If you’d prefer a fully written, non video (who could blame you), uncharacteristically formal introduction to bivariate and multivariate maps, you’ll find that in this article from the GIS&T Body of Knowledge which is a somewhat encyclopedic resource for GIS educators and enthusiasts.
If you are ok with delving into a rip roaring wild meandering breathless journey through just a small slice of the information-dense world of bivariate and multivariate mapping then press play my friend…
Part 1
What is a bivariate or multivariate map? Why would you want to make one? What do the legends look like? What are some alternatives to a multivariate map?
0:00 Nerd alert!
0:22 Getting some obligatory definitions out of the way. Bi (two) variate (data variables).
0:33 Show a relationship. Each layer of data give the others context. Where are they in/out of synch? Connecting the dots. That sort of thing.
1:10 Careful not to blow out your audience’s cognitive load. Find a balance.
1:31 Is it ingenious or burdensome? More complex maps mean more effort for your audience.
2:17 Building a simple bivariate map by just stacking two symbology dimensions on top of each other.
2:55 Applying two data variables to a single symbol.
3:21 Value by alpha. Using transparency to represent an underlying value (like population) while using color to show another value (like health measures).
4:08 Bivariate choropleth maps (aka relationship maps). Calculating a grid of values (and colors) for two variables. Great for showing where phenomena are low or high overall and where the two phenomena differ.
4:40 Some tips about bivariate choropleth legends. The corners are the most interesting parts.
5:03 Disassembling a bivariate choropleth map into just its extremes. Extreme!
6:00 Always choropleth? No way! You can get reallll crazy. Here’s a Chernoff map.
6:26 What about smooshing a value by alpha method into a bivariate choropleth method? An abomination? Nah, it just might work! The important thing is to try it and see.
6:42 The same data can be multivariat-ized in lots of ways. Literally a kajillion.
7:15 If you stack up some graduated symbol layers for a multivariate map you definitely will have issues with occlusion (stuff covering up other stuff). Well not if you offset them radially like the petals of a flower.
7:25 More thoughts on making understanding your multivariate map easy for your audience by disclosing it narratively like in a StoryMap.
8:51 Stuck with a static legend? Think about some extreme cases and use them as a handful of for instance style legends.
9:24 Sometimes you throw caution to the wind to make a bit of multivariate map…art.
9:54 Doesn’t always have to be multivariate! There are alternatives, like animation or a small multiple.
10:47 Getting a bit crazy…with a trivariate choropleth map. I told you we were diving deep!
11:35 Even weirdly complex multivariate maps can be craftily communicated to an audience, if you keep in mind all the options you have for explaining them:
-a narrative structure like a StoryMap
-layers that are revealed progressively to build up the concept
-use plenty of helpful legend-like graphics
-label interesting parts of the map directly
12:01 Interesting relationships is the name of the game when it comes to multivariate maps.
Part 2
That was a lot of examples and chatter. What about digging in and actually making some of these things? We learn best by doing, so here are a handful of ways to make multivariate maps in ArcGIS Pro.
0:00 Who dresses this person?
0:24 We’ll use one layer to make all these multivariate maps. The UN World Happiness Index (more info here; I’m using old data from 2017 which you can steal).
0:59 Did you know you could tie thematic symbology to a polygon’s fill or its outline? Or both? That could be interesting!
2:00 Check out the “vary symbology by attribute” tab in the symbology panel if you seek multivariate powers.
2:29 Making a value by alpha map to include uncertainty in your map.
4:17 Stacking up many instances of graduated symbology, each mapping a different variable.
5:57 Radially offsetting graduated symbol layers avoids problems of overlap and creates an interesting symbol bouquet (here’s how to use the symbol’s x/y offset option).
6:44 Chart symbology is a ready-to-rock-and-roll multivariate option.
-Bar charts at 7:00
-Stacked bar charts at 7:31
– Pie charts at 7:44
8:05 I tricked you into learning about redundant cues! That means using multiple visual dimensions showing the same variable. Like waving your hands while you raise your voice. While this is somewhat the opposite of a multivariate map, you can use redundant cues in a more complex multivariate map to really lock in a particular variable.
9:12 Using the visual dimension of color to add another variable to a graduated symbol. Again, good old “vary symbology by attribute.”
10:10 Bivariate choropleth mapping in ArcGIS Pro (called “bivariate colors” in the primary symbology droplist)
11:28 Are you ready to have your mind blown? “Symbol property connections” let you tie virtually any visual option of a symbol to data or a formula. It’s bananas. Warren Davison even tied a hachure graphic’s size and thickness to things like steepness and angle in this masterpiece of digital reconstruction (half way down).
Ok, there you go, just a small sampling of multivariate mapping options in ArcGIS Pro. Seriously, when you activate symbol property connections you just plug right into the Matrix; you’ll be bending spoons and learning kung-fu and stuff, but like in a cartographic way. Also check out the work of Ken Field, who manages to sneak at least two visual dimensions into nearly ever map he makes.
But what about the web?
Part 3
Can you make multivariate maps in ArcGIS Online? Come on.
0:00 Couldn’t be bothered to comb that hair?
0:46 Overlaying two layer copies (of labor force data from Living Atlas) of the same enumeration unit (counties, in this case), but using different symbology.
1:50 Baking color for one variable into a graduated symbol of another variable. Underlying population drives size, labor force drives color.
3:16 Including transparency as another visual dimension, to show uncertainty. This is important, people!
4:33 Making a relationship map (trusty old bivariate choropleth).
Annnnnd if you dig into writing expressions, there is no end to the visual dimensions that you can harness in ArcGIS Online. Check out Lisa Berry’s work for a smorgasbord of multivariate mapping in ArcGIS Online.
Happy Multivariate Mapping! John
Article Discussion: