It was a data nerd kind of Sunday.
I flew to Washington D.C. for an Edward Tufte training course. I brought my laptop because I finally finished all the #MakeoverMonday Challenges I missed. I couldn’t fall behind again! I didn’t feel too bad about hanging out in the hotel room because I lived outside of D.C. in the past and it was a pretty rainy day.
After doing my Global Arms Trading #MakeoverMonday, I headed down to the hotel bar. There was a group of Korean War Veterans there on an Honor Flight trip. I had several beers with 88-yr old Peter from Fargo, Minnesota. Awesome guy!
I went back to the room and was inspired to finish my first #RevizProject viz. You can check them out here. The data set was Tornadoes in the U.S. over 66 years.
I started off with one of my current favorites, the panel/trellis chart.
I really liked this start to my process. You could look at the # of tornadoes, fatalities, or injuries (via a parameter) over 66 years. I could immediately see the spring pattern of tornadoes. I could see there were some major storms out there if I wanted to focus on just the story of one of those, but Matt already did that with his Reviz.
This stayed on my dashboard for quite awhile! Ultimately, I needed more room for other charts so this got ditched in the end.
Then, I found a YouTube video of an F5 tornado in Tuscaloosa, Alabama. I knew I wanted to include this. I used colors from the static video image as the inspiration for the colors on the dashboard.
Now, it was time to add some interest. I confess this part is not dataviz best practice, but then again I am not at work, why not branch out a little? I wanted to make a “tornado” chart. I thought I could modify a bikini chart a little to get this effect. I thought the data would suit this well.
This is a dual axis line chart. State is on rows and the measure is on columns. State was also used for the path. The 2nd measure pill is on a reversed axis. There is a single reference line with fill below to fill between the lines. I dialed the line size up to cover the end of the reference bars. If you download the workbook, this one is still on there if you want to play with it.
It kind of worked. I saw a tornado.
I tried a hover action to display the labels.. I tried a filter action to display a filled State map – I didn’t like this in the end – so it too got ditched.
My next thought was maybe there is enough data here to do a radar/spider chart. Andy Kriebel just made a video on how to make one of those. I watched it. I can do that, now.
It kind of worked. No, let’s be honest. This doesn’t work at all!
I got the the radar to work (top half), but I just couldn’t see all the detail I wanted.
I briefly experimented with a vertical radar chart (bottom half) using index along year to try and create the look of a tornado.
It did not even kind of work. As you would expect, you can only see part of the ring.
This is how it usually goes with any viz I create. I make about ten views that never make it into the final dashboard.
I ended up creating a heat map calendar that took up way less room than the panel chart. It also more clearly communicated the patterns I saw on the line chart. So I pointed that out with a clever title.
I recently saw Matt Chamber’s viz using hexbins on this same dataset. I thought it was a great idea and I wanted to learn those functions so I stole Matt’s hexbin map. Thanks Matt! Check the original out.
I think this is awesome. It clearly shows the patterns based on lat/long where the tornadoes are occurring.
To round out the dashboard, I added some context around tornadoes, the Fujita Scale that rates their intensity, the YouTube video and an area chart. I had several iterations of the line/area charts. I ended up with 10 year moving averages based on the Fujita scale rating (0-5).
How do you think it turned out? Let me know in the comments.
Click the image below for the interactive version. Thanks for reading.