Whole Earth View

[FarGetaNik]

It seems then I roughly calibrated my textures correctly then. BMNG seems to feature forests that are a bit too dark, this can be easily corrected by boosting the dark values up a bit with water still remaining in its albedo range. My lunar texture seems to have a slightly higher contrast than supposed.

[scalbers]

Sounds good. As a quick followup on the blue appearance of my renderings, there is a second factor at play in addition to the white point I mentioned earlier. This is ozone. It absorbs about 3 percent of the red light and transmits almost all the blue. Since the light path travels twice in the atmosphere, the red in the land is attenuated relative to the blue by about 6%.

[FarGetaNik]

Oh I also thought that a bit of absorption could make my Earth rendereing more blueish. I assumed the Mie scattering in Celestia is just wrong with its brown hue or that I am using an inaccurate Rayleigh intenstity. Who would have thought ozone does the trick. Thanks for pointing that out.

[scalbers]

Yes, it's often underappreciated what ozone does in the sky. The effect is kind of subtle during mid-day and becomes a major effect in twilight areas.

Here's what I have so far for a calibrated BMNG albedo image for March. It is gamma corrected.

http://stevealbers.net/albers/allsky/bm/bm_albedo_03.png

[FarGetaNik]

I'll work on a setting for the Celestia absorption parameter to recreate this. Though it is really unpredictable, if I was to set the absorption of red light to 0.06 Earth probably will be a cyan sphere... is only red light absorbed or also a bit of green?

The texture looks quite nice. Contrast is unfamiliarly low (I suppose you used a gamma factor of 2.2?) but I guess this is actually close to reality. The BMNG textures seemed a bit extreme to me anyways. The colors are less saturated, but isn't the ocean a bit too bright?

I was using the July data to get rid of most of the snow. I wanted to blend it with Juanary data on the southern hemisphre to create an albedo map, but I am not able to handle such big textures (86k!). I managed to create a 32k virtual texture though.

[scalbers]

OK and good point about green for ozone, here are my optical thicknesses for ozone and the wavelengths:

real ext_o(nc) ! od per airmass
data ext_o /.037,.029,.001/ ! interp from gflash.bas (300DU)

! real wa(nc) ! wavelength (um)
! data wa /.615,.546,.450/

Yes I used a gamma of 2.2, though at the very low end a gamma correction should deviate from this. I will try to update my images with this taken into account. Perhaps this aspect of the gamma correction will dim the ocean. We do see more of the sedimentation in certain areas that seems reasonable.

Here are July and January BMNGs:
http://stevealbers.net/albers/allsky/bm/bm_albedo_07.png
http://stevealbers.net/albers/allsky/bm/bm_albedo_01.png

[FarGetaNik]

Thanks so much for the texture! I'm a bit confused, optical thickness is different from absorption. I assumed you posted the absoption values and scaled them by 0.2 to get a set of parameters for use in Celestia that give nice results. Using your calibrated BMNG July texture, here my view of Earth:

The code I'm using:

Code: Select all

   Atmosphere
   {
      Height      100
      CloudHeight   8
      CloudSpeed   65
      CloudMap   "earth-clouds.*"
      CloudShadowDepth 0.9

      Mie      0.015
      Rayleigh   [ 0.0016 0.0029 0.0060 ]
      MieScaleHeight   8.5
      Absorption   [ 0.0074 0.0058 0.0002 ]   # 0.037, 0.029, 0.001
   }


[scalbers]

Sounds good - we're getting there! I've now updated the gamma corrections to do the low end better at the links in post #26, so hopefully the oceans and dark forests will look better now. The gamma correction presently is the minimum of albedo^0.45 and albedo*4.5.

To clarify things, an optical depth of .037 is about 3.6% one way absorption and 7.1% two way absorption for red. For green, an optical depth of .029 is about 2.9% one way absorption and 5.6% two way absorption. So the conversion from optical depth to two-way absorption should be approximately a factor of 2.

Absorption [ 0.071 0.056 0.002 ] # 0.037, 0.029, 0.001

[FarGetaNik]

To clarify things, an optical depth of .037 is about 3.6% one way absorption and 7.1% two way absorption for red. For green, an optical depth of .029 is about 2.9% one way absorption and 5.6% two way absorption. So the conversion from optical depth to two-way absorption should be approximately a factor of 2.

I thought optical depth is proportional to transmission, not absoption. But regardless, I have no idea how Celestia's absoption parameter works, so it is a subjective decision anyways.

Looking great, the oceans are now roughly at a brightness I also calculated, the forests are greener, but temperate areas brighter. Great balance!

Added after 49 minutes 3 seconds:
Here a render of Earth using the new textures - same atmosphere settings. While these setting seem to improve the space view of Earth, the ground view is ruined. Dawn is now pretty much grey - increasing Rayleigh scattering makes matters worse.

This is my take on a mostly snow-free Earth combining the July and January data:

https://drive.google.com/open?id=0B14NSu0PnA5xTUQ5RTgtX3l2aHc

[scalbers]

Looks good from space. We can note that sea ice isn't in the BMNG in case you are over the north pole. It could be added I suppose.

At ground view ozone colors much more strongly at twilight, so that would ideally be coded in somehow. WIthout ozone early twilights would actually be closer to gray.

[FarGetaNik]

Hm maybe the Mie effect in Celestia is not gamma-corrected after all... or the fact it doesn't distinguish between Rayleigh scale height and Mie scale height results in these poor renders. I feel the Mie effect is too much already, but the haze isn't bright enough. It compares to uncorrected space images though.

So apart from the Rayleigh-red at dawn, the blueish sky after sunset is due to ozone? Interesting.

Sea ice is part of a realistic look I suppose... is there an easy way to add it? The only BMNG texture with sea ice is this one: http://eoimages.gsfc.nasa.gov/images/imagerecords/57000/57730/land_ocean_ice_2048.png. The ocean also has some structure there, but the south polar ice looks wrong to me.

[scalbers]

Yes, the "skyrenderer" website has a good illustration of the effect of ozone at sunset:

http://skyrenderer.blogspot.com/2013/05/the-importance-of-ozone.html

The sea ice in the image you posted (thanks) may be taken from July (summer) data, when Antarctic sea ice is near maximum. Now I've merged the sea ice into the textures that were set up previously, in this web directory: http://stevealbers.net/albers/allsky/bm/

If the sea ice is too bright I can reduce it a little.

[FarGetaNik]

Yes, the "skyrenderer" website has a good illustration of the effect of ozone at sunset:

There is an evident difference. Essential to consider for Earth renders. So much to hope for in Celestia 1.7...

If the sea ice is too bright I can reduce it a little.

Thanks for your efforts. Adjusting the brighness to the ice of continental antarctica might be appropriate... Most of the arctic ice is bourdered by land, but the antarctic sea ice is all over the place. Honestly, I think I'll use a version without antarctic sea ice, just arctic. Too bad the resolution of this map is not as high as the rest of the BMNG set.

[scalbers]

Sounds like you have a good strategy to approximate the minimum sea ice extent in both hemispheres, since Antarctic sea ice can get to a pretty small extent in their melt season. Another thing to hope for in Celestia 1.7 would be ability to use a seasonal BMNG image for each month.

Continental Antarctica seems like a good reference. On the other hand I can note that the Greenland Ice Sheet has a brighter count value compared with Antarctica. Perhaps something just a little bit brighter than Antarctica would work?

[FarGetaNik]

I also thought of a seasonal change in textures. For now this is probably only possible with scripts, I have no idea how advanced they are already. Having this coded into Celestia itself certainly would be neat.

I assumed the Greenland ice just was just off, because it is pretty much white without significant texture. I wonder how bright the ice really is, because it should depend on geometry and we don't know under which conditions the data was aquired. Anyways, the sea ice should just match the brightness of nearby continental ice. This is the cas for the standard BMNG textures, where there is still a bit of sea ice left at antarctica. The sea ice seems to be a crude overlay without texture.

[scalbers]

I think Greenland ice albedo is about 90% in winter and drops to 80% (or even less) in summer. If my spline function is right at the high albedo values, then the BMNG values are around 87%. Since most of the visible season it would be lower perhaps I can show it at about 80%, closer to the value for Antarctic land. I'm thus uploading an updated version to my web link (see post #32) as I write this.

Meanwhile, here is what I have so far for a crescent Earth with clouds, to compare with the Rosetta image posted earlier (post #8). I'm still checking on the sun glint.

An animation showing various phases is here: http://stevealbers.net/outerspace/animated_phases.gif

[scalbers]

And here's a hopefully improved calculation of the sun glint. The animation just above is also being updated.

[FarGetaNik]

This already looks very realistic. Good job! A must have in next Celestia... no way to do this right now.

The textures you updated are also improved. The difference at greenland is subtle, but the continental ice and sea ice match in brightness.

[scalbers]

Thanks FarGetaNik. For reference the core of the sun glint for a fully lit Earth corresponds to a reflectance of about 20% (in addition to what was there before). This increases for a crescent near the limb.

[Alexell]

scalbers, Great work. Realistic and very beautiful. Can you realize this in Celestia?