Uncertainty-aware geospatial visual analytics
Name
ica-abs-10-182-2025.pdf
Size
1.07 MB
Format
Adobe PDF
Checksum (MD5)
59d4fb040216c60d8ec33af6c61748c7
Author(s)
Xiayin Lou
Peng Luo
Liqiu Meng
Date Issued
December 12, 2025
Journal
Abstracts of the ICA
Publisher
Copernicus GmbH
Citation
Lou, X., Luo, P., and Meng, L.: Uncertainty-aware geospatial visual analytics, Abstr. Int. Cartogr. Assoc., 10, 182.
Version
Final published version
Abstract
Visual analytics (VA) is a paradigm for extracting novel insights through visual perception, cognition, and reasoning. It plays an important role in various applications, including medicine, digital humanities, urban planning, environmental sciences, etc. In the field of geography, visual analytics has already become a powerful tool for identifying the spatial patterns of geographic variables. Geospatial visual analytics (GeoVA) offers an intuitive insight into spatial phenomena, enabling people to understand the spatial processes deeply. However, geospatial visual analytics can be misled without an awareness of uncertainty. On the one hand, the existence of uncertainty can result in a biased interpretation of geospatial data. For example, for each data point, the importance scores of its explanatory variables, including uncertainty values, can be calculated using explanation methods (e.g., BayesSHAP) (Slack et al., 2021). Figure 1a illustrates the data points with the variable contributing the most, which are represented by different colors. When the uncertainty is low, it is easy to distinguish which variable has a higher contribution. Nevertheless, in the case of high uncertainty, the overlapping uncertainty intervals in the explanation make it confusing which variable is more important. On the other hand, the uneven spatial distribution of uncertainty will also cause geographic bias in analyzing geospatial data. The uncertainty can vary spatially in different regions, leading to different interpretations for similar spatial patterns. As shown in Figure 1b, the data points are divided into four areas according to their uncertainty levels. In the regions with low uncertainty, the visualization reflects the actual situation about a model, so the viewer is able to make a correct reasoning. In contrast, in the regions with high uncertainty, the spatial patterns perceived visually are not reliable, and viewers will misjudge. The misinterpretation caused by spatially varying uncertainty is a so-called geographic bias problem. For this reason, no geospatial visual analytics can be effectively conducted without considering the uncertainty problem (Andrienko et al., 2010).
MIT Department
Senseable City Laboratory
Terms of Use
Creative Commons Attribution
Persistent DSpace Link
DOI of Published Version
https://doi.org/10.5194/ica-abs-10-182-2025
