Physically Based Earth

2020-10-28 Update: Extended Grand Canyon

Lo-res terrain increased to 150,000km[SUP]2[/SUP]. Hi-res terrain increased to 615km[SUP]2[/SUP].

2020-10-19 Update: ISS and the Grand Canyon

To give an indication of scale I have added a model of the ISS at an altitude of 409km and positioned it above a section of the Grand Canyon. [HR][/HR]

Hello UE4 Users!

We create 3D models of real-world buildings to help telecoms engineers design roof-top sites. Some of our customers want the sites placed in geographically accurate terrain. Because this terrain is used to assess point-to-point microwave transmission we have an unusual “curved earth” system. The improved Sun/Sky functionality in UE4 4.25 encouraged us to see what our terrain would look like if we positioned it on an representation of the Earth at 1:1 scale.

There is nothing particularly clever about our system. There are no fancy shaders or materials. After a fair bit of number crunching we create a series of meshes that fit together to give detailed terrain close to the point of interest but become gradually less detailed as you zoom out to see the entire planet.

The main feature that makes this useful from a GIS / mapping / engineering perspective is that our Earth is based on the WGS84 geoid, i.e. our Earth is ellipsoid, not spherical. This means that if you have (or can get) data with a WGS84 latitude and longitude it is possible to visualise this data on a physically based Earth without worrying about map projections.

We’re just not sure if this level of accuracy is useful or needed outside our rather specialist area and that’s why we are opening it up to the UE4 development community for feedback. We are painfully aware that our terrain is embarassingly dull. We have used just two materials: a green one for the land and a blue one for the ocean.

If anyone would like some sample data to play with please send me a descritive name and the latitude and longitude (decimal degrees please!) and I will create a bare-bones UE4 4.25 project for you, with detailed terrain centred on your location and a much lower resolution Earth.

The images on this page are of a set of remote islands in the South Atlantic called Tristan da Cunha, lat/lon -37.25 -12.25.

One thing to note that is that if you turn off the ocean surface you can see the sea bed. If anyone has a model of the RMS Titanic it would be great to see that in it’s correct geographic location!

More info on our development blog.

Cheers,
Derek Hunter
Technical Director
dhp11 Ltd
www.dhp11.com [HR][/HR]Tristan da Cunha is invisible in the centre of this image.

](filedata/fetch?id=1819138&d=1601995934)

You can just about see it now.

](filedata/fetch?id=1819125&d=1601995934)

Becoming clearer. Nightingale Island is visible.

](filedata/fetch?id=1819127&d=1601995934)

More detail on Nightingale Island.

](filedata/fetch?id=1819130&d=1601995934)

Ocean Surface “On”.

](filedata/fetch?id=1819128&d=1601995935)

If you turn the Ocean surface off you can see the sea-bed. The conical shape of Tristan da Cunha, an active volcano, is clear.

](filedata/fetch?id=1819140&d=1601995934)

To give an idea of scale I have added a 1m cylinder.

](filedata/fetch?id=1819131&d=1601995935)

Very interesting project. It will be interesting to see what can be done to the terrain to get textures etc added


This image shows three types of mesh:

  1. An ellipsoid World Mesh with vertices every 0.5 degrees and a diameter of approximately 12,720km. The semi-major axis is 6378137.0m and the semi-minor axis is 6356752.3141m.
  2. 15 x M30 tiles. These are 0.5 degrees across (0.5 degrees = 30 minutes, hence M30). Vertices are 15" (15 arc-seconds) apart. Each tile has 121 x 121 = 14,641 vertices.
  3. 743 x AS30 tiles. These are 30 arc-seconds across (hence AS30). Vertices are 1" (1 arc-second apart). Each AS30 tile has 31 x 31 = 961 vertices.