Confirmed Claim: disputed satelite imagery showing "changes in vegetation"

Herman Aven

Member
[Edit by @Mick West] The original post here suggested that comparisons of images of vegetation vs. no vegetation might be explained by difference in the type of imaging, and could actually show little or no change. However inspection of the chronological sequence of satellite images shows beyond all doubt that the change in vegetation is real, hence the claim is marked as "confirmed".


Location: 48.099246°, 37.750438°
For details see:
https://www.metabunk.org/confirmed-...wing-changes-in-vegetation.t6430/#post-156990

-----------------------------------------------------------------------------------
Original Post Follows:
-----------------------------------------------------------------------------------

Already a for a while I've been wondering about a couple of specific claims made in Bellingcat's article New July 17th Satellite Imagery Confirms Russia Produced Fake MH17 Evidence. This is not about the "missing BUK" issue but about the following claims which seem intended as some kind of important circumstantial evidence.
  • interactive photo 2. "note the changes in terrain, vegetation"
  • interactive photo 3. "note the changes in the terrain"
  • interactive photo 4: "note the changes in vegetation"
  • interactive photo 5: "note the changes in the field"
It's easy to note changes but it's also easy to note the incredible amount of changes overall. The Russian photo shows in higher contrast the various pathways and some particular features are more pronounced while other features less so. Here's an example from photo 3, note how the MoD picture shows a lot of changes to the right in soil texture, and high contrast roads. Hard to explain this would be all done in a few weeks but easy if you think the task for the MoD is reconnaissance, not sight seeing or any "true" color representations.

dg-sample-1.png sample-mod-1.png

This might be related to specification. The Russian "Persona-2", which is the suggested satellite, is claimed to have 0.33 meter max ground resolution using panchromatic (source). Compare this to Worldview-2 with 0.46-0.50 for panchromatic and 1.84-2.4 for multi-spectral. This means the two satellites might disagree on the presence and coloration of certain details which will be most problematic when it comes to soil structure, smaller relief and vegetation because of the finer textures and mixed compositions.

What is also interesting and possible relevant is the change in shadows: the observation zenits and azimuth angles of both satellites. Here's a sample of image 2 to check the shadows. The MoD is the image to the right with short, high-contrasted shadows pointing north-west. WV2 at the lest appears to have a more north-east oriented shading here but in the other images way less so.

sample-dg-2.png sample-mod-2.png

Now the most stunning example in the research article: the "changes in vegetation". First the original from the article:

trench-dg-version.png trench-mod-version.png

Now what I ask here is to let go for a moment of the concept of looking at "tree tops" but instead look for some of the terrain features. With some hesitation I made a quick attempt and also corrected the vertical alignment (just focus on the main "knob" feature halfway the "passage"). It's just for orientation and the realization that we'd might be looking at a bunch of different colorations, different elevation levels perhaps, some shadows and/or higher contrast added to some shadows at 10 o'clock. So it's certainly possible we don't have "changes in vegetation" at all! Or at least we just don't have the resolution here to know for sure if it's all a matter of just vegetation or rugged terrain features! Keep in mind that many smaller features below 0.5m might become erroneously defined or colored.

trench-dg-version-marked.png trench-mod-version-marked.png

Note that if Bellingcat's claim for changed vegetation would be invalid, the Russian analysis claiming the "tree belt" being "deliberately distorted" in the Ukrainian image (which probably also was taken as well by WV2) would also become invalid.

One other interesting issue is the following news:

"DigitalGlobe Inc. on Wednesday secured approval from U.S. regulators to sell satellite images of objects as small as a shoebox to overseas and commercial clients " 2014-06-11 Wall Street Journal
"During June 2014 DigitalGlobe received permission from the US Department of Commerce to collect and sell imagery at the best available resolutions" Satellite Imaging Corporation

This is possibly a factor when comparing old and new DigitalGlobe imagery for example through Google Earth, which appears to have indeed way higher detail since July 2014. Keep in mind that the color details are provided by merging multi-spectral images with a resolution between 1.84m and 2.4 m, meaning that finer features will receive a calculated hue when the multi-spectral is merged with the high-res panchromatic. In other words (and this can be easily seen browsing through GE history) one easily gets too way much green in an area or if we tune some parameters or use higher resolutions, we'll now get earth toned colored features!

My conclusion however is not that there were no changes in landscape at all or that one party claims a wrong date and the other not. But the claim of "changes in vegetation" is not convincing based purely on comparing a high-res panchromatic of one satellite with generated pan-sharpened, probably slightly less high-res image from another satellite and angle. The most clear demonstration can be seen with interactive photo 5 with the claimed "changes in the field". It just takes a bit of careful looking to see the same field lines are present in both images but only in the MoD version extremely faint. This is a good example of misreading landscape features based on various wrong assumptions.

The Bellingcat claims about the landscape appear therefore to be largely unsubstantiated but it would be great if other views and critiques could be presented here.
 
Last edited by a moderator:
upload_2015-6-21_20-34-3.png upload_2015-6-21_20-34-26.png
Is there an extra tree overhanging that road on the left in the mod image?
...

Unfortuantely I think overlaying the lines to make your case that the features are present in both, obscure those details and prime the eye (tell it what to see) too much to useful. Perhaps a gif overlay of the two or a slider image would be better.
 
Last edited:
Pete, I think those are the same trees with large shadows. But I see what you mean, the two center trees appear to have "jumped" from one side of the road to the other and in the MoD image one appears even smaller. I think these are very typical distortions for these type of comparisons simply because I've seen too many of them to be explained by busy landscaping around the airbase.

You are right about the overlay. But we're talking about hard to see stuff. Actually, I'm making the case that one should not rely too much on certain appearances at all but appreciate that elevation, shadows and differences in rendering and coloring textures could distort surprisingly much. But it's very hard to make that case without some very clear examples. It would be so much easier if I had the panchromatic or separate bands of the Wordview-2 image. But I'm not going to order them and work on those since I don't even know if I'd be allowed to distribute the results with the copyrights! My hope is that more qualified people will be enticed to make an effort, although I do have some experience with the technical background, making a believable presentation is another thing. And I don't have really a motive either apart from the question if using this type of imagery for the purpose of proving anything at all is valid. This is since the WMD claims in Iraq where many "experts" easily went the wrong way by some group think combined with politics.
 


Now what I ask here is to let go for a moment of the concept of looking at "tree tops" but instead look for some of the terrain features.

Why? It's very clear they are tree tops in the greyscale image. It's equally clear the trees have been removed in the color image. You simply have to look at the history of images in that area to see this, use the context of the surrounding vegetation and exposed soil.

In the images below, the date is in the slider on the top left. The location is 48.099246°, 37.750438°







several similar images follow, then the most recent:


It's incredibly clear that the two photos have trees in the black and white image, but no trees in the color image. That is a significant change in vegetation, and hence the claim is confirmed.
 
The passage cut through the brush really seals the deal. It's there on Jun 18. It (and all the trees) are gone July 1st, but the MoD claims it's there on July 14th.


(Correction: this originally had the middle image as claimed July 17th, when it was claimed July 14th. This does not make any difference to the fact that the claim is false).
 
Last edited:
Yeah, while your post was informative technically @Herman Aven I don't think a credible case can be made that those trees are there and just obscured by the nature of the photo, they are physically gone.
 
The whole point of this thread was to demonstrate how the idea of looking at "trees" or "brush" was possibly wholly unsubstantiated. To respond with "look at those trees" as some argument is not a proper critique but just a simple denial and a rather convenient assumption that it's now therefore all "confirmed". But it should have been clear by my first example that significant changes in soil texture and features are plainly visible in all the examples and much more likely caused by the differences in how the various spectral ranges are rendered. It should be clear Russian images show a different spectral response to various soil differences ignoring sometimes the thin partial surface cover. On top of that come all the differences in resolution and shadowing. It's surprising to me that it's not being appreciated by anyone that "trees" or "bushes" are easily rendered by the mind's eye like for example a "face" is rendered when looking at an older image of the Cydonia region on Mars. Higher resolutions (in this case pan-sharpening and difference angles with shadows) can change that perception obviously. But it doesn't prove anything about dates or changes made to objects.

But I guess I'll need more examples and better presentation to make a more convincing case against what appears to me as a potentially serious confirmation bias. In the end all I've read above as counter-argument is the doubtful assumption that "green" shading means "covered with vegetation" in some older lower-res pan-sharpened images and also what seem to be two small trees, appearing or disappearing when comparing it to images two months before. It might just as well be some spectral distortion. If anything that image from 5/29 should tell you something about the way the landscape is rendered suspiciously green in the whole region. This is a known resolution problem with all pan-sharpening processing, a case I've been trying to make.

Of course I understand I obviously failed to make a convincing case but I'm in the process of acquiring more evidence since comparisons are drawn up here and inferences being made without using proper sources and also some way better examples are required. So I'll hope to be back and continue this or come back and admit I was completely wrong!
 
Last edited:
So I'll hope to be back and continue this or come back and admit I was completely wrong!

I'm quite baffled as to how you can continue to argue this. I notice you disagree with my two posts above, so clearly you looked at the photos. Do you think the passage cut though the brush in the MoD photo is some kind of optical illusion?


Is the change in vegetation in winter also some optical illusion?
 
This is not quite correct, the claimed date of Russian MoD image in the middle was July 14.

Apologies, I'll fix that. Of course it does not change the conclusion.
(I've updated the image, cached copies might reflect the old date for a while)
 
Last edited:
There are obvious shadows/shade cast by the trees and brush in the May and June images. Those shadows are gone in the July 1st image while shadows of the trees that were not removed are still there. How can it be possible that they are still there and not casting any shadows? Spectral differences can't account for that.

If anything that image from 5/29 should tell you something about the way the landscape is rendered suspiciously green in the whole region.

My yard and the whole neighborhood become pretty vividly green in the spring. I've never thought of it as suspicious.
 
I'm quite baffled as to how you can continue to argue this. I notice you disagree with my two posts above, so clearly you looked at the photos. Do you think the passage cut though the brush in the MoD photo is some kind of optical illusion?

It's a possibility which I don't believe has been ruled out properly unless "I think I'm looking at" counts as complete evidence these days when dealing with various degraded JPG's distilled from pan-sharpened blends computed from other sets of satellite images. It's just strange to see this presented as being hard evidence for determining various more subtle landscape features and especially when comparing with other degraded JPG's from another rather undocumented source in Russia. That kind of research can never be more than "interesting"!

Here's your image again with the contrast as high as I could get it, although I'm aware this is dubious as enhancement, it will have to do for now. And I do think it helps a bit to see what I'm thinking about. Also, I'll not draw on it this time :) Please keep in mind the obvious fact that any Google "green land" doesn't render dark at all on the MoD images like the many roads which appear so much brighter: they certainly appear to render certain types of soil lighter, where WV2 at times almost obscures the same features.

20150621-093847-jsmll-highlighted.png
To me all the potential trees and shrubs are present in all versions but what is claimed to be some "smutted" out sand pit looks like mostly rugged terrain to me which on the MoD and some earlier Google Earth just happens to show as being way more patchy and renders darker (but so much does). You can see even on Google Earth July 1 still some small vegetation (or other features but most likely some growth) here and there on most of the dirt, many visible at exactly the same locations on the Russian MoD version. Do I really have to add arrows? So did they leave some green standing and shaved around those? Any other "dark round patches" in the grayscale panchromatic could easily be different soil colorization and texture highlighting because of how certain wave lengths are being recorded and processed. If it really used to be woodlands before, the top soil will be almost certainly different!

In all the supplied imagery of the MoD so far many underlying soil textures appear to be rendered differently, brighter respectively darker depending on for example water content or temperatures of the ground, which is common for many panchromatic images. These aspects can easily change in any pan-sharpened version because that deals with only visual ranges. When someone is serious about remote sensing they will flip through some other multi-spectral renderings to make sure what the upper soil exactly contains in terms of granularity.

It's not like I'm saying that I'm presenting a tight shut case here, not at all, but I believe this option cannot be completely ruled out either. It simply needs more material not made available yet. Or in other words, I simply require higher standards of evidence before I accept that Russia would use some seemingly random image from the recent past, then remove vehicles on that older photo using some editing tool (why not use a recent picture, didn't they have it?) and then submit the full versions to the Joint Investigation Team who obviously will have access to a more extensive range of satellite sources of the same location and dates. So to me that doesn't make any logical sense, unless the fabrication is only supposed to influence a domestic audience on the short term. This last bit is off topic, I realize but I wanted to explain a bit of my reasoning behind my skepticism when looking at all these pretty pictures.
 
It's not like I'm saying that I'm presenting a tight shut case here, not at all, but I believe this option cannot be completely ruled out either.

I think it can, and it has. You just are not seeing it.

What do you think the feature that appears to be a cleared path though brush actually is? Why is it soil colored in the Google Earth image. Why are there shadows there in the MoD image?
 
Back
Top