More than 30,000 fires in non-residential areas per year happen in Ukraine in average and this is largest number for Europe. To show a scale of a problem and to prove most of them are artificial we have created a program that automatically finds and shows satellite photos of places on fire. The visualization currently covers the years 2018-2021 and contains around one thousand such photos. Over time, each week the program adds new photos with fires.
The fire kills animals, insects, and bacteria, it destroys fertile black soils. Millions of tons of carbon dioxide and other harmful gasses from fires worsen the air we breathe and makes global climate crysis even worth.
Ukraine have the largest number of fires among all neighboring countries. Checking out our interactive visualization users can see what these fires look like. Satellite images in our project show that many such fires are created intentionally (fires appear simultaneously in many parts of the field).
Farmers who burn stubble are involved in targeted arson. In fact, this method is guaranteed to quickly kill blacksoils. In addition, these fires can often spread to nearby forests or settlements. And they are a direct threat to human life.
Our visualisation clearly shows the scale of the fires, partly created intentionally, and so points out the harmful effect from fires on arable land. Together with visualisation we add to the article comments about legal grounds of such activity and comparison to situation in other countries. Ukrainian society have to pay more attention to the problem in order to avoid harmful and moreover irreversible effect on arable land, which provide a huge part of national economy.
We obtain data for fires using the FIRMS service, which uses the results of monitoring the earth’s surface by NASA / NOAA satellites. The satellites fly over every point on the Earth’s surface twice a day, and their infrared sensors can detect heat sources in an area of about 400 by 400 meters.
After receiving the data, we group the points with fire and select only clusters larger than a certain size to cut off minor / accidental heat sources and sensor errors. We also reject heat sources in industrial areas, which are present there almost constantly. We use the coordinates of each selected cluster as the coordinates of the fire, along with its date. All processing is done with a python script.
After that, knowing the coordinates of the fires, our program tries to find and download a photo for this area, taken at the right time. To do this, we use Sentinel-2 satellite data from the European Space Agency. If a needed data for a place and time is available, we process it with a special algorithm to “highlight” the areas with flames in the pictures and to create composite RGB images to show in our project.
All data are obtained, processed and served with a custom-build python scripts. Sentinel-2 data obtained thru Google Earth Engine cloud service. Processed images and metadata are served trough API created in Django web framework. D3.js library is used to create interactive visualization.
What was the hardest part of this project?
The hardest part of a project was a method to find and process thousands of files with raw satellite data from which images created later. Although to obtain one such “image” is a relatively easy task, it’s much more harder to do it for such a batch processing as in our case. After many attempts — we tried at least five different services and libraries, at last we choosed Google Earth Engine which in our opinion currently is the best platform for custom mass processing of satellite data, ready to use by teams with limited budgets.
Second problem: in the original data there are no such zones of flame that you can see on visualization. We use special algorithm to highlight flames on images to quickly locate the fire. However, the zone of flame, which is completed by the algorithm, clearly corresponds to the places where the combustion takes place (this can be seen in the pictures).
As far as we know this is a first such a project with constantly updating large collection of recent satellite images with fires.
What can others learn from this project?
Using open source data (in this case from European Space Agency) and different open source tools with a little bit of creativity allow journalists to visualize almost any story where remote sencing is involved, even on very low budget.