Updated to a Tableau based interactive dashboard that now combines both the 1950-2015 US dataset from the Storm Prediction Center, and the 1980-2009 data set from Environment Canada!
Click on the image below to bring up the visualization in an new tab!
While Microsoft’s Power BI continues to mature as a visualization tool (with some very impressive capabilities), the monthly updates have not addressed many of the issues with mapping aesthetics/readability that were hampering my original dashboard.
Tableau suits this project very nicely, so it is now the visualization tool. This update also merges the Canadian Tornado dataset, recently complied/released by Dr. David Sills of Environment Canada.
I believe this is the first visualization of the combined dataset available on the web! I should point out that during the data-wrangling process I needed to discard about a hundred records from the Canadian dataset that did not have a day as part of the date information.
Tableau allowed me to customize the mapping to maximize readability and interactivity. Key points for users are:
- A combination of dropdown boxes and exposed filters support selection of Year, Month, or Day, as well as Path Length and Path Width groupings. Multiple selections can be made.
- Each dot represents the starting lat/lon for the event and is colour coded by F-Scale rating. Cling on the colour-based legend to the left of the map will highlight that category (and holding <CTRL> will allow selection of multiple categories).
- Mousing over any data point brings up a box with further details on the event, including the date, the lat/lon starting point assigned, and actual path length and width in both imperial and metric units.
Lots of opportunity for interesting investigations/insights from the data. Some I find useful include comparing specific weeks or clumps of days (or even months) from year-to year, or season-to-season comparisons of regional shifts in higher end events (filtering on path length/width or damage rating). I’m eager to hear in anyone has suggestions for looking at other dimensions in the data (grouping by season/decade/other).
Image below is a visualization of all F5 events in both the US and Canadian data sets highlighted (but all other events still depicted). Even in this extreme example the visualization is easy to look at and easy to digest – the relevant information stands out much better with Tableau! Also note that the mouse over data is still active for non-highlighted events, which can be handy with a less cluttered selection mix! The mouse over data in the image below is from the Super Outbreak event at Xenia, Ohio.