It’s a faaaake… — Or not?
When is a satellite image fake? Are there easy ways to tell?
A guest blog post by Pierre Markuse
With easily accessible satellite data, more and more satellite images are being processed by individual experts, institutions, interested members of the public, as well as by journalists and media outlets.
But with more people creating satellite images you will also get more people trying to misuse satellite images to better align with their agenda. It will, therefore, become more important for everybody to be able to use common sense and fact-check images.
So wait a moment… Is that real?
You have to differentiate between three different ways an image could be understood (or really debunked) as being a fake.
The first would be images that are only perceived as fake. One reason could be, that the image is looking very unusual because of the bands chosen and the processing, therefore, making it hard for people to interpret what can be seen. Another reason could be different images, maybe even from the same data, showing a scene differently albeit the same bands were used. This might lead to the feeling that both images can’t be real and are “doctored” in some way.
Second would be images that are perceived as fake but which are actually real images that have been used out of context, be it because of a simple mistake caused by a lack of expertise, or with malicious intent — to misrepresent what can be seen on purpose and to act as some kind of proof for a story.
Finally, for the third way, we have the actually faked images. Those could be completely made-up images or real images that have been altered to show something that wasn’t really in the data.
So let us take a look at these three ways.
1. Perceived as fake but in fact just a different representation of the data
Every now and then some people have doubts about the validity of satellite images. And given that not all people have the same agenda and interests, it stands to reason that images will be processed in very different ways, albeit from the same data, which might lead to confusion in people not actively processing images themselves but just looking at them. Those images are not really faked though, they just focus on different aspects of the data or use different techniques to show that data.
Since most satellites deliver data in different bands of certain wavelengths, some images are different because they use different bands. The often-used natural color view uses data from bands that correspond to red, green, and blue light, mapped to the respective channels of the image. This combination is probably the one most viewers will be familiar with, as it more or less looks like a normal photograph, just from very high altitude. However, given the number of bands (Sentinel-2, for example, offers 13 bands), there are countless combinations the person processing the image could choose from to highlight specific things. But even when using the same bands, and therefore the exact same data, images can look different because they were processed differently.
People find just one image, but it is very different from a natural color view and maybe counter-intuitive
Sometimes, especially when the image is given without much context and unusual in the type of how it is using the data, people believe that an actual image made from satellite data is a painting or other type of visualization.
Here is an example:
This image was processed to show snow and ice in red, rock is pretty much black, and water appears from light-green-yellowish (lots of rock flour) to very dark-green (little rock flour).
Imagine the image cropped without the annotation and posted with little context and it’s understandable that some people might not see it as a real satellite image.
And a little side note to the image, while the red-green combination is looking punchy, I would not use it anymore as it is not color-blind-proof.
People find multiple images and get confused
Take a look at the example below. Two images of the Camp Fire in California on November 18th, 2018. The first one processed by Joshua Stevens, data visualization and cartography lead at NASA Earth, the second one by myself.
As you can see, Joshua’s and my own version are similar, but not identical. We both used a natural color image and then added data from the shortwave infrared bands to highlight hot spots and possible active fires. Even though we used the same data, our natural color backdrops, as well as the added infrared data, are looking different.
None of these versions is “wrong”, we just visualized the data available in a different way. And since we are talking about images being perceived as fake, it should be noted that neither his nor my visualization really depicts a real-life view as perceived by the human eye, since we both added shortwave infrared data, which the human eyes can’t see.
Here is another example, the Drake bushfire in New South Wales, Australia, on September 7th, 2019. I asked a few people to process images from Sentinel-2 data and also processed some myself, and as you can see the results may look similar in some cases but not identical. The reason is again that we used different techniques to process those images and that we might focus on different things. Click to enlarge images.
We all chose to crop the image slightly different, partly used different band combinations and techniques to finally process the image. Yet, all images are perfectly real and made from the data available.
A lot of the confusion some people may experience in those cases could be prevented with some more explanations regarding the images being used.
The good side of this
So even if those different versions of the same scene might be confusing to some people, it is also cool to see that many versions. Satellite images are helpful in telling many stories and their easy availability is leading to them being featured more prominently in the media.
The number of different versions available will in the end only help to show the validity of the images and make it easier to focus on different things.
2. Perceived as fake but just a misrepresentation of facts
Also not a fake, but surely wrong, are real satellite images being misrepresented, trying to convince people of something that can’t really be seen in the image. This could happen by accident, or worse, intentionally with the image acting as fake evidence. This technique of using an image as fake evidence can be especially irritating for people when the image being used as fake evidence comes from (or is just attributed to) a reputable source such as ESA, Copernicus, NOAA, or NASA. A lot of people will just focus on the reputable source and then be more willing to believe the misrepresentation of the image in the context it is used in.
Now, imagine to see the following image, which has been processed and published by ESA in a different context, in a Tweet or Facebook post by someone with the text “Horrible spill of corrosive chemicals is reaching the Tyrrhenian Sea through the Tiber River after a chemical plant’s tanks leaked into the river for hours before the drainage could be stopped.”
People used to see satellite images as well as surely a big chunk of the general population will immediately see that something is fishy and that it looks more like sediments getting carried into the Tyrrhenian Sea (which indeed was the case here, following heavy rainfalls for multiple days, see ESA’s original post here). However, seeing this image and the headline mentioned in Tweets and Facebook posts will also convince some people of it being true.
As with the next type, actually faked images, professional media outlets will usually fact-check images, although sometimes a lack of resources and/or expertise, paired with the need to be quick to get the views for a particular story, could lead them to omit that step. Regarding this, you may also want to read this blog post I wrote some time ago “Why Newsrooms need People with Expertise in Remote Sensing”.
The downsides of social media
However, as with actually faked images, the real problem will be social media, where those images, with false claims accompanying them, could go viral and reach millions. Since there is little or no control you could claim whatever and try to bolster that claim, with a real — albeit out of context — satellite image.
With some media competence (and here I believe schools are really behind on teaching media education that would lead to better media competence) most of those false claims could be uncovered as such within seconds or minutes of doing some research. However, since a lot of people lack that media competence those false claims can go viral and have real impacts on people’s believes about things.
3. Actually faked satellite images
So let us have a look at the worst category, actual fakes. Satellite images that have been completely made-up, or altered the real data to go along with whatever somebody wants to falsely claim as true fact.
As an example see here this image of the devastating September 2019 Central Park fire — which of course never happened and is totally made-up.
The quality of this fake isn’t even that great and still, I’m certain a lot of people would be convinced by it. It wasn’t even hard to create it. I used the EO Browser to generate an enhanced natural color image of the real September 3rd, 2019, satellite data and then used Photoshop to add a fire front and smoke from a different fire. This took me about 10 minutes and, with more time, talent, and ill will, it could be made even more convincing.
Now, a fire as big as this one in Central Park would surely be all over the media, so fact-checking this image would be really easy, even for people not used to usually fact-check. Just the absence of major news outlets reporting on it should tell you all you need to know, as surely none of the bigger media outlets would just omit a story that big and affecting that many people in New York City. Going the extra mile, to be absolutely certain, you could just take a look at the September 3rd, 2019, satellite data yourself and you would see there was no fire.
But here is also where it is getting harder. Not all people know how to access and look for satellite data. So a satellite picture showing what looks like tanks rolling through a city, or burning and destroyed buildings after an alleged airstrike, is relatively easy to fabricate but then hard to fact-check for a big chunk of the population. Here media outlets have to step in and make sure what they publish are real images. But even if the media — usually, but not always — gets it right, in today’s world of social media it can be easy to start rumors and make false allegations, and with some luck and effort (and maybe an army of social bots to help it), those rumors can then spread like wildfire, go viral, and reach millions of people.
How to tell an image might be faked or out of context
So let us take a look at what you can do to make sure you don’t fall for a faked or used-out-of-context image. Here are some things you can — and should — do when you have doubts about the validity of a satellite image:
- Take a look at the publisher! Is it a reputable news outlet or a shady Twitter account with lots of odd tweets? This is not to say news outlets can’t make mistakes or that every Twitter account posting a somewhat unique image should always considered to be wrong, but it can help you in getting a feel for the validity of the claims made.
- Check other sources! As many as you can and include the local media of the area shown. Do they agree with the story? Are they using the same or maybe a different satellite image (or maybe even photographs) also showing what has been claimed? Sometimes looking at Twitter, Facebook, or Instagram accounts of people living in the area can be helpful.
- Try to verify it! If you know how to get satellite images yourself and that particular image is from a satellite provider with an open-data policy get the data yourself and take a look. This can also be helpful if the original image is of an unknown source but you can find an acquisition of the same time and date at an open-data-policy provider, possibly showing the same.
- Ask experts! On Twitter, you will find many experts in different fields of expertise and experts in remote sensing. If you ask nicely, most of them will certainly be willing to give you at least an opinion on a particular image.
- Validate the source of the image when given! Should you be doubtful of an image claiming to come from a reputable source such as ESA, Copernicus, NOAA, or NASA try to contact them via email or tweet and ask them if they have published a specific image and about the context. You can also use Google to find the image on their sites and maybe verify the image yourself.
- Check location and time! Be careful with images not giving exact dates and locations. It will be hard to verify those.
The Sentinel Hub Custom Script Contest
A quick reminder that the next round of the Sentinel Hub Custom Script Contest will start on October 15th, 2019.
Pierre Markuse is a freelance remote sensing expert and blogger with a special interest in optical remote sensing, his images being used by numerous media outlets.
How to make satellite images more approachable to the general public, communicating scientific backgrounds, and the usage of satellite images in journalism are additional fields of interest.