The pandemic disruption of air traffic — as seen from Space
Monitoring flying airplanes with satellite data
The COVID-19 pandemic led to a massive drop in air traffic across European skies. ‘RACE’, a tool resulting from the joint cooperation between ESA and the European Commission, allows users to monitor the change of flying airplanes relatively to pre-pandemic periods using satellite data.
COVID-19 lockdown restrictions, have impacted the operations of air carriers, airports and air navigation service providers putting many aviation industry jobs at stake. Air traffic is thus a good indicator to unveil human and economic activities affected by the ongoing pandemic. This is one of the indicators that can be accessed at the RACE (Rapid Action coronavirus and Earth observation) online platform which allows anyone to access data collected from satellites without registration needed.
This indicator can be selected on the left side panel — which will show the different airports marked in blue🔵 or red🔴 dots depending on their current status of air traffic comparatively to reference values (July 2018 to December 2019).
Airport traffic
The decrease of ‘airport traffic’, is displayed on the right side panel, upon clicking on each airport. These graphs, present the average monthly airport count detected on an area of about 6,000 km2 centered on each airport. The values go back to mid 2018, and these are used as reference values to understand and compare the amount of flying airplanes in respects to pre-pandemic times. On each graph, the time of ‘high restrictions’ — shaded in light red, show a clear match with a strong decreased on flying airplanes — marked in red bars.
Detected airplanes
The ‘detected airplanes’ can be accessed clicking on the right tab above the air traffic graphs. This allows to visualize the location of detected airplanes (marked in yellow circles) by date of observation. But how are the airplanes detected?
The detection is made based on Copernicus Sentinel-2 satellite data. This satellite carries onboard a multispectral sensor which collects images from our Planets’ surface on different bands of the electromagnetic spectrum. When looking at a true-color image of a flying airplane, we will see it in three different colors (red, green and blue) like the image below.
What happens is that, an high-speed and high altitude object is seen in those three colors and apparent different locations — as the three bands of the visible light will see the airplane from different viewing angles. This is what is called the ‘parallax’ effect — brilliantly explained on this medium post by Tyler Erickson. The detection of flying airplanes, was then automatized by deep learning algorithms. In other words, images of Sentinel-2, were individually analyzed and pixels belonging to the center of an airplane classified as such. This segmentation problem was then carried out with a Fully Convolutional Network (FCN). This approach, developed by Mauricio Pamplona, Rodrigo Minetto and Sudeep Sarkar, USA, was one of the winning entries of the Sentinel Hub contest which was then upscaled to entire Europe.
