Machine and memory I can only work with areas 150 x 150 miles, so merging them properly is a key part of being able to do this. After years of trial and error it is still a struggle, but the results are getting better and better. I’ve now created the Shield Land tile, and all its neighboring tiles and blend them together. The results are 90% satisfactory right away, but with a few key issues needing manual editing to fix.
Rivers are the big thing with this since each tile calculates them independently, the border overlapping areas will have two conflicting sets of river systems and manual editing is needed to soft out what will be the final remaining rivers and lakes. This might sound easy but is tricky and VERY tedious to make sure that the result is believable and flows downhill. Lots of back and forth between 2D in Photoshop and 3D in World Machine to double check. The benefit is that you get intimate with the terrain and what each area looks like.
The other issue requiring manual editing is harder to spot at a glance, but once you see it, it can’t be unseen. Large straight lines of slightly differing slope angles stretching sometimes for miles. These come from the masking away differences in the overlapping areas. I’ve back fed areas into the neighboring areas when working on the to try and make sure they match, but inconsistencies creep in. Fractals outcomes vary depending on how centered they are in the render, what they are surrounded by and many other things, hard or impossible to hide or compensate for. The solution is a combination of editing the heightmap in Photoshop and going back into World Machine and patching over the faulty parts. This inevitably introduces new errors, so the hard part here is to settle with the doable rather than perfect.
Error in the terrain mirrors the issues real world cartographers are facing. 3D and image data of the real world is also dotted with faults and missing bits, and a lot of the tricks I use are taken straight from real world ways to fix things. In some ways doing fantasy cartography is easier, making stuff up is the game, and there is no real world to consider. Working on famous areas of Greyhawk is the hardest, they are full of detail everyone knows well and getting them right requires precision that can be hard to get.
Almost done with the fixes to the main Shield Lands area and I have already tested out texturing to make sure that the data quality was enough to get the results I’ shooting for. Below is a test with vegetation for reference. The colors are a bit off but I’m very happy with the detail. With a final tweak, added farmlands and settlements it will look great.
Combining and fixing errors is tedious and boring; texturing is tedious and fun. Fixing render errors feels like working your way out of a hole, texturing is more like bringing life to a place. Once you have defined a set of textures and see the landscape rendered it is like magic, and I instantly start pouring over it dreaming up adventures and trying to figure out who lives there and what they are up to.
After having spent a few weeks in the weeds fixing errors it is good to know that I have learned a lot and a lot more terrain is already well under way. The downside is that fixing errors is where a lot of my time spent re-mapping Greyhawk again will be spent. I’m going to try and see if going from 50 to 100 feet or more per pixel will be the way forward to map Oerth in its entirety. Once I’ve done a few more areas I will make a comparison and see which might be the better way. The key bit is how well the data can be used to create more detailed area maps of key locations. Another consideration is that areas away from the Flanaess gives me a lot more freedom, it doesn’t matter if a river are a few miles one way or the other.
