Astro Art

[geckzilla]

I created a pretty cool simulated Earth using Blender and some freebie NASA images from the Blue Marble collection. I might make some more visualizations later too so I figured I'd start a thread about it. Feel free to join in if you make any art yourself.

The Earth model is a rather simple one consisting of five icospheres: 1 land, 1 cloud, and 3 atmosphere layers all nestled very close to one another. I was afraid Blender wouldn't be able to render the polygons accurately enough and that they would clip into one another because the land sphere has a radius of 6.378 units and the cloud layer is 6.379, which realistically places the cloud layer about 1 km above the surface. Turning off approximate shadows resulted in an accurate rendering without weird polygon edges showing up. The 3 atmosphere layers are what create the subtle continuation of the atmosphere above and below the terminator, which actually might not be subtle enough. When astronauts say the atmosphere is paper thin they really mean it. The bumpiness of the land is also probably a little too bumpy. Oh well, it was fun.

One thing which is very interesting to me is how much changing the colors of the Blue Marble image drastically improved the quality of the illustration. At the top you can see the original image as supplied from NASA Visible Earth and below are the adjustments I made, which were quite a bit more radical than I anticipated. Using some Apollo imagery for reference and just by remembering some photos taken by ISS astronauts I've noticed that the plant life that looks so green up close looks more muddy and brownish from far above. The sea ice is still pretty bad at literally pure white in color but you don't see much of it in the render.

Note that if you want to try it yourself, Blender is a free, open source program available from blender.org and the tutorial I used is here:
http://www.blenderguru.com/tutorials/cr ... tic-earth/

[Ann]

Beautiful, Geck.

Ann

[Chris Peterson]

I created a pretty cool simulated Earth using Blender and some freebie NASA images from the Blue Marble collection. I might make some more visualizations later too so I figured I'd start a thread about it. Feel free to join in if you make any art yourself.

The Earth model is a rather simple one consisting of five icospheres: 1 land, 1 cloud, and 3 atmosphere layers all nestled very close to one another. I was afraid Blender wouldn't be able to render the polygons accurately enough and that they would clip into one another because the land sphere has a radius of 6.378 units and the cloud layer is 6.379, which realistically places the cloud layer about 1 km above the surface. Turning off approximate shadows resulted in an accurate rendering without weird polygon edges showing up. The 3 atmosphere layers are what create the subtle continuation of the atmosphere above and below the terminator, which actually might not be subtle enough. When astronauts say the atmosphere is paper thin they really mean it. The bumpiness of the land is also probably a little too bumpy. Oh well, it was fun.

It does look like a fun project, and the results are excellent. I'd think that 1 km might be a little low for the cloud layer. Not sure doubling or tripling this would make any visible difference, although perhaps it might affect how shadows contribute.

[geckzilla]

It does look like a fun project, and the results are excellent. I'd think that 1 km might be a little low for the cloud layer. Not sure doubling or tripling this would make any visible difference, although perhaps it might affect how shadows contribute.

It does have a minor effect when rendering it at a very high resolution. See for yourself. I had it too high in an earlier version which I had guessed at instead of using real values (it was probably at ~5km) and it is very tempting to increase the height to a very unrealistic one because the cast shadows look very interesting. Some people have posted their examples in the comments below that tutorial and they almost all do this. It is interesting to compare what is perceived as realistic against reality.

Even already having some basic knowledge about the height of Earth's atmosphere it still surprises me just how thin it really is. To us, the clouds are so high and we have this expectation that they must appear to be floating above Earth when viewed from space but from a high orbit where we can see the entire disk of Earth, the cloud layer may as well be glued to the surface. The people on the ISS are in an enviable position at low Earth orbit so they can still see the layers but also get a fair look at how thin it is.

Last night I worked on getting a moon put together. It has its own challenges. The LRO data is very nice but it is difficult to separate the actual color of the moon from the shadows and highlights the craters create. I had to manually remove some edges using a combination of digital painting and some Photoshop filters. I also had to tediously align my final color map with the height map. A few pixels off here and there made the surface look blurred. I still don't think it's perfect but it's still pretty darn good looking.

The Moon is at an average distance from Earth in this picture. The camera is far away, nearly 2 lunar distances from Earth and it is looking telescopically at the simulated syzygy.

Is the moon dark enough?

[geckzilla]

Putting an ocean and life on Mars. Fun stuff.

http://www.geckzilla.com/astro/mars_ocean_render.jpg
http://www.geckzilla.com/astro/mars_oce ... render.jpg

Don't click the thumbnails below if you want to see bigger ones. Click the links above.

[rstevenson]

Lovely!

Rob

[geckzilla]

Thank you for the compliment, Rob.

Here are two versions of Earth: The left is close to real values for atmosphere thickness while the right is an exaggerated one showing how thick people imagine it to be. You see representations of Earth all the time with that blue fuzzy atmosphere extending far out into space.

For this one I have figured out a much better way of rendering light scattering. It turns out that spherical volumetric mapping is not that hard relatively easy. It would be cool if I could figure out how to do it with other shapes or if I could at least warp a sphere map into, say, things like cones. I might be able to model some planetary nebulas.

[geckzilla]

Trying to simulate sunset refraction. That went well enough but then I decided I needed to get some (greatly emphasized) dispersion going on and it didn't turn out quite how I thought it would. When the Sun sphere is near the top there is a blue rim on the bottom and red on the top. That doesn't seem right. I also had to set the index of refraction slightly higher for red and slightly lower for blue for it to end up like this. I think that might be the opposite of reality but it looks more like the "green flash" style sunset with it like this.
http://www.geckzilla.com/astro/sunset_dispersion.gif

[geckzilla]

Better one. Fixed the math to be in the right order. Added a thin atmosphere block to make the green flash actually green.
http://www.geckzilla.com/astro/sunset_dispersion2.gif

[rstevenson]

Very nice. I find it fascinating that we can analyse what's happening in a real sunset, using the mathematics of diffraction applied to our understanding of the atmosphere, then you can turn the math around to create a simulation of a sunset. There's some symmetry in that.

Rob

[Chris Peterson]

Very nice. I find it fascinating that we can analyse what's happening in a real sunset, using the mathematics of diffraction applied to our understanding of the atmosphere, then you can turn the math around to create a simulation of a sunset. There's some symmetry in that.

In fact, it's an example of the modeling concept at the heart of much modern science. We understand most of the basic principles, the things described by straightforward theories and mathematics. We're moving on to complex systems, that can only be modeled numerically, integrating multiple theories and allowing them to interact.

There is no "theory of sunsets". We can't take all the pieces and turn them into a closed equation that describes the visual phenomenon. But we can take all the pieces, put them into a model, and test the results against reality. If they match, it gives us confidence that we understand how the pieces interact.

The sunset model here is a fairly simple example, but it's not qualitatively different than the models we develop to understand climate, to understand n-body systems, to understand ecological systems, to understand how cells work.

[geckzilla]

Blender is very fun. Cycles is very simple compared to ray tracing renderers I have tried to use previously. But even though it is simple, Nodes allow you to insert as much complexity and as arbitrarily as you wish. Here is the Node setup for the refractive plane:
http://www.geckzilla.com/astro/sun_dispersion_node.png
It got pretty messy splitting the three colors but there's not much you can do about it since dispersion isn't something that's built in. I can tell you I was very happy that it has a built in blackbody converter so I never have to think even for a moment what color to make the sun. I just type 5778 into the blackbody converter, plug it into the sun's color value and forget about it. (It has very little effect beyond giving me the satisfaction that it's "right")

[geckzilla]

Thoughts meandering without direction eventually lead to the creation of this silly little trophy. I'm not sure what it's for or who you would give it to. This is a model of comet 67P that came off ESA's website. I even added jets.

[geckzilla]

Various attempts at rendering something akin to a light echo. #2 looks the coolest but is arguably the worst one. #3 has the same debris sphere as #2 but the light falloff is tweaked to be very quick. #4 has a diffuse debris sphere and the tweaked light. The renders are super noisy because the volumetric rendering is tanking my computer so I used a low sample size.

http://www.geckzilla.com/astro/light_echo.gif
http://www.geckzilla.com/astro/light_echo2.gif
http://www.geckzilla.com/astro/light_echo3.gif
http://www.geckzilla.com/astro/light_echo4.gif

[Nitpicker]

Various attempts at rendering something akin to a light echo. #2 looks the coolest but is arguably the worst one. #3 has the same debris sphere as #2 but the light falloff is tweaked to be very quick. #4 has a diffuse debris sphere and the tweaked light. The renders are super noisy because the volumetric rendering is tanking my computer so I used a low sample size.

http://www.geckzilla.com/astro/light_echo.gif
http://www.geckzilla.com/astro/light_echo2.gif
http://www.geckzilla.com/astro/light_echo3.gif
http://www.geckzilla.com/astro/light_echo4.gif

In #3, assuming the debris is spread within a sphere surrounding the star, I don't understand why the illuminated echo diameter doesn't appear to contract in the latter half. It doesn't look right, or I don't understand what I'm looking at.

[geckzilla]

#3 and #2 have the exact same debris field, which is to say it has a spherical fog which is thicker closer to the star and decreases in density on its way out and there are several "blobs" of thicker material randomly generated in a sphere of a smaller radius around the star. I don't know why you would expect the diameter to contract. It doesn't contract in any of them. The only difference between #3 and #2 is that the light rays are much shorter in #3 and hence only a small shell is illuminated instead of the whole scene.

It's very disorienting so I wouldn't be surprised if you are disoriented.

Btw, here is yet another attempt:
http://www.geckzilla.com/astro/light_echo5.gif

FYI, because it's not clear unless I say so, the lights never fade away or turn off. They simply run into an empty space and have nothing to illuminate.

[rstevenson]

Those all seem to be pretty good illustrations of a light echo. Because they're happening so fast in your GIFs, it's easy for our minds to hold onto the shapes of what the light is running into as they are revealed. Shame we can't do that with astronomical light echoes. I guess we will some day have enough imagery to create a good animation of a real light echo from beginning to end. For now we have to interpolate and extrapolate from the small time-slices we can see.

Rob

[Nitpicker]

I don't know why you would expect the diameter to contract.

Because the "pointy bit" of the ever-expanding ellipsoid/paraboloid light echoes, moves further away from the observer and the star, at a speed of c/2. By the time it gets to the furthest part of the debris sphere at the very back, it should only be illuminating a small spot. (I described this in more detail in the infamous light echo topic, and even provided an image of the contours.)

[geckzilla]

You'll have to elaborate on "pointy bit" some more because this is a perfect sphere.

[Nitpicker]

You'll have to elaborate on "pointy bit" some more because this is a perfect sphere.

The apex or vertex of each ellipsoid shell. The debris is in a perfect sphere. But the light echo forms an ellipsoid in space at each time step.

[geckzilla]

It's just a sphere in my animation. I don't understand why you think the light echo should be anything but a sphere.

[Nitpicker]

It's just a sphere in my animation. I don't understand why you think the light echo should be anything but a sphere.

It's not just what I think it should be. It's what it would be. The pulse of light from the star propagates outwards as a sphere as viewed from the star, but what we observe (from outside the sphere, due to the extra distance that the reflected light has to travel compared with direct starlight) are illuminated ellipsoidal shells of debris. Assuming a debris sphere of 3 light years radius, an outside observer would see the front most part of the sphere 6 years before rear most part. A uniform debris field would appear as a disc which expands from nothing to the radius of the sphere, then contracts back down to nothing.

[geckzilla]

Dunno how to do that.

[geckzilla]

If I totally hack the shape into what I think you mean, I get something like this:

Maths needed. Please see manager for available positions.

I can hit the render button tonight before I go to sleep but the little preview of this shape propagating through the dust already doesn't look quite like what you might expect. Conceptually I can think of how to hack a better sphere into such a way that the front vertex is twice as far ahead of the center and the rear vertex is half as far out but I'm having difficulty telling Blender exactly how to model that since its proportional editing tool isn't letting me define the parameters very well.

[Nitpicker]

That shape doesn't look right to me.

The parameters for the ellipsoids are tabulated here:
http://asterisk.apod.com/viewtopic.php? ... 1&start=38

I'm afraid I don't know enough about your simulation software, to comment on what parameters it needs.