Geolocation means finding the real world location of an object, such as the place where a photograph or a video was taken. Determining the exact location of where and when an image was taken can be critical for an investigation as it can often provide useful evidence to verify and corroborate other pieces of your story's puzzle. The geolocation process is rarely straightforward but there are various methods - some more creative than others - of doing so. This case-based guide shows you how to mix and match methods, tools and a curious mindset to address a geolocation challenge.
By: Robin Taylor
Editor: Tyler McBrien / Illustration: Yiorgos Bagakis
Determining the exact location of where an image or video was taken can be critical for an investigation. This practice is called geolocation.
The difficulty of geolocation can be based on several factors, such as what information is stored in the image’s metadata, by whom and where it was posted, the content of the image itself, or the quality and resolution of an image – to name a few. The variety of factors that affect the amount of information contextualising your image means there are no specific templates to help you solve the question of where it was taken.
There is, however, a basic methodology I follow that works for me in many cases.
Photo by Exposing the Invisible
- NOTE: During an Exposing the Invisible (ETI) Institute for Emerging Investigators in May 2021, I – Robin Taylor - attended several sessions on using OSINT (Open Source Intelligence) techniques to uncover and verify various information traces online. The exercise and method I’ll be sharing here is based on an image we were given to geolocate during one of the sessions. Feel free to attempt to geolocate this image yourself before reading the following. This methodology isn’t one-size-fits-all, and there could be easier ways to identify the location that I may have overlooked or not considered below.
While there are tutorials and examples out there using similar methodologies for geolocation, more often than not their case study examples are either very specialised (think Bellingcat, or Forensic Architecture) or often simplified for the purpose of the tutorial and might not accurately reflect the types of images you come across regularly in your investigation. In this post, I will showcase my thought process in geolocating an image with little additional information by adding a dash of creativity.
The common theme throughout my methodology below is that each point aims to reduce your geographical search area. Sometimes, if you are lucky, many of the steps below might not be needed. For example, if your image’s metadata, also known as Exchangeable image file format (EXIF), has GPS information, like latitude or longitude points, accompanying your image could reduce the time it takes to geolocate your image significantly.
An image’s metadata is text information belonging to your image, such as GPS information, but also time and date or exposure of the image, to name a few. However, you should not count on metadata alone, always corroborate it with other details and sources of information you can get from the image itself, the source of the image - if available – or other possible sources. Metadata of an image can easily be manipulated and in some cases even intended to be misleading.
Part of the geolocation process is verifying numerous points of interest in your image, corroborating it with your satellite map or street view in order to be certain, beyond doubt, that you have found your location. My approach to geolocation is similar to many others, and I have plenty of sources of inspiration - Benjamin Strick is a big one - and I will add a few others at the very end. Other people might have similar methodologies with slight variations, I just happen to like this one, and it works well for me. I encourage those interested to sit down for an afternoon and look at the various methods, tutorials, and tools used to get a better understanding of what is out there and see what works for you.
The following list includes a number of factors I take into account when attempting to geolocate an image.
Context refers to the information that complements the image either internally (e.g., EXIF/metadata) or externally (e.g., a person X posted this from city Y on Facebook). In some cases, you might be fortunate enough to get the exact geographical coordinates (latitude/longitude). For this case study, I used Foto Forensics to find the metadata. Foto Forensics is a website where you can simply upload your image or paste a link to the image, and it will quickly provide you with the metadata in addition to other information. There are plenty of other websites you can use to view an image’s metadata, such as Jeffrey's Image Metadata Viewer. Alternatively, you can download ExifTool as a command-line application. Discovering or noticing an image’s context can sometimes be key to geolocation. However, as already mentioned, you still need to corroborate the context with the following steps.
Foreground and background analysis
Foreground and background analysis relate to finding points of interest in your image. It can be buildings, trees, car models and license plates, people, any writing such as street signs, and much more. Commonly, it will be your foreground and background analysis that helps you figure out which country or city you are in. For example, a license plate number could tell you what country you are in or a store’s phone number area code might help indicate a specific city or area. In addition, you might attempt to reverse image search points of interest that stand out – such as a specific building in the skyline or street artwork. Reverse image searching is the process of searching the web using an image instead of text. You can do this on most search engines such as Google, Bing, or the Russian equivalent Yandex. Here is a short introduction guide on reverse image searching, it is important to note that some services are better at recognising faces or buildings than others. These services are also frequently updated and thus their effectiveness can be improved or worsened.
Map markings are anything that should be easily identifiable on a map, such as mountains, sea, rivers, etc. If you have managed to figure out what city or country you are in, the identified map markings will help you further narrow down your area of interest. For example, knowing you are close to the sea or mountains could eliminate large sections of a country. Alternatively, if you know what city you are in and you are right next to a river could help you quickly pinpoint your image’s location.
Trial and error
This is the process in which you try and unsuccessfully attempt to locate an image and have to go back to the drawing board. Perhaps you have pre-emptively limited your search area based on an assumption and you need to revisit your reasons or try a different avenue. For example, if you assumed your image was in a country based on a license plate, but have failed to find a location, perhaps the car is in a different country and you need to go back to the drawing board.
This might sound like the odd one out in comparison to the others in this list, but geolocation sometimes requires creative exploration of various sources and databases on the internet. I am constantly surprised by the vast array of public data out there. For instance, garbage bins in the Netherlands are colour coded and the city of Amsterdam has an interactive map that shows where all the different types of garbage bins are located in the city. If you know you are in Amsterdam, such information could be key to successfully geolocating an image but might not be a type of resource that immediately springs to mind. Being creative implies exploring the different avenues and resources the internet and other alternative public resources – like libraries – have to offer!
Finally, every point above should reduce your area of interest. I have mostly discussed analysing what is visible in a given image, it is also important to take into consideration what is not – I call this deductive reasoning. For example, in the image (above) we will geolocate in this article, the lack of trees or bushes on the sidewalk suggests our street is quite narrow and small, and not a boulevard or a two-way street. Likewise, the lack of mountains in an image can be just as useful as the inclusion of mountains. It can be useful to repeat the points above after your initial analysis and try to recognise map markings (anything like trees, pillars, sand, sun, etc.) or foreground-background elements that are not visible in your image.
Geolocating our image
More often than not, you will be attempting to geolocate an image that has significantly reduced contextual information. For example, most social media platforms tend to remove the metadata from any image and video posted to their sites. Other times you might have the time an image was taken, but not by whom or any accompanying text that could help narrow down your area of interest. Nonetheless, there is usually more information stored in the image than first meets the eye. To make sure you do not miss these pieces of information, conducting a simple foreground and background analysis goes a long way.
In this image, we can see rows of small apartment buildings with several black balconies and one white balcony. The buildings are of different heights and design which to me, suggests we are in a European city of some sort. In the background, you have one larger building, blue sky, and aeroplane trails. In the foreground of the image, we can further identify a white car with a visible license plate number, real estate signage, house number 62, and a sidewalk made of some kind of stone. On a side note, the street does not seem to be a large street due to the lack of street plants or trees and the angle at which the picture was taken. Through trial and error, I quickly attempted to reverse image search the entire image and noticeable features, such as the fancy door in the bottom right, but it resulted in nothing useful. It was at this point I had realised I was going to have to limit my area of interest with whatever small hints I could put together from the image before running around virtually in Google Street View.
While hard to see, the visible license plate number has a shade of brown-red, indicating that we are in Belgium. Belgium is the only European country, that I know of, that uses dark-red lettering on their license plates. World License Plates has a helpful database of most license plates around the world. However, cars, as they are designed to do, travel. So, while this is a great indicator, a car registered in Belgium can often be found on the streets of its neighbouring countries. Thus, before I was confident enough to limit my search area, I attempted to find more information to corroborate this finding.
Following a quick Google text search on another key element in the image – the advertising plaque next to a window (seen here and marked with a red square in the initial image above), I was able to find the real estate agency "Calao Consult", a Brussels-based agency, by googling the phone number at the bottom of the plaque (without the forward dash and punctuations). We can also see that the company logo from their website matches that of the one in our image.
Image: Calao Consult logo
As luck would have it (and you always need some luck), it would seem Calao Consult only have one office in the city district of Forest. This helps limit our search area as the chances that they would be operating on the opposite side of town, are slim. The images metadata tells us the image was taken on the 18th of April 2018 at 09:15am, assuming it has not been manipulated.
Image: metadata of the photo obtained by uploading the image to Foto Forensics: https://fotoforensics.com/
Since we know the time and date the image was taken, I attempted to see if I could find any Calao Consult listings six months before and after, but to no prevail. I also attempted to check their website - calaoconsult.be/ - in Internet Archive’s Way Back Machine, but throughout most of 2018, their website seemed to be down for maintenance.
Nonetheless, I decided to map out a few of their current listings just to get a sense of the areas in which they operate. In the image below, we can see they actually cover quite some ground: noticeably the areas of Uccle in red, Forest in Blue, and between Anderlecht and Sint-Jans Molenbeek in green. I first proceeded by briefly looking into each of these areas in Google Street View to see if any place or street looked similar to something in our photo (and perhaps in the hopes of getting lucky).
Image: Calao Consult coverage area
Unfortunately, the search demonstrated that there are a lot of places in Brussels that look like the street in our photo; lots of dark-coloured balconies, similar brick and building style, etc. However, it did reaffirm a couple of suspicions. Firstly, we are on a small street. Many broader streets in Brussels will have a great number of trees and other items on the sidewalks. More importantly, I noticed that not all areas had similarly styled streetlamps. In fact, in some places they varied from street to street or block to block. Streetlamps in Brussels vary so much they have an open-air museum dedicated to the evolution and variety of streetlamps in their city.
This was an important finding as it would help limit the area of interest. I went back into the areas defined by the Calao Consult listings and this time only looked at the various streetlamps in Google Street View. As it would happen, in neighbourhoods like Forest, where the streets are much broader and houses are newer, streetlamps were not placed on the side of buildings, and the models are subsequently more modern (though arguably less aesthetic). More densely built areas have them on the side of buildings while old and dense areas have streetlamps akin to the one we are on the lookout for. To me, this suggested we are closer to downtown Brussels and became an important factor in, again, limiting my area of interest.
To briefly recap, the real estate sign limited the search to Brussels, their area of operations and office to the west side of Brussels, and finally, the streetlamp indicates the odds are we are in an older, more densely populated, part of town. To confirm my suspicion, the district of, and the area around Barrière de St Gilles had many narrow streets and similar streetlamps – suggesting it would be a good place to start. This is not to say that certain streets in say, Uccle, did not have the type of lamppost we were on the lookout for, but they were a rare sight.
Had I attempted to walk down every single street in Barrière de St Gilles in Google’s Street View, I might have spent days looking. So, to limit my area of interest one final time, I attempted to try and establish the angle of the street by using the shadow of the sun. We are fortunate enough to have plenty of sun and shade in our image. The hope was to roughly figure out the cardinal direction (e.g. northwest or southeast) that could then be projected onto a map, some might call this a form of map orientation.
When you orient in relation to a map, you are positioning yourself and the map pointing north. In this case, if I manage to successfully figure out what cardinal direction the street is; I would be able to go into my map and look at all streets that corroborate with my established cardinal direction. This would, in my head, exclude a great number of streets that did not match the established direction. For example, if my street runs east to west, then I would not have to look at any street that runs north to south.
As to how I came to this idea, I do not have a straight answer, but I blame the nearly perfect triangle on the rooftop as my source of inspiration when I was at loss and did not know what else to do. It also seemed like a fun exercise. In this case, the basic assumption would be that the rooftop runs parallel to the road below. By finding what direction the rooftop is pointed, I would also find the direction of our street. I decided to use the largest, most clearly defined triangle. However, I would not be surprised if you could have done the same in another image with the rooftiles or any other point of interest that has a clearly defined shadow and positioned parallel to the street.
Complete image showing the marked sun-shade triangle on the rooftop
Cropped and zoomed-in image showing the marked sun-shade triangle on the rooftop
To establish the position of the sun, there is a useful tool called SunCalc. SunCalc is a webpage that shows you the sun’s position and any given time or place. Usually, SunCalc is used in establishing (or verifying) the time an image was taken, a skill known as chronolocation. So, by pasting in the time stamp in SunCalc for the 18th of April 2018 we get the position of the sun. Now, I could have measured the lines and calculated the angles in our triangle. Instead, I made a PNG of the SunCalc, lined up the shadow from the rooftop triangle, resulting in an image that roughly showcases the direction of the street in relation to any north orienting map. The reason I made the PNG was that I knew I could later overlay the image on the map and visualising it would be quicker and easier than knowing roughly the point of direction.
I now went back into Google Maps with my SunCalc PNG and started looking in streets that a) were not too wide b) travelled in the same direction as my image c) in and around the area of Barrière de St Gilles.
I used a tool called PiP, which you can download on GitHub, that allowed me to overlay any image on top of any open applications. It is quite handy as it also allows you to adjust the opacity (PiP is an application for Mac but Windows options exist as well). As you can see in the image below, while I had eliminated most streets, there were still quite a few. In my case, it took me roughly 5 minutes to find the street we were looking for, but I was prepared to look for at least another 20-30 minutes. Expanding the search area as I went along. Lucky for me that was not needed. The angle of SunCalc PNG did not align exactly with the street, but as shown in the image above, it was pretty darn close. However, the image also shows how many streets the PNG excluded from my search and our assumptions about how big the street excludes all major roads.
In the end I was able to find the exact street and location: 50.8242838,4.3489451. As you can see, we have the fancy door, the black balconies, and house number 62.
To briefly conclude, I was able to reduce my area of interest through several rounds of trial and error, deductive reasoning and with a little help of SunCalc. All in all, it took me about 1.5-2 hours.
There are no templates for geolocation, sometimes there is more than one way, and sometimes there are easier ways. It is important to say that had I not managed to find something within the hour, I would have revisited some of my earlier assumptions or tried a different route. Perhaps I would have been able to find the listing of the apartment for sale. Perhaps, had my French been better, I would have attempted to find a French/Flemish database of all types of lampposts in Brussels, as there is for types of trees in London. In fact, in doing additional research for this blog post, I found a collection of hundreds of databases for various things such as ATMs, street art, dog toilets, and public urinals in Brussels – but not for streetlamps to the best of my French. This challenge took me roughly two hours, but they can sometimes take days or weeks, or minuets.
I hope this case study was useful and I would be interested in knowing whether anyone reading was able to find it through other methods. You can find me at @robintayyy on Twitter.
First Draft’s Geolocation Challenge – a game-like set of activities meant to engage viewers in finding out where certain images were taken. It also provides useful tips, advice and feedback on your findings. Website: https://ftp.firstdraftnews.org/articulate/glch318/story_html5.html
GeoGuessr – is a fun web-based game where you get dropped randomly in google street view, almost anywhere in the world and have to ‘geolocate’ yourself! It is a great game to learn different license plates, street signage, climate, the list goes on, to help you become even better at geolocating and problem solving. I can highly recommend the many detective rounds created by community members. Website: https://www.geoguessr.com/; detective rounds by user Druv: https://www.geoguessr.com/game/miqnNoGggNPABJTg.
OsintCurious - is a great website to find quick tutorials on anything OSINT, therefore they have plenty of geolocation guides that I have used and found very helpful. OsintCurious are some of the friendliest and experienced OSINT people out there. They also regularly host live streams where you can ask questions – and even better, play GeoGuessr with them! Geolocation guides: https://osintcurio.us/category/geolocation/.
OSINT Tutorials by Benjamin Strick: https://www.youtube.com/channel/UCW2WOgSiMr216a27KWG_aqg.
Quiztime on Twitter - is a great source for practicing your geolocation skills. From Monday to Friday, OSINT professionals post a verification quiz, often geolocation related. You can ask for help from the community or even the experts themselves. It is a fantastic way to learn, one I enjoy regularly, and a great way to become familiar with the twitter community. Quiztime: https://twitter.com/quiztime.
Examples of advanced or difficult geolocation:
How to Crack Complex Geolocation Challenges: A Case Study of the Mahibere Dego Massacre, from Amnesty International Citizen Evidence Lab, by Martyna Marciniak and Sam Dubberley, 9 April 2021.
Mahbere Dego: Clues to a Clifftop Massacre in Ethiopia by Bellingcat, BBC Africa Eye, and Newsy.
John Doe 29: Image From FBI Child Exploitation Case Geolocated to Turkey by Carlos Gonzales.
Built to Last: China Secretly Built A Vast New Infrastructure To Imprison Muslims - A Pulitzer prize-winning BuzzFeed News investigation based on thousands of satellite images reveals a vast, growing infrastructure for long-term detention and incarceration, by Megha Rajagopalan, Alison Killing, Christo Buschek.
This article is part of a series of resources and publications produced by Exposing the Invisible during a one-year project (September 2020 - August 2021) supported by the European Commission (DG CONNECT)
This text reflects the author’s view and the Commission is not responsible for any use that may be made of the information it contains.