Hydrological risk is a growing challenge for renewable power systems, as droughts reduce generation, increase costs, and expose gaps in risk-transfer mechanisms. We examine this problem in Brazil, home to the world's sixth-largest electricity system and highly dependent on hydroelectricity, which accounted for 43% of installed generation capacity in 2026. We develop and evaluate a parametric insurance scheme for run-of-river hydroelectric generators indexed to the National System Operator's affluent flow measure. Using monthly energy and climate data from 151 powerplants between 2006 and 2022, we incorporate hydrological network dependence into insurance design. We model generation through a time-varying upstream-downstream spatial network and estimate it using a two-way fixed-effects SAR-IV-GMM-HAC specification. We then use vine copulas to assess the dependence structure among climatic variables, hydrological conditions, and electricity generation, supporting the use of affluent flow as the insurance trigger. Results reveal significant hydrological propagation across connected plants and indicate that parametric insurance can mitigate drought-related losses. However, feasibility depends on plant-specific trigger design, diversification, reinsurance support, and payout magnitude. Water-storage plants exhibit higher basis risk because of their operational flexibility.
Decision and risk analysis tools must be accurate and transparent. Classification trees, especially sparse ones, and other visual models, such as scorecards or risk tables, have been claimed to strike this balance. There is, however, little empirical evidence for the transparency of such models. We derive relevant hypotheses and test them in a controlled laboratory experiment with an ecologically valid, high-risk, critical task: threat classification in peacekeeping. Three classification models are studied: a complete tree, a sparse (specifically, fast-and-frugal) tree, and a risk table. To focus on transparency, all three models make identical classifications and thus have equal accuracy. We assess and score three aspects of transparency for each model: time required to learn to a strict criterion, accuracy of application under time pressure, and accuracy in a delayed surprise memory recall test. In a between-participants design, the fast-and-frugal tree is learned more quickly, applied more accurately, and recalled more accurately than the complete tree and the risk table; all statistical effect sizes are large. The recall accuracy of the fast-and-frugal tree is, in contrast to the other two models, robust to individual differences in statistical numeracy and risk literacy. In sum, the results of the experiment, together with reflection on limitations and challenges, plus theoretical arguments, suggest that sparse trees might serve as a reasonable benchmark and starting point for designing transparent support for risk assessment.
In practice, many climate change risk assessments fail to capture the true depth of uncertainty. Despite widespread scientific recognition of deep uncertainty (large ranges of possibility) in climate conditions, this nuance is often lost in translation to policy and planning contexts; instead, conditions are presented as single projections. This means that communities are making large-scale infrastructure investments and long-term policy commitments based on false precision, leaving them unprepared for climate surprises or potentially wasting resources and disrupting communities unnecessarily by overadapting. To examine how climate uncertainties are represented and accounted for in adaptation planning, we conducted a structured review of 39 climate risk and vulnerability assessments from across the world, using sea level rise as a case study. These documents inform policy that guides billions of dollars in infrastructure investments and shape community preparedness strategies. Our analysis reveals that only 54% of these documents correctly represent sea level rise as deeply uncertain. This issue is compounded when making decisions; 71% of decisions were made by misapplying scenarios as individual projections to plan for, rather than as a tool for exploring potential future conditions, directly contradicting their intended use. This demonstrates a gap between scientific understanding of climate uncertainty and planning practice. Addressing this gap requires improved uncertainty communication, moving beyond just quantifying uncertainty to also characterizing uncertainty. This must be done in conjunction with the support of decision makers to incorporate a stronger understanding of uncertainty into planning by using tools designed specifically for decision making in deeply uncertain environments.
Power outages are a substantial global issue across both advanced and developing countries, affecting economic productivity and growth. Consequently, numerous studies have examined the potential macroeconomic impacts of these disruptions, employing a wide variety of modeling methods and data parameterization techniques. A frequent approach is the use of input-output macroeconomic modeling, yet there is a lack of clarity about how ex ante parameterization and other methodological decisions affect output estimates, warranting further investigation. In this paper, we quantify the macroeconomic effects of three significant natural hazard US power outages: Hurricane Ian (2022), the 2021 Texas Blackouts, and Tropical Storm Isaias (2020). Our analysis evaluates the sensitivity of three commonly used data parameterization techniques (household interruptions, kWh lost, and satellite luminosity), along with three static models (Leontief and Ghosh, critical input, and inoperability input-output). We find the mean domestic loss estimates for these three blackout events to be $2.42 Bn, $3.24 Bn, and $2.27 Bn, respectively. However, data parameterization techniques can alter estimated losses by up to 52.8% of the mean. Consistent with the wide range of outputs, we find that risk analysis stemming from gross output estimate severity is highly sensitive to model architecture, data parameterization, and analyst assumptions. Results sensitivity is not uniform across models and arises from important a priori analyst decisions, demonstrated by data parameterization techniques yielding up to 55.9% differences from empircal results within a model. To our knowledge, we contribute to the literature the first systematic comparison of multiple IO models and parameterizations across several natural hazard long-duration power outages, providing guidance and insights for analysts.
Natural and technological disasters often have transboundary impacts, affecting multiple countries simultaneously, while the lack of joint preparedness and coordinated response mechanisms can significantly amplify damages and complicate recovery efforts. This study proposes an integrated methodology to assess transboundary hazards and population vulnerability across European cross-border regions. The framework combines georeferenced data on extreme weather events, wildfires, industrial areas, and nuclear power plants with a population-based vulnerability analysis. The results revealed pronounced spatial heterogeneity in transboundary hazards distribution across Europe. Extreme weather events affected most European border regions, with the highest values observed in central Europe and along the Portugal-Spain border. Wildfire hazards were strongly concentrated in southern and Mediterranean regions, particularly in the Iberian Peninsula and the Balkans. Technological hazards displayed a more localized pattern, with nuclear power plant-related hazards mainly concentrated in central European border regions and industrial hazards clustered in highly industrialized areas, notably in Belgium, the Netherlands, Germany, Italy, and Switzerland. The combined hazard analysis showed that the highest transboundary hazard levels occurred in regions where multiple hazards spatially co-occurred, such as the Italy-Switzerland, Belgium-the Netherlands, and Portugal-Spain borders, highlighting the importance of multi-hazard interactions. The assessment found that vulnerability was largely determined by population density and demographic structure, with the highest levels occurring in densely populated areas along the Belgium-Germany border. Overall, the proposed framework provides reproducible and policy-relevant evidence to support cross-border risk governance, coordinated prevention strategies, and emergency preparedness planning in Europe.
Malicious and negligent insiders pose significant security risks to mission-critical systems like electric power grids due to their high privileges in increasingly digitized infrastructures. This paper investigates the vulnerability of smart power grids to load redistribution (LR) attacks in the presence of insider threats. We introduce a stochastic optimization model to minimize the expected risk of high operation costs due to LR attacks by protecting critical grid components and deploying detection technologies, such as honeypots, to detect insider threats and prevent information leakage. Our model accounts for uncertainties in insider presence, honeypot effectiveness, and attack targets, and uses the conditional value-at-risk (CVaR) measure, which can be adjusted based on the decision-maker's conservatism. In addition, it accounts for real-time power demand variations and dynamic false-data injection by attackers. To enhance tractability, we transform our model, originally formulated as a trilevel mixed-integer nonlinear programming (Tri-MINLP) problem, into an approximate single-level mixed-integer linear programming (MILP) formulation. We apply our proposed model to the IEEE 14-bus test system, and our results highlight the effectiveness of our approach in lowering the risk of high operation costs due to LR attacks. In addition, we present several insights by assessing the impact of key factors on the expected financial risk of attacks, including the protection budget, insider-threat likelihood, honeypot-detection effectiveness, and the defender's decision-making conservatism.
Understanding virus occurrence and reduction at advanced treatment facilities for potable water reuse constitutes a high-priority research need to protect human health and to enhance available water supply alternatives. The objectives were to (1) determine log reduction values (LRVs) of 15 viruses by advanced wastewater treatment; (2) evaluate suitability of the final wastewater effluent for potable reuse; (3) evaluate viruses or groups of viruses as indicators of wastewater treatment efficiency. Reduction data of seven human enteric and eight surrogate viruses were obtained from three potable reuse facilities. LRVs were estimated using a Bayesian model that can handle nondetects and identifies groups of viruses with similar LRVs. Quantitative microbial risk assessments were conducted for adenoviruses, enteroviruses, and noroviruses GI and GII. Mean total LRVs ranged from 5.4 to 11 log10. Required mean total LRVs for the four pathogenic viruses ranged from 11.3 to 13.4 log10. Both male-specific and somatic coliphages, detected by classical enumeration of infectious virions, are recommended as indicator viruses. Of the pathogenic viruses, adenovirus was found to be the most effective indicator for virus reduction. At all facilities, infection risks were higher than 10-4 per person per year, implying that the finished water may not comply with existing safety standards. Infection risks may have been overestimated by 2-4 log10 because only a fraction of the detected virus was infectious. Nevertheless, achieved LRVs were still too low and given the uncertainty on infectious virus fraction, one may accept overestimation of risks to stay on the safe side.
Urban agglomerations increasingly confront disaster risks that overwhelm individual cities' reserve capacities. Collaborative emergency reserves offer a potential remedy but also introduce coordination burdens and distributional concerns. To address this, this study formulates a multi-objective optimization problem that integrates economic utility with perceived fairness. The model explicitly accounts for residents' shortage and delay experiences alongside managers' perceptions of coordination costs, reserve burdens, and cross-city disparities. It is solved using an enhanced multi-objective particle swarm optimization algorithm coupled with the ε-constraint method and driven by Monte Carlo disaster scenario generation. A numerical illustration based on 21 mainland prefecture-level cities in Guangdong Province evaluates the framework under heterogeneous risk, budget, and coordination conditions. Sobol's global sensitivity analysis identifies regional budget, inter-city collaboration experience, disaster intensity, and nonlinear perceived-loss parameters as dominant drivers of collaborative advantage, whereas transportation costs and standalone city resilience exert weaker effects. The results demonstrate that collaboration generally outperforms independent reserves, especially under severe disasters and fiscal constraints. This advantage diminishes when disasters are widespread or when transaction and coordination costs are prohibitive. Distributional stress tests further reveal that resource-abundant hub cities may perceive lower fairness when acting primarily as providers yet stand to gain substantially when severe shocks occur locally. This research provides a decision-making framework for jointly evaluating efficiency and perceived fairness. It highlights mutual-aid agreements, fiscal compensation, contribution-credit mechanisms, and joint exercises as institutional complements to optimized reserve strategies.
The Fermi paradox may partly reflect technological self-destruction acting as a Great Filter during post-industrial development. Because the empirical record consists of a single civilization observed over an eighty-year nuclear age, the exercise reported here is best understood as a structured, historically anchored scenario analysis rather than a data-driven statistical estimate. We develop a time-varying hazard model to estimate civilizational survival under nuclear risk, using Earth's documented close calls as an illustrative case. A Bayesian survival-analysis framework with regime-switching hazards captures temporal non-stationarity between Cold War and post-Cold War periods and propagates parameter uncertainty. Under high-alert nuclear postures, estimated annual hazards range from roughly 0.1% to 2%, implying median survival times on the order of decades to millennia conditional on institutional configuration. These hazards plausibly exceed natural extinction risks but may decline substantially with improved governance and control systems. The results bound detectable civilizational longevity and offer conditional implications for existential-risk prioritization and SETI expectations under deep uncertainty.
Maritime risk is a fundamental component of operational cost and decision-making in global container shipping, yet quantitative, route-comparable risk estimates, especially in monetary terms and under future climate conditions, remain limited. This study develops a climate-driven Bayesian network (BN)-based Global Maritime Incident Risk Assessment and Prediction tool to quantify how projected future climate conditions may alter maritime incident risks at the grid-cell, link, and route levels. Global open waters are partitioned into 2° × 2° grid cells, and route risk is computed by aggregating cell-level risk along a path. The framework combines: (1) an estimate of incident occurrence probability derived from historical traffic and incident patterns, (2) two tree augmented BN models that predict incident type and then severity using discretized climate and bathymetry variables, and (3) monetary consequence values to express risk in real units ($ per voyage). Using monthly climate projections for 2025-2069, the approach is demonstrated on container routes in the 2M alliance network. In addition to grid cell, link and route-based incident probabilities and their projections, a key contribution is interpretable, global monetary risk values and predictions that enable direct comparison of competing routes.
In this article, we model the endogenous boundary of informativeness of uncertainty indices (UIs) using a panel threshold framework, employing both static LS and dynamic system-GMM specifications, and gauge the predictive power of these thresholds on economic outcomes. Specifically, we identified critical thresholds where UI shifts from background noise (uninformative) to a decisive factor (informative) in renewable energy (RE) adoption decisions. We find that the negative relationship between the UIs and energy transition is only significant and identified in the regime where uncertainty is relatively low, whereas the prediction power fails (becomes insignificant) in the high-uncertainty regimes. This suggests that the informational utility of UIs in driving (or forecasting) RE adoption operates conditionally upon moderate uncertainty levels. To explain the heterogeneous informativeness of UIs and validate the variations in our identified thresholds, we further conduct channel analyses. Relying on theoretical frameworks from risk literature, we show that risk preference, information overload, and priority concern theories can explain the level of threshold effects (informativeness) of UIs on economic outcomes. These findings provide an actionable insight for policymakers towards designing preemptive risk triggers when uncertainty crosses a critical threshold by frontloading investment activities in clean energy technologies to safeguard green energy transition pathways from headwinds of uncertainties.
A critical gap in disaster response persists between optimized logistical plans and the complex, often unpredictable evacuation behaviors of affected populations. Traditional location-allocation models often prove insufficient in practice, as they typically treat evacuees as homogeneous, rational actors, thereby overlooking the dynamic, hierarchical needs that govern decision-making on the ground. This study addresses this gap by introducing a hybrid optimization framework that systematically translates data-driven behavioral insights into a multi-objective model for shelter location and material assignment. Rather than modeling microscopic routing behaviors, our methodological innovation lies in a two-stage pipeline. First, we integrate thematic analysis of large-scale social media data (3400 relevant posts) from recent earthquakes to extract and quantify time-varying demand patterns for three core needs: essential survival, medical treatment, and psychological needs. Second, these empirically derived priority dynamics are embedded as dynamic parameters within the optimization model. Validated against the 2013 Ya'an earthquake case study and compared with classical baselines (e.g., the p-median model), the framework reduces the average evacuation distance while ensuring feasible access for heterogeneous vulnerable groups. Concurrently, it enhances resource equity by dynamically prioritizing material allocation based on the evolving urgency of evacuee needs. This research thus contributes not only a specific decision-support tool but also a generalizable paradigm for bridging descriptive behavioral data with prescriptive operations models, offering a more realistic foundation for risk analysis in humanitarian logistics.
Essential services are necessary for individuals to recover from and adapt to disruptive events, yet the relationship between access to such services and recovery trajectory is poorly understood. Data on household recovery times are typically only available through surveys, which are limited in scale and scope. Further, the availability of all open and accessible essential services facilities in a region may not be recorded anywhere. Location-based services (LBS) data from cell phones offers new opportunities for estimating this relationship. Using LBS data, we approximate facility availability and household recovery times following Hurricane Irma in Southwest Florida in 2017 and then statistically model the importance of access for recovery incorporating social vulnerability, local storm parameters, and infrastructure outage variables. We show that power, cell service, and school outages rank highest in importance, followed by measures of access to essential services. These results underscore the importance of including access metrics for predicting community recovery and evaluating resilience as well as the need for planning and policies that improve access to essential services both in times of stability and disruption.
Elevated healthcare strain during the COVID-19 pandemic increased patient mortality rates and prompted costly non-pharmaceutical interventions. This work examines how the size of a hospital's service population, which can be expanded by linking hospitals through patient transfer networks, influences healthcare strain through two diversification mechanisms: (i) volatility dampening, which reduces the volatility of demand for healthcare resources by aggregating uncorrelated individual patient needs, and (ii) epidemic phase averaging, which flattens demand peaks by aggregating patients from asynchronous local epidemics. Using facility-level intensive care unit (ICU) data across three COVID waves, we analyze how service population size affects the size of peaks in ICU occupancy, evaluate the two proposed mechanisms, and explore the health impacts of pooling through scenario simulations grounded in observed occupancy levels. We find that volatility dampening effects reduce variability in ICU occupancy at a rate proportional to the square root of population size, whereas the effectiveness of epidemic phase averaging is dependent on the transmission dynamics of the pathogen. Increasing service population size also reduces the frequency and magnitude of demand spikes, lowering the likelihood of costly short-term interventions and increasing demand predictability. In the winter 2020 wave, our simulations find that randomly constructed local pools (2-6 hospitals) reduce COVID ICU mortality by a median of 2%-7%, with reductions exceeding 6%-18% in the top decile of outcomes. These findings suggest that coordinated patient transfer strategies could meaningfully reduce mortality and costs during future pandemics.
Maximizing the public benefits of flood insurance in high-risk areas requires premiums that are affordable and measures that minimize the concentrated demand that causes adverse selection. Governments make policy choices for flood insurance based on their national interests, and these choices lead to variation in how this balance is achieved. This article evaluates the influence of different design elements on the policy objectives identified for the Canadian Flood Insurance Program (CFIP), which include affordability, adequate compensation, and market participation. Combinations of deductibles, limits, exclusions, and subsidies were evaluated using a novel dataset of flood losses from 700,000 simulated flood events to determine how they shape the trade-offs between these objectives. The optimal design includes a progressive subsidy, premium caps, a moderate deductible, a coverage limit, and exclusions for the highest risk households. These findings offer a useful comparison with international practices and insights for policymakers seeking to maximize the benefits of flood insurance in Canada.
This study proposes a systems-based analysis approach for analyzing interorganizational coordination in complex sociotechnical systems through the interorganizational risk management coordination taxonomy (IRM-CT). Unlike traditional risk-analysis approaches that primarily focus on intraorganizational controls, probabilistic assessment, or component reliability, the IRM-CT operationalizes interorganizational coordination as a systemic control problem embedded within distributed governance and feedback structures. Building on the STAMP-based Port Risk Control Structure (PRCS), the IRM-CT distinguishes between operational adaptors, which enact coordination in practice, and enabling adaptors, which provide the structural, procedural, and relational conditions that sustain coordination across organizational boundaries. The approach integrates coordination theory, sociotechnical systems thinking, and resilience engineering, embedding coordination within control and feedback loops and evaluating adaptor configurations through three system value conditions: accountability, predictability, and common understanding. A proof-of-concept analysis using the Halifax Port Authority Port Information Guide demonstrates how the IRM-CT identifies adaptor configurations, diagnoses "Supported," "Fragile," and "Orphan" coordination states, and evaluates their contribution to accountability, predictability, and common understanding. The IRM-CT advances systemic risk governance by revealing how authority, codification, and trust interact to sustain safe control. Beyond ports, the proposed approach may be applicable to high-risk sectors such as mining, aviation, and energy, offering a structured means for diagnosing and strengthening interorganizational coordination.
This study examines the impact of epistemic uncertainties on the seismic risk of a critical transport system, the Campania railway network, during the ongoing bradyseismic unrest at Campi Flegrei, south of Italy. Seismic risk is quantified as the length of the railway requiring post-event inspection, an operational proxy for service disruption. A scenario-based risk formulation is adopted, integrating fragility functions for tunnels and masonry arch bridges with a ground-motion model tailored explicitly to volcanic areas. Peak ground acceleration serves as the intensity measure, whereas the annual acceptable probability of failure from the Eurocode informs safety decisions. Epistemic inputs comprise epicenter-location error and azimuth, duration-magnitude error, central-value variability in the ground-motion model, a soil term, and median fragility parameters. A regional sensitivity analysis (SA) using Latin hypercube sampling reveals that uncertainties in epicenter location have a negligible impact on the length of the inspected network. In contrast, hazard- and vulnerability-related variables, including magnitude, ground-motion median, soil class, and fragility medians, exert a strong influence. Increases in magnitude or ground-motion median, or softer soils, markedly expand the network to be inspected, whereas lower component vulnerability, such as bridges and tunnels, reduces it. A variance-based global SA extends the assessment to capture first-order and interaction effects, supporting the prioritization of data collection and model refinement to achieve the greatest reductions in disruption. The methodology provides actionable guidance for infrastructure managers to target and enhance monitoring and mitigation strategies under bradyseismic crises.
Microbial contamination of drinking water poses a significant public health concern, yet age-specific risks in Islamabad, Pakistan, remain poorly characterized. This study applied a Monte Carlo simulation-based quantitative microbial risk assessment to estimate the annual probability of infection (P(a)inf) and disease burden (DB) associated with indicator organisms Escherichia coli (E. coli) and total coliforms across four age groups and one overall population category at three microbiologically contaminated drinking-water locations in Islamabad. Hydrochemical analysis, including principal component analysis, was conducted to characterize groundwater chemistry and explore links between physicochemical conditions and contamination vulnerability. Estimated risks were evaluated against U.S. EPA and WHO health-based benchmarks, and sensitivity analysis was used to identify the main drivers of DB. The results showed that estimated P(a)inf exceeded the U.S. EPA benchmark (1E-4 pppy) and DB exceeded the WHO benchmark (1E-6 DALYs pppy) across all modeled age groups at the three microbiologically contaminated drinking-water locations. Risk estimates were highest among middle-aged adults and elderly individuals and lowest among children, primarily due to differences in assumed drinking-water intake volumes. The total coliforms-associated risks were generally higher than those associated with E. coli. Sensitivity analysis identified exposure concentration as the dominant contributor to DB variability, accounting for 73%-75% of total variance. Some microbiologically unsafe locations showed hydrochemical characteristics associated with higher mineralization, including elevated EC, TDS, and HCO3 - concentrations; however, these associations represent statistical patterns rather than evidence of direct causal relationships between hydrochemical conditions and microbial contamination. These findings suggest that routinely measured physicochemical parameters may serve as practical screening indicators for identifying contamination-prone zones and support targeted drinking water monitoring and intervention in Islamabad.
Human-artificial intelligence (AI) teams are increasingly embedded in risk analysis, yet news reports repeatedly portray meaningful oversight as fragile when review conditions are poorly designed. This study examines how public discourse has portrayed the roles, failure modes, and governance mechanisms of human-AI teaming in risk analysis. Drawing on 184,282 AI-related Factiva news articles (1956-2025), we use a funnel-shaped mixed-methods design to identify 3571 articles describing human-AI interaction in risk-analytic contexts. Structural topic modeling identifies 20 thematic clusters and their temporal and cross-domain dynamics, whereas LLM-assisted inductive thematic analysis produces five mirror-mapped pairs of failure modes and governance mechanisms: opacity and explainability, bias and auditing, agency erosion and cognitive friction, systemic fragility and oversight and liability, and accountability vacuums and institutional governance. The analysis shows that media discourse often follows an AI-predicts-human-decides paradigm and portrays formal human-in-the-loop requirements as insufficient without deliberate interaction design.
