Architectural Paradigms for Spatiotemporal Event Visualization: A Taxonomy of Presentation Modes

Full report

On this page

The expanding deployment of multidimensional sensor systems, ubiquitous mobile telemetry, and complex digital logging architectures has fundamentally transformed the volume, density, and complexity of spatiotemporal data1. As phenomena occur concurrently in space and time, translating these raw, distributed data points into actionable, cognitive assets requires highly sophisticated visual analytics frameworks1. The foundational challenge lies in the seamless integration of spatial and temporal information without overwhelming the cognitive bandwidth of the user, as the establishment of computational methods for comprehensive spatiotemporal analysis has historically proven elusive and prone to spatial-temporal decoupling2. This report establishes an exhaustive architectural framework for selecting the optimal primary presentation mode for complex events. The taxonomy of presentation modes includes animated globes, static maps, timelines, network graphs, before-and-after comparisons, document readers, statistical panels, narrative sequences, and mixed views. The selection of these modes is governed by a multidimensional calculus encompassing the event's geographic scale, chronological flow, inherent uncertainty, source data density, and the specific analytical user task5.

Theoretical Framework for Mode Selection and Configuration

The classification of spatiotemporal data relies on understanding the "event" as a foundational unit of analysis—an occurrence possessing spatial, temporal, social, and thematic dimensions2. Lower-level events describe localized, single-stream point occurrences, whereas higher-level events describe aggregated spatial deformations and collective reactions over time2. Visual analytics abstractions categorize the user tasks into high-level goals: anomaly detection, stage progression, pattern exploration, prediction and recommendation, correlation and causality analysis, and cohort comparison5.

The Calculus of Uncertainty and Presentation

Uncertainty visualization is a critical factor; it is defined as the difference between a real geographic phenomenon and the user's understanding of that recorded data6. Presenting uncertainty is essential for transparent decision-making, yet it introduces the psychological risk of dichotomous thinking—often referred to as the "cliff effect"—where users misinterpret statistical thresholds (such as confidence intervals or p-values), causing sudden, irrational drops in confidence7. To mitigate this, visual encoding relies on intrinsic variables (such as clarity, crispness, transparency, or symbol fuzziness) and extrinsic overlays (such as reliability inset maps, glyphs, or hatched fills)6. High-uncertainty data demands presentation modes that can natively integrate these variables without exacerbating visual clutter or compromising the underlying basemap topography10.

Cognitive Load, Spatial Decoupling, and Immersive Arrays

The configuration of the visualization dictates how users navigate and update data layouts11. Linked views (juxtaposition) display temporal and spatial data in separate but synchronized panels. While this prevents the occlusion of useful map information, it incurs a high context-switching cost for the user, known as spatial decoupling, which hinders comparative judgments of slight variations between data instances3. Conversely, embedded views (such as the 3D space-time cube) merge spatial and temporal aspects into a single continuous view, allowing users to perceive time and space simultaneously3. However, dense embedded volumetric data risks severe occlusion, necessitating advanced interaction models such as volume slicing or spotlight brushing directly within the 3D space3. Furthermore, modern immersive presentation systems in augmented and virtual reality (AR/VR) offer new spatial layout mechanisms, demanding careful attention to symmetric versus asymmetric presenter control, gaze awareness, and intuitive pointing references to maintain audience engagement13.

Accessibility and Scrollytelling Mechanics

Adherence to web accessibility standards (WCAG 2.2 AA) requires that visual formats do not restrict functionality to single sensory characteristics or unmanageable physical operations15. The presentation of structural data must rely on programmatic semantics rather than layout formatting. For instance, data tables must strictly utilize \<th\> elements with defined scope attributes (col or row) while prohibiting merged cells (colspan or rowspan) that cause screen readers to lose tracking16. Furthermore, narrative-driven modes such as "scrollytelling" leverage progressive disclosure to guide analytical interpretation19. Scrollytelling increases information recall and time-on-page by parsing complex data into manageable, step-by-step moments where the user controls the pace19. However, it inherently introduces severe failure risks, such as scrolljacking, layout thrashing, or vestibular triggering via layered parallax effects19. Consequently, robust fallbacks utilizing the prefers-reduced-motion media query and localized, semantically structured tables are absolutely mandatory21. When generating interactive charts, utilizing frameworks that default to accessible semantic ARIA outputs and keyboard navigation (e.g., Observable Plot or Vega-Lite) provides a critical baseline for compliance23.

Analysis of Representative Events

The following analysis categorizes 30 distinct spatiotemporal events into thematic clusters, determining the optimal primary presentation mode based on geographic scale, chronicity, uncertainty, density, and task abstraction5.

Cluster 1: Macro-Spatiotemporal and Global Phenomena

Events spanning planetary scales with dense, continuous chronological flow demand visualizations that handle sheer volumetric scale while conveying unbroken global continuity. Spatial abstraction is less desirable here than topological accuracy. The global tracing of a highly contagious viral pathogen over an 18-month period exemplifies a continuous, massive-scale tracking requirement. Because the data spans physical international borders and operates on millions of daily nodes, the primary task abstraction is identifying macro-stage progression and anomaly detection of novel outbreaks5. The animated globe is superior to a 2D Mercator projection, which heavily distorts high-latitude landmasses and disrupts the visual continuity of trans-polar or trans-Pacific transmission vectors.

To mitigate the half-space occlusion inherent to 3D globes, analysts must rely on the temporal scrubber to continuously update the visual state, creating a mental model of the unseen hemisphere. For accessibility, since the HTML \<canvas\> element drops the native accessibility tree, a parallel hidden DOM containing standard \<table\> structures must update synchronously with the globe's state, satisfying WCAG 4.1.2 Name, Role, Value requirements15. Simulating oceanic microplastic debris accumulation over a decade involves observing millions of individual data points governed by fluid dynamics. This continuous, cyclic chronological event contains high uncertainty due to the predictive nature of ocean current modeling. The user task is fundamentally pattern exploration—identifying where macro-gyres force density accumulation5.

The visual encoding of this event heavily leverages intrinsic uncertainty variables. By utilizing crispness and resolution (MacEachren's visual variables), areas of high predictive confidence feature sharply defined vector flow lines, while highly uncertain peripheries blur into the background6. This directly signals honesty regarding data limitations without eroding trust8. Tracking the localized but globally distributed recession of major ice sheets over 50 years requires comparing discrete snapshots across decades. The source density is sparse in temporal frequency but exceptionally high in optical fidelity (satellite imagery). The primary analytical task is cohort comparison and stage progression5, where the visual evidence of physical loss is paramount.

The psychological impact of glacial retreat is best communicated through unadulterated visual evidence rather than aggregated statistical lines. The split-screen interactive swipe loupe leverages cognitive spatial memory, allowing the user to rapidly context-switch between two states without losing the exact geographic anchor points4. This direct manipulation fulfills the "Configure Visualization" intent, keeping the user in the loop5. Identifying single points of failure within the international submarine internet cable network prioritizes topological relationships over strict geographic fidelity. The chronological aspect is effectively static (current state), and uncertainty is low, yet the vector density is extremely high where lines converge at major landing stations. The task involves anomaly detection and dependency mapping5.

To counteract the hairball effect, the graph must employ non-linear zooming algorithms or edge-bundling techniques that broaden landing zones when the user zooms in, dynamically decluttering the space4. Furthermore, for accessibility compliance, a sibling \<table\> element rendered from the exact same dataset must accompany the canvas to ensure screen readers can systematically read the connection topology without relying on visual interpretation23. Analyzing the ripple effect of a major canal blockage (e.g., the Suez Canal) on global manufacturing supply chains over a 90-day period demands a synthesis of physical geography and abstract chronological delays. The uncertainty is moderate regarding specific ship arrivals, and the analytical task is fundamentally correlation and causality analysis5.

The primary challenge of a mixed view is spatial decoupling, which taxes working memory4. By enforcing strict color synchronization between the spatial map and the temporal Gantt chart, the cognitive load is reduced. Interactions must focus on direct manipulation of the temporal brushing tool, instantaneously reflecting changes in the spatial domain to prove the causal link between the geographic blockage and the temporal manufacturing delay3.

Cluster 2: Localized Spatiotemporal Evolutions & Environmental Shifts

Events restricted to local or regional scales where precise volumetric deformation, multi-layered depth, or localized temporal changes are the primary focus of the analysis. The physical unearthing of a 14th-century castle spanning decades of archeological excavation involves highly localized, macro-topological data. The chronology represents centuries of historical strata, characterized by high uncertainty regarding missing artifacts and dense point-cloud source data. The task requires assessing volumetric stage progression5.

The space-time cube (STC) allows the evolution of spatiotemporal phenomena to be effectively conveyed through temporal narratives, linking geographic information along a continuous timeline3. However, this mode heavily risks visual clutter. The volume spotlight interaction is essential, replacing cumbersome 2D lassos with a mechanism for users to select spatial ranges directly in the cube space, maintaining context without switching to secondary linked views3. The deformation of a hurricane track—specifically its shape, intensity, and projected path over 14 days—is a high-stakes predictive trajectory analysis2. The geography is regional, the chronology is a continuous short-term projection, and the uncertainty is exceptionally high due to the volatile nature of meteorological models.

The visualization of uncertainty here is literally a matter of life and death. If the boundary of the predictive cone is drawn with a crisp, solid line, viewers fall victim to dichotomous thinking, assuming absolute safety if they reside one mile outside the line7. By employing MacEachren's variables of fuzzy crispness and gradient transparency, the visualization actively teaches the user appropriate modes of probabilistic reasoning, smoothing out the psychological "cliff effect"7. Analyzing the construction impact of a mega-dam, which involves flooding a massive river valley over a 5-year reservoir filling period, is a deterministic event with zero uncertainty regarding the historical record. The data is heavily reliant on high-resolution multi-spectral imagery, and the task is observing localized stage progression and environmental degradation5.

Because this event is fundamentally about demonstrating a permanent topological shift, false-color encoding strips away irrelevant visual noise (like cloud cover or seasonal foliage changes) to isolate the exact dependent variables: water and land mass. The direct comparison overrides the need for complex charting, providing immediate visceral comprehension. Tracking the hour-by-hour boundary deformation of a localized wildfire under shifting wind conditions requires meticulous temporal sequencing without losing the spatial reference frame. The uncertainty is moderate due to dense smoke occlusion blocking satellite telemetry, and the core task is boundary tracking and predictive routing5.

Small multiples (juxtaposition) display multiple data instances simultaneously, eliminating the need for the user to rely on short-term memory4. The spatial uncoupling is minimized because the basemap coordinates remain rigidly locked across all panels, allowing the eye to track the growth of the fire polygon organically. Incorporating wind glyphs directly on top of the polygons provides instant causal analysis for the fire's directional expansion. Post-earthquake damage assessment and emergency responder routing is an event governed by immediate, static snapshots of catastrophic change. Uncertainty is critically high due to unverified citizen reports and severed communication lines. The task is rapid anomaly detection and triage routing5.

Visualizing both damage and uncertainty on a single plane requires prioritizing one visual channel over the other so as not to exceed the "visual budget"10. By utilizing a discrete bivariate scale—for example, using color hue to represent damage severity and color saturation/opacity to represent uncertainty—responders can immediately filter out highly uncertain data points that might waste valuable triage time, optimizing their geographic routing.

Cluster 3: Networked Dispersions and Relational Dynamics

When the causal links, contagion pathways, and relational topologies supersede Cartesian geographic accuracy, visualization must prioritize node-edge relationships to reveal structural truths. Patient zero epidemiological contact tracing involves mapping the interpersonal transmission network of a localized viral outbreak. Geography is abstracted to topological connections rather than physical coordinates, the chronology is discrete and sequential, and uncertainty revolves around unreported contacts. The primary task is correlation and causality analysis5.

In tracing human interaction, standard maps fail because infection jumps across cities via air travel instantly. The network graph successfully flattens this geography. However, providing accessibility for dynamic network graphs is notoriously complex; frameworks like D3 require a hand-written accessibility layer including \<title\> and \<desc\> at the SVG root, and per-mark role="img" with aria-label tags to ensure the interactive focus model functions correctly23. The lateral movement of malware through a corporate intranet—tracking a cyberattack origin and server compromise—operates in an entirely virtual topology. Chronology is tracked at millisecond precision, and uncertainty is functionally zero based on immutable server logs. The task is anomaly detection and stage progression modeling.

By locking the network into a radial or hierarchical structure rather than a physics-based force-directed graph, the analyst can trace the exact chain of command the malware exploited. Filtering edges via the temporal slider allows the user to perform post-mortem causality analysis, replaying the attack vector step-by-step. Tracking the viral propagation and mutation of misinformation on social media platforms involves massive data density (millions of posts) and highly abstract geography. Chronology is rapid and continuous. Uncertainty is high regarding the classification of human versus bot accounts. The core task is pattern exploration5.

The sheer volume of social media cascades mandates aggressive data simplification techniques (aggregation and summarization) prior to visual encoding5. The mixed view is essential here: the statistical panel alerts the analyst when the anomaly occurred, while cross-filtering the network graph reveals who orchestrated the structural propagation. The long-term regional dispersal of refugee migration routes across international borders relies on analyzing flow volume over a decade. Uncertainty is high due to undocumented border crossings. The primary task is cohort comparison between different origin nations and evaluating stage progression over the years5.

This mode perfectly bridges the gap between pure topological networks and strict geographic maps. By anchoring the nodes to real-world coordinates but drawing the edges as sweeping curves rather than literal road networks, the visualization correctly prioritizes the volume of the relationship over the exact physical path taken, aligning perfectly with the cohort comparison task. Analyzing the cascade of collapsing asset prices across global financial exchanges during a market contagion event involves discrete geographic points but extreme, millisecond-level data density. Uncertainty is zero, as the ledger is absolute. The task is uncovering micro-correlations and causality during the crash sequence5.

When correlation at the microsecond level is the objective, spatial geography (the physical location of the server in New York vs. London) is irrelevant. The statistical panel allows multiple financial instruments to be juxtaposed simultaneously. This avoids the memory burden of context switching, provided the cross-filtering interaction is instantaneous and latency-free4.

Cluster 4: High-Uncertainty and Predictive Models

When an event has not yet occurred, or the data is highly probabilistic, the presentation mode must explicitly encode doubt, limitations, and variability without eroding the reader's trust in the underlying scientific models8. Projecting multi-decade global climate change temperature anomalies up to the year 2100 involves continuous global geography and long-term future chronology. Uncertainty is inherently massive due to competing scenario models (e.g., emission mitigation vs. business-as-usual). The density is high, and the analytical task is prediction and policy recommendation5.

Visualizing climate uncertainty requires a delicate balance. Using intrinsic visualization techniques like value-by-alpha (transparency), highly uncertain regions seamlessly fade into the background, preventing them from dominating the visual hierarchy6. This prevents the user from making falsely confident geographic assessments while still engaging with the macro-level predictive trends. Real-time modeling of election night voting returns based on partial precinct reporting involves localized discrete geography (precincts) and rapid, real-time chronology. Uncertainty starts critically high and tapers to zero as the night progresses. The task is immediate prediction and anomaly detection based on historical benchmarks.

Election maps are notorious for misrepresenting data if they map geography rather than population. The mixed view grounds the spatial map with a statistical panel that provides the actual numerical weight. The bivariate color scale ensures that a precinct only appears visually dominant (highly saturated) if both the margin is wide and the reporting percentage is near completion, accurately reflecting confidence. Probabilistic modeling of tsunami inundation zones following a seismic event focuses on localized coastal geography and short-term predictive chronology (hours). Uncertainty is moderate, highly dependent on off-shore wave height telemetry. The core task is emergency anomaly detection and safe-routing recommendations.

In crisis scenarios, cognitive load must be minimized. Presenting uncertainty through extrinsic hatched fills ensures the data is easily interpreted without requiring the parsing of complex color gradients or probability density functions6. The visual contract is straightforward: solid color implies certain inundation; hatched overlay implies potential risk. Mapping the prehistoric migration routes of hominids out of Africa spans global macro-geography and deep-time continuous chronology. Uncertainty is exceptionally high, extrapolated from sparse and fragmented archaeological digs. Density is sparse point data. The task is broad pattern exploration and educational stage progression2.

This event relies heavily on scrollytelling to bridge the massive gaps in the data. By guiding the user step-by-step through the narrative, the visualization fills in the sparse data points with contextual storytelling. The intentional blurring (uncrispness) of the migration vectors serves as a constant, subtle reminder of the inherent scientific uncertainty, preventing the viewer from assuming the paths were highly defined highways6. Charting the trajectory of a deep-space probe utilizing gravitational assists over a decade involves Cartesian 3D interplanetary geography and long-term continuous chronology. Uncertainty is remarkably low due to orbital mechanics. Density is continuous line data, and the task is monitoring stage progression.

Navigating 3D space on a 2D screen requires constant spatial awareness. To prevent disorientation, the visualization must embed dynamic scale bars and reference grids. The interaction model allows users to decouple their viewing angle from the chronological timeline, isolating specific maneuver nodes for detailed inspection without losing the macro-trajectory context.

Cluster 5: Narrative and Sequence-Driven Chronicles

When the objective is not open-ended exploration but rather guided public engagement, educational synthesis, or journalistic storytelling, presentation modes must prioritize narrative sequencing and progressive disclosure over raw analytical toolsets19. The chronological breakdown of historical troop flanking maneuvers (e.g., the WWII Battle of the Bulge) operates on regional geography and discrete daily intervals. Uncertainty is low, density is moderate, and the primary objective is sequential storytelling to explain complex causal battlefield shifts.

Scrollytelling thrives here because it breaks a highly complex spatial conflict into digestible, chronological beats20. Instead of dumping the user into a massive map with a slider, the author curates the journey, ensuring the viewer's eye is directed precisely to the breakthrough points on the map at the exact moment they read about it in the text. Tracking the evolution of the 100m Olympic sprint record over a century requires tracking a sequence of highly sparse point events. Geography is largely irrelevant (the track surface is standard), uncertainty is zero, and the primary task is communicating narrative stage progression and human achievement5.

Because improvements in sprinting are now measured in hundredths of a second, the visualization must manipulate the Y-axis dynamically to make microscopic changes visible20. Scrollytelling allows the axis to seamlessly zoom in during specific eras, anchoring the viewer's understanding of scale before introducing the next data point. Investigative journalism tracing a complex web of corporate fraud, shell companies, and illicit transnational transactions operates across multi-national geography with discrete, documented event chronology. Uncertainty is low due to hard evidence, but the density and topological complexity are massive. The task is simplifying causality for a lay audience.

This event requires a radical layout update mid-analysis, shifting from spatial presentation to abstract relational presentation11. By wrapping this transition in a tightly controlled narrative sequence, the author prepares the reader's mental model for the shift, explicitly linking the physical offshore location to the abstract corporate node before dissolving the map entirely. A product tour explaining the inner workings of a localized targeted medical radiation device involves micro-spatial geography (the human body) and a minute-by-minute protocol chronology. Uncertainty is zero, density is low, and the task is educational augmentation and product familiarization.

Parallax scrolling is highly effective at creating depth and explaining spatial relationships between nested objects (like a machine inside a body)21. However, it is an accessibility minefield. The design pattern must gracefully degrade to a linear, static layout for any user indicating motion sensitivity or utilizing a touch interface where scroll events are inherently asynchronous22. Visualizing urban gentrification and demographic displacement in a metropolitan neighborhood over 30 years utilizes localized urban block geography and decadal census chronology. Uncertainty is low, data density is high, and the task is cohort comparison tied to a narrative exploration of zoning policies.

This approach leverages the emotional impact of geographic familiarity. The user sees their own neighborhood physically recolored over time. The primary technical hurdle is ensuring that the floating narrative boxes do not occlude the very data points they are describing, requiring intelligent, dynamic placement of the text overlays based on viewport dimensions.

Cluster 6: High-Density Statistical and Document-Centric Occurrences

Events defined by rigid logging, legal frameworks, or complex numerical tracking require presentation modes that prioritize absolute data integrity, literal reading order, and advanced filtering over spatial geographic mapping. The analysis of a leaked political dossier comprising 10,000 pages of diplomatic cables spans multi-national geography over years of chronology. Uncertainty is moderate due to heavy redactions. The sheer mass of unstructured text dictates that the primary task is text-mining, entity extraction, and anomaly detection.

In this scenario, spatial mapping is secondary to the primary source material. Attempting to map 10,000 documents geographically results in an unreadable cluster of points on capital cities. The Document Reader respects the provenance of the data5, allowing the analyst to configure the data via natural language search and semantic linking, revealing the hidden spatiotemporal connections within the text itself. Investigating an autonomous vehicle sensor crash log requires breaking down millisecond telemetry leading to a physical collision. Geography is micro-local (an intersection), chronology is microsecond-precise, and uncertainty is zero. Density is ultra-high multidimensional scalar data. The task is rigid root-causality analysis.

When data represents a linear mechanical failure, the temporal dimension heavily outweighs the spatial one. The vertically stacked timeline configuration allows the analyst to cross-reference multiple independent variables (speed vs. braking) at the exact moment of failure, identifying the precise causal anomaly without spatial distractions. Tracking a complex corporate merger timeline involving the legal, financial, and personnel integration of two global conglomerates is an abstract geographic event driven by multi-year, milestone-based chronology. Uncertainty is low, density is moderate, and the task is stage progression tracking and project management.

Timelines are fundamentally sequential, which aligns perfectly with semantic HTML structures. The visual encoding of a Gantt chart is highly intuitive for management tasks, but the underlying code must reflect a linear reading order15. This ensures that visual layout choices do not obstruct the programmatic determination of the data's meaning. Monitoring a real-time algorithmic trading log that executes high-frequency buy/sell orders across dark pools involves discrete server geography and microsecond chronology. Uncertainty is zero, but the data density is overwhelmingly extreme. The task is continuous algorithmic monitoring and rapid anomaly detection.

The statistical panel excels when velocity and volume are paramount. For an algorithmic trader, spatial location is irrelevant compared to execution speed. The architecture must prioritize performance above all, utilizing technologies like Apache ECharts with canvas renderers to handle hundreds of points per second without layout thrashing23. Logging the minute-by-minute scientific observations during a deep-sea hydrothermal vent submarine dive relies on localized bathymetric depth geography (operating entirely on the Z-axis). The chronology is a linear 6-hour dive. Uncertainty is low, but density spans multiple media types (video, telemetry, text). The task is rich documentation and provenance management5.

This mixed view serves as an archival tool. By tightly coupling the video timeline to the text transcript and the spatial depth gauge, the analyst can rapidly locate anomalies (e.g., discovering a new biological species) by searching the text, which instantly provides the visual context and spatial location, seamlessly unifying space, time, and theme.

Conclusion

The effective translation of spatiotemporal data into cognitive insights relies not upon a monolithic visual standard, but upon a rigorous, highly contextual architectural calculus. As demonstrated across the analysis of 30 diverse representative events, the selection of a primary presentation mode—be it an immersive space-time cube, a highly orchestrated scrollytelling narrative, or a dense real-time statistical panel—must be inexorably linked to the event's geographical bounds, temporal continuity, and the analyst's specific task abstractions3. Crucially, the visualization of uncertainty must be handled with precise calibration of intrinsic visual variables (such as fuzziness and transparency) to prevent dichotomous interpretations and the psychological "cliff effect," actively teaching users to navigate probabilistic data6. Furthermore, the deployment of these visual modes must flatly reject the false dichotomy between aesthetic complexity and digital accessibility. Adherence to WCAG 2.2 AA mandates that fallback mechanisms are not mere afterthoughts, but foundational structural layers—utilizing strict semantic table tagging, prefers-reduced-motion queries to prevent vestibular harm, and robust ARIA roles to support complex interactive DOMs16. By synthesizing these constraints, visualization architects can construct spatiotemporal analytical environments that are as cognitively resonant as they are universally accessible.

Works cited

1. INTERACTIVE EXPLORATORY VISUAL ANALYTICS APPROACH FOR DISTRIBUTED SPATIOTEMPORAL DATA by Abdullah-Al-Raihan Nayeem A dissertatio \- Niner Commons \- UNC Charlotte, https://ninercommons.charlotte.edu/record/2401/files/Nayeem\_uncc\_0694D\_13399.pdf

2. A Framework for Visualization and Exploration of Events | Request PDF \- ResearchGate, https://www.researchgate.net/publication/32027543\_A\_Framework\_for\_Visualization\_and\_Exploration\_of\_Events

3. Volume-Based Space-Time Cube for Large-Scale Continuous Spatial Time Series \- arXiv, https://arxiv.org/html/2507.09917v1

4. A Comparison of Spatiotemporal Visualizations for 3D Urban Analytics, https://www.computer.org/csdl/journal/tg/2023/01/09904453/1H1giUQajSM

5. A Multi-Level Task Framework for Event Sequence Analysis \- arXiv, https://arxiv.org/html/2408.04752v1

6. The Visualization of Uncertainty – Digital Cartography \- Open Books, https://wustl.pressbooks.pub/digitalcartography/chapter/the-visualization-of-uncertainty/

7. Can Visualization Alleviate Dichotomous Thinking? Effects of Visual Representations on the Cliff Effect \- ResearchGate, https://www.researchgate.net/publication/350910248\_Can\_Visualization\_Alleviate\_Dichotomous\_Thinking\_Effects\_of\_Visual\_Representations\_on\_the\_Cliff\_Effect

8. The Impact of Uncertainty Visualization on Trust in Thematic Maps \- arXiv, https://arxiv.org/html/2602.00248v1

9. (PDF) Visualizing geospatial information uncertainty what we know and what we need to know. Cartography and geographic \- ResearchGate, https://www.researchgate.net/publication/288959978\_Visualizing\_geospatial\_information\_uncertainty\_what\_we\_know\_and\_what\_we\_need\_to\_know\_Cartography\_and\_geographic

10. Visualizing 3D Terrain, Geo-Spatial Data, and Uncertainty \- A Rostock, https://vca.informatik.uni-rostock.de/\~ct/publications/Duebel173DGeoUncertainty.pdf

11. A Multi-Level Task Framework for Event Sequence Analysis \- IEEE Computer Society, https://www.computer.org/csdl/journal/tg/2025/01/10684003/20lmfDTPRio

12. Feasibility of the Space–Time Cube in Temporal Cultural Landscape Visualization \- MDPI, https://www.mdpi.com/2220-9964/7/6/209

13. Examining the Effects of Immersive and Non-Immersive Presenter Modalities on Engagement and Social Interaction in Co \- matt gottsacker, https://mattgottsacker.space/docs/papers/CHI25\_\_Paper\_\_Immersive\_Presentation\_Study\_preprint.pdf

14. Examining the Effects of Immersive and Non-Immersive Presenter Modalities on Engagement and Social Interaction in Co-located Augmented Presentations \- arXiv, https://arxiv.org/html/2503.13174v1

15. Web Content Accessibility Guidelines (WCAG) 2.2 \- W3C, https://www.w3.org/TR/WCAG22/

16. How to Keep Tables Accessible and WCAG 2.2 AA Compliant | Aubergine 262, https://www.aubergine262.com/how-to-keep-tables-accessible-and-wcag-2-2-aa-compliant

17. Tables \- Digital Accessibility \- UC Berkeley, https://dap.berkeley.edu/websites/web-accessibility-basics/tables

18. Tables – Accessible Technology \- University of Washington, https://www.washington.edu/accesstech/guide/tables/

19. scrollytelling | Skills Marketplace \- LobeHub, https://lobehub.com/skills/doodledood-claude-code-plugins-scrollytelling

20. What great scrollytelling looks like — and how to build it yourself \- Flourish, https://flourish.studio/blog/scrollytelling-examples/

21. scrollytelling-and-parallax-data-visualization | Agent Skills Library \- Awesome MCP Servers, https://mcpservers.org/agent-skills/openai/scrollytelling-and-parallax-data-visualization

22. Scrolling Designs: 8 Patterns and When to Use Each (2026) | Lovable, https://lovable.dev/guides/scrolling-designs-patterns-when-to-use

23. Accessible Data-Visualisation Tooling in 2026 \- Disability World, https://www.disabilityworld.org/articles/accessible-data-viz-tooling-2026/

24. Oregon Health Authority : Accessible Data Visualizations Quick Guide, https://www.oregon.gov/oha/digital-accessibility/pages/accessible-data-visualizations-quick-guide.aspx

25. Visualising uncertainty \- EU Open Data Portal, https://data.europa.eu/sites/default/files/course/7.2\_VisualizingUncertainty.pdf

26. From Quantification to Visualization: A Taxonomy of Uncertainty Visualization Approaches \- PMC, https://pmc.ncbi.nlm.nih.gov/articles/PMC4319674/