I was watching a video about spatial statistics this week and I learned that the beloved heat map is not the same thing as a hotspot map. I don’t expect a kid to ever ask me the difference between a heat map and a hotspot map. But for me, learning about their difference was another cool way to understand how statistics, GIS, and cartography work together to help eliminate the subjectivity of maps.

Who knew?

WHAT ARE HEAT MAPS?

• Heat maps illustrate the clustering (density) of a common variable, or variable with a common value. Heat maps look for these patterns and show us where they occur the most (i.e. red=most intense, blue=least intense).
• Heat maps can hold different values, such as temperature, location of ATM machines, or traffic delays on roads.
• You can visualize heat maps as a gridded chart too (though not as fun). Ex: Think hours of the day on your X axis and number of customers entering your coffee shop on your Y.
• Heat maps are sensitive to scale. As you zoom in, the less density per land area thus less intensity is illustrated. Zoom out and the opposite occurs. Go ahead and try it with a Google traffic map.
• While heat maps can usually result in something that is transferable into an isoline map, we typically call those Isarithmic maps instead. They use nice wavy lines to connect areas of common value. I love me a good isoline map.
• In the case of heat maps, the cartographer has greater ability to manipulate the visual representation by determining the number of classes that the values are depicted (i.e. red=20 ATM machines or red=5 ATM machines, it is ultimately my choice).

SO WHAT ARE HOT SPOT MAPS?

• I can plot the same type of variables using a hot-spot map, but for a different purpose. Why would we do this? To prove that our clusters are not random.
• Hot spot maps are meant to show the level of variance from an area’s expected values, given a random distribution. GIS Lounge breaks it down more.
  • This one gets a bit more techy, but heat maps are meant to find densities of a phenomenon compared to a completely random distribution. It unfortunately deals with probability statistics, but maps help you see it better.
  • Basically I would be looking for areas that show high or low standard deviations from the mean of the study area.
  • In other words, if I see a clustering of coffee shops in my city and they are surrounded by red (intensity again), my hot-spot map is most likely telling me that there is statistical significance that those coffee shop clusters are not there at random, and basically yes, that the clustering I see is in fact real and purposeful. On the flip-side, areas of low clustering when compared to a randomize distribution would show me blue because there is low statistical significance of a cluster happening there.
• Because each plotted spot is given a significance level and the purpose was not to illustrate density, hot-spot maps are not sensitive to scaling.
• Because I am running my plots against a completely random distribution (that amazing geo software does for you now), I eliminate some of the subjectivity that goes into what my end visualization looks like.

Other sources:

• ESRI has a decent little story map comparing the two.
• ESRI: Finding Hot Spots in ArcGIS Online: Minimizing the Subjectivity of Visual Analysis. PDF Notes

Anyway, did you know that? I didn’t.

Let me know if I got something wrong, can fix, or you want to add 🙂

H.I.