ClusterRadar: an Interactive Web-Tool for the Multi-Method Exploration of Spatial Clusters Over Time

TL;DR

ClusterRadar presents a client-side web-tool for multi-method spatial clustering over time, prioritizing temporal analysis and cross-method comparison. It implements Local Moran's I, Geary's C, Getis-Ord Gi/Gi*, and permutation-based significance entirely in-browser, with an interactive five-panel dashboard that links geospatial maps, time-series, density plots, and cell plots. The design emphasizes usability for non-experts, visual analytics principles, and privacy-preserving in-browser execution, demonstrated on US cancer mortality data and supported by user feedback. The work contributes a practical, extensible framework for exploratory spatiotemporal clustering, addressing visualization of large results, multi-variate spatial results, and temporal dynamics in a unified interface.

Abstract

Spatial cluster analysis, the detection of localized patterns of similarity in geospatial data, has a wide-range of applications for scientific discovery and practical decision making. One way to detect spatial clusters is by using local indicators of spatial association, such as Local Moran's I or Getis-Ord Gi*. However, different indicators tend to produce substantially different results due to their distinct operational characteristics. Choosing a suitable method or comparing results from multiple methods is a complex task. Furthermore, spatial clusters are dynamic and it is often useful to track their evolution over time, which adds an additional layer of complexity. ClusterRadar is a web-tool designed to address these analytical challenges. The tool allows users to easily perform spatial clustering and analyze the results in an interactive environment, uniquely prioritizing temporal analysis and the comparison of multiple methods. The tool's interactive dashboard presents several visualizations, each offering a distinct perspective of the temporal and methodological aspects of the spatial clustering results. ClusterRadar has several features designed to maximize its utility to a broad user-base, including support for various geospatial formats, and a fully in-browser execution environment to preserve the privacy of sensitive data. Feedback from a varied set of researchers suggests ClusterRadar's potential for enhancing the temporal analysis of spatial clusters.

Figures (6)

• Figure 1: Examples of how colors are assigned using the individual and aggregate color schemes. This plot shows examples of how the individual cluster assignments are aggregated into a single aggregate color. The aggregate color scheme visually represents the level of agreement between clustering methods. Locations are assigned a core color: red for "high-high" or "hot-spot", blue for "low-low" or "cold-spot". The final color is adjusted based on the presence of non-conflicting assignments: assignments that don't match the core assignment but don't directly contradict it either. A purple color is used for conflicts: a deeper purple for more significant conflicts. A yellow color is used for other assignments that do not fall naturally into the high or low cluster groups. Examples in this plot include total agreement in the main positive groups (a and b), partial agreement (c, d, e), minor conflict (f), major conflict (g), other (h,i) and not-significant (j).
• Figure 2: Examples of the density plot panel. The content of the density plot panel depends on whether a specific location is in focus. If so, the panel shows the local indicators (b), otherwise it shows the global indicators (a). Each density plot shows the empirical distribution of its respective statistic in light grey, the upper and lower significance cut-offs as dashed grey lines, and the statistic's actual value as a solid red line. If the red line is outside the dashed lines, then the statistic's value is significant. In the examples here, all global statistics are significant, and the Local Geary's C statistic is also significant. The Local Moran's I statistic is close to significant.
• Figure 3: Examples of the cluster assignments cell plot panel. The content of this panel depends on whether a specific location is in focus. If no location is in focus, the panel shows a) the global cluster assignments over time. If a location is in focus, the panel shows b) the local cluster assignments over time. In the latter case, an additional row is shown at the bottom representing the color assignments from the aggregate color scheme.
• Figure 4: Examples of the time-series plot panel. The content of the panel depends on whether a specific location is in focus. If so, the panel shows b) the global statistics, otherwise it shows a) the local statistics. Each time-series plot shows the value of the respective statistic over-time as a solid red line, with the currently visualized time point marked with a black dot. A dashed horizontal pink line shows the first value as a reference. The dashed grey lines show the upper and lower significance boundaries.
• Figure 5: The graphical tooltip, showing an example of the user hovering over a location. If the user has configured a geospatial name field, the top of the tooltip will show the locations name, otherwise it will show the location's geospatial ID. Below the name field is a time-series plot showing the location's value over time. Attached to the right axis of the time-series is a density plot showing the density of values across all locations and time points, with the current time point and location's value represented by a red line and text label. Below the density plot is a single-row cell plot showing the color assignments over time for the selected color mode. In this example, the color mode is the aggregate color scheme.