@article{li2025incorporating,
  title = {Incorporating environmental considerations into infrastructure inequality evaluation using interpretable machine learning},
  author = {Li, Bo and Mostafavi, Ali},
  journal = {Computers, Environment and Urban Systems},
  volume = {120},
  pages = {102301},
  year = {2025},
  publisher = {Elsevier},
  doi = {10.1016/j.compenvurbsys.2025.102301}
}

A growing body of literature has recognized the importance of characterizing infrastructure inequality in cities and provided quantified metrics to inform urban development plans. However, the majority of existing approaches suffered from two limitations. First, prior research has provided empirical evidence of negative environmental impacts that infrastructure can incur, while infrastructure provision inequality assessment has not taken those environmental concerns into consideration. Second, comprehensive provision assessment for multi-infrastructure system calls for a proper weight assignment, while current studies either determine the infrastructure components as equal weights or rely on subjective methods (e.g. AHP), which may be affected by potential biases. This study proposes a novel approach for incorporating environmental considerations into quantifying and assessing infrastructure provision in cities based on a data-driven method. We applied an interpretable machine learning method (XGBoost + SHAP) to capture the relationship between infrastructure features and environmental hazards (i.e., air pollution and urban heat), and then determined feature weights as their relative contributions towards environmental hazards when calculating infrastructure provision. The implementation of the model in five metropolitan areas in the U.S. demonstrates the capability of the proposed approach in characterizing inequality in infrastructure. Further the study reveals both spatial and income inequality regarding infrastructure provision. Environmentally integrated infrastructure provision proposed in this study can better capture the intersection of infrastructure development and environmental justice in measuring and characterizing infrastructure inequality in cities. This study could be used effectively to inform integrated urban design strategies to promote infrastructure equity and environmental justice based on data-driven and machine learning-based insights.

@article{li2025revealing,
  title = {Revealing Growing and Disparate Vulnerability in the US Power System: A Spatiotemporal Analysis of Nationwide Outages from 2014 to 2023},
  author = {Li, Bo and Ma, Junwei and Omitaomu, Femi and Mostafavi, Ali},
  journal = {International Journal of Disaster Risk Reduction},
  pages = {105980},
  year = {2025},
  publisher = {Elsevier},
  doi = {10.1016/j.ijdrr.2025.105980}
}

Power systems are increasingly challenged by a range of external and internal threats that undermine their reliability and resilience. Power system vulnerability, the proneness of the power system to disruptions, can be empirically characterized through the observable manifestations of power outages. However, existing research remains limited in spatial coverage, temporal scope, and analytical consistency, and lacks comprehensive, longitudinal, large-scale, and fine-grained analyses to capture the spatiotemporal dynamics of vulnerability. Recognizing this, we analyzed 179,053,397 county-level power outage records with a 15-min interval across 3,022 US counties during 2014–2023. Applying a framework encompassing frequency, duration and intensity, we systematically assessed the dynamics of U.S. power system vulnerability. Results reveal an escalating trend over the past decade, with outages becoming more frequent, prolonged, and intense. Nationally, cumulative customer outage time reached 7.86 billion customer-hours, with a median of 0.64 million per-county over the past decade, underscoring significant service disruptions. Coastal regions, especially in California, Florida, and New Jersey, experienced more frequent and longer outages, while some inland areas exhibited higher outage intensity relative to their customer base. Moreover, we observed a strengthening association between social vulnerability and outage metrics over time, indicating that counties with higher social vulnerability experienced more severe and frequent outages, creating “dual-burden” regions where social disadvantage and infrastructural vulnerability compound each other. These findings provide a nationwide and longitudinal characterization of power system vulnerability in the U.S., offering empirical insights to inform practitioners in prioritizing investments for a more reliable, resilient and equitable energy infrastructure.

@article{ma2025establishing,
  title = {Establishing nationwide power system vulnerability index across US counties using interpretable machine learning},
  author = {Ma, Junwei and Li, Bo and Omitaomu, Olufemi A and Mostafavi, Ali},
  journal = {Applied Energy},
  volume = {397},
  pages = {126360},
  year = {2025},
  publisher = {Elsevier},
  doi = {10.1016/j.apenergy.2025.126360}
}

Power outages have become increasingly frequent, intense, and prolonged in the US due to climate change, aging electrical grids, and rising energy demand. However, largely due to the absence of granular spatiotemporal outage data, we lack data-driven evidence and analytics-based metrics to quantify power system vulnerability. This limitation has hindered the ability to effectively evaluate and address vulnerability to power outages in US communities. Here, we collected ∼179 million power outage records at 15-min intervals across 3022 US contiguous counties (96.15 % of the area) from 2014 to 2023. We developed a power system vulnerability assessment framework based on three dimensions (intensity, frequency, and duration) and applied interpretable machine learning models (XGBoost and SHAP) to compute Power System Vulnerability Index (PSVI) at the county level. Our analysis reveals a consistent increase in power system vulnerability across the US counties over the past decade. We identified 318 counties across 45 states as hotspots for high power system vulnerability, particularly in the West Coast (California and Washington), the East Coast (Florida and the Northeast area), the Great Lakes megalopolis (Chicago-Detroit metropolitan areas), and the Gulf of Mexico (Texas). Our heterogeneity analysis indicates that urban counties and those located along regional transmission boundaries tend to exhibit significantly higher vulnerability. Our results highlight the significance of the proposed PSVI for evaluating the vulnerability of communities to power outages. The findings underscore the widespread and pervasive impact of power outages across the country and offer crucial insights to support infrastructure operators, policymakers, and emergency managers in formulating policies and programs aimed at enhancing the resilience of the US power infrastructure.

@article{esparza2025ai,
  title = {AI Meets Natural Hazard Risk: A Nationwide Vulnerability Assessment of Data Centers to Natural Hazards and Power Outages},
  author = {Esparza, Miguel and Li, Bo and Ma, Junwei and Mostafavi, Ali},
  journal = {International Journal of Disaster Risk Reduction},
  pages = {105583},
  year = {2025},
  publisher = {Elsevier},
  doi = {10.1016/j.ijdrr.2025.105583}
}

With increasing advancements in artificial intelligence (AI), there is a growing expansion in data centers (DCs) serving as critical infrastructure for this new wave of technologies. This technological wave is also on a collision course with exacerbating climate hazards which raises the need for evaluating the vulnerability of DCs to various hazards. Hence, the objective of this research is to conduct a nationwide vulnerability assessment of (DCs) in the United States of America (USA). There is an increasing importance to maintain new infrastructure to support the digital age thanks to the emergence of these innovative technologies being public. DCs provide such support; however, if an unplanned disruption (like a natural hazard or power outage) occurs, the functionality of DCs are in jeopardy. Unplanned downtime in DCs cause severe economic and social repercussions. Therefore, this research uses spatial analysis methods to assess the current vulnerability of DCs toward natural hazard and power outages. With the Local Indicator of Spatial Association (LISA) test, the research found that there are a large percentage of DCs that are in non-vulnerable areas of disruption; however, there is still a notable percentage in disruption prone areas. For example, out of the natural hazards investigated, areas vulnerable to earthquakes, hurricanes, and tornadoes exhibit the most DCs. When examining power outages, DCs reside in areas that have faced frequent power outages during 2014–2022 and experience long durations without power. Next, the research identified areas within the USA that have minimal vulnerabilities to both the aforementioned natural hazards and power outages with the BI-LISA test. Then a composite vulnerability score on the Cold-Spots from the BI-LISA analysis was conducted based on natural hazards risk, power outages, physical features, and social features. Two weighting schemes are implemented, and a Monte Carlo simulation is used to analyze the sensitivity of the composite vulnerability scores. Ouray, Colorado (0.238) and Kandiyohi, Minnesota (0.239) have the top two scores in scheme one. White Pine, Nevada (0.099) and Garfield, Montana (0.115) have top scores in scheme two. McCone, Montana and Ouray, Colorado are the only counties to appear in both schemes. The contribution of this research is to provide infrastructure managers with interpretable maps to guide their decision-making and by understanding the current vulnerabilities, they can develop specific solutions to ensure the functionality of DCs.

@article{li2024human,
  title = {Human mobility disproportionately extends pm2. 5 emission exposure for low income populations},
  author = {Li, Bo and Fan, Chao and Chien, Yu-Heng and Mostafavi, Ali},
  journal = {Sustainable Cities and Society},
  pages = {106063},
  year = {2024},
  publisher = {Elsevier},
  doi = {10.1016/j.scs.2024.106063},
  cover = {li2024human_cover.jpg}
}

Ambient exposure to fine particulate matter with diameters smaller than 2.5 μm (PM2.5) has been identified as one critical cause for respiratory disease. Disparities in exposure to PM2.5 among income groups at individual residences are known to exist and are easy to calculate. Existing approaches for exposure assessment, however, do not capture the exposure implied by the dynamic mobility of city dwellers that accounts for a large proportion of the exposure outside homes. To overcome the challenge of gauging the exposure to PM2.5 for city dwellers, we analyzed billions of anonymized and privacy-enhanced location-based data generated by mobile phone users in Harris County, Texas, to characterize the mobility patterns of the populations and associated exposure. We introduce the metric for exposure extent based on the time people spent at places with the air pollutant and examine the disparities in mobility-based exposure across income groups. Our results show that PM2.5 concentrations disproportionately expose low-income populations due to their mobility activities. People with higher-than-average income are exposed to lower levels of PM2.5 concentrations. These disparities in mobility-based exposure are the result of frequent visits of low-income people to the industrial sectors of urban areas with high PM2.5 concentrations, and the larger mobility scale of these people for life needs. The results inform environmental justice and public health strategies, not only to reduce the overall PM2.5 exposure but also to mitigate the disproportional impacts on low-income populations. The findings also suggest that an integration of extensive fine-scale population mobility and pollution concentration data can unveil new insights into inequality in air pollution exposures at urban scale.

@article{li2024unraveling,
  title = {Unraveling fundamental properties of power system resilience curves using unsupervised machine learning},
  author = {Li, Bo and Mostafavi, Ali},
  journal = {Energy and AI},
  volume = {16},
  pages = {100351},
  year = {2024},
  publisher = {Elsevier},
  doi = {10.1016/j.egyai.2024.100351},
  cover = {li2024unraveling_cover.jpg}
}

Power system is vital to modern societies, while it is susceptible to hazard events. Thus, analyzing resilience characteristics of power system is important. The standard model of infrastructure resilience, the resilience triangle, has been the primary way of characterizing and quantifying resilience in infrastructure systems for more than two decades. However, the theoretical model provides a one-size-fits-all framework for all infrastructure systems and specifies general characteristics of resilience curves (e.g., residual performance and duration of recovery). Little empirical work has been done to delineate infrastructure resilience curve archetypes and their fundamental properties based on observational data. Most of the existing studies examine the characteristics of infrastructure resilience curves based on analytical models constructed upon simulated system performance. There is a dire dearth of empirical studies in the field, which hindered our ability to fully understand and predict resilience characteristics in infrastructure systems. To address this gap, this study examined more than two hundred power-grid resilience curves related to power outages in three major extreme weather events in the United States. Through the use of unsupervised machine learning, we examined different curve archetypes, as well as the fundamental properties of each resilience curve archetype. The results show two primary archetypes for power grid resilience curves, triangular curves, and trapezoidal curves. Triangular curves characterize resilience behavior based on three fundamental properties: 1. critical functionality threshold, 2. critical functionality recovery rate, and 3. recovery pivot point. Trapezoidal archetypes explain resilience curves based on 1. duration of sustained function loss and 2. constant recovery rate. The longer the duration of sustained function loss, the slower the constant rate of recovery. The findings of this study provide novel perspectives enabling better understanding and prediction of resilience performance of power system infrastructure in extreme weather events.

@article{li2024mobility,
  title = {Mobility behaviors shift disparity in flood exposure in US population groups},
  author = {Li, Bo and Fan, Chao and Chien, Yu-Heng and Hsu, Chia-Wei and Mostafavi, Ali},
  journal = {International Journal of Disaster Risk Reduction},
  volume = {108},
  pages = {104545},
  year = {2024},
  publisher = {Elsevier},
  doi = {10.1016/j.ijdrr.2024.104545},
  cover = {li2024mobility_cover.jpg}
}

The current characterization of flood exposure is largely based on residential location of populations; however, location of residence only partially captures the extent to which populations are exposed to flood hazards. An important, yet under-recognized aspect of flood exposure is associated with human mobility patterns and population visitation to places located in flood prone areas. In this study, we analyzed large-scale, high-resolution location-intelligence data collected from anonymous mobile phone users to characterize human mobility patterns and the resulting flood exposure in coastal counties of the United States. We developed the metric of mobility-based exposure based on dwell time in places located in the 100-year floodplain. The results of examining the extent of mobility-based flood exposure across demographic groups reveal significant disparities across race, income, and education level groups. The results show that Black and Asian, economically disadvantaged, and undereducated populations in US coastal cities are disproportionally exposed to flood due to their daily mobility activities, indicating a pattern contrary to that of residential flood exposure. The results suggest that mobility behaviors play an important role in extending flood exposure reach disproportionally among different socio-demographic groups. The results highlight that urban flood risk assessments should not only focus on the level of flood exposure to residences, but also should consider mobility-based exposure to better learn the disparities in flood exposure among social groups. Mobility-based flood exposure provides a new perspective regarding the extent to which floods could disrupt people’s life activities and enable a better characterization of disparity in populations’ exposure to flood hazards beyond their place of residence. The findings of this study have important implications for urban planners, flood managers, and city officials in terms of accounting for mobility-based flood exposure in flood risk management plans and actions.

@article{ma2024characterizing,
  title = {Characterizing urban lifestyle signatures using motif properties in network of places},
  author = {Ma, Junwei and Li, Bo and Mostafavi, Ali},
  journal = {Environment and Planning B: Urban Analytics and City Science},
  volume = {51},
  number = {4},
  pages = {889--903},
  year = {2024},
  publisher = {SAGE Publications Sage UK: London, England},
  doi = {10.1177/23998083231206171},
  cover = {ma2024characterizing_cover.jpg}
}

The lifestyles of urban dwellers could reveal important insights regarding the dynamics and complexity of cities. The availability of human movement data captured from cell phones enables characterization of distinct and recurrent human daily visitation patterns. Despite growing research on analysis of lifestyle patterns in cities, little is known about the characteristics of people’s lifestyle patterns at urban scale. This limitation is primarily due to challenges in restriction of human movement data to protect the privacy of users. To address this gap, this study constructed networks of places to model cities based on location-based human visitation data. We examined the motifs in the networks of places to map and characterize lifestyle patterns at urban scale. The results show that (1) people’s lifestyles in cities can be well depicted and quantified based on distribution and attributes of motifs in networks of places; (2) motifs show stability in quantity and distance as well as periodicity on weekends and weekdays indicating the stability of lifestyle patterns in cities; (3) networks of places and lifestyle patterns show similarities across different metropolitan areas implying the universality of lifestyle signatures across cities; (4) lifestyles represented by attributed motifs are spatially heterogeneous suggesting variations of lifestyle patterns within different population groups based on where they live in a city. The findings provide deeper insights into urban lifestyle signatures and significant implications for data-informed urban planning and management.

@article{ma2024attributed,
  title = {Attributed network embedding model for exposing COVID-19 spread trajectory archetypes},
  author = {Ma, Junwei and Li, Bo and Li, Qingchun and Fan, Chao and Mostafavi, Ali},
  journal = {International Journal of Data Science and Analytics},
  pages = {1--18},
  year = {2024},
  publisher = {Springer},
  doi = {10.1007/s41060-024-00627-5},
  cover = {ma2024attributed_cover.jpg}
}

The spread of COVID-19 revealed that transmission risk patterns are not homogenous across different cities and communities, and various heterogeneous features can influence the spread trajectories. Hence, for predictive pandemic monitoring, it is essential to explore latent heterogeneous features in cities and communities that distinguish their specific pandemic spread trajectories. To this end, this study creates a network embedding model capturing cross-county visitation networks, as well as heterogeneous features related to population activities, human mobility, socio-demographic features, disease attribute, and social interaction to uncover clusters of counties in the USA based on their pandemic spread transmission trajectories. We collected and computed location intelligence features from 2787 counties from March 3 to June 29, 2020 (initial wave). Second, we constructed a human visitation network, which incorporated county features as node attributes, and visits between counties as network edges. Our attributed network embeddings approach integrates both typological characteristics of the cross-county visitation network, as well as heterogeneous features. We conducted clustering analysis on the attributed network embeddings to reveal four archetypes of spread risk trajectories corresponding to four clusters of counties. Subsequently, we identified four features—population density, GDP, minority status, and POI visits—as important features underlying the distinctive transmission risk patterns among the archetypes. The attributed network embedding approach and the findings identify and explain the non-homogenous pandemic risk trajectories across counties for predictive pandemic monitoring. The study also contributes to data-driven and deep learning-based approaches for pandemic analytics to complement the standard epidemiological models for policy analysis in pandemics.

@article{li2022location,
  title = {Location intelligence reveals the extent, timing, and spatial variation of hurricane preparedness},
  author = {Li, Bo and Mostafavi, Ali},
  journal = {Scientific reports},
  volume = {12},
  number = {1},
  pages = {16121},
  year = {2022},
  publisher = {Nature Publishing Group UK London},
  doi = {10.1038/s41598-022-20571-3},
  cover = {li2022location_cover.jpg}
}

Hurricanes are one of the most catastrophic natural hazards faced by residents of the United States. Improving the public’s hurricane preparedness is essential to reduce the impact and disruption of hurricanes on households. Inherent in traditional methods for quantifying and monitoring hurricane preparedness are significant lags, which hinder effective monitoring of residents’ preparedness in advance of an impending hurricane. This study establishes a methodological framework to quantify the extent, timing, and spatial variation of hurricane preparedness at the census block group level using high-resolution location intelligence data. Anonymized cell phone data on visits to points of interest for each census block group in Harris County before 2017 Hurricane Harvey were used to examine residents’ hurricane preparedness. Four categories of points of interest, grocery stores, gas stations, pharmacies, and home improvement stores, were identified as they have a close relationship with hurricane preparedness, and the daily number of visits from each census block group to these four categories of points of interest were calculated during the preparation period. Two metrics, extent of preparedness and proactivity, were calculated based on the daily visit percentage change compared to the baseline period. The results show that peak visits to pharmacies often occurred in the early stage of preparation, whereas the peak of visits to gas stations happened closer to hurricane landfall. The spatial and temporal patterns of visits to grocery stores and home improvement stores were quite similar. However, correlation analysis demonstrates that extent of preparedness and proactivity are independent of each other. Combined with synchronous evacuation data, census block groups in Harris County were divided into four clusters in terms of extent of preparedness and evacuation rate. The clusters with low preparedness and low evacuation rate were identified as hotspots of vulnerability for shelter-in-place households that would need urgent attention during response. Hence, the research findings provide a new data-driven approach to quantify and monitor the extent, timing, and spatial variations of hurricane preparedness. Accordingly, the study advances data-driven understanding of human protective actions during disasters. The study outcomes also provide emergency response managers and public officials with novel data-driven insights to more proactively monitor residents’ disaster preparedness, making it possible to identify under-prepared areas and better allocate resources in a timely manner.

@article{wang2022trust,
  title = {Trust repair in the aftermath of conflict occurrence in construction subcontracting: an intergroup contact perspective},
  author = {Wang, Conghan and Zhang, Shuibo and Gao, Ying and Li, Bo},
  journal = {Construction management and economics},
  volume = {40},
  number = {10},
  pages = {781--795},
  year = {2022},
  publisher = {Taylor \& Francis},
  doi = {10.1080/01446193.2022.2110272},
  cover = {wang2022trust_cover.jpg}
}

Although trust repair after conflict occurrence is significant for effective work and sustainable relationships in construction subcontracting and contact is inevitable, researchers have yet to examine the relationship between contact quality and trust repair after conflict occurrence. This study investigates the effect of contact quality on Party A’s trust repair after conflict occurrence and the mediating mechanisms of that effect. The authors conducted a questionnaire survey to collect data for hypotheses testing, receiving 310 valid questionnaires from general contractors and subcontractors engaged in construction projects. The results reveal the positive effect of contact quality on Party A’s trust repair after conflict occurrence and the mediating roles of Party A’s feeling of threat and Party B’s self-disclosure in that effect. This study contributes to the trust research and intergroup contact theory. It also offers suggestions for construction subcontracting practitioners to facilitate trust repair after conflict occurrence.

@article{lee2022quantitative,
  title = {Quantitative measures for integrating resilience into transportation planning practice: Study in Texas},
  author = {Lee, Cheng-Chun and Rajput, Akhil Anil and Hsu, Chia-Wei and Fan, Chao and Yuan, Faxi and Dong, Shangjia and Esmalian, Amir and Farahmand, Hamed and Patrascu, Flavia Ioana and Liu, Chia-Fu and Li, Bo and Ma, Junwei and Mostafavi, Ali},
  journal = {Transportation Research Part D: Transport and Environment},
  volume = {113},
  pages = {103496},
  year = {2022},
  publisher = {Elsevier},
  doi = {10.1016/j.trd.2022.103496},
  cover = {lee2022quantitative_cover.jpg}
}

Using quantitative measures to assess road network resilience, this study proposes a system-level framework that offers insights into existing road network resilience that could inform the planning and development processes of transportation agencies. This study identified and implemented four quantitative metrics to classify the criticality of road segments based on dimensions of road network resilience. Integrating the four metrics of classification using two mathematical approaches, this study arrived at overall resilience performance metrics for assessing road segment criticality. A case study was conducted on Texas road networks to demonstrate the effectiveness of implementing this framework in a practical scenario. The data used in this study is available to other states and countries; thus, the framework presented in this study can be adopted by transportation agencies for regional transportation resilience assessments.

@article{li2021understanding,
  title = {Understanding the effects of trust and conflict event criticality on conflict resolution behavior in construction projects: Mediating role of social motives},
  author = {Li, Bo and Gao, Ying and Zhang, Shuibo and Wang, Conghan},
  journal = {Journal of Management in Engineering},
  volume = {37},
  number = {6},
  pages = {04021066},
  year = {2021},
  publisher = {American Society of Civil Engineers},
  doi = {10.1061/(ASCE)ME.1943-5479.0000962},
  cover = {li2021understanding_cover.jpg}
}

Because conflicts are unavoidable in construction projects, resolving them constructively is essential. The purpose of this paper is to examine whether trust and conflict event criticality can affect conflict resolution behaviors, and, if so, how. This research proposes that trust and conflict event criticality influence conflict resolution behaviors mediated by prosocial motive and proself motive, respectively. An empirical study using 253 valid questionnaires was conducted to test the hypotheses. The results verify that trust has positive impacts on integrating, compromising, and obliging behaviors mediated by a prosocial motive, while conflict event criticality positively affects obliging and dominating behaviors via a proself motive. Regarding the direct effects, trust and conflict event criticality are positively associated with integrating behavior, while trust also has a negative effect on dominating behavior. This study highlights how trust and conflict event criticality affects conflict resolution behaviors and uncovers the underlying psychological mechanisms. This paper also provides implications for practitioners.

@unpublished{chen2026disastqa,
  title = {DisastQA: A Comprehensive Benchmark for Evaluating Question Answering in Disaster Management},
  author = {Chen, Zhitong and Yin, Kai and Dong, Xiangjue and Liu, Chengkai and Li, Xiangpeng and Xiao, Yiming and Li, Bo and Ma, Junwei and Mostafavi, Ali and Caverlee, James},
  journal = {arXiv preprint arXiv:2601.03670},
  year = {2026},
  doi = {arXiv:2601.03670}
}

Accurate question answering (QA) in disaster management requires reasoning over uncertain and conflicting information, a setting poorly captured by existing benchmarks built on clean evidence. We introduce DisastQA, a large-scale benchmark of 3,000 rigorously verified questions (2,000 multiple-choice and 1,000 open-ended) spanning eight disaster types. The benchmark is constructed via a human-LLM collaboration pipeline with stratified sampling to ensure balanced coverage. Models are evaluated under varying evidence conditions, from closed-book to noisy evidence integration, enabling separation of internal knowledge from reasoning under imperfect information. For open-ended QA, we propose a human-verified keypoint-based evaluation protocol emphasizing factual completeness over verbosity. Experiments with 20 models reveal substantial divergences from general-purpose leaderboards such as MMLU-Pro. While recent open-weight models approach proprietary systems in clean settings, performance degrades sharply under realistic noise, exposing critical reliability gaps for disaster response. All code, data, and evaluation resources are available at this https URL.

@unpublished{li2025disaster,
  title = {Disaster Management in the Era of Agentic AI Systems: A Vision for Collective Human-Machine Intelligence for Augmented Resilience},
  author = {Li, Bo and Ma, Junwei and Yin, Kai and Xiao, Yiming and Hsu, Chia-Wei and Mostafavi, Ali},
  journal = {arXiv preprint arXiv:2510.16034},
  year = {2025},
  doi = {arXiv:2510.16034}
}

The escalating frequency and severity of disasters routinely overwhelm traditional response capabilities, exposing critical vulnerability in disaster management. Current practices are hindered by fragmented data streams, siloed technologies, resource constraints, and the erosion of institutional memory, which collectively impede timely and effective decision making. This study introduces Disaster Copilot, a vision for a multi-agent artificial intelligence system designed to overcome these systemic challenges by unifying specialized AI tools within a collaborative framework. The proposed architecture utilizes a central orchestrator to coordinate diverse sub-agents, each specializing in critical domains such as predictive risk analytics, situational awareness, and impact assessment. By integrating multi-modal data, the system delivers a holistic, real-time operational picture and serve as the essential AI backbone required to advance Disaster Digital Twins from passive models to active, intelligent environments. Furthermore, it ensures functionality in resource-limited environments through on-device orchestration and incorporates mechanisms to capture institutional knowledge, mitigating the impact of staff turnover. We detail the system architecture and propose a three-phased roadmap emphasizing the parallel growth of technology, organizational capacity, and human-AI teaming. Disaster Copilot offers a transformative vision, fostering collective human-machine intelligence to build more adaptive, data-driven and resilient communities.

@unpublished{li2025evac,
  title = {Evac-Cast: An Interpretable Machine-Learning Framework for Evacuation Forecasts Across Hurricanes and Wildfires},
  author = {Li, Bo and Liu, Chenyue and Mostafavi, Ali},
  journal = {arXiv preprint arXiv:2508.00650},
  year = {2025},
  doi = {arXiv:2508.00650}
}

Evacuation is critical for disaster safety, yet agencies lack timely, accurate, and transparent tools for evacuation prediction. This study introduces Evac-Cast, an interpretable machine learning framework that predicts tract-level evacuation rates using over 20 features derived from four dimensions: hazard intensity, community vulnerability, evacuation readiness, and built environment. Using an XGBoost model trained on multi-source, large-scale datasets for two hurricanes (Ian 2022, Milton 2024) and two wildfires (Kincade 2019, Palisades–Eaton 2025), Evac-Cast achieves mean absolute errors of 4.5% and 3.5% for hurricane and wildfire events, respectively. SHAP analysis reveals a consistent feature importance hierarchy across hazards, led by hazard intensity. Notably, the models perform well without explicit psychosocial variables, suggesting that macro-level proxies effectively encode behavioral signals traditionally captured through time-consuming surveys. This work offers a survey-free, high-resolution approach for predicting and understanding evacuation in hazard events, which could serve as a data-driven tool to support decision-making in emergency management.

@unpublished{ma2025quantifying,
  title = {Quantifying Functional Criticality of Lifelines Through Mobility-Derived Population-Facility Dependence for Human-Centered Resilience},
  author = {Ma, Junwei and Li, Bo and Li, Xiangpeng and Liu, Chenyue and Mostafavi, Ali},
  journal = {arXiv preprint arXiv:2512.16228},
  year = {2025},
  doi = {arXiv:2512.16228}
}

Lifeline infrastructure underpins the continuity of daily life, yet conventional criticality assessments remain largely asset-centric, inferring importance from physical capacity or network topology rather than actual behavioral reliance. This disconnect frequently obscures the true societal cost of disruption, particularly in underserved communities where residents lack service alternatives. This study bridges the gap between engineering risk analysis and human mobility analysis by introducing functional criticality, a human-centered metric that quantifies the behavioral indispensability of specific facilities based on large-scale visitation patterns. Leveraging 1.02 million anonymized mobility records for Harris County, Texas, we operationalize lifeline criticality as a function of visitation intensity, catchment breadth, and origin-specific substitutability. Results reveal that dependence is highly concentrated: a small subset of super-critical facilities (2.8% of grocery stores and 14.8% of hospitals) supports a disproportionate share of routine access. By coupling these behavioral scores with probabilistic flood hazard models for 2020 and 2060, we demonstrate a pronounced risk-multiplier effect. While physical flood depths increase only moderately under future climate scenarios, the population-weighted vulnerability of access networks surges by 67.6%. This sharp divergence establishes that future hazards will disproportionately intersect with the specific nodes communities rely on most. The proposed framework advances resilience assessment by embedding human behavior directly into the definition of infrastructure importance, providing a scalable foundation for equitable, adaptive, and human-centered resilience planning.

@unpublished{ma2025decoupling,
  title = {Decoupling Urban Food Accessibility Resilience during Disasters through Time-Series Analysis of Human Mobility and Power Outages},
  author = {Ma, Junwei and Li, Bo and Li, Xiangpeng and Mostafavi, Ali},
  journal = {arXiv preprint arXiv:2511.14706},
  year = {2025},
  doi = {arXiv:2511.14706}
}

Disaster-induced power outages create cascading disruptions across urban lifelines, yet the timed coupling between grid failure and essential service access remains poorly quantified. Focusing on Hurricane Beryl in Houston (2024), this study integrates approximately 173000 15-minute outage records with over 1.25 million visits to 3187 food facilities to quantify how infrastructure performance and human access co-evolve. We construct daily indices for outage characteristics (intensity, duration) and food access metrics (redundancy, frequency, proximity), estimate cross-system lags through lagged correlations over zero to seven days, and identify recovery patterns using DTW k-means clustering. Overlaying these clusters yields compound power-access typologies and enables facility-level criticality screening. The analysis reveals a consistent two-day lag: food access reaches its nadir on July 8 at landfall while outage severity peaks around July 10, with negative correlations strongest at a two-day lag and losing significance by day four. We identify four compound typologies from high/low outage crossed with high/low access disruption levels. Road network sparsity, more than income, determines the depth and persistence of access loss. Through this analysis, we enumerate 294 critical food facilities in the study area requiring targeted continuity measures including backup power, microgrids, and feeder prioritization. The novelty lies in measuring interdependency at daily operational resolution while bridging scales from communities to individual facilities, converting dynamic coupling patterns into actionable interventions for phase-sensitive restoration and equity-aware preparedness. The framework is transferable to other lifelines and hazards, offering a generalizable template for diagnosing and mitigating cascading effects on community access during disaster recovery.

@unpublished{li2025quantifying,
  title = {Quantifying the Social Costs of Power Outages and Restoration Disparities Across Four US Hurricanes},
  author = {Li, Xiangpeng and Ma, Junwei and Li, Bo and Mostafavi, Ali},
  journal = {arXiv preprint arXiv:2509.02653},
  year = {2025},
  doi = {arXiv:2509.02653}
}

The multifaceted nature of disaster impact shows that densely populated areas contribute more to aggregate burden, while sparsely populated but heavily affected regions suffer disproportionately at the individual level. This study introduces a framework for quantifying the societal impacts of power outages by translating customer weighted outage exposure into deprivation measures, integrating welfare metrics with three recovery indicators, average outage days per customer, restoration duration, and relative restoration rate, computed from sequential EAGLE I observations and linked to Zip Code Tabulation Area demographics. Applied to four United States hurricanes, Beryl 2024 Texas, Helene 2024 Florida, Milton 2024 Florida, and Ida 2021 Louisiana, this standardized pipeline provides the first cross event, fine scale evaluation of outage impacts and their drivers. Results demonstrate regressive patterns with greater burdens in lower income areas, mechanistic analysis shows deprivation increases with longer restoration durations and decreases with faster restoration rates, explainable modeling identifies restoration duration as the dominant driver, and clustering reveals distinct recovery typologies not captured by conventional reliability metrics. This framework delivers a transferable method for assessing outage impacts and equity, comparative cross event evidence linking restoration dynamics to social outcomes, and actionable spatial analyses that support equity informed restoration planning and resilience investment.

@unpublished{yin2024crisissense,
  title = {Crisissense-llm: Instruction fine-tuned large language model for multi-label social media text classification in disaster informatics},
  author = {Yin, Kai and Li, Bo and Liu, Chengkai and Mostafavi, Ali and Hu, Xia},
  journal = {arXiv preprint arXiv:2406.15477},
  year = {2024},
  doi = {arXiv:2406.15477}
}

In the field of crisis/disaster informatics, social media is increasingly being used for improving situational awareness to inform response and relief efforts. Efficient and accurate text classification tools have been a focal area of investigation in crisis informatics. However, current methods mostly rely on single-label text classification models, which fails to capture different insights embedded in dynamic and multifaceted disaster-related social media data. This study introduces a novel approach to disaster text classification by enhancing a pre-trained Large Language Model (LLM) through instruction fine-tuning targeted for multi-label classification of disaster-related tweets. Our methodology involves creating a comprehensive instruction dataset from disaster-related tweets, which is then used to fine-tune an open-source LLM, thereby embedding it with disaster-specific knowledge. This fine-tuned model can classify multiple aspects of disaster-related information simultaneously, such as the type of event, informativeness, and involvement of human aid, significantly improving the utility of social media data for situational awareness in disasters. The results demonstrate that this approach enhances the categorization of critical information from social media posts, thereby facilitating a more effective deployment for situational awareness during emergencies. This research paves the way for more advanced, adaptable, and robust disaster management tools, leveraging the capabilities of LLMs to improve real-time situational awareness and response strategies in disaster scenarios.