Scientists tell us that the world is ending, but when you look around, things don’t seem so bad! There may have been wildfires and droughts in California, sure, but Whole Foods still has my avocados, orange juice, and almonds. Maybe life, to paraphrase Jeff Goldblum, will just “ find a way ” no matter what we do to it.

It’s a nice thought, until you watch Network Earth–and realize that in nature, things tend to get only a little bit bad before, suddenly, the whole system collapses.

Based upon an esoterically titled research paper Universal resilience patterns in complex networks, Network Earth is a five-minute data visualization by Mauro Martino, head of the Cognitive Visualization Lab for IBM Watson, and Jianxi Gao, a researcher at Northeastern University.

It presents its staggering thesis visually: that network relationships connect all species on Earth. Remove one or two of those species, and the network compensates as if nothing has happened. Remove enough, and an entire ecosystem dies.

The project is what Martino calls a “data-film,” which is able to visualize complicated scientific and statistical ideas through animation.

“From my ‘director’ point of view, the paper is the script from which to develop a film,” Martino says. The ensuing challenges range from writing the voice over to coding the visualizations themselves. While Martino can produce impressive 2-D charts in a few days or even hours, as a director of the fully animated video he worked an undisclosed number of hours, eventually cutting half of all the content he generated.

“One of the biggest problems is always simplification,” he says, referring to both the script itself and the graphs he included, working in tandem to make the data approachable, but not crop the numbers out of the piece. So while, at 2:00 in, any data scientist will notice the bipartite network–a complicated graph showing the interrelationship of two groups–the narrative doesn’t attempt to label or classify the graph itself.