The Imperfect Storm


You can play with a full-screen interactive storm map in the blog post here. Read on for details.
As I write this on a Saturday morning in mid-March, a major storm is bearing on New Zealand. The intense depression that was formerly Tropical Cyclone Lusi made landfall late last night and will spend the day passing south across the country. For the last 24 hours folk have been sharing EarthWindMap screenshots through social media. EarthWindMap transforms US National Weather Service’s Global Forecast System data into dancing DayGlo particle streams. The results are extraordinary, enlightening and beautiful.

EarthWindMap March 15 2014

I love how these maps convey the nature of complex atmospheric systems through stillness and motion. Conventional weather maps are difficult to interpret until you learn the visual conventions of meteorological cartography. In thinking through how these maps so successfully communicate complexity, I am reminded of something Paul King wrote recently in response to a Quora question on what UI/UX designers can learn from neuroscientists.

“While we appear to take in the whole visual field as a single complete experience, neuroscience and cognitive science shows that it doesn’t work that way. What we actually process is a narrow region that moves around, and these regions are processed in sequence through visual exploration. So thinking about this narrow window of attention, how it moves, and how the user will interpret and organize information sequentially can help to design more intuitive and effective UIs.”

As EarthWindMap’s swirling currents guide my eyes over familiar landmasses, my mind starts weaving stories about what my eyes see. A  narrative about the state of the world’s weather develops as my eyes sweep through different parts of the map. Country outlines are known reference points. Streams of activity are narrative arcs. Areas of stillness offer stable environments where I can catch my breath. The cartographic design promotes visual exploration gives me a sense of the whole even though I am only ever looking at one small part of the map.


EarthWindMap offers an unparalleled realtime view of the world’s weather. However, I wanted to get a sense of historical storms layered on top of one another. What would it look like if we overlaid many decades of tropical cyclone paths one of top of the other. After a bit of hunting and experimenting I found a Historical Tropical Cyclone Storm Tracks dataset from the Pacific Disaster Center.

“Tropical cyclone storm tracks are charts and projections of severe storm and hurricane/typhoon movement. This layer is updated annually and depicts historical storm tracks starting from 1940 through 2012 for the Pacific (Western, Eastern, and Southern), Indian and Atlantic Ocean regions. Only storms that were classified above a tropical depression in strength, wind speeds > 30 mph for 2 or more 6-hour periods, have been included. Information includes charts on the track of the storm, tracking position in latitude and longitude, maximum sustained winds in knots, and central pressure in millibars.”

I reviewed the data and realised that it is incomplete. For example, Cyclone Bola, a tropical storm that devastated New Zealand’s East Coast in 1988, is conspicuously absent. I created an interactive map in any case with TileMill. Despite the data being imperfect—what data isn’t—I think that the resulting map is interesting enough to share for the questions it raises. For example, why are there so few tropical cyclones in South America?

Full-screen version here.

One Response to “The Imperfect Storm”

  1. Ed Summers says:

    Thanks for this Chris. I hadn’t seen EarthWindMap (wow) and the connection with that connection with UI/Neuroscience is really interesting. The narrow & roaming field of attention kind of reminds me of what reading text is like as well.

    I wonder if the lack of any storm activity around South America is the result of missing data from Historical Tropical Cyclone Storm Tracks? It does seem hella odd. It’s interesting to see how many storms track similarly. It would neat to be able to hover over a particular track, and have it get highlighted somehow, with an overlay with information about the storm. Or perhaps, color code them by the speed, or intensity of the storm.

Leave a Reply

Your email address will not be published. Required fields are marked *