Reframing Spatial Uncertainty Visualisation with a Layered Grammar of Graphics

Abstract

Communicating both estimated values and their associated uncertainties presents a particular challenge in spatial visualisation because spatial coordinates already occupy the primary visual channels, leaving relatively few channels available for encoding uncertainty alongside the estimated values. A range of visualisation techniques has been developed to address this challenge, including bivariate choropleth maps, pixel maps, and glyph maps. However, existing implementations typically treat these techniques as dedicated visualisations rather than expressing them through the composable elements of a grammar of graphics. We reframe uncertainty visualisation as an extension of the layered grammar of graphics, allowing uncertainty-specific representations to be composed, customised, and extended alongside other graphical components. We demonstrate this perspective by integrating uncertainty visualisation into the layered architecture of ggplot2, accommodating these techniques while also motivating new approaches such as the dual map for selective attention between estimated values and their associated uncertainties. We implement the framework in the R package ggincerta and demonstrate its application through three case studies.