Disaster Management

======================================================================== LECTURE GUIDE: Disaster Management ======================================================================== TOTAL TIME: ~50 minutes (without exercise) | ~60 minutes (with exercise) READING: McBean, G.A. — "Role of Prediction in Sustainable Development and Disaster Management" Key thesis from reading: prediction is a fundamental public-good role of government; the gap between prediction capability and political will to act on predictions is where disaster governance fails. LEARNING GOALS: 1. Define disaster through a human security lens using the ISDR formula (hazard + vulnerability + insufficient capacity) 2. Describe the disaster cycle and explain why governments systematically underinvest in mitigation 3. Explain the role of prediction — and its limits — in linking hazard to human response 4. Analyze AI's role in prediction, response, and recovery — and its limits 5. Evaluate how vulnerability and equity shape who bears disaster's costs 6. Connect disaster governance to international frameworks (Sendai) and domestic coordination failures PREPARATION CHECKLIST: - [ ] Open https://ourworldindata.org/natural-disasters before class — have the deaths bubble chart ready - [ ] Open https://www.ncei.noaa.gov/maps/hazards/ for the NOAA Natural Hazards Viewer demo - [ ] Open https://coast.noaa.gov/hurricanes and search "Virginia, USA" — Helene 2024 will appear - [ ] Open https://worldmonitor.app (or self-hosted instance) — have the disaster layer and 3D globe ready for live demo; filter to active events - [ ] Check FEMA.gov or ReliefWeb.int for a recent disaster event to reference as a live case - [ ] Review the Sendai Framework for Disaster Risk Reduction (2015–2030) key targets - [ ] Identify 1-2 recent AI-in-disaster examples (e.g. Google Flood Hub, wildfire spread modeling) - [ ] Note: Simulation Phase 5 is Thursday (Mar 19) — close by connecting the prediction-action gap to simulation decision-making SECTION BREAKDOWN (TARGET TIMES): 1. Opening + framing disasters as human security: 5 min 2. The scale of the problem (EM-DAT data): 3 min 3. Types of disasters + the disaster cycle: 5 min 4. Prediction — the core government tool: 7 min 5. AI in disaster management: 6 min 6. Case — Hurricane Katrina (governance failure + equity): 6 min 7. Case — Complex emergency / information failures: 5 min 8. Discussion — who is responsible for preparedness?: 4 min 9. Vulnerability and equity: 5 min 10. International frameworks and governance gaps: 4 min 11. Key takeaways + simulation bridge: 2 min ADAPTATION FOR 30-MIN VERSION: - Skip the scale/data slide — open directly with definition - Skip the COVID/wildfire case — use Katrina only - Reduce discussions to 90 seconds - Merge equity and governance into one combined slide - Skip the optional exercise entirely KEY MESSAGES TO REINFORCE THROUGHOUT: 1. Disasters are not "natural" — they are the intersection of hazard and vulnerability; human choices determine who suffers 2. Prediction is necessary but not sufficient: a forecast that no one acts on is just information, not protection 3. Governments systematically underinvest in mitigation because prevention's benefits are deferred and uncertain ========================================================================

OPENING (4 min) - The ISDR formula is from the McBean reading — use it verbatim; it anchors the rest of the lecture - The "exceed ability to cope" clause is doing important work: the same hazard is a manageable event in a high-capacity community and a disaster in a low-capacity one - The "natural disaster" framing is a critical thinking hook: use it to anchor the political and equity dimensions early KEY MOVE: - Connect immediately to Week 1: "Pick any of the seven human security dimensions — I'll show you how a major disaster attacks it." - Food → crops destroyed, supply chains disrupted - Health → injury, disease outbreak, mental health - Economic → job loss, infrastructure destruction, insurance gaps - Community → displacement, social fragmentation - Environmental → pollution, ecosystem damage - Personal → violence often increases post-disaster - Political → governance legitimacy challenged THEORY NOTE (from reading): - Bogardi and Brauch (2005) proposed adding "freedom from hazard impacts" as a third pillar of sustainable development alongside "freedom from want" and "freedom from fear" — prediction and early warning are what make that freedom possible - This is directly relevant to the human security framework students know from Week 1 TRANSITION: "Before we get to governance, let's look at the scale of what we're talking about."

SOURCE: McBean (reading); EM-DAT: OFDA/CRED International Disaster Database; MunichRe 2005-2006

SCALE SLIDE (3 min) - These numbers are from the McBean reading — reference it directly: "Your reading this week establishes the baseline" - The trend is more important than the specific numbers: frequency is rising, costs are rising, and the burden falls inequitably - The "deaths per disaster" comparison is striking and sets up the equity dimension: 23 vs. 1,000+ - The 75% weather-related figure connects back directly to Week 7 (Environmental Security) KEY EMPHASIS: - "Rising frequency" is partly better data collection, but IPCC analysis confirms genuine increase in hydrometeorological extremes - The economic loss numbers are in absolute dollars; as % of GDP the developing country burden is even heavier (this sets up the Sendai "relative to GDP" target later) TRANSITION: "The frequency trend is real — but the deaths trend tells a more complicated story. Let's look at the data."

OWID CHART SLIDE (2 min) - Switch to the live Our World in Data page — the bubble chart showing deaths 1900–2016 by disaster type is visually striking - The key interpretive move: frequency is rising, but deaths have fallen dramatically for most categories - This is because early warning systems, building codes, evacuation infrastructure, and healthcare have improved — exactly the "capacity to respond" dimension of the ISDR definition - Don't let students conclude "disasters aren't a problem anymore" — pivot to the economic costs, which are still rising, and to the persistent vulnerability of the developing world KEY QUESTIONS TO ASK: - "What explains the dramatic drop in drought deaths since the 1930s?" → famine early warning systems, food aid mechanisms, Green Revolution productivity gains, better logistics - "Why do earthquake deaths still spike so high?" → because earthquake mortality is almost entirely determined by building quality; you cannot out-warn a building collapse in 30 seconds THE CRITICAL NUANCE: - The deaths decline is real — AND it is fragile: it depends on sustained investment in the mitigation and preparedness phases that governments systematically underinvest in - Climate change may reverse these gains by intensifying hazards faster than our coping capacity can adapt LIVE DEMO NOTE: - The Our World in Data page lets you filter by disaster type and country — show a comparison between a wealthy country and a developing country for any given hazard type TRANSITION: "So prediction and preparedness work — when they're funded. Let's look at the taxonomy of what we're dealing with."

DISASTER TYPES (3 min) - Use the IFRC taxonomy (from the old slides) — it's more precise than a consolidated table and is the actual international standard - Source: https://www.ifrc.org/en/what-we-do/disaster-management/about-disasters/definition-of-hazard/ - The "blurring" point is the key intellectual contribution of the old slides' discussion questions — introduce it here and return to it as a discussion GOOD EXAMPLES FOR EACH: - Geophysical: 2010 Haiti earthquake (200k deaths — due to building quality, not hazard magnitude alone) - Hydrological: 2023 Libya floods (dam failure — technological + hydrological hybrid) - Climatological: 2021 Pacific Northwest heat dome (extreme temp event unprecedented in recorded history) - Meteorological: Hurricane Helene 2024 (directly relevant to VT students — see next discussion) - Biological: COVID-19; H5N1 avian flu (current concern) - Technological: Bhopal 1984; East Palestine, OH 2023 train derailment THE BLURRING QUESTION (preview for later discussion): - "If a wildfire is intensified by climate change, is it natural or man-made?" - "If a cyberattack disables a dam's control system and the dam fails, is the resulting flood a technological disaster?" NOAA TOOLS DEMO NOTE: - Consider pulling up https://www.ncei.noaa.gov/maps/hazards/ to show real hazard data by type and location - Or https://coast.noaa.gov/hurricanes to show historical hurricane tracks near Virginia — Helene 2024 appears prominently TRANSITION: "How do we manage all of this systematically? Through the disaster cycle."

DISASTER CYCLE (3 min) - The four-phase framework is from FEMA and widely used internationally - The political economy point is directly from the McBean reading: "Investments in mitigation where the benefits are only gained later if there is a disaster (which is uncertain for any location), compete for political priority with items of more certain and immediate political return." - This is a critical insight: the underinvestment in mitigation is not irrational at the individual political level — it's a collective action problem KEY FRAMING: - In federal systems, this gets worse: the government that pays for mitigation often isn't the government that receives the benefit of reduced response costs (state vs. federal, or jurisdiction A funds levees that protect jurisdiction B) - Katrina is the textbook example of this: federal mitigation investment in New Orleans levees was repeatedly underfunded CONNECT TO COURSE THEMES: - Ethics (Week 3): if mitigation is underfunded and people die, who is accountable? - Economic security (Week 5): disaster recovery can widen wealth gaps — wealthier homeowners rebuild; renters relocate and don't return TRANSITION: "If there's one tool that cuts across all four phases of the cycle, it's prediction. Let's look at how that works — and what its limits are."

PREDICTION SLIDE (7 min) — This is the core analytical contribution of the reading - This slide is the bridge between the disaster cycle and AI — prediction is what links hazard awareness to human action - The chain diagram (McBean Figure 74.1) is the conceptual heart of the reading — walk through it slowly THE THREE-PART PREDICTION CHAIN: 1. Natural system prediction: What will happen physically? (weather, hydrology) 2. Impact prediction: What will the physical event do to people and infrastructure? 3. Social response prediction: How will people and institutions respond to warnings? - "There are no laws in social science" (Galtung, quoted in McBean) — this is where prediction gets hard KEY EXAMPLES FROM READING: - Tornado: deterministic prediction only for minutes to hours; probabilistic risk for "elevated risk tomorrow" - Major weather system / flood: skill for several days, decreasing with time; Lorenz showed theoretical limit is ~2 weeks for deterministic weather prediction - Earthquake/volcano: very limited timing/magnitude prediction; once event starts, consequences (tsunami, ash plume) can be predicted - Climate: scenario-based, not deterministic; useful for planning horizons but not operational warning CASCADE OF UNCERTAINTY: - Start with the atmospheric model (uncertainty #1) - Run through hydrological prediction (uncertainty #2) - Translate to infrastructure impact (uncertainty #3) - Predict how people will respond to the warning (uncertainty #4) - Each step multiplies uncertainty — this is why "AI predicted a flood" ≠ "people were safe" THE CRITICAL INSIGHT FROM READING: - "Fate can become a choice and choices can make the prediction wrong" — accurate prediction of an upcoming disaster can change human behavior enough to prevent the disaster; but this only works if warnings reach people who have the capacity to act - A prediction that no one acts on is just information, not protection CONNECT TO AI: - AI improves step 1 and step 2 significantly; it does very little for step 4 (the social response problem) - This is exactly why prediction is necessary but not sufficient TRANSITION: "So where does AI fit in this prediction-action chain?"

AI IN DISASTER MANAGEMENT (5 min) - Connect directly to the prediction typology from the previous slide: AI primarily improves the natural and impact prediction steps; the social response gap remains - The Google Flood Hub example is genuinely impressive: 7-day flood forecasts via ML for countries with sparse gauge infrastructure, delivered via SMS KEY EXAMPLES: 1. Google Flood Forecasting Initiative: probabilistic ML models using satellite altimetry + river gauge data + terrain data; SMS alerts to millions in Bangladesh, India, parts of Africa 2. Wildfire spread modeling: CAL FIRE and USFS use physics-informed ML to project fire behavior; gains are real but contested by incident commanders who distrust model confidence intervals 3. Damage assessment: UNHCR and WFP use satellite + ML to map building damage in conflict/disaster zones; faster than human assessment but accuracy varies by context 4. World Monitor (https://github.com/koala73/worldmonitor): open-source, 39k+ GitHub stars; aggregates 435+ feeds across disaster, military, economic, and escalation indicators using local AI (Ollama) — no API keys required; 45 geospatial data layers on a 3D globe; Country Intelligence Index with composite risk scoring across 12 categories; shows how AI-driven situational awareness tools are increasingly accessible outside government/military contexts LIVE DEMO OPTION (World Monitor): - Pull up the disaster/geophysical layer on the globe — show currently active events - Switch to the Country Intelligence Index for a country recently affected by a major disaster — show what the composite risk score captures - Ask: "This tool runs on a laptop with local AI. What does that mean for who can now build situational awareness systems?" → democratization of intelligence tooling; also raises questions about accuracy, data sourcing, and potential misuse - Note: self-hosted version (Docker/Vercel) lets organizations run it with their own data; the public app draws on curated commercial feeds - "What's NOT in this dashboard?" → local community knowledge, indigenous early warning systems, informal sector data; the marginalized communities most at risk are often least represented in the data feeds CRITICAL THINKING PUSH: - "If AI can predict a flood 7 days out, what determines whether that prediction saves lives?" → Trust in the alert, ability to evacuate, financial capacity to leave, whether warnings are in local languages, whether marginalized communities are in the model's training data - "AI improves what McBean calls the 'natural system prediction.' What about the 'human response prediction'?" → This is where AI is weakest: predicting how communities will respond to warnings is still poorly modeled CONNECT TO AI ETHICS (Week 3): - If an AI flood model is trained on data from high-income countries, how accurate is it for informal settlements in low-income countries? - False positive: evacuate unnecessarily → economic loss + erosion of trust in future warnings - False negative: no warning → deaths; potential liability TRANSITION: "Let's look at a case where prediction worked — and governance failed anyway."

KATRINA CASE (5 min) - This is the canonical US disaster governance failure — McBean specifically cites it as a "failure to invest" case - The key analytical move: the storm was forecast accurately; the flooding was foreseeable; people died because mitigation was underfunded for political reasons - Quote from McBean reading (paraphrased): "Investments in mitigation where the benefits are only gained later if there is a disaster (which is uncertain for any location), compete for political priority with items of more certain and immediate political return... The recent incident of Hurricane Katrina demonstrates the possible impact of failure to invest." KEY NUMBERS: - 1,800 deaths (about 40% in New Orleans; rest in surrounding areas) - $125 billion in damages (2005 dollars) - 400,000 people displaced long-term - In the Lower 9th Ward, 98% of residents were Black; recovery was slower and less complete than in wealthier white neighborhoods FACILITATION: - "What phase of the disaster cycle failed most critically?" — the answer is MITIGATION; the levees were the preventable failure - "The NWS issued accurate hurricane forecasts days in advance. People still died. What does that tell us about the prediction-action chain?" → It wasn't prediction that failed — it was capacity (can't evacuate without transportation), mitigation (levees), and response coordination CONNECT TO COURSE: - Political security (Week 6): FEMA was reorganized post-9/11 to focus on terrorism; critics argue this degraded natural disaster capacity — this is a governance priority choice - Community security: diaspora communities never returned; the social fabric of some neighborhoods was permanently destroyed - Economic security: the "failure to invest" in levees reflects exactly the political economy McBean describes: deferred benefits, uncertain event TRANSITION: "Katrina is 20 years ago. Let's look at more recent cases that add the information dimension."

HELENE CASE (4 min) - This case is directly relevant to VT students — many students had family or community connections affected by Helene; Appalachian Virginia was in the impact zone - The prediction worked: NWS track and rainfall forecasts were accurate well in advance; this is not a prediction failure - The failure was in the warning → response chain: inland communities don't have hurricane evacuation culture, hurricane-rated structures, or the infrastructure to receive aid after roads wash out - This connects directly to McBean: "Appropriate preparedness strategies will vary with the event. In the case of a tornado... For a river cresting within the next five days, preparations for evacuation can be taken." - The disconnect: FEMA/state emergency managers were focused on coastal impacts; inland flooding was underestimated in public communication KEY FACTS: - Helene dropped 20–30 inches of rain in some NC mountain areas in ~48 hours - Chimney Rock, NC largely destroyed; Asheville lost water infrastructure for weeks - Cell towers, roads, and bridges destroyed simultaneously — "connectivity collapse" made response extremely difficult - National Guard and private helicopters were the primary access for many communities FACILITATION: - "The forecast was accurate. People still died in huge numbers in landlocked states. Where did the prediction-action chain break?" → Mental model mismatch, lack of inland flood preparedness, infrastructure vulnerability - If any students are from western VA/NC/TN — invite their perspective, don't put them on the spot HISTORICAL HURRICANE TRACKS DEMO: - Pull up https://coast.noaa.gov/hurricanes and search "Virginia, USA" - Show Helene's track — it came far inland; show it in context of other recent storms - "197 storms have passed within 60 nautical miles of Virginia. You are not immune from hurricanes in Blacksburg." TRANSITION: "Helene shows that the prediction-action chain can break even when the technical forecast is perfect. Now let's look at a category of failure that makes this even harder: when digital systems fail."

CYBER-DISASTER DISCUSSION (4 min) Format: 1 min individual think, 3 min discussion FRAMING (from the old slides): - The IFRC definition of man-made hazards includes "events caused by humans that occur in or close to human settlements" - A cyberattack that causes a dam failure, or that corrupts the prediction system, fits this definition — but it was the hurricane that prompted the attack opportunity - These are "cascading" or "compound" disasters: multiple hazard triggers interacting CONNECT TO WEEK 2 (AI/Cyber): - Critical infrastructure attacks are increasingly a disaster risk multiplier — they target the systems that enable disaster response - The 2021 Colonial Pipeline ransomware attack caused fuel shortages in the Southeast during hurricane season; this is not hypothetical - Ukraine: Russian attacks on power infrastructure in winter = compound disaster (infrastructure failure + extreme cold) POSSIBLE STUDENT ANSWERS: - "The cyberattack is the real disaster" → push: "What's the hazard? What's the vulnerability? What's the insufficient capacity?" - "This is terrorism, not disaster management" → push: "FEMA's mandate explicitly includes 'technological disasters' and critical infrastructure failures — how do you respond to a flood when your coordination system is compromised?" - "You just harden the systems" → push: "Every security system has vulnerabilities; the question is what happens when they fail during a disaster" KEY INSIGHT: - Digital systems are now critical infrastructure for disaster management itself: early warning systems, emergency dispatch, hospital coordination, supply chain logistics, FEMA aid applications — all digital - Compromise any of these during a disaster and you've attacked the response capacity, not just the warning TRANSITION: "This blurring of categories shows up in vulnerability too — let's look at who gets hit hardest."

DISCUSSION (4 min) Format: 1 min individual think, 3 min open discussion or chat waterfall FRAMING (connect to reading): - McBean argues prediction is "the ultimate public good role of government — to protect its citizens" - If that's true, then failing to act on a prediction is a failure of public duty — not just bad luck FACILITATION TIPS: - The evacuation question is concrete: there are currently millions of people in FEMA flood zones; some can't afford to move; some were never given adequate warning - Push for specific accountability: not "the government" but "which agency, which decision-maker, which moment of choice" - The AI legal obligation question is genuinely unsettled: if a government uses an AI flood prediction and ignores its output, does that create negligence liability? (There is ongoing litigation in Maui over the siren decision) POSSIBLE STUDENT ANSWERS: - "Government should protect everyone" → follow-up: "At what cost? Who decides how to allocate limited DRR resources across mitigation, preparedness, and response?" - "Personal responsibility to prepare" → follow-up: "What about people who can't afford emergency supplies or evacuation? What about renters vs. homeowners?" - "Companies should be liable" → follow-up: "Which companies? The developer? The insurer who underwrote it? The company whose AI predicted it and didn't issue a public warning?" IF DISCUSSION STALLS: - "Should FEMA be required to publish AI confidence intervals on its flood maps — like nutritional labels on food?" - "After Katrina, New Orleans rebuilt in largely the same location. Was that a good decision?" TRANSITION: "The responsibility question always runs into the same obstacle: vulnerability is not evenly distributed."

EQUITY SLIDE (4 min) - This is one of the most important slides — don't rush it - The deaths-per-disaster figure comes directly from McBean: "23 deaths per disaster in highly developed countries" vs. "over 1,000 deaths per disaster in less developed countries" — this is a stunning inequality - The Katrina age statistic is striking and often surprises students KEY EXAMPLES: - FEMA flood maps: critics argue they systematically underestimate risk in poor communities where political pressure to avoid flood zone designation is stronger — a governance capture problem - AI damage assessment: models trained on data from wealthy areas may misclassify damage patterns in informal settlements (tin roofs read differently than tile; satellite imagery resolution affects poor-area data more) - Evacuation compliance: research shows lower compliance in low-income and minority communities — not because of poor judgment but because of structural barriers (no car, no ability to leave work, no money for hotel) CONNECT TO COURSE: - Economic security (Week 5): insurance is the primary recovery mechanism; uninsured losses fall entirely on individuals; the National Flood Insurance Program has debt problems that subsidize risk-taking in wealthy coastal areas - Community security: which communities recover fastest after a disaster? Research consistently shows it tracks income and social capital CRITICAL MOVE: - "The AI can tell you the flood is coming. But if you don't have a car, that forecast doesn't save your life." - This is McBean's prediction-action chain again: prediction is necessary, capacity to respond is what makes it sufficient TRANSITION: "What does international governance look like for this problem?"

SENDAI FRAMEWORK (4 min) - Sendai is the current global reference point for DRR; it explicitly reframes disaster risk as a development issue, not just a humanitarian one - The early warning target is directly aligned with McBean's thesis: prediction and warning as a public good - "Freedom from hazard impacts" (Bogardi and Brauch, cited in reading) aligns with Sendai's framing — this is now mainstream international development language KEY POINTS: - Sendai's enforcement mechanism is voluntary — it creates shared metrics and diplomatic reference points, not binding obligations - UN OCHA (Office for the Coordination of Humanitarian Affairs) is the main operational coordination body for complex emergencies — it has coordination authority, not command authority; sovereign states still decide - The "relative to GDP" framing in the economic losses target means poor countries are held to a proportional standard — but they have the least capacity to meet it - Progress on early warning access has been uneven: SMALL ISLAND DEVELOPING STATES and landlocked least-developed countries have the least coverage despite facing the highest climate hazard exposure CONNECT TO PREVIOUS LECTURES: - Political security (Week 6): sovereignty means each country controls its own disaster response; international actors can only assist, not command - Environmental security (Week 7): same voluntary/non-binding governance architecture as Paris Agreement and other international environmental frameworks CRITICAL QUESTION: - "If Sendai is non-binding, what gives it any force?" → Shared metrics, diplomatic pressure, donor funding conditionality - "Is voluntary coordination better than nothing, or does it substitute for real governance?" → This connects back to Week 3's accountability discussions TRANSITION: "Let's close with what this all means together."

--- ## Mini-Exercise (Optional, 5 min): Disaster Governance Audit Pick one scenario (or assign groups): 1) **City Emergency Manager** — design an equitable early warning system for a flood-prone city with a large non-English-speaking population 2) **UN Coordinator** — develop an AI-assisted response plan for a complex emergency where two sovereign states are contesting jurisdiction over aid corridors 3) **Tech Company** — your disaster prediction AI just flagged a 90% probability major flood event; city officials are skeptical; you have 6 hours Deliverable (1 slide / whiteboard): - Where does the prediction-action chain break in your scenario? - Who has decision authority — and who doesn't? - What would responsible AI use look like here?

MINI-EXERCISE (5 min total: 3 min work + 2 min share) [OPTIONAL] - SKIP THIS if running short on time — the Key Takeaways and simulation bridge matter more - This exercise directly primes students for Simulation Phase 5 (Thursday) - Updated deliverable now explicitly references the prediction-action chain from McBean SETUP (30 sec): - Assign groups or self-select; 3 min work, 2 min share-outs - Point to the three deliverable questions — paste in chat if online USE CASE NOTES: 1) City Emergency Manager: - Expected: multilingual alerts, accessible for disabled, transportation for those without cars - Governance challenge: who has authority to issue mandatory evacuation? What's the liability if you call it wrong (false positive = lost trust; false negative = deaths)? - Push: "How do you build trust with communities that have been burned by false alarms or ignored warnings before?" - McBean connection: this is the social response step in the cascade — communication is as technically complex as the meteorological model 2) UN Coordinator: - Expected: identify who controls territory/aid corridors, how to coordinate multiple NGOs with competing mandates - AI challenge: whose data does the AI access? Can it cross-reference refugee records with displacement data across sovereign borders? - Push: "What happens when your AI recommendation conflicts with what the host government wants?" - McBean connection: sovereignty limits prediction's usefulness — you can know what's happening but be unable to act 3) Tech Company: - Expected: notification to authorities, transparency about confidence interval, documentation of communication - Governance challenge: false positive → panic + liability; false negative → deaths + liability; how do you issue a probabilistic warning as a company? - Push: "McBean says predictions should include guidance on how to respond. How does a private company do that responsibly?" - Connection to Week 3: who is accountable if the AI is wrong? SHARE-OUTS (2 min): - Ask each group: "Where does the prediction-action chain break in your scenario?" - Highlight: the break is almost never in the technical prediction step — it's in the authority, trust, or capacity-to-respond step DEBRIEF: "The hardest part of disaster management is not the forecast. It's getting the forecast to people who have the authority and capacity to act on it. That's a governance problem, not a technical one."

KEY TAKEAWAYS (2 min) - Read each slowly; one sentence of elaboration per point - The "prediction is necessary, not sufficient" takeaway is the direct contribution of the reading — make sure it lands - The "governments underinvest in mitigation" takeaway is the political economy insight from McBean — it reframes disaster governance as a structural problem, not a competence problem CLOSING THOUGHT (from reading): "For sustainable development and disaster management, there is need for an integrated all-hazards information and warning system... These systems must also include information to guide citizens in how they should respond to the information. This is the ultimate public good role of government — to protect its citizens." — McBean SIMULATION BRIDGE (important — next class is Simulation Phase 5): - "Thursday, you'll be running a simulation scenario. Think about where the prediction-action chain sits in your scenario. What information does your team have? Who has the authority to act on it? What happens when the prediction is uncertain?" - Leave students with that concrete framing to carry into the simulation ONLINE EXIT TICKET (paste in chat): "Which step in the prediction-action chain fails most often in the cases we studied? (a) the natural system prediction itself; (b) getting warnings to the right people; (c) people having the capacity to act; (d) governments having political will to invest in advance)" Read distribution, note what students identify, connect to the governance conclusion.

======================================================================== POST-LECTURE NOTES ======================================================================== COMMON STUDENT QUESTIONS: - "Why does FEMA always seem so slow?" — Address: FEMA is a coordination and funding agency, not a first-responder agency; response is primarily a state and local function; FEMA activates only when states formally request a federal disaster declaration; this is by constitutional design - "Can AI actually predict earthquakes?" — Address: No — seismic prediction of timing and magnitude remains a hard scientific problem (McBean addresses this directly); AI helps with early warning (seconds to minutes after a quake starts using seismic wave propagation) and with post-event damage assessment, but not with pre-event prediction; probabilistic hazard mapping is possible but not operational event prediction - "Why do people rebuild in flood zones?" — Address: combination of attachment to place, lack of alternatives (many can't afford to move), insurance subsidies (National Flood Insurance Program), regulatory failures, and in some cases no choice (renters, manufactured housing); the economic incentive structure currently subsidizes risk-taking - "If prediction is getting better, why are disaster costs still rising?" — Address: exposure is rising (more people and infrastructure in hazard zones); economic value of exposed assets is rising; climate change is changing the baseline; this is the "rising vulnerability" dimension McBean emphasizes CONNECTIONS TO OTHER COURSE CONTENT: - Week 1 (Human Security): all seven dimensions are affected by major disasters; ISDR formula (hazard + vulnerability + insufficient capacity) is an applied test of the human security framework; "freedom from hazard impacts" as a proposed third pillar of sustainable development - Week 2 (AI/Cyber): AI systems as disaster response infrastructure — and as failure points; same critical infrastructure vulnerabilities - Week 3 (Ethics): accountability for mitigation failures; AI bias in risk modeling; who is liable when AI predictions are wrong - Week 5 (Economic): insurance as recovery mechanism; uninsured losses; NFIP; economic dimensions of disaster loss and recovery inequality - Week 6 (Political): governance legitimacy in crisis; multi-level coordination failures; sovereignty limiting international response; the political economy of prevention - Week 7 (Environmental): climate change as disaster multiplier; hydrometeorological events at 75% of all disasters; cascading hazard scenarios SIMULATION CONNECTION (Week 9, Mar 19 — Simulation Phase 5): - This lecture provides the analytical frame for the simulation - Ask students to identify: where is the prediction-action chain in their scenario? Who has decision authority? What are the cascade of uncertainty steps? Where does vulnerability intersect with hazard? ASSESSMENT CONNECTION: - The ISDR "hazard + vulnerability + insufficient capacity" formula is directly applicable to project case analyses - The prediction-action chain (McBean Figure 74.1) is a useful analytical tool for any crisis scenario - The political economy of mitigation underinvestment is a recurring theme applicable to infrastructure, public health, and cybersecurity cases RESOURCES FOR DEEPER EXPLORATION: - McBean, G.A. (2006) — course reading; full text covers prediction typology, cascade of uncertainty, and the political economy of prevention - IFRC Hazard Taxonomy: https://www.ifrc.org/en/what-we-do/disaster-management/about-disasters/definition-of-hazard/ - Our World in Data — Natural Disasters: https://ourworldindata.org/natural-disasters - NOAA Natural Hazards Viewer: https://www.ncei.noaa.gov/maps/hazards/ - NOAA Historical Hurricane Tracks: https://coast.noaa.gov/hurricanes - Sendai Framework for Disaster Risk Reduction 2015–2030: https://www.undrr.org/publication/sendai-framework-disaster-risk-reduction-2015-2030 - UN OCHA: https://www.unocha.org/ - FEMA National Preparedness Report (annual) - Google Flood Forecasting Initiative: https://sites.research.google/floods/ - GDACS (Global Disaster Alert and Coordination System): https://www.gdacs.org/ - World Monitor (open-source global intelligence dashboard): https://github.com/koala73/worldmonitor — AI-aggregated feeds, 45 geospatial layers, disaster/geopolitical/financial monitoring; self-hostable with local AI via Ollama - Mileti, D. (1999) "Disasters by Design" — foundational text on how human choices shape disaster impacts - "A Paradise Built in Hell" by Rebecca Solnit — community resilience and mutual aid in disaster response ========================================================================