Procedural Generation of Physical Geography for Palaeogeographic Reconstruction

[h_trismegistus]

Hi Everyone,

New here - I am a life long map enthusiast. I began by collecting every map I could find in my grandparents' National Geographic collection starting from the 30s, and today I work in GIS and Data Visualization and Web Development, doing interactive web-based geospatial visualizations and the like. However, another one of my lifelong interests is Geology, and I am looking to transition my career more directly into the Earth Sciences, and I am specifically interested in Plate Tectonics History and Palaeogeographic Reconstruction.

I have started a project to create highly detailed palaeogeographic maps based on Plate Tectonics reconstructions. I don't know if anyone is familiar with the history of this endeavor, but proper attempts at reconstruction of palaeogeography go back about 30 years, and since then many great tools have been developed, and the amount of data available for reconstruction is ever expanding. There are a number of good reconstructions out there, specifically Christopher Scotese's palaeo map animations, and Ron Blakey's physical geographic 'art' maps of palaeogeography, but Scotese's maps, while rigorous in their scientific underpinning, are merely diagrammatic in style and lack detail, while Blakey's maps are highly detailed, but essentially arbitrary artistic products - he basically takes the plate tectonics reconstructions that people like Scotese does and photoshops in, using hundreds and hundreds of layers, modern-day geomorphological features onto the palaeo geometry of the terrestrial land mass and ocean floor.

My project is to use the data available in the plate tectonics models, which simulate geological processes such as orogenesis, rifting, seafloor spreading, subduction, etc as well as the host of geological and palaeoclimate data out there, to procedurally generate physical detail on to the plate tectonics reconstructions to generate maps with a level of physical detail that approaches Ron Blakey's maps, but that are based on actual data and physical laws as opposed to artistic license, modeled by computer processes to simulate deformation, erosion, river incision, subsidence, uplift, orographic preciptation, climate, etc.

I have a lot of experience in programming for the web (full stack development) and also for data visualization (r, d3, javascript), but this project will be new territory for me, and basically I am looking for help on where to begin. I would like to begin by conducting several experiments and trying to develop models and ways to simulate geological processes, but I really have no idea where to begin. So far I have been inspired actually by the plethora of Fantasy Map generators online, since a Palaeogeographic Reconstruction of the world in geological history is essentially a kind of fantasy map, except that it has real world data as inputs.

Perhaps I should start by asking a more specific question to help me get started: Let's say I have a geometry of shorelines as a vector shape file, and perhaps another vector shape file that traces out the deformation fronts of mountain ranges. How could I generate a realistic looking physical geography for the land masses? I have the elevation model - I mean in terms of graphics programming, how do i go from the elevation and basic geometry to a rendered physical map with rivers, erosion, hillshading, etc - I will figure out the physics, crustal and lithospheric properties, isostatic adjustment, etc. Though I suppose I could use some direction in how to filter the basic elevation model (based on uplift, crustal thickness and isostasy) to procedurally generate realistic mountain terrain taking into account erosion, river incision, orographic precipitation, etc. Really I just need to know what tools I should be using and what are the essential techniques, and I can figure out the way to model it.

Thanks for your help, and if anyone is interested in this project, please let me know.

[Redrobes]

I'll keep an eye on your progress. I wrote a geomorphological erosion app and some others, like say, Waldronate, also write apps which do erosion and other geological processes. I wont comment too much at this stage as id like to see more where you are heading. But I would say that I think its hard to go from shorelines to 3D terrain. Its very easy to go from 3D to shorelines however. So I think you may have to model your plate tectonics in 3D and do your simulation like that and then export maps in 2D from them rather than to do it in 2D shapes to create 3D elevation models. But I am up for seeing new ways of doing things.

[h_trismegistus]

I'll keep an eye on your progress. I wrote a geomorphological erosion app and some others, like say, Waldronate, also write apps which do erosion and other geological processes. I wont comment too much at this stage as id like to see more where you are heading. But I would say that I think its hard to go from shorelines to 3D terrain. Its very easy to go from 3D to shorelines however. So I think you may have to model your plate tectonics in 3D and do your simulation like that and then export maps in 2D from them rather than to do it in 2D shapes to create 3D elevation models. But I am up for seeing new ways of doing things.

Ok, yeah I only mentioned shorelines as an example of one piece of what I need to do, I will have collisional and extensional information as well and as I mentioned a model for uplift of mountains and isostasy. For the Mesozoic and beyond there is additionally dynamic topography but the jury is out on the ultimate effect of that relative to isostasy (you're talking maybe 500m of dynamic topography max in either direction and often quite less, compared to say 5000m isostatic uplift of thickened crust in say the Andean Altiplano or Tibetan Plateau. And what I'm not modeling from plate tectonics movements, there is a load of geological information out there for palaeo mountain ranges.

So the 3D component will be there. Basically I am going to have a grid of points associated with individual pieces of the earth's crust, which at any given point in time will also have a z-dimension which changes over time. A digital elevation model is essentially 3D, since you have x-y coordinates and then the height or 'color' is the third dimension. In my case, it will be a '4D' DEM.

I am only going to go back as far as perhaps 2 Ga, because there is no evidence plate tectonics existed in the way that it does further than that, and some people believe it only started in earnest in the late Neoproterozoic. I don't want to guess at models for the Archean, for which there is very little geological data anyway. Even 2 Ga may be overambitious, but I have a personal interest in Mesoproterozoic North America, so I am going to do my best.

Because erosion and the 'form' of a given landform changes so rapidly in comparison to the long march of geological time, it's unrealistic to produce a 'real-time' physical geography that changes at a realistic rate compared to say, plate motion and deformation. (Imagine a 10-minute animation of plate tectonics animations starting from say 2 Ga, the modern form of many mountain ranges, for example, has only been such since the Pleistocene and the last Glacial Maximum, which is approximately just 0.1% of the whole animation representing 2 Ga, or 6/10 of a second! If mountains had to be sculpted continuously every half second, it would be a very distracting animation to say the least, although it may be pretty interesting to see the effect). So I will basically be exaggerating the duration of the form of a given landscape in order to showcase the bigger picture of tectonic history and the quality of the land over a longer duration of time. (same goes for climate, which if I was incorporating realistic changes by milankovitch cycles, would be changing faster than the frame rate of the animation). And rivers for instance change even more rapidly - if their change was scaled realistically relative to geological time, they would but be an ephermal blur on the landscape, a vibrating, dendritic hallucination.

To put it simply, I am going to be exaggerating the duration of landforms relative to geological time, for the sake of clarity, so I suppose the physical features will be mostly symbolic than purely simulated, that is it will communicate something more like 'there were river systems here going from the NE to the SW for ~20 Ma', or 'a mountain range uplifted and then was peneplained over the course of 60 Ma' instead of 'this is a simulation of what the area looked like this specific time'. I am really just interested in telling the bigger picture story, but with realistic looking physical geography.

Thinking about it, though, if I develop these land-sculpting processes in such a way that I can control the rate at which they occur with any given run of the model, it would be interesting to render it both at the 'big picture' rate that I describe above as well as a 'realistic' rate just to see what happens, maybe the earth will seem like it is breathing or decaying like a timelapse video of decomposing fruit. :p