I have been attempting to make computer generated images of the Martian atmosphere, both as seen from the surface and from space. To check the results I have been looking for spacecraft images to use as ground truth. I have found lots of images - by far the best ones I have found are from UMSF in this thread: http://www.unmannedspaceflight.com/index.php?showtopic=3324

However, I'm always looking for more ;-). So if anyone knows of more and/or better images I'm interested in them. What would be best are mosaics showing the sky from the horizon (with the horizon/surface visible) and towards the zenith.

The sky varies a lot because of variable amount of dust but the general impression I get is that the sky is bright near the horizon (usually brighter than the surface) but gets much darker higher in the sky. There is probably a fairly large, bright area in the sky near the sun, possibly less reddish (lower R/B ratio) than parts of the sky farther from the sun.

I'm already getting fairly interesting results, this one has a field of view of 90 degrees:

Click to view attachment
(needless to say this one is 'overexposed' near the horizon; dynamic range is sometimes a problem)

The problem is that even though this may not be bad the limb currently appears far too bright as seen from space :

Click to view attachment

This shows that my atmospheric model is erroneous in some way - I suspect that as seen from the surface the Martian sky is darker high above the horizon than I have been assuming.

That first image is surprisingly (and pleasingly) similar to what Dreamscape ( a terrain and sky plugin for 3ds max) produces when I'm rendering my HiRISE DEM flyovers.

I expect you've seen this paper (one of my favourite ones), but just in case not:
http://marswatch.astro.cornell.edu/Bell_etal_SkyColor_06.pdf

The first image seems very good for me. Don't know if you know, but with Terragen you can made very good sky, with various parameters like intensity of atmospheric scattering, or dust opacity (and color).

Don't know if it's accurate but I like make "artificial" skies like here :




You can also have a look on this galery of picture from Olivier : http://www.planete-mars.com/goursac/2006/vision2.html

According to Mark Lemmon -- http://www.unmannedspaceflight.com/index.p...l=&pid=9656 -- for typical dust loadings the brightness variation from horizon to zenith is no more than a factor of two and wouldn't be very visible to the naked eye.

I've dabbled with Terragen, but not much. It would save a LOT of rendering time to pre-render a nice sky like that in Terragen, then map that to a sky hemisphere in 3ds max rather than using Dreamscape. Do you have a save file for that Terragen render as a starting point (or a list of parameters)

Thanks - the paper mentioned by Mark Lemmon turned out to be extremely useful. Seems I might end up using a more sophisticated atmospheric model than I'm currently using. For the dust I'm currently using Mie scattering only and have been experimenting with varying the amount of scattering, optical depth and adjusting a parameter similar to the Henyey-Greenstein asymmetry parameter. This has been largely a process of trial and error.

I still suspect the relative brightness of different parts of the sky varies a lot depending on atmospheric conditions and the sun's position in the sky. For example the brightness near the horizon can be highly variable:

http://photojournal.jpl.nasa.gov/catalog/PIA06915

There are many more similar examples, including images/mosaics from the pancam (this one is from the navcam). In this one the brightness of the sky varies by a factor of more than three. Here is another one, this time from the pancam. It was taken over a period of 8 days which might mess things up a bit:

http://photojournal.jpl.nasa.gov/catalog/PIA07334

Then there are these images where the sky darkens a lot when you get higher above from the horizon:

http://photojournal.jpl.nasa.gov/catalog/PIA06765 and
http://photojournal.jpl.nasa.gov/catalog/PIA06850

I suspect the sun was fairly low in the sky which probably contributes to this effect.

Another aspect of the Martian sky: Interestingly the sky is bluish near the sun, at least when the sun sets:

http://photojournal.jpl.nasa.gov/catalog/PIA07997

I'm fairly happy with the sunsets I'm getting from my renderer, especially because I should be able to improve it significantly:

Click to view attachment

Far from perfect but not too bad. As these are experimental images they are rather crude to speed things up (for example no antialiasing). The parameters I'm using now are different from the ones I used for the images in the first message of this thread. For example I decreased the scattering and increased the optical depth.

...And here is a photo controlled mosaic showing the Martian Sky at VL2 site... with quite accurate color.
Enjoy also !
Click to view attachment

As you suggest, the phase function (or scattering angle) and optical depth, are the primary variables in explaining the array of sky brightness profiles in your list of example images. As such, the assumed form of the aerosol phase function will be very important in reproducing observations. Simple functions like the Henyey-Greenstein are notoriously poor for the particle sizes and wavelengths involved here, particularly if the range of photometric angles sampled in an image (or mosaic of images) is large. Recent analyses using MEX/OMEGA and MRO/CRISM have substantially improved aerosol scattering properties in the Pancam filter range such that you should be able to reduce the number of ad hoc parameters in your rendering. At a first approximation, the shape of the Tomasko et al. (1999) phase functions does quite a nice job...and avoid the inevitable backscatter enhancement that simple analytical (nonspherical) shapes can produce, such as the D/L=1 cylinders that i typically use. Of course, one can empirically remove this enhancement without much effort...and easily capture the behavior seen in the Tomasko et al. work.

I am curious about the level of radiative transfer that you are using in your sunset images...they look quite nice.

All those words and 'sky hemisphere' didn't get mentioned once. Good work Mike.

Sorry - who's Mike? - and who mentioned 'sky hemisphere'? I'm beginning to feel like a cat in a shoe factory. What is going on?

I'm extremely interested in the colour of the Martian sky but I can't make head nor tail of this. Please be a little more explicit.

Check the bottom of http://pancam.astro.cornell.edu/pancam_instrument/team.html

The Sky Hemisphere is the unlikely-to-ever-happen idea of doing a full pan of the sky, with Pancam.

Much appreciated, thanks Doug. I'll start agitating for a sky hemisphere now.

I wish they would place a hemispherical mirror somewhere on Martian lander/rover spacecraft within camera sight, over somewhere on the spacecraft that would be blocking the scene anyway. it should be large and close enough to provide a mini fisheye view of the sky to the camera, providing at a glance (at least early in the mission) a visual summary of brightness variations, clouds, halos, etc.

Hum… I will better prefer a camera with a fisheye lens like "8mm fisheye pelang" pointed toward zenith, with a color CCD, somewhere on the deck of the lander/rover.

So would I, but the mirror should be easier to arrange in the competition for room etc.

It would also be a lot lighter. Beagle 2 used a wide angle mirror that would pop up infront of its camera to get a complete 360 very quickly and dirtily.

Here is the settings of the atmosphere in TG2, sets by default to have a classical blue sky. Maybe you can give me some advise to have a martian sky.
And the settings to hava "martian sky", with the result rendering (notice that the fov is 104° with a sun at 45° high in the sky).

I assume "exp height" is the scale height. If that is the case it is incorrect - it should be ~11 km for both haze (dust) and bluesky.

Thanks to everyone for useful information - mwolff's message was especially helpful. As I was starting to suspect, using a Henyey-Greenstein (or similar) phase function didn't work well (before seeing mwolff's message I was even considering using the sum of two or more HG-functions). I was able to get nice sunset images but in that case images with the sun high in the sky looked bad or vice versa. So I replaced it with the Tomasko et al. functions (444 nm and 671 nm) mentioned by mwolff. Actually I simply measured the function values from a graph in the Tomasko et al. paper and construced a lookup table. The function isn't perfectly smooth yet - this shows up in some of the images where the sun is visible. Dynamic range is a big problem so I processed the function in Excel to greatly reduce its dynamic range while preserving the color ratio (R/B) as a function of scattering angle. The results are very promising. I still need to do some tweaks, for example I probably need to increase the optical depth (the normal optical depth is currently ~0.18) and I'm currently assuming it doesn't vary with wavelength which I'm not sure is correct.

Some test images:

The sun as seen from the surface, the field of view (FOV) is 90 degrees. Interestingly, the sky near the sun has a bright bluish color. I don't know how accurate this is:
Click to view attachment

Similar viewing geometry but looking in roughly the opposite direction:
Click to view attachment

Sunset, the upper one has a FOV of 90 degrees and the lower one a FOV of 30 degrees. From these images I suspect I need to change some parameters because if I slightly increase the solar elevation a large area around the sun gets completely saturated. Because this didn't happen with the sun high in the sky (see above) I suspect the amount of inscattering is too high relative to absorption. From a comparison with actual images from the surface I suspect the sky is also too bright near the sun here which also implies I need to increase the optical depth (i.e. absorption).
Click to view attachment
Click to view attachment

Flying above the surface, altitude 7 km (upper one) and 607 km (lower one). FOV 40 degrees:Click to view attachment
Click to view attachment

As previously, these are rather crude test images without antialiasing. Also to speed things up I'm using a low resolution texture map and no DEM.

I might eventually make my renderer available one day - this has been a very helpful discussion. However, a lot of work remains before that's even possible.

Bjorn, your test images and simulations are really interesting
Here are 2 color samples to show the color of the Martian sky (close to the horizon + at zenith) that is considered to be the closest to the "moderate yellowish brown" (although it's not looking so "yellowish brown" close to the horizon) as seen by the VL landers and confirmed later by MPF and MER. It's the average color, useful for computing, that is NOT taking into consideration the dust opacity variations over the Martian seasons. Enjoy
Click to view attachment Click to view attachment

I would be interested in whether realistic forward scattering properties (i.e., derived from CRISM EPF sequences) would help. In the tabulation below, you can directly apply the single scattering albedo (Csca/Cext) and the wavelength dependence of the extinction (normalize Cext to your favorite wavelength, say 0.88 or 0.90 microns to scale MER Pancam optical depths). The model below assumes one is interested in an effective particle size (r_{eff}) of 1.5 microns (v_{eff} = 0.3).

wave = wavelength in micrometers
Cext = extinction cross section (arbitrary units)
Csca = scattering cross section (same units as Cext)
kappa = ignore
g = asymmetry parameter

wave Cext Csca kappa g
0.4400 4.4040e+00 3.4910e+00 1.8460e+16 7.6870e-01
0.4600 4.4230e+00 3.5760e+00 1.8540e+16 7.6200e-01
0.4800 4.4360e+00 3.6780e+00 1.8600e+16 7.4980e-01
0.5000 4.4660e+00 3.7980e+00 1.8720e+16 7.4040e-01
0.5250 4.4680e+00 3.9270e+00 1.8730e+16 7.2790e-01
0.5500 4.5060e+00 4.0800e+00 1.8890e+16 7.1330e-01
0.5750 4.5150e+00 4.2010e+00 1.8930e+16 7.0650e-01
0.6000 4.5420e+00 4.3260e+00 1.9040e+16 6.9290e-01
0.6300 4.5830e+00 4.4250e+00 1.9210e+16 6.8760e-01
0.7000 4.6190e+00 4.5100e+00 1.9360e+16 6.8060e-01
0.7500 4.6630e+00 4.5580e+00 1.9540e+16 6.7820e-01
0.8000 4.7110e+00 4.6030e+00 1.9750e+16 6.7700e-01
0.8500 4.7480e+00 4.6350e+00 1.9900e+16 6.7730e-01
0.9000 4.7700e+00 4.6510e+00 1.9990e+16 6.7870e-01
0.9500 4.7850e+00 4.6640e+00 2.0060e+16 6.7990e-01
1.0000 4.8350e+00 4.6940e+00 2.0270e+16 6.7920e-01

One question that is relevant to the appearance of the Martian sky is whether one would expect to see stars during the daytime.
My impression is that the answer to this question would be heavily dependent on the amount of dust in the atmosphere.
Perhaps the more technically-oriented members of this forum can help..

Regards to All,

Tolis.

Unfortunately, NO stars are visible during daytime : this fact was confirmed after the VL1 landing in 1976, because the dust makes the sky too bright for stars to be visible even at zenith.
Maybe the only place where the brightest stars (and planets) could be visible from the ground on Mars AND only when dust opacity is at its lowest, is at the summit of the highest volcanoes... But to see them you would have to concentrate your vision a bit, avoiding ambiant sunlight, staying in the shadow of a big rock for example, as the Apollo astronauts did on the Moon by looking at the brightest stars from the shadow of the LM...

Thanks - I'll try this and see what happens. Might take some time though.

Meanwhile, I've been having a look at some more images. This one is particularly interesting because of the caption:

http://photojournal.jpl.nasa.gov/catalog/PIA00917

The image caption says:


This is interesting because I was not sure the bluish color near the sun when it is high in the sky in my images is correct. However, according to the Tomasko et al. phase function much more blue light than red gets scattered at low scattering angles and I have been having problems 'suppressing' the bluish color with the sun high in the sky without losing that color at sunset too (I might be able to fix this using data from mwolff's table though). Does anyone know of true color images showing the Martian sky relatively near the sun and with the sun high in the sky? I haven't found any that I'm happy with.


And the 'star' easiest to see might be the Earth...

I have now made significant improvements to my rendering software. It's about 6 times faster now, meaning I can do nice animations without keeping my computer running for something like 100 hours (it's an old one BTW). More importantly, I'm getting results that I consider much better. I'm using the parameters posted by mwolff in an earlier message and cannot completely rule out bugs but using these parameters definitely resulted in a big improvement. I had to modify the phase function to reduce the dynamic range though - I'm using somewhat lower values for g because of this. Also I'm using an improved Henye-Greenstein function devised by Cornette but the resulting difference is in this case negligible compared to a normal HG function.

Now some test renders. In these two the sun is high in the sky. The difference is that in the left one I reduced the dynamic range as described above but in the right one I'm using the g values posted by mwolff (and a regular HG function and not Cornette's version). I'm posting the right one solely to show the problems with dynamic range; all of the other renders I'm posting are rendered using the same parameters as in the left one. These renders have a field of view (FOV) of 90°.
Click to view attachment Click to view attachment

In the renders above the ground is brighter than the sky. As the sun gets lower in the sky the sky/ground brightness ratio increases when looking towards the sun. Actually I still haven't made any attempts to get the sky/ground brightness ratio correct - it's clear I need to take a careful look at the MER images to do this. The FOV is 90°.Click to view attachment

The sky is much darker when looking away from the sun. In this one the sun is equally high in the sky as in the first two renders but the sky is much darker since we are looking away from the sun. The FOV is 90°.
Click to view attachment

I think I'm getting more realistic sunsets now, compared to the MER and MPF images they not perfect but I think they are fairly realistic. In the following one the sun is visible through the dust (actually it should be brighter). The FOV is 90°.
Click to view attachment

Here's another version of the sunset image, this time with a FOV of 45°, making it roughly comparable to some of the MER sunset mosaics:
Click to view attachment

I don't know how realistic this one is, the altitude is 10 km and the FOV 45°. I enhanced the brightness a bit so unlike all of the other renders here it has been post-processed:
Click to view attachment

The limb is darker now as seen from space than it used to be but probably still too bright, also it's too uniform (there should probably be visible haze layers). The altitude is 607 km and the FOV 35°:
Click to view attachment

And Mars as seen from space. The texture map I'm using isn't of high quality and the color is a bit strange but this is intended to show the atmosphere near the limb. As in the previous render I think it's too bright:
Click to view attachment

An interesting result is that the sky is bluish near the sun even when the sun is high in the sky (the color is less pronounced than near sunset though). Looking at the parameters posted by mwolff it seems obvious why: The g parameter is higher at short wavelengths. This actually may be consistent with the image caption I mentioned in my above message ("The blue color only becomes apparent near sunrise and sunset") - the color is less obvious with the sun high in the sky but it still is there. So I'm now leaning towards thinking the sky really has a subtle bluish color near the sun when it is high in the sky. However, I still haven't searched the MER images to see if there are any images that might show this effect.

Finally I'll end this big message with an experimental Martian sunset animation. It has a FOV of 70°. In this animation I darkened the sky a bit compared to the renders above.
Click to view attachment

Any feedback appreciated. These test renders have improved greatly since I started this thread, largely thanks to suggestions and feedback received here.

Really interesting Bjorn ! My 1st impression is that it looks like a good a render close to the horizon with the Sun having its bluish halo, but close to zenith there should be no blue halo visible around the Sun and the sky should be darker and less gray (i.e. a little more dark pink). Anyway, what a great & nice work you did. Congrats !

Which reminds me - there really should be a test render showing the sun at the zenith. Here it is:
Click to view attachment

It has a FOV of 90°.

Needless to say I now have become really interested in seeing some kind of a 'sky camera' (as discussed briefly earlier in the thread) on a future lander mission .

Don't encourage him Bjorn

I would like to encourage a small hemispheric mirror be placed on the deck of a future lander/rover to allow a modest fisheye view of the sky to be had by a camera looking down on it.

I have imagined a modified MER hazcam camera, but with 180 deg FOV optics, and with a color CCD chip which is mounted on the top of the camera mast of a lander/rover. That way, it is not limited to just pointing up .

Don,
Some experiments and deconvolution software might help encourage the proper folks to consider doing that.

An all-sky pan with the 34mm Mastcam on MSL wouldn't be that many frames (about 100? Have to think about the geometry a bit) and would only take a few minutes to acquire, assuming that the remote sensing mast can actually point straight up. I'd be interested in taking it, bandwidth permitting.

--- unnecessary quoting removed ---

That would be great! As with MER, one could do some basic science (for me, scattering phase functions of dust) using just the calibrated "thumbnails", so bandwidth wouldn't be a concern. Of course, higher resolution would be interesting from the point of view of water ice cloud detection (and full resolution obtained periodically would be an excellent source for flatfield monitoring). I would suspect that Mark L. would be enthusiastic about such a thing as well...

I have now done an improved version of the sunset animation. The new one has more than two times as many frames (things really happened too fast in the old one) and the sun now gets well below the horizon. There was a bug in my renderer when I made the old one that resulted in a 'contoured' sky when the sun was below the horizon so the old sunset animation ended with the sun on the horizon. In the new one the sun gets well below the horizon.

Click to view attachment

I can probably still make some improvements to my Martian sky renders but probably nothing major. The next step is to combine this Martian atmospheric model with the MRO DEMs which should result in interesting (and hopefully realistic) test renders.

Edit: The area around the sun looks somewhat 'contoured' in the animation, especially when the sky is getting dark. This contouring is not present in the original frames used to make the animation file, it's an artifact of the compression used for making the AVI file. It might be possible to fix this by using a different codec.

Nice animation above by Bjorn from 2009. Here's what I get so far for an animated whole sky simulation with the sun dropping lower in the sky.

Click to view attachment
Full Resolution Animation

The animation shows just the scattered light so I'll want to add the actual sun in as well. Some increased blueness shows up at low phase angles. I'm basing this (wavelength dependent extinction, single scattering albedo and asymmetry parameter) on a Pollack et al paper and using a Henyey-Greenstein (HG) function. A second HG function is used to add some back-scattering.

Dynamic range of brightness is compressed a bit for display.

This post is a continuation of a discussion I started in Earth Observations, with Mars related results now being posted in this thread.

The latest version shown here has somewhat improved handling of multiple scattering and has the sun's disk added in.

To help with rendering, phase functions for different wavelengths (red/blue) can be checked in a 2014 paper mentioned in the post linked below. It's interesting that the phase function near 90 degrees scattering angle is about the same for red and blue implying the sky should be gray at right angles to the sun, and only turning reddish at roughly 130 degrees. The blueness in the glow around the sun is predicted to be showing up at scattering angles less than about 28 degrees.

http://www.unmannedspaceflight.com/index.p...mp;#entry212746

Perhaps though a fuller consideration of reflection from the ground would shift things more to the red at somewhat smaller phase angles.

Here is a more direct link to the recent paper that explores the question of why the sunset on Mars is blue:

https://www.osapublishing.org/ao/fulltext.c...8&id=281919

This comes to mind as the recent APOD with Damia's Mars sunset mentions the reasons for the blue sunset aren't completely understood. This paper though seems to explain things pretty well, in that the blue sun is from a negative Angstrom coefficient (associated with a specific dust size near the wavelength of light), and the blue surrounding the sun is related to the Mie scattering with the dust where the blue Airy diffraction disk is more concentrated than the red one. The composition of the dust is only important (and producing red) at greater scattering angles in other parts of the sky. These colors are there to some extent at other times of the day as well.

Yes, I was a bit annoyed by that APOD comment. The reasons are not understood by a great many people. But they are quite clearly understood by many on this forum and any scientist working with Martian atmospheric images professionally. Interesting link, thanks.

Here's a simulation with my latest software for the Mars sky, sun is 14 degrees up:

Click to view attachment

An animation of the frames every 2 degrees of solar elevation is here (animation updated 2100UTC).

And in this directory the individual frames (8 / 16 bit) are located.

Amazing work Scalbers, that's a really nice product of radiative transfer!
I'll use your skies for my Curiosity panos to make full 360x180 panos, with appropriate credits of course.
One thing: when downloading your skies on your website, I noticed the 78° solar elevation is all blank.

Thanks Thomas, and glad they can be used in your mosaics. I worked through a few code glitches to get the 78 degree frame to work.

I have a new version including some additional improvements, such as an expanded solar altitude range from -6 to +90 degrees animated here. The individual frames are here.

Click to view attachment

Greetings,

Considering that light can reflect off the colored ground and then back off of the dust in the sky, I made an updated version that looks slightly redder away from the sun. Additional refinements might still be needed, and of course it's an interesting question how well we can objectively characterize the Martian sky color. I'm considering whether I can easily add a generic surface to the images so we can see the relative color and intensity.

Click to view attachment

http://stevealbers.net/ast/mars/sky_2016 (frames)

Below we can see how this looks using the CRISM scattering parameters from earlier in this thread (instead of the Viking ones). At first glance this looks pretty good with less of a purple cast (farther from the sun).

Click to view attachment

The CRISM frames can be found here: http://stevealbers.net/ast/mars/sky_2016_crism

I can report on some progress with a Monte Carlo planetary atmosphere simulation program. The first image shows the Monte Carlo result compared with the second image made with my "regular" ray-tracing program. The sun is 51 degrees high. Aesthetically the Monte Carlo one has a bit more brownish tint.

Click to view attachment

Click to view attachment

I'm curious what are the differences between the techniques - I guess MC uses some random element...

Indeed both techniques involve tracing of light rays from the observer and from the sun, and they both have mechanisms to speed up the code. Both are pretty similar in situations with just single scattering. The MC excels when there is multiple scattering and this is where the random scattering and redirection of the light rays really comes into play. The other version parameterizes multiple scattering by an equivalent single scattering phase function. MC should also be doing a better job of handling light reflection off the ground, though this is something I'm still working on. This is a backward MC implementation with forced scattering and local estimation.

One difference I see in the images is near the horizon at 90 degrees elongation from the sun the MC appears darker. I'm unsure whether that is correct. It could relate to the preliminary nature of the ground reflection handling.

If you looked up on an average dusty day in Gale Crater, would you see any stars, or would the 'butterscotch scattering' extend all the way to the zenith?

P

Do you mean at night?

If so - yes - https://www.jpl.nasa.gov/news/news.php?feature=4121

I presume the OP meant in the daytime.

With all due respect to pre-spaceflight artist's conceptions, I don't think you can see stars in the daytime on Mars any more easily than you can on Earth.

Would Jupiter be visible to the naked eye in the daytime?

A quick Google search shows that some folks have seen Jupiter from Earth unaided before sunset/sunrise. It apparently helped to have the Moon nearby as a guide. I suppose it should be a cinch from Mars, being a few AU closer, especially if Phobos or Deimos have a conjunction with the giant planet.

I've seen Jupiter at 10:20am from a high desert location, right next to the moon (as documented in an ancient issue of Sky and Telescope). There is a case to be made that Sirius would be visible from a clean air location when it is at the zenith and I recall some sightings have been made (I can dig out a paper on this).

For Mars it seems that daytime stars might be similar to Earth. The atmospheric scattering (and optical depth) is quite a bit greater and more than compensates for the dimmer sun. Yet aerosols have a greater angular dependence than gases, so away from the sun the sky brightness would tone down a fair amount. Extinction is more of a factor on Mars.

Mars being closer to Jupiter mostly applies near opposition. Daytime signtings would be closer to quadrature when there can be both a distance and a phase angle disadvantage impacting the apparent magnitude. Based on a review of the above sky simulations (assuming an optical depth of 0.5) I may be able to calculate the zenithal limiting magnitude for various solar elevation angles. The best part of the sky to look would be somewhat lower than the zenith and opposite the sun in azimuth.