Ah, uncertainty. We don’t often talk about it, but no data is just absolutely rock-solid truth. That is not the nature of data. Data, and the maps we assemble from them, are models of a complex and multifaceted universe. Even the best most curated content has some amount of uncertainty due to collection unknowns, processing/editing/aggregation necessities, the problematic Boolean geometric nature of vector data representing soft or fuzzy phenomena, and just good old universal statistical honesty.
In this example, I’m using tornado paths as a catch-all phenomenon for illustrating uncertainty. I got this notion from Dr. Joseph Kerski, who uses tornadoes as an excellent example of a phenomenon who’s historic data has varying levels of…
- collection confidence (older tornadoes are based on approximations from eye witnesses vs the more precise radar-detected events of today)
- geometric confidence (we draw a simple straight line between the touchdown and dissipation points of a tornado but in reality they trace a sinuous and dangerously unpredictable meandering swath of destruction)
- scale confidence (the infinitely-precise line geometry representing a tornado’s path is not a perfect fit for its actually rather fuzzy zone of risk that dissipates with distance)
There are all sorts of reasons to consider uncertainty when communicating a spatial phenomenon. And there are lots of ways to visually represent that uncertainty in your map, if you choose. To dig deeply into this topic and learn about ways of modeling uncertainty, check out the related work of Ashton Shortridge.
In the meantime, let’s dig into just a handful of ways to brew some visual uncertainty into overconfident vector line features.
- 0:00 Introducing uncertainty. Notice the poor video quality and weird angle? Lightning struck the power line a while back and it zapped my laptop’s docking station (via the ethernet cable! who thinks of the ethernet cable!) so here I am via the grainy little webcam built into the laptop’s cover. We’ll both just have to live with that for a while.
- 0:24 Passing reference to Firefly as a method for point-feature uncertainty or fuzzy location. How about that cute little animation?
- 0:32 Passing reference to making polygons look uncertain or fuzzy.
- 0:38 The data! This is an extract of the tornado track data available in Living Atlas.
- 0:49 Dashed lines. Classic. Broken lines like this are a simple and intuitive indication of uncertainty. Think disputed borders.
- 1:11 Dotted lines. Just a dashed line with less ink. Often used for representing a transient phenomenon like a ferry path (less consistent and precise than a road) or the general path of a journey (remember those Family Circus comics where a dotted line showed Billy’s absurdly meandering path all over the neighborhood! I loved those).
- 1:46 Wavy lines. A wiggly line (ideally a somewhat random wave) indicates a lack of spatial certainty.
- 2:57 Blurry lines. Here’s how to make a fuzzy, glowy, blurry, firefly-ish sort of line effect.
- 4:06 Blurry/glowing things are a nice representation of Tobler’s First Law of Geography.
- 4:57 Lines with a broken edge or hatched fill. Filling the body of a line with a hatch, or dotted hatch, and giving it a wavy edge, can provide a wide range of visual representations of uncertainty that you can dial in to whatever visual indication of uncertainty works for your mapped phenomenon.
- 8:03 Using feature-specific confidence data as a variable driver of visual uncertainty.
- 8:48 Graduated transparency as in indication of uncertainty: more transparent = more uncertain.
- 9:22 Data-driven waviness using “symbol property connections.” The world is now your uncertain oyster.
- 12:09 Coupling transparency and waviness concurrently.
- 13:18 Hey look at that, the new docking station came in the mail in the middle of making this video so the closing section has a bit better resolution. Given the subject I don’t know if that’s a good thing or a bad thing.
- 13:21 An appropriately uncertain ending.