Cast a reddish tint over the landscape when sun is low. #3858
Conversation
Unfinished: may need better balancing
gzotti
added
feature
|
Great PR! Please pay attention to the following items before merging: Files matching
This is an automatically generated QA checklist based on modified files. |
|
The solution is not complete yet...
|
|
I can't currently test; how does it work with an eclipsed Sun? I guess we wouldn't want a red-tinted landscape even with a low eclipsed Sun. |
|
The coloration should be a mixture of the illuminance by the sky and by the Sun. This would make the tint change with the solar altitude non-monotonically. Moreover, the amount of reddening would ideally also depend on the angle between the Sun direction and the normal to the surface being lit. |
Yes, this missing sky component is why I feel the tint is too strong. A cheap solution can be some sqrt of the color factor which pulls it closer to 1/1/1. This is still a cheap solution, no global illumination...
This may be possible to achieve in Scenery3D. However, the landscape pano has no usable normals. |
The tint is just taken from the solar color. During eclipse, the landscape is first and foremost darkened. I assume a partially eclipsed low sun will still be reddened enough to tint the landscape. However, such tests can be done when we are happy about the commonday situation. |
also fix old_style landscape
- May just look smoother after all.
|
Stupid: tint changes only when sun is in frame... Must do that elsewhere. |
Also activate tint in Scenery3D
|
I think the second of the options is the best.
The one currently in use in this PR is too weak, barely altering the color of the scene even with the sun at the horizon: |
|
For some reason the tint in Scenery3D appears much weaker than the tint in regular landscapes. I have not found why so far. Of course, Scenery3D has ambient/diffuse/specular components, and ambient is still white (neutral gray) to simulate the all-sky contribution. For landscapes, one of the two last solutions looks best to me. I don't want to become too kitschy. |
|
You're right, I hadn't even noticed. Indeed one of the two latter looks best. I suppose we can't simply split them, so that 3D Sceneries use one tint, and 2D landscapes another? But that might look weird if someone wants to use a landscape photo as a background for their 3D Scenery. |
|
Of course some better solution is needed, therefore this is still a draft. Now you see where endless late hours of finetuning go... Where are my weekends? |
- now reddening is drawn down to -3° solar altitude with a smooth curve - Scenery3D: ambient also receives a slight reddening - Scenery3D: cleanup outdated code (night mode)
|
I think behaviour looks OK now. Maybe the color hue itself needs some more work. |
|
Hmm... I see no effect in comparison with version 24.2 |
The difference is quite subtle compared to the early stages of the PR. Do you see any difference using a snowy landscape panorama, such as Hurricane? |
|
It's just a slight change, as I don't want to become too kitschy. Use e.g. any grassy ground, and see reddening with the sun higher than -3°, max. on the horizon and tint fading away as the sun climbs higher. With Scenery3D, I now had also to tint the ambient term slightly. If you really think it is not strong enough, I can step up one level. |
|
Wouldn't it be better to query Atmosphere for illuminance (including chromaticity)? Maybe even split between the direct solar component and the diffuse sky component. |
In the long run, sure. This is just a quick first solution. In Scenery3D, someone with enough time could think about global illumination, although we never initially aimed that high. (Earth curvature correction would be more important in its primary field of use.) The color of the solar "halo" [yes, it's not a halo in the meteorological sense. It's just a variable name] and reddening of sun and moon should take atmospheric extinction factor into account. (If SMS is used, even spectral differences are possible, we had talked about Mars and many other incredible possibilities once somebody could explain and document the generation parameters of SMS and write a guideline how to create other atmosphere settings, including several real-world examples.) For now, this provides a first improvement. I am open for better ones, now or next year. |
It's all documented at https://10110111.github.io/CalcMySky.
The problem here is that there's not a lot of comprehensive models. Even for the Earth I still haven't found any model of the stratospheric aerosols that produce afterglow, which is present virtually everywhere in clear weather. For Mars there have been some papers that described the efforts to model its atmosphere, but the description there appears to be self-inconsistent (e.g. a formula that contradicts the corresponding chart), so I couldn't reproduce their results yet. And surely, the parameters are not just turbidity, humidity etc. like in Preetham model, atmospheres are more complicated than that.
This should work automatically if the scenery is large enough and not created to be flat. But I haven't seen any sceneries so large that they'd cover hundreds of kilometers. |
|
These large scenes unfortunately cannot be made publicly available, my largest just shown in our paper. Once they extend to 80km or so, a difference is notable, and it depends on the application whether it is to be regarded relevant. |
@10110111 where could I ask the atmosphere for chromaticity? If this is not available now, I think this PR should be OK as-is, but as soon as atmosphere can deliver such data, we can further improve here. |
There's no such API currently.
Well it doesn't look realistic to me. The colors look a bit weird, particularly for Guereins the grass becomes too red (I'd expect more of an orange tint than red), and this redness continues into the twilight—why? |
|
It just looked bad to just cut off at solar altitude 0 as originally planned. So, now I fade it out 0...-3°. Tell me your favourite solar altitude (or transition altitude) for the cutoff. Can you quote your favourite RGB values (1, G, B) for max discoloration? |
|
Just tried with SMS. Looks even better. Horizon is reddish (sun side)/pink (opposite) with the sun just below horizon. Anywhere in-between the color that is applied on the grass. |
The illuminance due to the setting sun compared to the illuminance by the diffuse sky glow decreases with decreasing solar elevation. Especially so for the horizontal surfaces. So the tinting should have a maximum somewhere between 0° and the upper cutoff, not at 0°.
Or the afterglow, but we don't simulate it either (though in real life it has a profound effect on the twilight color of mountains).
I don't necessarily think the issue is in the RGB, rather in the way the tinting is applied (see above). In the absence of afterglow or red clouds we shouldn't have too much redness after the sun has set. |
|
Note though that in 3D scenery the reddening of the sunlight should be maximum at 0° (or rather -0.25°), since there, IIUC, the sunlight is separate from ambient skylight. |
|
There is no upper cutoff. Reddening goes with airmass. The lower cutoff is currently set arbitrarily to -3°, and looks pretty good IMHO. Definitely better than 0. Scenery3D: You are cordially invited to improve my attempts. Yes, the ambient term should ideally receive its information from the atmosphere. This is not possible yet. Therefore I have to mix in some of the reddened light, or else the 3D surfaces are not perceivably colored. As soon as the atmosphere can be asked for global chromaticity, this can of course be improved. So, we can have this effect as simple painting effect now, or this PR has to wait until it can immediately be presented physically correct in X years, or have the simple solution now AND the better solution whenever atmosphere API is available. For a test, setting transition to -1° only. The message is "there is reddening around sunrise". Too much then depends on real-world conditions which we fail to simulate in any case. |
|
This is acceptable solution for me at the moment |
I'm worried that when the physically-correct solution appears, people will complain that the effect is not as noticeable as it used to be. And then, if we revert to this ad hoc implementation in response, this will mean that we prefer an exaggerated drawing to a simulation. And from my experiments, the effect on horizontal surfaces is indeed very subtle.
The API to query illuminance is not so hard to create, it can be done in a few weeks. |
|
I also think this is good enough. Not everything needs to be physically correct to be usable. |
- a bit more green (yellow) - docfix
|
I added a bit more green to the color, does not change too much. The effect is noticeable during 20 minutes of sun near horizon, but not too strong. Good enough now, but as everywhere, there will be room for improvement based on atmosphere tuning later. A cloudy sky (whenever this may come) will certainly also have to influence colours. |
alex-w
added
the
state: published
|
Hello @gzotti! Please check the fresh version (development snapshot) of Stellarium: |
alex-w
removed
the
state: published
|
Hello @gzotti! Please check the latest stable version of Stellarium: |
Description
This may add a yellowish-orange tint to the landscape when the sun is low.
When sun is -3...0° below horizon, tint is smoothly transitioning.
Also when atmosphere is off, there is no tint.
Fixes #3855 (issue)
Screenshots (if appropriate):
Type of change
How Has This Been Tested?
Test Configuration:
Checklist: