Science Wednesday: visualizing uncertainties across space

December 1st, 2021

Across the information products of CSIRO’s Digiscape Future Science Platform, it’s probably fair to say that we’ve found our users to be more focussed on getting some kind of answer than on assessing the uncertainty in the answer they obtain. That state of affairs won’t last.

The first step in assessing uncertainties is being able to see them: and for spatial data, that is a non-trivial problem. Today’s paper describes Vizumap, an R package developed by an Lydia Lucchese of the Australian National University, Petra Kuhnert of CSIRO and Chris Wikle of the University of Missouri. Vizumap uses a range of different visual cues (hue, intensity, glyph rotation and animation) to show users the uncertainties of spatial predictions simultaneously with their central estimates.

The paper is here, and here is the Git.

Bivariate map: hue and intensity show different variables

Pixel map: regions with higher uncertainty appear more pixellated

Glyph map: colour shows the estimate, rotation the uncertainty

Probability of exceedance: colour shows a probability