After having tried to map using a “Bottom Up” approach, with a 5 feet resolution very detailed map covering nearly a hundred square miles at a time, found out that they maps looks great, but the approach scaled badly. Lakes and rivers that covered multiple areas where nearly impossible to get right, and those subtle elevation differences that stretch out over long distances impossible to get right. Add to these problems with tiling data import problems, it forced me to rethink my approach.
After a couple of years of experimenting, testing, and better tools with bug fixes, I have finally worked out a top-down approach to replace my previous bottom up way of doing things. It requires no tiling, well suited to cover huge areas and regions can be seamlessly merged. The resolution I settled with is roughly 50 feet per pixel. Not good enough for battle maps, settlements and the like, but well suited to give great overview maps that scale all the way up to planetary size. With the new workflow each 16K area covers 24,000 square miles.
The data from this set of maps can be scaled up, so it will be fairly straight forward to take an area covering a hundred square miles or so, scale it up to "Porta-Potty Scale" and at the detail as a separate project and since they are based on the large lower res data they will match. This means a first set of overview maps covering whole countries and more and then zoom in and make more detailed maps of areas of special interest.
It has taken me a decade of World Machine and 3D terrain making to get to this point. With several thousands of hours honing my skills, better computers and improved versions of the software, it is my pleasure to present a first sneak peek at my vision of what the world of Greyhawk looks like. This is till a work in progress with nothing final yet, but all the hard bits are solved. I’ve managed to render well over a million square miles of terrain covering the Shield Lands, White Plume Mountain, parts of the Bandit Kingdoms with Rift Canyon, half of Furyondy and the Horned Lands as well.
The work will be divided into several steps:
1. Terrain creation and rendering
2. Area merging and correction
3. Texturing and rendering final data
The first half decade doing this I spend figuring out how to try and get it to look the way I wanted, the next half decade I spent simplifying the process enough that it became doable. Thanks to both my increased skills and more computing power, I can now crank out an impressive amount of terrain of a quality that is good enough. To render an 16K area have gone from around 20 hours down to around 3 hours. I upgraded my old desktop with a new 12 core AMD CPU and a Nvidia Quadro GPU so I can render and work at the same time which means I can do at least two iterations of an area per day, and another two overnight.
In this first step I create two height maps, one with water surfaces, and one without. A special height map depicting water depth and masks for lakes, rivers, ocean, an elevation color texture and a shadow map so I can look at things in Photoshop if needed. The height maps are 32bit EXR files for maximum precision. For masks, shadow and texture files I’m using 8bit PNG’s which are good enough, and here storage needs are considered. When you start to ram up things, the number of files pile up and storage need can skyrocket if you use files with too much precision. Several tens of Terabytes will be needed for this project while working on it, thankfully a lot of the work files can be deleted when final data is rendered for each area. This will hopefully keep the overall amount of data small enough to be possible to share with the world. I will also keep the World Machine files which are like the source code for recreating things again if needed. They are only about 100 KB each so they can be kept.
I’ve used World Machine 3 for this step, but the latest update of WM4 made the new version stable and good enough to be used which pleases me immensely. WM4 has a whole slew of improvements to erosion and rendering which will make the terrain better, so I’m really glad to be able to start using it at all the steps of the workflow. I will also use Gaea for some areas and then import the data into WM, this so I can make use of Gaea’s special erosion and terrain creation tricks. This will mean more variety and higher quality.
This workflow scales well so I can see this project spanning the entire globe. With QGIS and georeferencing paired with this mapping even polar regions should work seamlessly. In fact mapping other areas of the planet is easier, the less that is known of an area, the easier it becomes. The reason for this is that things like coastlines and river placements don’t need to be so precise if there are no previous sources to match. Algorithms are great for producing fine details, but it can be very hard to order it to place the details exactly whare you need them to be. So, mapping a large area of Western Oerik will be way easier than the Flanaess, and my way to tackling problems is to try and go the most difficult bits first, knowing that if I can get that done right the rest will be easier.
Merging areas together is the key bit to getting right, and where countless of my attempts have failed before. The fact that lake water surfaces must be at the same height even when they span over more than a single area, and rivers must flow downhill the same way and be of a similar width when they flow across area, has caused me countless headaches.
Over time I have figured out how much overlap is needed for different types of terrain, and where to split things into different regions. Main rivers and highlands are most suited features for splits, and with plenty of overlap you hopefully have enough to play with in the merging process. Key here is hot managing data and masks properly, and it was when I realized you can create a common mask that can then be used for multiple types of data to simplify the work needed as well as improving the result.
Even after doing this for a decade, it is still a great feeling when you see things work the way you want them too. It also never seizes to surprise me how long a simple solution evades me, why the heck couldn’t I figure out how to do it earlier. It is often the easy straight forward solutions that are the hardest for me to see, but I guess I must be happy I managed to figure it out eventually, even if it takes a long time.
This phase is also where I try to make sure what I have done is good enough, and one trick is to place the camera low to see if the landscape looks believable from a ground perspective. Below is a view of the Shield Lands that shows large part of it stretching out a hundred miles or more, and the distant blue is Nyr Dyv and Walworth Isle in it.
Here is a view from the eastern slopes of the White Plume Mountain looking southeast, with the Yellowflow River valley beneath the foothills.
A part of the interior of Shield Lands with the White Plume foothills to the left and Walworth Isle to the right in the far distance.
I’ve merged the Shield Lands area in the north and west, left to do is east, which I will hopefully get done tomorrow, then it is time for the next step.
I haven’t taken any of the new Shield Land terrain through this phase yet, still tweaking the details as I write this. Texturing is both the most fun, but it can also be the most frustrating part. Fun because it is so cool to get a first glimpse at the near final result. I’m not worried about this bit, with heightmaps that are detailed enough there are few pitfalls to get a good result, just a lot of trial and error. Another good thing about texturing is that a lot of post processing is possible, and often necessary to do in Photoshop, so with enough work you get good results.
The key bit wit this step is to figure out what makes the terrain convincing at the resolution it has. Here is an example of my Critwall map down sampled to 10% of its original size.
This is close to what I want the quality of the new large area maps to be. Detailed enough to get a basic idea of the landscape, and larger settlements. Not enough to see houses and roads properly but hinting at all the col stuff that is there. I will need to experiment with how to best depict forests in the form of bumps to create the right chawing, and color variations. On this down-sampled image, shading might be a bit too harsh. The good ting is I can adjust that more on the new maps done the whole way at the same resolution. The same goes for building height, I need to test that too to see what gives the best result.
Compared to my now over 20 years old Bryce based maps this will be a fundamental leap forward. The Midgard maps I have done for Kobold Press represent a kind of halfway between the two, not in results, but in technology used and approach. For my Kobold Press commissions, due to time and other constraints I had to cut some corners which meant that true 3D maps with proper rivers and lakes had to be abandoned for what could be done in time that could do the job. My Midgard maps are something I will always be proud of, but now the time has come to leap ahead again, this time for Greyhawk and Oerth. My goal with this project is to set a new benchmark for fantasy cartography, to prove what can be done with dedication, skills and most importantly, the support of you guys. Thank you so much for making this come true!
A technology deep dive with World Machine files will come soon or top tier members.
