Y. Drocourt, R. Borgo, K. Scharrer, T. Murray, S. I. Bevan, and M. Chen. Temporal visualization of boundary-based geo-information using radial projection. Computer Graphics Forum, 30(3):981-990, 2011. (Presented in EuroVis 2011.) DOI.
This Case Study Report was authored by Rita Borgo, King's College London, UK
This case study analyses the visual analytics approach presented in [1]. The work addresses the challenge of visualizing a ten-year record of seasonal and inter-annual changes in the frontal position of nearly 200 marine-terminating glaciers distributed along the Greenland coastline. The dataset was derived from Landsat satellite imagery and contains irregular time series describing glacier advance and retreat relative to a reference terminus position. The goal of the visualization is to enable glaciologists to analyse spatial patterns and temporal changes in glacier behaviour simultaneously, facilitating scientific understanding of glacier dynamics and their contribution to sea-level rise.
Prior to the development of the proposed visualization, glaciologists relied primarily on two forms of visual analysis. The first approach consisted of time-series plots representing the relative frontal positions of individual glaciers over time (see Figura1-(a)). While effective for examining temporal change of a single glacier, the plots did not support the preservation of spatial relationships (geographical position, proximity etc.) making spatial comparisons difficult. The second approach used color coding in the form of map-based glyph visualizations where glacier changes were represented using colour-coded symbols on geographic maps (see Figure 1-(b)). Although these provides spatial context, it requires multiple juxtaposed views to analyse temporal variation, moreover it is prone to visual clutter when several glaciers are displayed. Furthermore, empirical studies have shown how evaluating temporal changes from colour-coded pixel-based maps imposes a high cognitive load on human observers, leading to reduced accuracy and slower response times [2].
Figure 1. Two classic visualizations in geography: (a) plotting the relative frontal positions of 10 calving glaciers as time series, and (b) using colour-coding glyphs to represent different scales of relative frontal positions, and visualizing temporal changes by juxtapositions.
Here, we revisit our design process retrospectively using the SCORE* framework. The baseline workflow exhibits a key symptom: scientists (glaciologists) experienced high cognitive effort when attempting to analyse both spatial and temporal patterns in glacier dynamics (Vis-High-Ct, Int-High-Ct). The underlying cause is that existing visualizations separate the spatial and temporal dimensions into different representations, forcing analysts to integrate information mentally. This results in low alphabet compression and increased cognitive cost during the analytical process (Vis-Low-AC).
To address this issue, the design team introduced a remedy in the form of a radial visualization that integrates spatial and temporal information into a single representation (Vis-High-AC). The design leverages the geographical characteristic that glacier termini lie along the coastal boundary of Greenland. By treating the coastline as a one-dimensional boundary embedded in two-dimensional space, the visualization reduces the spatial dimension and maps glacier locations onto angular coordinates around a circle. Time is represented as concentric rings radiating outward from the centre, where each ring corresponds to a different year. This radial projection allows the visualization to display nearly 200 glacier time series simultaneously while preserving their spatial ordering along the coastline. This approach makes a shift from one suffering from Vis-Low-AC to one that increases Vis-High-AC to reduce Vis-High-Ct in the analytical process.
A naive radial mapping would project glaciers to a circle according to their original x-y positions. Because the coastal boundary is not convex, the correct order alone the coastal line is not assured. The cause of this issue is Vis-high-AC in terms of ordering. A side-effect of the remedy is thus that increasing Vis-High-AC risks introducing Vis-High-PD. This is likely due to the abstraction potentially impacting geographic relationships. In [3], the authors highlight how increases in alphabet compression often come with increased potential distortion, this is possibly an example of this trade-off.
This introduces a new symptom related to loss of geographic context. The cause is the distortion introduced by uniform angular mapping. To mitigate this issue, we developed a spatial mapping algorithm that preserves the natural ordering of glacier termini along the coastline. In a first step the algorithm snaps glacier positions to the coastal boundary, then computes traversal distances between neighbouring glaciers. These distances are then used to distribute the glaciers in angular space while maintaining neighbourhood relationships.
Despite this improvement, glaciers spatially tightly clustered can lead to densely packed radial axes and introduce overlapping or visual clutter (Vis-High-Ct, Vis-High-PD). To address this symptom, we introduced an angular relaxation algorithm enforcing a minimum spacing between neighbouring axes while maintaining key spatial reference points (Vis-Low-PD). These reference points correspond to glaciers located near cardinal directions (north, south, east, west), which serve as anchors for maintaining geographic orientation. While this relaxation process improves readability and reduces clutter, it introduces a small degree of spatial distortion, representing a typical trade-off between alphabet compression and pattern distortion within the framework [3].
A further challenge concerns the effective representation of glacier advance and retreat values. Colour-only encoding can make it difficult to distinguish magnitude and sign of changes, especially when galciologists compare multiple time steps (Vis-High-Ct). To overcome this limitation, the final visualization design, as shown in Figure 2, uses diverging colour schemes combined with geometric encoding. Positive values representing glacier advance are shown using blue tones, while negative values representing retreat are shown using red tones. The magnitude of change is represented through amplitude or thickness in radial stream-like plots. Two final visual styles were proposed: an area-based radial plot and a tube-style representation that emphasises magnitude through thickness (reducing Vis-High-Ct, while still maintaining sufficient Vis-High-AC for overview).
Figure 2. The final visual design, supported by several visualization algorithms.
The resulting visualization enables scientists to observe the behaviour of all 199 glaciers across ten years within a single integrated display. Domain experts reported that the visualization provided a clear overview of glacier dynamics and helped identify large retreat events and regional patterns that were previously difficult to detect. The visualization also helped identify anomalies in the processed data, demonstrating its usefulness not only for analysis but also for data validation.
This case study illustrates how a visualization design approach can optimize a scientific workflow by improving the balance between alphabet compression, pattern distortion, and cognitive cost. By exploiting the boundary-based spatial structure of the dataset and integrating spatial and temporal information into a radial representation, the visualization significantly improves the efficiency of glacier behaviour analysis compared to traditional methods. Fig.ure 3 summarizes the workflow for improving the domain experts' workflow by introducing a new visual design supported by visualization algorithms.
Figure 3. The workflow, which has three iterations in visual design and algorithm development, for optimizing the glaciologists' routine workflow for gaining knowledge from spatiotemporal data. In some of our writings, the three iterations are sometimes combined into a single process.
Y. Drocourt, R. Borgo, K. Scharrer, T. Murray, S. I. Bevan, and M. Chen. Temporal visualization of boundary-based geo-information using radial projection. Computer Graphics Forum, 30(3):981-990, 2011. (Presented in EuroVis 2011.) DOI.
R. Borgo, K. Proctor, M. Chen, H. Jaenicke, T. Murray, and I. M. Thornton. Evaluating the impact of task demands and block resolution on the effectiveness of pixel-based visualization. IEEE Transactions on Visualization and Computer Graphics, 16(6):963-972, 2010. DOI.
M. Chen and D. S. Ebert. An ontological framework for supporting the design and evaluation of visual analytics systems. Computer Graphics Forum, 38(3):131-144, 2019. DOI.
* During an action research project, the ontology framework is named in 2026 as SCORE (Symptom, Cause, Optimize, Remedy, side-Effect).