What doesn't help readers (on the chart) and what does help (off the chart)
A Tufte fix for austerity

Raining, data art, if it ain't broke

Via Twitter, reader Joe D. asked a few of us to comment on the SparkRadar graphic by WeatherSpark.

At the time of writing, the picture for Baltimore is very pretty:


The picture for New York is not as pretty but still intriguing. We are having a bout of summer and hence the white space (no precipitation):


Interpreting this innovative chart is a tough task - this is a given with any innovative chart. Explaining the chart requires all the text on this page.

The difficulty of interpreting the SparkRadar chart is twofold.

Firstly, the axes are unnatural. Time runs vertically, defying the horizontal convention. Also, "now" - the most recent time depicted - is at the very bottom, which tempts readers to read bottom to top, meaning we are reading time running backwards into the past. In most charts, time run left to right from past to present (at least in the left-right-centric part of the world that I live in.)

Location has been reduced to one dimension. The labels "Distance Inside" and "Distance from Storm" confuse me - perhaps those who follow weather more closely can justify the labels. Conventionally, location is shown in two dimensions.

The second difficulty is created by the inclusion of irrelevant data (aka noise). The square grid prescribes a fixed box inside which all data are depicted. In the New York graphic, something is going on in the top right corner - far away in both time and space - how does it help the reader?


Now, contrast this chart to the more standard one, a map showing rain "clouds" moving through space.


(From Bing search result)

The standard one wins because it matches our intuition better.

Location is shown in two dimensions.

Distance from the city is shown on the map as scaled distance.

Time is shown as motion.

Speed is shown as speed of the motion. (In SparkRadar, speed is shown by the slope of imaginary lines.)

Severity is shown by density and color.

Nonetheless, a panel of the new charts make great data art.




Feed You can follow this conversation by subscribing to the comment feed for this post.


I've never seen these charts, but I like them much more than the video screenshot you posted. These SparkRadar charts seems to do a good job of plotting the relevant parts of a radar chart in a single chart, while the alternative requires a video. The key trick is replacing location with distance, which is generally what we care about. Basically, is a storm nearby? and is it coming closer? are much more important that asking which direction the storm is coming from.

You also mentioned that the New York includes something "far away in both time and space", but your alternative Baltimore chart actually includes the weather over Ottawa, Canada over 500 miles away!

Most importantly, there are challenges with viewing and interpreting video charts. Even here, you've posted a screenshot of the video without a link to the video itself. Thus, the reader is forced to look at location as the two dimensions, with no sense of time. Similarly, these are hard to include in reports and presentations. Even when they work, I don't like needing to watch N loops of a video trying to piece together the key information I need: how far is the storm from me and how does the distance change over time.

John Roth

I agree with Brian. What I want to know is how close the storm is and when I need to worry about getting wet. Flipping the time axis would be helpful - I'm used to seeing time from top to bottom on schedule charts, for example, rather than left to right.


Yeah, I have to say that I find these intriguing.

They are somewhat disorienting at first, but not terribly complicated to understand.

They desperately need some more finesse in the design choices.

I think it's clear that I can get more information at a glance from the spark radar than I can the map, but I can obviously get a more complete understanding by watching the animated map.

Neither display fills the role of the other, and I think at the very least there is room for both methods of looking at this information.

Either way, they're an interesting concept.


"Firstly, the axes are unnatural. Time runs vertically, defying the horizontal convention."

I use my scheduling calendar in outlook with a vertical axis every day and I don't find it unnatural.

What I find lacking in the spark radar is the ability to tell if the storm is actually going to reach my location or just get close and then pass on one side or the other.

The second challenge is that storms move from west to east (in my geographical location). It would have taken me less time to understand this chart if the horizontal axis had been reversed.

As jlbriggs noted, more finesse in the design choices is also needed. But overall this is innovative and should not be dismissed as merely "art".

The comments to this entry are closed.