The idea of Amherst WorldView is not new, its first iteration was called Amherst SndBx (Sandbox). Which came from the need to learn and understanding geographic features and how it shapes the development of infrastructure. This quest on finding this answer comes from a project I was working on late spring / early summer of 2015. A 3D rendering of a fictional town I called Mistic.
The town would be located on the Western side of the state of Massachusetts and would be a center of planned stories would take place. This town would be the medium to introduce the audience to more explicit themes and stories. Once the first renders were created I noticed a large unrealistic factor; which was that everything was placed flat within the axis. This made the area look unrealistic. I tried fixing up a bit but it’d still look unrealistic to me. I noticed that I needed to learn more about topography, and how we as humans shape it to our desire with the occasional compromise.
I then started working on smaller scale render in which the topography of the terrain was the main focus. I did physical surveys of areas across western Massachusetts that I wanted to recreate. I used pictures as references as well. The end results were satisfactory but I wanted to experience a more diverse settings.
To do so I had to choose a location in which I was able to surveys routinely and close by, I selected the Center of Amherst as my starting point. The idea of Amherst SndBx comes from a childhood dream of being able to explore my local community freely, virtually in a sandbox-like mechanic. This came from my early exposure to Open-world Sandbox games. Which excited joy on just exploring the map to the point I had grew acquainted to the landmarks of map of places I’ve never visited before. What I would want to come out of the Amherst WorldView project, is to bring a sense of exploration to people and act as a learning venue of the history and culture of an area they’ve never visited.
The Amherst Sandbox was created in blender within Blender using the Blender Game Engine (BGE). The desired size of the map was around 3 square miles. One of the initial issues I encountered was on how to obtain all of this measurements as I wanted the world to be on a 1:1 scale as much as possible. I found a useful tool in the Amherst’s Town Website. A GIS-like map viewer in which you could draw your measurements.
This viewer turned out to be one of the most critical tools in the development of both Amherst SndBx and Amherst WorldView. To the point that if the site was down I couldn’t do much as I would be working blind. The process that I used to transform this 2D representation of the world was a learning process. I started by drawing the measurements of manmade structures under the assumption that manmade structures are:
- Straight lines within a rectilinear shape.
- Corners in general conditions are assumed to be 90 degrees.
This process is done by placing cubes as markers of key places were a wall, sidewalk or road would be placed. This process grows in complexity as more measurements are added essentially making all markers connected to each other. The terrain height is done by placing planes on their respective heights on the Z axis. The zero-height is set as zero and the following planes are set accordingly to the addition or subtraction of their real life height.
During the creation and testing of Amherst SndBx I learned the limit of my skills and of the software I was using. The BGE is a really outdated engine that doesn’t receive much support and had limitations like materials or a variety of export options. Graphic limitations which would required shadows and light effects to be baked on the texture through the Cycles engine for a realistic result. This process was cumbersome and I avoided it during the testing of the main map. Programming wise it used python and blocks to represent inputs or events/actions. My skills where also low. I made rookie mistakes like creating a sidewalk as multiple individual blocks rather than as a single piece textured or created trees as hundred of individual pieces. This and my lack of knowledge of the workings of LODs in BGE made the project highly inefficient memory wise. I ended the Amherst SndBx project Fall of 2015 with inconclusive results.
I dedicated the following months in researching on how do artists conduct surveys within the context of game design. What artists did like taking pictures, doing research and how they used this information to create a virtual world. I decided to try recreating a new environment, this time I picked the Umass Amherst Campus. I did it because the campus has a diverse architecture on different geographic locations which they could be potentially reused for future projects. I was also going to be on the campus for multiple years as a transferring sophomore so I had the opportunity and time to do research.
On late October of 2015 I decided to start a new project this time it’d concentrate on the Umass campus. The startling place that I choose were all the axis meet was the Haigis Mall as it could be considered a central place of the campus. Because z height is zero and the grade of the staring location is around 130 feet. All following measurements would be the product of a subtraction or addition to the staring place.
The process of laying down the measurements one by one is a relatively slow process. There’s a margin of error this error is compounded from the tracing done by the maps cartographer and my own. Making the margin or error as low as 1-5 feet a rather acceptable margin on my own standards in dense areas while a 5-10 feet margin on bigger/open space areas. This is due to the lack of nearby structures that would aid with measurements. This can lead to “stretching” or “warping” which would make the map imprecise when compared to height markers. To solve this I placed the items on the average of the difference between the two measurements. The nature of this process with its human error and approximation makes this project more of an artistic representation of the area rather than a direct copy of the area.
The usage of this technique hasn’t been limited to just this project. I have applied to scene tests and other animations to a less intense extent. During the 2018 Spring semester I decided to work on the animation Planet 9-x that used different landscapes as it’s set pieces. Creating a detailed landscape for each scene by hand would’ve been impossible with the time allocated for its creation.
I found a perfect tool for this project a website that generates satellite based height maps for a city simulator game. The website terrain.party enables you to choose virtually any place in the world and generates satellite imagery from both the ASTER & SRTM satellites. Which facilitates the creation of the terrain, the images come a bit dark which can be fixed in any image manipulation software and then placed into blender. By placing a 18km by 18km plane, adding at least 8 subdivisions. Then placing a displacement modifier with the image attached will create a rough idea of how the terrain would look like. You can add more subdivision to create a more finished look, but be aware of memory constraints. The same can be do with a subsurface modifier, it is still recommend doing a pass by hand with the modeling tool to soften any edges.
I see using satellite imagery as a good place to start a new scene, now any area of the world is at within reach in terms of creation. From the deserts of Egypt to the Green Mountains of Vermont all of their topographic data is available for imagination to come about. I will still use hand method in the Amherst WorldView Project but in future projects this tool will cut time on production and help focusing on other details like Buildings, Characters or other props.
As of February of 2019 around 68% of the campus has been measured and placed on the map. Just finished the “Grasslands” a large area of woods and hills that are located North East on campus. It took around 4 months to measure and model, this area was considered the “big piece” which would enable more accurate measurements on the outskirts of the map after its creation. After February 20 the process of creating the buildings will begin and more testing within the engine will follow. The next update will be by the end of this year.