When are planes most likely to crash? And why? Our interactive visualization explores two decades of commercial plane crashes to reveal the patterns hidden in the data.
Flying is one of the safest ways to travel. Airplanes cause significantly fewer deaths per million people than sea, rail or road. But things do go wrong.
So what is the most common cause of air crashes? What is the emerging trend among the causes? And what phase of a flight is the most dangerous?
To explore the subject for BBC Future, we looked at every commercial plane crash from 1993 onwards and created an interactive visualization that revealed the patterns and connections hidden in the data.
We delved into the data from the last 20 years of commercial plane crashes. We researched everything from types of aircraft and location to cause of crash and even the certainty of cause.
This is an exploratory visualization that rewards time spent – you can find your own stories by diving into the data sets.
Using the controls, people can filter, sort and rearrange the data by five different causes of incident (for example, ‘Human error’ or ‘Mechanic’) and six different phases of flight (from ‘Grounded’ to ‘Landing’).
We needed to pack in a lot of on-screen information and overlapping bubbles make for an efficient use of space.
Clicking on a bubble reveals more background information for each incident such as the type of plane and the country the crash occurred.
This is a serious subject, but we wanted to offset a sombre background with bold colours that made the experience feel inviting and contemporary. Similarly, we spent a lot of time adding smooth, organic physics to the movements of the bubbles to provide a pleasing user experience.
When the visualization was complete, stories began to emerge from the data points. Some were expected, such as the two huge bubbles that appeared over the year 2001.
But intriguing patterns were also revealed. When we visualized causes of crash per year, the breakdown showed the colour yellow shrinking massively over the past two decades.
It highlighted the fact that human error is largely being removed by automation. Something that could have interesting implications for the fast-growing driverless car industry.