Consequence Briefing, Explainability, and Comparative Product Benchmark

Full report

On this page

The design and implementation of a public-facing, educational planetary simulation dealing with nuclear-event consequences requires an intricate balance between scientific fidelity, user accessibility, and stringent ethical boundaries. When the interaction model restricts users to minimal inputs—specifically, a geographical site and a minimal nuclear-event or weapon classification—the simulation engine and its user interface must assume the entirety of the explanatory burden. The outputs, which range from prompt physical destruction to multidecadal systemic stressors, must be communicated without conveying false scientific authority, without inducing unwarranted panic, and without facilitating operational military targeting. This report details an exhaustive, evidence-based strategy for presenting these generated results. It defines how the system must guide the user to understand causal mechanisms, the temporal evolution of the disaster, and the inherent uncertainties of the underlying models through deterministic boundaries and structured narrative design.

1. Comparative Benchmark

To derive foundational principles for the simulation’s interface, causal explainability, and uncertainty architecture, a comprehensive analysis of twelve distinct product categories, analytical frameworks, and interface patterns was conducted. The analysis extracts critical lessons in spatial navigation, temporal progression, uncertainty communication, and cognitive accessibility. Understanding the successes and failures of these existing paradigms is paramount for designing a simulation that minimizes user inputs while maximizing educational output. The landscape of disaster and strategy simulators reveals a stark dichotomy: tools are either highly accessible but analytically shallow, or analytically exhaustive but completely inaccessible to the layperson. Educational visualizers like Alex Wellerstein’s NUKEMAP1 have achieved massive public reach by overlaying deterministic blast radii on familiar geographic interfaces, relying heavily on the physical equations defined by Glasstone and Dolan4. However, such tools typically present a static snapshot of prompt effects, lacking the temporal cascading of systemic infrastructure collapse. Conversely, professional platforms like FEMA’s Hazus-MH6 and HURREVAC9 provide rigorous, phased temporal tracking and economic loss estimations based on geographic information systems (GIS), but require exhaustive data inputs and domain expertise to operate, making them unsuitable for public educational tools. To bridge this gap, the simulation must incorporate cognitive frameworks from intelligence and scientific communities. The Analysis of Competing Hypotheses (ACH), developed by Richards Heuer for the CIA, provides a structural basis for comparing divergent outcomes based on identical evidence, emphasizing the "diagnosticity" of variables12. In parallel, the Intergovernmental Panel on Climate Change (IPCC) and risk-communication researchers have demonstrated that qualitative uncertainty (e.g., "Likely") is routinely misinterpreted by the public unless explicitly paired with numerical probability bounds15. Finally, spatial rendering engines like Cesium and Mapbox18 and W3C web accessibility guidelines21 dictate how visual and auditory information must be controlled to prevent user disorientation during complex data visualizations.

The synthesis of these twelve benchmarks establishes that input minimization must be counterbalanced by transparent, phased explainability. The simulation must guide the user through a temporal narrative that respects the physical realities of the event while explicitly acknowledging model limitations, utilizing structured spatial control and dual-scale uncertainty mechanics.

2. Recommended Result Hierarchy

To prevent cognitive overload while satisfying the requirement for exhaustive detail, the post-commit result must be presented in a progressive disclosure hierarchy. When the user commits the site and event class, the simulation transitions from an open exploratory state to a highly structured briefing. This layering strategy ensures that the user absorbs the prompt physical reality before grappling with complex secondary systemic cascades. Immediate Confirmation: The system immediately acknowledges the committed coordinates and event class. The map transitions to a locked viewport focused on the site, establishing spatial context without requiring further navigation18. This confirms the user's minimal input before rendering the output. What Happened: A plain-language, high-level summary of the deterministic physical event. This layer translates the raw kiloton or megaton inputs into immediate geographical and physical realities, utilizing the established principles of thermal radiation, air blast phenomena, and initial nuclear radiation5. Why the Model Produced This Outcome: An accessible causal explanation detailing the specific relationships between the selected event class, the geographical characteristics of the chosen site, and the resultant physical destruction. This step demystifies the simulation engine, proving the outcome is a mathematically calculated result rather than a pre-authored narrative. What Changes Now: An analysis of immediate systemic pressures, such as infrastructure collapse, displaced populations, and emergency response degradation. This transitions the focus from raw physics to human and systemic geography, echoing the operational focus of platforms like Hazus-MH7. What Develops Over Time: A phased timeline illustrating the progression of the event from the initial milliseconds of the thermal flash to the weeks and months of secondary cascades and recovery efforts. What Remains Uncertain: A dedicated module identifying the variables that the model cannot predict with absolute certainty, such as exact meteorological conditions, localized structural resilience, or human behavioral responses. This aligns with the IPCC's mandate to make uncertainty a primary feature of the communication, not an afterthought17. What Assumptions Were Used: An explicit disclosure of the baseline data driving the simulation, including ambient population density, generalized infrastructure resilience scores, and idealized terrain models. How to Compare Another Branch: An interface mechanism allowing the user to view how the identical event at the identical site would unfold under altered initiating conditions (e.g., daytime versus nighttime population distributions), utilizing the principles of ACH12. How to Inspect the Globe: Controls that unlock the camera, allowing the user to pan and observe the broader geographical impact, with visual effects scaled to the timeline's current phase. How to Revise the Initiating Facts: A clear pathway to reset the simulation, returning to the initial minimal input state to explore a completely new geographic site or event class.

3. Analytical Briefing Template

The analytical briefing bridges the gap between raw algorithmic output and human comprehension. It is designed to be reusable, modular, and strictly deterministic. This structural rigidity ensures that even if optional artificial intelligence components experience timeouts or fail safety validations, the core educational value remains perfectly intact via template-driven text.

Template Structure

Fictional Example Implementation

Event: Class III Atmospheric Detonation (Approx. 100kt yield equivalent) Site and Coordinates: Lat: 45.1234, Lon: \-90.5678 (Simulated Mid-Latitude Urban Proxy) Simulation Date: 14 October 2026 Branch: Baseline Terrestrial Initiating Facts: Mid-autumn ambient temperatures, standard atmospheric pressure, daytime population distribution. Model Assumptions: Idealized uniform terrain elevation, generalized urban infrastructure resilience, static wind velocity (10 knots easterly). Immediate Consequence Summary: The event generated a high-temperature plasma fireball and subsequent overpressure wave consistent with a 100kt airburst5. Within a 3.2-kilometer radius, structural integrity of commercial and residential buildings suffered catastrophic failure due to extreme air blast loading. The thermal radiation extended sufficiently to ignite exposed combustible materials across a 5.5-kilometer radius, initiating secondary fires. Major System Pressures: The immediate destruction of the localized electrical grid and the severing of primary transportation arteries have isolated the primary impact zone. First responder staging areas within the 5-kilometer radius are rendered non-operational, severely degrading local emergency capacity. Timeline: Phase 1 (0-1 Hour): Initial blast and thermal flash; prompt nuclear radiation exposure; primary structural collapse; toroid cloud formation5. Phase 2 (1-24 Hours): Propagation of secondary fires; atmospheric loading of particulate matter; initiation of spontaneous civilian evacuation. Phase 3 (1-7 Days): Systemic failure of regional medical supply chains; stabilization of thermal events; early fallout deposition patterns materialize. Confidence and Uncertainty: The structural failure radii carry a High Confidence (85-95% probability) based on established overpressure equations16. The extent of secondary fires carries Low Confidence (20-40% probability), as the simulation cannot account for localized combustible fuel loads or spontaneous firebreak creation. Visual-Layer Status: The map currently displays the 5 psi overpressure radius (solid orange boundary) and the thermal ignition probability zone (textured yellow gradient)40. Secondary Cascades: The loss of the regional water purification facility at the edge of the impact zone will result in a rapid depletion of potable water for the surrounding 50-kilometer metropolitan periphery within 72 hours. Recovery Factors: Restoration of systemic stability depends heavily on the rapid deployment of external logistics convoys via intact southern highway corridors and the establishment of peripheral triage centers. Alternative Branch Comparison: Under a Nighttime Population branch, immediate residential casualty rates increase by an estimated 22%, while commercial and industrial district impacts fall proportionately, shifting the required focus of peripheral medical recovery efforts from urban centers to suburban rings. Educational Boundary: This simulation utilizes generalized physics models for educational illustration. It does not predict precise real-world casualties and must not be utilized for operational military targeting, device construction, or emergency management resource allocation. Provenance and Model Version: Simulation Engine v2.4, utilizing generalized algorithms adapted from Glasstone and Dolan (1977) and modern systemic risk frameworks44.

4. Causal Explanation Design

Explaining the "why" behind the simulation results requires translating complex, non-linear system dynamics into comprehensible cognitive pathways. While system dynamics dashboards often rely on node-link graphs and causal loop diagrams, empirical research demonstrates that these complex visual webs frequently overwhelm lay users, inducing cognitive fatigue and reducing comprehension34. Therefore, the causal explanation must utilize a strictly linear, expandable narrative structure supported by visual evidence badges, drawing upon the methodological rigor of the Analysis of Competing Hypotheses (ACH)27. The architecture of the causal chain follows a linear progression that the user can expand for deeper technical granularity: Selected event → Immediate local disruption → Infrastructure and service pressure → Population movement → Institutional response → Regional economic effects → Escalation or recovery To implement this, the interface will utilize an accordion-style textual narrative. Each stage of the chain acts as a plain-language summary sentence. When a user interacts with a stage to expand it, the system reveals the specific deterministic logic and quantitative data driving that link. Within the expanded text, specific claims are adorned with interactive Evidence and Assumption Badges. For example, if the narrative states that "Regional medical capacity collapsed within 4 hours," the word "collapsed" is underlined with a badge indicating the model assumption: Assumption: Baseline hospital bed capacity of 2.5 per 1,000 residents was exceeded by 400%. Confidence Indicators are placed adjacent to every causal link. Utilizing the dual-scale approach required by climate communication research, a badge might read "High Confidence (80-100%)" for physical blast damage, but "Low Confidence (0-20%)" for the speed of population evacuation15. Furthermore, "Why this happened" Disclosures are provided as non-obtrusive tooltips for physical phenomena. If the user hovers over the term "thermal radiation," a deterministic disclosure explains the rapid conversion of weapon mass into a high-temperature plasma fireball and the subsequent emission of electromagnetic radiation, referencing established unclassified physics principles without venturing into weapon engineering5. Counterfactual Comparisons are lightly integrated into the causal chain. At critical junctures, a small prompt indicates how the chain might have broken or altered under different circumstances. This lays the cognitive groundwork for the user to explore the dedicated Branch Comparison view, teaching the user that the outcome is highly dependent on initial contextual conditions rather than just the explosive yield.

5. Timeline Design

The temporal evolution of a nuclear event is as critical to educational comprehension as its spatial footprint. The interface must seamlessly harmonize the textual result summary, the visual effects on the globe, and the simulated event-age phase. The timeline is governed by a horizontal date scrubber anchored to the bottom of the viewport. As the user manipulates the scrubber, the globe visual effects and the analytical briefing dynamically and synchronously update to reflect the selected temporal phase. The relationship is strictly synchronized across the following components:

1. Result Summary: Text dynamically updates. If the scrubber is set to "Phase 3: 72 Hours," the summary emphasizes systemic pressures, displacement, and economic halt rather than the immediate thermal flash.

2. Globe Visual Effects: Prompt effects (fireball, initial radiation) fade out of the rendering engine, replaced by representations of atmospheric loading, displaced population heatmaps, or infrastructural darkness.

3. Phase Markers: Distinct visual notches on the scrubber denote critical transition periods (e.g., T+0, T+1 Hour, T+24 Hours, T+1 Week).

4. Historical Replay: The scrubber can be dragged backward to observe the physical expansion of the prompt effects in slow motion.

5. Future Branch Projections: Beyond the deterministic baseline, the timeline fades into a dashed line representing highly uncertain long-term recovery or escalation projections.

To ensure spatial continuity, manipulating the timeline does not unlock or wildly swing the camera. The viewport remains constrained using SDK features (e.g., enableRotate: false, lookAtTransform) to ensure the user can observe the geographic changes from a stable vantage point18.

Desktop Wireframe Structure

Mobile Wireframe Structure

6. Uncertainty and Model Humility

Simulations of this magnitude possess inherent, unavoidable margins of error. Presenting point estimates without contextualizing their uncertainty violates core risk communication principles and risks imparting a false sense of scientific authority38. Furthermore, research into the IPCC's communication strategies reveals that the public consistently misinterprets qualitative probability terms (e.g., assuming "Likely" means exactly 50%) unless they are paired with numerical boundaries16. The simulation will adopt a standardized, dual-scale language for uncertainty, combining qualitative terms with quantitative probability bounds. Crucially, uncertainty must not be hidden in a footer disclaimer; it must be presented immediately adjacent to the relevant claim.

Exact Language Patterns

On the map, visual uncertainty is represented extrinsically, adhering to cartographic best practices established by MacEachren38. Rather than drawing hard, distinct circles for blast radii or thermal zones, the edges must utilize gradient blurring or textured hatching. This visual language intuitively communicates that the boundary is a transition zone of probability rather than an absolute line of safety, mitigating the risk of the user misinterpreting the map as a precise operational forecast.

7. AI Explanation Boundary

To enhance the narrative flow of the causal explanation and the analytical briefing, an optional Large Language Model (LLM) may be utilized to synthesize the deterministic outputs into fluid, readable prose. However, the AI must act strictly as a translator of the simulation's data, never as the authoritative source of the data itself. Allowing an LLM to freely generate consequences risks hallucination, thereby destroying the educational integrity of the simulation. Structured Input to the Explainer: The LLM receives a strict, structured JSON payload containing the locked inputs, the deterministic calculation results, and the allowed physical relationships. The prompt strictly instructs the LLM to only synthesize the provided data and forbids the introduction of external factual claims. Schema-Constrained Output: The LLM must output its response in a validated JSON schema (e.g., utilizing frameworks like Pydantic or Outlines) ensuring the response maps exactly to the required sections of the analytical briefing36. If the output breaks the schema, it is rejected. Claim Grounding and Rejection: A secondary validation layer checks the LLM output against the deterministic payload. If the LLM generates a claim unsupported by the payload (e.g., hallucinating a specific casualty figure not provided by the math engine), the output is flagged and rejected before it reaches the user interface. Citation and Provenance Links: The LLM is instructed to append citation tags to its sentences linking back to the specific line of deterministic data or the static assumption that generated it. This allows the user to audit the AI's reasoning. Timeout and Outage Fallback: If the LLM fails to generate a validated response within 4.5 seconds, or if the external API is unreachable, the system must instantly and seamlessly fall back to a purely deterministic, template-driven text rendering. The user experience must never break due to an AI failure. Disclosure of AI-Generated Text: Any prose generated by the LLM must be accompanied by a subtle but distinct visual indicator (e.g., a specific background tint or an icon) and the explicit uncertainty pattern: "This causal explanation was dynamically generated by an AI model utilizing the deterministic simulation data." Versioning and Reproducibility: Because LLMs introduce non-determinism, reproducing exact text across different sessions is challenging. Therefore, the system must log the exact seed, model version, and deterministic payload used for generation. Users sharing a simulation result will share the saved deterministic output, ensuring reproducibility regardless of subsequent LLM updates.

8. Branch Comparison

A core educational objective of the simulation is demonstrating how sensitive consequence models are to initial conditions. To achieve this without allowing the user to actively optimize or target specific outcomes (e.g., manually increasing the yield to achieve a desired casualty count), the simulation utilizes a controlled Branch Comparison view. This methodology draws heavily from RAND wargaming techniques, where the value lies in exploring alternate contextual realities rather than changing the fundamental adversary action47. The user cannot change the site or the weapon class during comparison. Instead, they are presented with the same initiating event under different contextual states pre-calculated by the system, mirroring the mechanics of the Analysis of Competing Hypotheses (ACH) to reveal which variables possess the highest diagnosticity13. The comparison is presented as a side-by-side matrix analyzing:

• Selected Site: Constant.
• Event Class: Constant.
• Current-World Conditions (Baseline): E.g., Daytime population distribution, standard weather, current infrastructure resilience.
• Alternative Assumptions (Branch): E.g., Nighttime population distribution, severe weather patterns, severely degraded infrastructure.
• Calculated Results: Highlighted differences in prompt effects and systemic pressures based on the assumption delta.
• Time to Peak: Variations in the speed at which systemic failure occurs under the alternative branch.
• Recovery Capacity: How the altered context degrades or enhances hypothetical stabilization efforts.
• Uncertainty: Shifts in confidence levels based on the altered assumptions.

By locking the aggressive inputs and only allowing manipulation of the contextual defense or environmental variables, the simulation remains firmly educational and avoids acting as a weapon-employment simulator.

9. Accessibility and Comprehension

The simulation must be entirely navigable and comprehensible for users utilizing assistive technologies, adhering to modern web accessibility standards (WCAG) and prioritizing the correct implementation of WAI-ARIA guidelines. Plain-Language Summary: The primary 'What Happened' section is written at an eighth-grade reading level, avoiding dense physical jargon where possible to ensure immediate comprehension during high-stress information absorption42. Expert Detail: Technical users can access exact overpressure PSI thresholds, thermal fluence (cal/cm²), and REM exposure rates by expanding the narrative accordions, ensuring the product scales to academic users without overwhelming lay audiences. Screen-Reader Ordering: The Document Object Model (DOM) is structured logically, ensuring screen readers announce the critical event parameters and the immediate summary before moving into the timeline controls or complex causal chains. Table Alternatives and Chart Descriptions: All visual data, including the branch comparison matrices and timeline scrubbers, are backed by hidden textual summaries and properly labeled ARIA components to ensure non-visual access to data trends. Color-Independent Legends: Map overlays do not rely solely on color hue to convey meaning. They utilize distinct patterns, hatching, and opacities, ensuring legibility for users with color vision deficiencies or situational glare49. Keyboard Navigation: All interactable elements, including the map camera controls, accordion toggles, and the timeline scrubber, are fully accessible via standard keyboard tabbing and arrow keys. Mobile Reflow: The interface utilizes flexible grids to ensure the analytical briefing stacks logically below the map on narrow viewports without requiring horizontal scrolling or awkward zooming. Cognitive Load: The progressive disclosure hierarchy prevents overwhelming the user. Information is chunked into logical units, and white space is heavily utilized to separate distinct phases of the event. ARIA-Live Regions for Dynamic Updates: When the timeline scrubber is moved, dynamic updates to the analytical briefing must be announced to screen readers. Crucially, the aria-live="polite" attribute is used to ensure the screen reader waits for a pause in the user's interaction before announcing the new timeline phase. The aria-live="assertive" or role="alert" attributes are strictly avoided for timeline updates, as they will disrupt the user's flow and create a chaotic auditory experience22. Assertive alerts are reserved solely for critical system failures (e.g., API timeouts). Numerical Formatting: Large numbers (e.g., populations, distances) use localized grouping separators, and coordinates are formatted consistently with cardinal directions for immediate legibility.

10. Ethical and Educational Boundaries

Simulating catastrophic events carries profound ethical responsibilities. The interface is designed to support serious educational inquiry while strictly preventing misuse or psychological harm. The tone and framing of the simulation rely heavily on the Centers for Disease Control and Prevention (CDC) Crisis and Emergency Risk Communication (CERC) guidelines41. Avoiding Operational Targeting: The simulation accepts only minimal inputs and deliberately obscures the ability to fine-tune height-of-burst, precise fission-fusion fractions, or specific delivery vehicles. It will not calculate target optimization, provide strike planning data, or offer device construction insights. Avoiding False Scientific Authority: By heavily prioritizing uncertainty disclosures and explicitly stating model assumptions, the simulation refuses to act as an infallible crystal ball. It presents a plausible outcome, not a certain one, maintaining humility regarding the limits of simulation46. Real Casualty Prediction: The simulation does not output precise, integer-level casualty figures (e.g., "14,532 dead"). Instead, it reports on "populations at extreme risk" or "severe casualty probabilities," maintaining the focus on systemic vulnerability rather than macabre accounting. Invisible AI Invention: The strict AI explanation boundaries and required visual disclosures ensure the user always knows when they are reading generated prose versus deterministic data, preventing the AI from usurping the authority of the physics model. CERC Principles Integration: Following CERC guidelines, the tone of the interface is respectful, empathetic to the gravity of the subject matter, and objective42. It avoids gamification, scoring systems, or sensationalized aesthetics. The goal is to preserve the chilling educational value of the physical realities—honoring the devastating nature of nuclear weapons—rather than reducing the result to a generic, easily dismissible warning or a detached video game.

11. Final Recommended Product Specification

The unified sequence of the product experience, integrating all previously defined systems, is specified as follows:

1. Creation Screen: A global map view. The user is prompted to select a site. The interface is clean, displaying only the map, a search/selection tool, and the overarching educational boundaries.

2. Locked Site: Upon clicking the globe, the camera centers on the coordinates. A minimal modal appears requesting the selection of a broad event or weapon class (e.g., "Minimal Fission Device," "Strategic Atmospheric Detonation"). No other inputs are permitted.

3. Calculation Transition: Upon commit, the modal disappears. The screen dims slightly. A deterministic loading state indicates that physics models, population data, and systemic risk assessments are being calculated.

4. Result Briefing: The interface resolves into the Desktop or Mobile Wireframe structure. The camera is locked to prevent spatial disorientation19. The Immediate Confirmation and What Happened summaries are visible. The map displays the Phase 1 visual overlays.

5. Globe Effects: The user manipulates the timeline scrubber at the bottom of the screen. The map and the sidebar dynamically update to show the progression from prompt physical effects to systemic fallout.

6. Uncertainty Integration: As the user expands the causal explanations in the sidebar, uncertainty badges and confidence intervals are displayed alongside every causal claim, forcing an acknowledgment of model limitations.

7. Branch Comparison: The user selects the Branch Comparison toggle. A matrix appears overlaying the briefing, showing how altered environmental assumptions would shift the current timeline, facilitating deep comparative analysis.

8. Revision: The user selects a reset function, clearing the current state and returning to the Creation Screen to explore a different geography or event class.

9. Failure States: If the AI explanation engine times out, the text boxes instantly populate with pre-formatted, deterministic string templates. If the map tiles fail to load, a stylized vector fallback maintains the spatial context without breaking the application.

12. Acceptance Criteria

To ensure the product meets these rigorous standards and fulfills its educational mandate, development must satisfy the following 60 measurable acceptance criteria:

13. Traceability Matrix

To ensure the recommendations provided in this report are grounded in established research, the following matrix connects the major design decisions to their supporting evidence and primary sources.

Works cited

1. Nukemap \- Wikipedia, https://en.wikipedia.org/wiki/Nukemap

2. NUKEMAP \- Alex Wellerstein, https://alexwellerstein.com/projects/nukemap/

3. Full article: NUKEMAP creator Alex Wellerstein puts nuclear risk on the radar, https://www.tandfonline.com/doi/full/10.1080/00963402.2017.1338001

4. The Effects of Nuclear Weapons \- Glasstone and Dolan | Historical Documents, https://www.atomicarchive.com/resources/documents/effects/glasstone-dolan.html

5. The Effects of Nuclear Weapons \- Glasstone and Dolan | Chapter II \- Atomic Archive, https://atomicarchive.com/resources/documents/effects/glasstone-dolan/chapter2.html

6. Hazus \- FEMA Flood Map Service Center, https://msc.fema.gov/portal/resources/hazus

7. Hazus | FEMA.gov, https://www.fema.gov/flood-maps/products-tools/hazus

8. HAZUS-MH Coastal Flood Model \- Overview \- ArcGIS Online, https://www.arcgis.com/home/item.html?id=79c14c02c87d427d945866b8730c6181

9. Guide to the Workspace | HURREVAC, https://www.hurrevac.com/wp-content/uploads/2026/03/2026-HURREVAC-workspace-guide.pdf

10. Key Features – HURREVAC | Home, https://www.hurrevac.com/program-information/key-features/

11. HURREVAC Quick Reference Guide, https://www.hurrevac.com/wp-content/uploads/2026/03/2026-HURREVAC-Quick-Reference-Guide.pdf

12. Analysis of Competing Hypotheses \- Rodolfo Santos Flaborea \- Medium, https://anticitizenone.medium.com/analysis-of-competing-hypotheses-176bd8147dbc

13. Analysis of competing hypotheses \- Wikipedia, https://en.wikipedia.org/wiki/Analysis\_of\_competing\_hypotheses

14. Analysis of Competing Hypotheses (ACH part 1\) \- SANS ISC, https://isc.sans.edu/diary/22460

15. How are uncertainties communicated? \- Climate-ADAPT, https://climate-adapt.eea.europa.eu/en/knowledge/tools/uncertainty-guidance/topic2/index\_html

16. The interpretation of IPCC probabilistic statements around the world, https://researchportalplus.anu.edu.au/en/publications/the-interpretation-of-ipcc-probabilistic-statements-around-the-wo/

17. Improving communication of uncertainty in the IPCC reports David V. Budescu, https://www.ipcc.ch/site/assets/uploads/2016/02/Budescu\_IPCC\_Communication\_Meeting\_OSLO\_February\_2016.pdf

18. ScreenSpaceCameraController \- Cesium Documentation, https://cesium.com/learn/ion-sdk/ref-doc/ScreenSpaceCameraController.html

19. Control the Camera \- Cesium, https://cesium.com/learn/cesiumjs-learn/cesiumjs-camera/

20. Viewport | Maps SDK | iOS | Mapbox, https://docs.mapbox.com/ios/maps/guides/camera-and-animation/viewport/

21. ARIA Authoring Practices Guide | APG | WAI \- W3C, https://www.w3.org/WAI/ARIA/apg/

22. ARIA live regions \- MDN Web Docs \- Mozilla, https://developer.mozilla.org/en-US/docs/Web/Accessibility/ARIA/Guides/Live\_regions

23. Map of the Week: NUKEMAP | Mappenstance. \- University of Richmond, https://blog.richmond.edu/livesofmaps/2022/03/29/nukemap-alex-wellersteins-mapping-tool/

24. NUKEMAP FAQ \- The Nuclear Secrecy Blog, https://nuclearsecrecy.com/nukemap/faq/

25. Hazus \- U.S. Climate Resilience Toolkit, https://toolkit.climate.gov/tool/hazus

26. HAZUS: FEMA's Flood Damage Estimation Tool Explained \- Continuuiti, https://continuuiti.com/blog/hazus/

27. Improving Intelligence Analysis with ACH \- Pherson, https://pherson.org/wp-content/uploads/2013/06/Improving-Intelligence-Analysis-with-ACH.pdf

28. Uncertainty (Chapter 17\) \- A Critical Assessment of the Intergovernmental Panel on Climate Change, https://www.cambridge.org/core/books/critical-assessment-of-the-intergovernmental-panel-on-climate-change/uncertainty/3B238E862AB873D1D746F8A594DC6DFD

29. Guidance Note for Lead Authors of the IPCC Fifth Assessment Report on Consistent Treatment of Uncertainties, https://www.ipcc.ch/site/assets/uploads/2017/08/AR5\_Uncertainty\_Guidance\_Note.pdf

30. Camera position | Maps SDK | Android Docs | Mapbox, https://docs.mapbox.com/android/maps/guides/camera-and-animation/camera/

31. Technique ARIA19:Using ARIA role=alert or Live Regions to Identify Errors \- W3C, https://www.w3.org/WAI/WCAG22/Techniques/aria/ARIA19

32. What Are ARIA Live Regions? \- Bureau of Internet Accessibility, https://www.boia.org/blog/what-are-aria-live-regions

33. Use ARIA to announce updates and messaging \- Centre for Excellence in Universal Design, https://universaldesign.ie/communications-digital/web-and-mobile-accessibility/web-accessibility-techniques/developers-introduction-and-index/use-aria-appropriately/use-aria-to-announce-updates-and-messaging

34. Causal loop diagrams | OpenLearn \- The Open University, https://www.open.edu/openlearn/science-maths-technology/engineering-technology/causal-loop-diagrams

35. Nuclear Weapons \- Outrider, https://outrider.org/nuclear-weapons

36. LLM Structured Output in 2026: Stop Parsing JSON with Regex and Do It Right, https://dev.to/pockit\_tools/llm-structured-output-in-2026-stop-parsing-json-with-regex-and-do-it-right-34pk

37. Generate structured output from LLMs with Dottxt Outlines in AWS | Artificial Intelligence, https://aws.amazon.com/blogs/machine-learning/generate-structured-output-from-llms-with-dottxt-outlines-in-aws/

38. Visualizing Uncertainty in GIS Data | PDF | Geographic Information System \- Scribd, https://www.scribd.com/document/895369887/11

39. VISUALIZING UNCERTAIN INFORMATION \- Semantic Scholar, https://pdfs.semanticscholar.org/f89c/70913c39be1b56445ccb5e11785eea010ec3.pdf

40. Perspectives on Uncertainty, Geovisualization, and Geographical Reasoning / Decision-making \- NCG, https://ncgeo.nl/wp-content/uploads/2024/05/MacEachren\_Uncertainty\_ITC.pdf

41. Crisis & Emergency Risk Communication (CERC) Manual \- CDC, https://www.cdc.gov/cerc/php/cerc-manual/index.html

42. Crisis and Emergency Risk Communications (CERC): Radiation Emergencies \- CDC, https://www.cdc.gov/radiation-emergencies/php/communication-resources/cerc-rad.html

43. Communicating Radiation Risks: Crisis Communications for Emergency Responders \- National Response Team, https://nrt.org/sites/2/files/Communication%20Radiation%20Risk.pdf

44. The Effects Nuclear Weapons \- UCSB Experimental Cosmology Group, https://www.deepspace.ucsb.edu/wp-content/uploads/2013/01/Effects-of-Nuclear-Weapons-1977-3rd-edition-complete.pdf

45. Effects of Nuclear Weapons \- Princeton University, https://www.princeton.edu/\~aglaser/lecture2007\_weaponeffects.pdf

46. glasstone-dolan-1977.pdf \- Stanford, http://large.stanford.edu/courses/2017/ph241/liang2/docs/glasstone-dolan-1977.pdf

47. War Gaming Methodology \- RAND, https://www.rand.org/content/dam/rand/pubs/research\_memoranda/2008/RM2413.pdf

48. Wargaming: The Surprisingly Effective Tool That Can Help Us Prepare for Modern Crises, https://www.rand.org/pubs/commentary/2025/11/wargaming-the-surprisingly-effective-tool-that-can.html

49. The Visualization of Uncertainty | GEOG 486 \- EMS Online Courses, https://courses.ems.psu.edu/geog486/node/906

50. Live regions & status messages — EqualWeb Academy, https://www.equalweb.com/academy/learn/lessons/live-regions.html