Generic selectors
Exact matches only
Search in title
Search in content
podcast
Filter by Categories
Galleries
Uncategorized

The Role Of Geospatial In Open Source Intelligence

Daniel:

Hi Michael. Welcome to the podcast. Thank you so much for taking the time to join me today. I really appreciate it. So in just a second, we’re going to talk about open-source geospatial journalism, and we’ll dive into exactly what that means in just a second for the listeners. Before we get there, can you just take the time to introduce yourself, and perhaps explain how you got involved in geospatial and journalism?

Michael Cruickshank:

I’m Mike Cruickshank, a journalist, originally from Australia, and I actually got into doing geospatial through journalism, which I suppose is a strange entry point to geospatial. This happened because I got involved in this kind of journalism called open-source intelligence. This uses a lot of geospatial data for cross-referencing things by using techniques such as geolocation. 

This first brought me in contact with satellite imagery through Google Earth and that sort of thing. I just became more and more interested in what this data is, how can I use it to do more than just have something to look at? What stories are there that kind of lie beneath the pixels, so to speak, of the data and how can I use it in more interesting ways? From there I taught myself how to program, and work more deeply with this data.

I’ve learned how to use all these different geospatial tools and programs.  I’m also doing academic research now, I’m doing a master’s thesis where I’m using this kind of data to investigate links between climate change and conflict. I’m also working professionally at a company in Berlin called LiveEO, where we use geospatial data to provide information for industries on things like environmental issues or forestry, as well as monitoring industrial assets. It’s quite a broad range of things that I’m involved in these days.

The Link Between Journalism and GIS

Daniel:

It sounds like you’re moving away from journalism and into geospatial. What kind of skills can you already see that are going to be really helpful in that move, that you’ll take with you from journalism and over to the geospatial side?

Michael:

I wouldn’t so much say I’m moving away from journalism as I am trying to tell a different kind of story to a different kind of person. In the past, my thinking was all about how do I inform people about what is going on? I came to the conclusion that, while I still want to inform as many people as much as I can about what is going on, I also want to be able to reach more of the kind of people who are actually making the decisions in the world, and provide them with information on how things will go into the future.

I’m particularly interested in looking at the effects of climate change and its security implications. I want to be able to provide insights on this prediction, analysis, etc., using geospatial, as well as some of my background in journalism on thinking about the real effects of these things. I want to present this kind of information in a way that’s engaging to people so they actually take it seriously, and just reach in a more targeted way the kind of people who can actually make a change in this world.]

Daniel:

You had this great sentence there, you said something like, “I want to tell different stories to a different kind of people.” What was it about traditional journalism that you felt like wasn’t working, or could be done better?

Michael:

I felt that in many ways, people were flooded by the same story. I was working in conflict journalism so perhaps you could go to a country, you can go to the front line of their wars, you can tell a story of, oh, it’s very awful. It definitely is awful what is going on in these places, but the unfortunate matter of fact, is that so many people are telling these kind of stories, people just tend to switch off. 

You need to find a different way of reaching them with a different kind of story, one that engages them in a different kind of way, and perhaps tells a bigger picture or a different picture.

 

Daniel:

I think you’re absolutely right with that. I think that’s an absolutely brilliant insight. I guess that the danger is when we think about engagement, we sort of fast forward this trend towards the short, incredibly clickable link bait.  I really do feel like we are a world trending shallow, as opposed to deep. We see this in all the social media trends, generally. It’s quick, short snippets. Do you feel like that’s the way to go? If we think about engagement, the numbers speak for themselves. That seems to be what people want, what people engage with. When I think about journalism, I kind of expect something deeper.

Michael:

You’re well within your rights to. I certainly think that going after the kind of engagement that you are talking about is a bad thing for publications and for journalists to do. I suppose when I talk about engagement, I don’t mean someone just reading something, clicking it, sharing it, liking it, reacting to it. What I mean is someone who’s really taking stock of what is written, what is seen, etc., and thinking about it and incorporating that knowledge into their life and their decision-making.

Daniel:

So I have to ask here, we are overwhelmed with media, we are flooded with information and data. How do you think your media, your style, this idea that you are going after here is going to cut through the noise?

Michael:

I’m not going to put myself up on some pedestal and say I could solve this problem on my own. What I think is it forms part of a body of information which is more verifiable, and forms more of a ground truth, or a reality of what has happened or what is going to happen, rather than just 1,000 different conflicting reports from different directions within the political sphere or different interested groups, etc. I think that part of the way that we combat this flood of information is simply providing people with better information, rather than just increasing the amount of it or telling them to consume less. Instead, we should suggest and provide the option for them to just consume better information.

The Tools of the Trade

Daniel:

Yeah. I totally agree. I feel like that volume game that you’re talking about inevitably leads to a lesser quality of information, or lesser quality of work because it’s a volume game, right? It’s about getting the amount out there as much as possible.

I know you’ve come a long way in your journey with geospatial, I know that you program in a couple of different languages now. Could you explain to me what it was like right at the start? Were you using tools like Google Earth just to look at the images and see what was on them and try and geolocate stuff there, or were you using something completely different than that?

Michael:

No, I was using Google Earth or Google Maps, those sorts of platforms, and just purely analyzing things on a visual level. There was no sort of looking at the data behind it, it was just all about identifying things that appeared in pictures, using video locations. I was using platforms like Google Earth, Wikimapia, and a few others to get different kinds of images or different dates of images. Beyond that, I hadn’t gone particularly deep into it.

Daniel:

Okay, so that’s where you came from. You started off using these very simple visualization tools like Google Maps, I believe you said. Where are you now? What tools are you using and where do you get your data from when you think about doing open-source geospatial journalistic work?

Michael:

These days I’m using a large number of open data sources, but I think the primary and most useful one is Google Earth Engine, which of course is very different from Google Earth itself. Google Earth Engine, as some of the listeners might be aware, is a multi-petabyte catalog of huge amounts of open-source satellite imagery from sources like the European Copernicus programme, Sentinel satellites, NASA’s Landsat program, and many others. The beauty of Earth Engine is that it allows people to basically program with the data. You don’t have to download these very large raster images and all that to your computer, you can work with it all remotely. The processing is done off your computer. So this is quite powerful because it means that you don’t have to mess around with huge amounts of files and huge amounts of processing locally.

Daniel:

So maybe I made a mistake. I’ve been saying open-source geospatial here, Google Earth Engine is not open-source, even though some of the data might be. Are there any other tools you’ve been using, or have I made a mistake on my end?

Michael:

By open-source, in this case, I was referring to the data itself. Earth Engine is certainly not open-source, and you can only use it for non-commercial applications, as far as I’m aware at least. In terms of fully open-source programs, I use QGIS as well, so when I’m doing more visual-related output, more than processing, more than programming, more than analytical stuff, when I’m doing visual, I will use QGIS. For the more programming side of things, I’ll usually use Google Earth Engine.

Daniel:

So let’s stay with Google Earth Engine for a second here. What are you doing? I know you can make these amazing time lapses in Google Earth Engine. What kind of analysis are you doing in there?

Michael:

The kind of analysis that I’m doing, really depends on the use case. Basically, you can do any kind of transformation of the data. If you imagine these satellite images are effectively raster arrays, or just data points, you can do any kind of mathematical operations on the data. You can compare bands or images over time. You can create custom visualizations, and programs that just access this data and then work with it in other ways. You can create front-end and backend applications of this data. Recently, for instance, I wrote a small web program that enabled users to work with Sentinel-1 imagery of some of these Russian military bases where they were building up troops before they invaded Ukraine.

This was taking data from Earth Engine, specifically from the Sentinel-1 satellites, and then creating time lapses out and creating different kinds of visualizations out of this that were presented in such a way that a user could do this without any background knowledge of how to actually interpret the data. Often, especially with synthetic aperture radar (SAR), it’s very difficult to get down to what you’re actually looking at unless you are experienced with working with that kind of data.

Using Geospatial Intelligence Techniques for Journalism

Daniel:

When we say Sentinel-1, are we talking about the SAR band?

Michael:

So Sentinel-1 is actually two satellites. One of them is currently non-operational. I’m not sure whether it’s dead or just not working currently. So it’s two satellites and they’re both synthetic aperture radar, SAR, satellites.

Daniel:

What were you doing with this data? What kind of analysis could you do on SAR data that would give you information about what’s happening with the Russian military in this case?

Michael:

The resolution of Sentinel-1 is quite low, it’s approximately 10 meters per pixel. Obviously, you can’t see something like a tank with this; a tank is smaller than a 10 meter square. If you know the areas where the tanks are being stored, at these bases, you have the polygons of the base. When more vehicles move into the base, the mean reflectance in the synthetic aperture radar increases. You can plot the mean of a time to get an idea of whether the number of vehicles in the base is increasing or decreasing. Moreover, if you have high resolution optical imagery, you can get a baseline of what a certain level of mean reflectance represents in terms of actual real numbers of vehicles.

Daniel:

How do you ground truth, something like that? Or can you use any other data sources to get an idea of perhaps a more precise number of how many tanks, in this case, that there are in an area? Can you incorporate other data sources into this kind of analysis?

Michael:

Absolutely. In my case, I have been using very high-resolution optical data as a ground truth that was taken on the same day. Obviously, this is not always possible because of clouds, especially in the more recent months in winter in Europe, it’s very cloudy. You have to kind of wait until you catch a break. This usually can be done with a very high-resolution optical image.

Daniel:

How do you document this, and what are the results of this kind of journalism? Has it been cited anywhere? Are people using it as evidence anywhere? I guess what I’m looking for is- are people referring back to this, almost like a peer review?

Michael:

Looking at these Russian military bases, this was getting a lot of play in the media. I wasn’t the only one doing it, so obviously I’m not going to take all of the credit for this. There were a large community of people within the open-source intelligence community who were sharing this kind of analysis. Many of the initial stories about this buildup were coming from within this community, and then were taken up later by the more mainstream media who were then tasking even higher resolution satellite imagery for their stories.

This really did get picked up in the lead up to the invasion. It helped in many ways to lend credence to what countries like the United States were saying when they were saying, “There’s definitely going to be an invasion. This is how it’s going to happen. They have all these troops here.” The advantage of having people doing this kind of open-source intelligence journalism, using things like geospatial data is that we can then have a second data point and say, “Well, they’re saying this, but now we can check if there really is data to back this up.” In this case, there certainly was, and we’ve seen what’s happened since then.

Using Social Media as a Data Source

Daniel:

Twitter is my social media drug of choice. At the moment, my account is flooded with images and short video clips of tanks and military operations in the Ukraine. At the same time, I’m constantly wondering, is this real, this thing that I’m looking at? Does this make sense? Is it misinformation? Is it disinformation? Can I believe this? Maybe before we dive into this as a topic, perhaps you could explain the difference between misinformation and disinformation.

Michael:

I’ll start with misinformation. Misinformation is when someone shares a piece of information that is untrue, however, they don’t necessarily know that that information is untrue. More often than not, they will probably believe that it is true. Disinformation is the case when someone is maliciously and with intent sharing something that they know to be untrue. The differentiation lies in whether the person sharing knows whether what they’re sharing is true or not.

Daniel:

So this has been a huge problem, obviously, not just in the current situation we are in now with Russia and Ukraine, but also in other political theaters around the world. Is there any opportunity here to sort of prove or disprove some of this kind of information that we are seeing in social media feeds, for example, by doing the kind of journalistic work that you are doing? Or can we use this data together with what you are doing?

Michael:

Absolutely. Using open-source intelligence is one of the best ways that we can actually prove or disprove whether something is indeed misinformation. The primary tool is geolocation, where you can compare elements within, say, a video or an image to satellite imagery to work out whether it really is where it is said to have happened. This often makes it very easy to filter out if something is misinformation or disinformation. You know something is misrepresented, because you can immediately see that, okay, this is indeed where it’s said to have happened, which then lends credence to the idea that that is indeed true.

You can take geolocation further and move on to chronolocation, where you start using things like elements of the shadows within the videos and backing this up with geospatial data on the exact position to calculate the exact time that a video or an image was taken. This further leads you down the path of whether this is true, or something which is misrepresented.

Daniel:

I can definitely see the power of geolocation here, but how do you do that? Are you looking for a place name in the video? Are you looking at some sort of metadata in the video, perhaps the geolocation tag, that kind of thing? Or are you doing something completely different to locate that image or that video?

Michael:

If there is a name, place name, shop name, or a business name within the video, this obviously makes it very easy. More often than not, that’s not present, or the video has such low quality that it’s difficult to actually read what’s written. The more common way this is done is by cross-referencing visual elements within the video to visual elements within satellite imagery. This would be things like the architecture of the buildings, the specific layout of streets, of plants, and that kind of thing. (Much like Geoguessr). You then think about what that same scene would look like from above, and then cross-reference that scene with satellite imagery of the location where the video or image is said to have taken place. You then see if you can work out exactly where this happened. More often than not, you can work out not just where, but what angle the camera was pointing in, what time of the day, what time of the year, etc..

Daniel:

This reminds me of trying to orientate myself to a map. Looking at a map, looking for features that I can identify on a map, and then looking up in the real world and saying, okay, where are those features relative to me? Can I use them to orientate myself if I’m out hiking, for example? So this makes sense to me, but we live in a time of deep fakes. You will have seen these videos on the internet before. Is it possible to fake a location, when we think about the kind of information we can glean from social media?

Michael:

It is possible to fake a location, or at least fake a plausible looking location. Geolocation is possible 99% of the time in any video. If you can’t find where somewhere is, if you can’t get a fix on it, you have a pretty strong indication that this may indeed be a fake location. I haven’t specifically heard any reports of videos that have been shared from things like conflict zones where the background has been completely faked using these kind of deep fake techniques, but it’s certainly possible.

There’s a website called This City Does Not Exist that sort of generates these deep fake landscapes, and there’s no reason someone couldn’t do something like this. It would take a lot of effort, and looking at some of the misinformation or disinformation that’s being spread around, most recently in the conflict between Russia and Ukraine, most of it is pretty low effort. This sort of stuff takes a lot more effort and I have yet to see it being used, especially at scale.

Doing the Analytics and Research

Daniel:

I’m imagining that the power of this kind of journalism is when you can start joining the dots. You see something happening over here, and you can connect it to an event somewhere else, or you can see that things are moving. How do you keep track of these events? How do you know for example that, oh, okay, I’m looking at tanks here now. I’ve used my SAR data. I can see that the values are changing over time, so there’s change over time. I’ve confirmed this one location using social media, and georeference some of the videos or images that I’ve seen there.This is sort of giving me a picture in my mind and understanding of what’s happening. Tomorrow, the tanks are gone, things have moved. How do you link that with other events, or how do you find where they’ve gone?

Michael:

I approach it kind of like a scientific, or a proof statement almost. You’re trying to prove that something happened or didn’t happen, or trying to work out what the limits of what you can know are. You try to create a chain of evidence going from the open data that you have, and then see how those things link together, how you bring this all together, and how each of these things mutually reinforce, or weaken the argument of the other. In terms of your question you asked, how would you know where the tanks have gone to? Well, maybe you’ve seen the mean SAR values at this base decrease, then maybe a day or two later, you start seeing videos pop up on TikTok of large columns of military vehicles moving around another town, somewhere else in the same region. So you think, okay, well maybe the vehicles are moving towards that area, but you don’t know where they’re going. 

In my case, recently, I created a map of an entire region, Oblast in Russia, just doing basically change detection in SAR and seeing which areas specifically had had large increases within the last week or two. Obviously you get a lot of false positives; you get things like snow melting, lakes melting, or different landscape changes. Some of these places, however, you might see a very large area of change; a polygon that looks perhaps somewhat artificial. Then you can sort of cross-reference this with other data and think, okay, well maybe there is something there. Then you can download very high-resolution satellite imagery, optical imagery of this area to try to cross reference and determine, is this really something which is interesting or suspicious?

Daniel:

I’ve got to tell you that this idea of using geospatial in this way for journalistic research is kind of new to me. It makes perfect sense though. It seems like a great tool for this kind of application. Is this different from what you were doing when you were working as a journalist? Is this completely different from the kind of work that journalists would generally be doing?

Michael:

I think in some ways, it is very different from traditional journalism. Traditional journalism is all about building human sources, and your strength of your argument, or of the story that you’re writing. It is almost based on your level of access that you have to reputable or trusted or well-positioned human sources. That is the skill; it’s a networking skill in many ways, it’s being in the right place at the right time, etc..This kind of journalism is about pulling together a whole lot of things that everyone knows, or everyone could know, and then seeing how those things all work together to prove something and drawing connections that other people aren’t.

This is unlike traditional journalism, where it’s all based on trusting that this is a reputable publication that wouldn’t lie to you. You’ve got to trust that this anonymous source is really a real person, etc.. The opposite is true with open-source intelligence. Everything is effectively a proof statement. You don’t need to trust the journalist. You can read through the way they’ve brought together their argument and their evidence and see if you agree with it. There’s nothing that’s hidden. Everything is in the open.

Daniel:

For me, this begs the question, why aren’t big publishing, media, journalistic companies doing this? Or are they doing this? Do they have their open-source and intelligence center within their magazine or their media company?

Michael:

Over the last few years, there have been more and more mainstream media outlets that have been starting to use these techniques. It is growing in popularity, but it’s also a relatively new thing. This really only started probably in 2012, 2013. That was when the ball started rolling. It has just grown more in importance over time, and in popularity and sort of the credibility that it has.

Taking the Story to the People

Daniel:

I’m pleased you brought up credibility. What do people say when they push back on this, when they don’t believe it’s a great idea? What kinds of arguments do they come with?

Michael:

It depends. A lot of the time they simply just don’t really understand what you are doing. They’ll just say, “Well, this isn’t forensic enough, this isn’t scientific. You’re letting your own biases get in the way. Your analysis is wrong.” They don’t really understand how these things prove each other and so they just say, “Well, you’re just hand waving things around and saying that this equals this,” and they don’t really understand why this equals this, so they dismiss it. To be fair, most of the criticism I’ve come across is more or less just bad faith criticism in the sense that these are people who have motivated reasoning and start with a position of “You’re definitely wrong, because you’re saying something I don’t like.” I find that there’s little good faith criticism, and it’s mostly just bad faith, politically-motivated criticism.

Daniel:

You brought up a great point there, that people didn’t understand the analysis you did; perhaps they didn’t understand the techniques, they didn’t understand the data that was being used. You’re telling people, “Yeah, I looked down from space and watched the change over time.” I mean, it sounds pretty fantastic. How do you explain that to people?

Michael:

This is sometimes the harder part, because you have to explain a bunch of things that aren’t always intuitive. I suppose for people who work in areas like geospatial, thinking in a spatial manner is very natural to us. For other people, sometimes it’s not. Imagining what a scene might look like from above, imagining how things could look over time, etc., or how things can be abstracted in a 2D or 3D format, can be difficult for some people. You need to make this easy, and so having really good visualizations is incredibly important for this. I’ve seen some groups that make really nice videos that are sort of linking things together with nice effects to sort of show how things transition over time. It’s all  in the kind of visual communication you use.

Daniel:

Oftentimes, I think that the people that work in geospatial, we get stuck in only doing analysis for big companies or municipalities, organizing and maintaining the data. When I meet people like you that are doing work like this, I think, wow, here is another opportunity, here is something that people should know about, because it makes so much sense. We can use these tools that we know about, that we’re comfortable with, and we can use them to tell a story. For me, the skill sounds like investigative journalism that tells a story, and backing it up with evidence. Could you imagine a time where new journalists need to learn these skills just in the same way they need to learn interviewing skills or other media skills, making that data analysis a part of their education?

Michael:

Absolutely. It will be critical going into the future. It’s the only way that journalism can push back against the kind of flood of misinformation and disinformation that we’re currently facing. This is a perfect way to combat it. Rather than just retreating back to ideas of, well, I’m a reputable outlet, or, you have to trust me at my word, instead, you’re saying, no, don’t trust me. You’re right to not trust me, but here, I can prove it. I think this will hold much more weight than simply asking people to take you at your word. If more and more stories are presented in this way, then it won’t be so much that people trust the media more, but the correct information will get out more. In the end, that’s what’s important.

Daniel:

If you do a great job of documenting this kind of work, do you think this will help people think critically about how people draw conclusions when they see other types of journalistic work?

Michael:

Absolutely. I think this idea of thinking very forensically, in terms of how you prove something, and what proof really means, and where the edges of what can be proven and can’t be proven lie really helps you narrow down where the gaps in our knowledge are, and where the area for debate lies. That way we don’t spend so much time talking about things that are kind of extraneous, or are red herrings within the conversation.

Daniel:

Michael, I think we could probably round things off here. I’m curious if you have any recommendations or references. If I want to learn more about this open-source intelligence, if I want to be a part of this, is there a community I can join? Is there a newsletter I can follow? Is there somewhere I can go?

Michael:

In terms of communities, there’s a great Discord community called Project Owl, which I think currently has about 25,000 users. It’s growing very rapidly with hundreds of people working on all sorts of different projects all around the world, all brought together by this kind of usage of collecting information from open-sources, especially about conflict zones, but also about many other different areas- bringing this together and synthesizing this information and seeing what can we prove? What can’t we prove? etc..

In terms of other sources online, I think the best group in the world who’s doing this is Bellingcat. They’re a UK-based publication started by Eliot Higgins, and they’ve done some really great investigative work. I believe they’ve also won a Pulitzer Prize, if my memory serves me correctly. They were looking at MH-17, some of the Novichok poisoning attacks in the UK, they’ve done all sorts of interesting looks at different conflict zones around the world; in Syria, in Yemen. I’ve also written a few articles for them myself, looking specifically at Yemen and Ukraine. There are lots of great people associated with this and they’re doing really good work all the time.

Daniel:

So I really hope you enjoyed that episode with Michael Cruickshank. I found a few other articles that I want to link to and share with you. And one of them is called The Growing Problem with Deep Fake Geography: How AI Falsifies Satellite Pictures, or Satellite Images. There’s another one here along the same lines, When is Satellite Imagery Fake? And the third one, Why Newsroom People Need Expertise in Remote Sensing. I’m also going to include a link to a newsletter on Substack that I found that I think you might find interesting. It’s called actualcontrol.substack.com. If I read the headline of this blog newsletter, it says A Blog About Satellite Imagery, Social Media, and Other open-source Information from All Corners of the Internet. There will also be a link to the publishing house that Michael mentioned, called Bellingcat.

Before I let you go, I just want to highlight one of Michael’s insights, and that was at some stage during the start of the conversation he said something like, “I guess I wanted to tell a different story to a different kind of person.” He used this idea of telling better stories. I think this is really important: telling better stories. We talk about data stories and we talk about customer journey stories, but we never talk about telling better stories. Michael wanted to tell better stories. He wanted to do that because he discovered that the old stories weren’t working anymore, they were being drowned out. They were too similar. Everyone was telling the same story so it wasn’t sinking in anymore. It was blending in, it was becoming part of the background noise. That sounds really simple, right? Tell a better story. What does a better story look like? I’m not completely convinced you need to know what a better story looks like, or what the right story looks like. I’m more convinced that you need to try a new story. If you try enough new stories, you’ll find the right story.

Okay. That’s it for me. That’s it for another episode of the Mapscaping podcast. I’ll be back again next week with a new story. I hope that you’ll join me then.