Department of Statistics
University of Connecticut
Philip E. Austin Building, Room 330
University of Connecticut
215 Glenbrook Road, U-4120
Storrs, CT 06269-4120
Phone: 860-486-2951
Email: marylai.salvana@uconn.edu
Website: UConn Faculty Profile
"Having lived through the devastations of Supertyphoon Sendong (Washi), I witnessed firsthand how a single event can unleash cascading disasters. Power grids failed, transportation systems collapsed, hospitals became overwhelmed, and communication networks went dark—each failure amplifying the next in a destructive chain reaction. What I learned is that these aren't random accidents but predictable outcomes when small shocks or extreme events hit an already fragile system. This experience drives my work advancing self-organized criticality models for next-generation early warning systems, shifting from crisis reaction to crisis prevention and building resilience before catastrophe strikes. My mission is simple:"
"A Self-Organized Criticality Model of Extreme Events and Cascading Disasters"
Presenting a novel SOC-based framework for modeling cascading failure dynamics in critical infrastructure networks, capturing power-law distributions of disruptions and long-tail risks of massive failures through analysis of the U.S. airline transportation system.
Key Finding: SOC models successfully reproduce empirical scaling laws in infrastructure networks, revealing how over-optimization for efficiency creates vulnerability to cascading disasters from climate shocks.
"Self-Organized Criticality: A Framework for Understanding Cascading Failures and Extreme Risks in Critical Infrastructure Networks"
Presenting novel applications of Self-Organized Criticality theory to critical infrastructure analysis, offering the first systematic framework for understanding how minor disruptions can trigger catastrophic cascading failures across entire networks.
Key Innovation: SOC-based early warning systems that detect approaching criticality through real-time monitoring of avalanche size distributions, providing mathematical tools for balancing operational efficiency with catastrophic resilience.
I will be one of the instructors for the Large Scale Data Science short course during the 2025 Joint Statistical Meetings (JSM 2025).
Learn More
Hi! I’m Mary Lai, an Assistant Professor in Statistics at University of Connecticut (UConn). Prior to joining UConn, I was a Postdoctoral Fellow at the Department of Mathematics at University of Houston. I received my Ph.D. at the King Abdullah University of Science and Technology (KAUST), Saudi Arabia.
My research interests include extreme events, risks, disasters, space-time statistics, high-dimensional and multivariate statistics, high performance computing, big data, machine learning, deep learning, artificial intelligence, environmental data science.
My mission is to understand the dynamics of the climate system using statistics and to communicate its laws through statistical models. My research has provided me the opportunity to analyze climate processes, to build models that describe how the components in the climate system interact, and to develop the high-performance computing skills to run big data experiments required for climate research. Through the years, my appreciation to how much story climate data can tell us about our planet has grown.
My research revolves around physics-motivated statistical modeling of the Earth to contribute to the scientific communities’ efforts to understand and tackle climate change. We are still facing significant gaps in understanding and predicting Earth’s climate and weather. Climate change is now outpacing climate models. Nowadays, the weather that we have prepared for is not the weather that we are getting. The systems and infrastructures built for managing disastrous events such as fires, floods, and droughts, were based on climate models which are rendered outdated by the new normal of extreme climate events. Climate models failed to predict the deadly heat wave in the cool and rainy Pacific Northwest region last June 2021, the megadrought in the Southwest US, the changing ecosystem of the Boreal forest (the world’s largest carbon sink), the rapid warming in the Arctic ocean, and the shifting jet stream behavior which caused simultaneously the deadly flooding in Germany and Belgium, the heatwave in Canada, and the Black Sea flooding in the summer of 2021. Unless we update these models, we are left vulnerable to these “unforeseen” climate events.
As an Assistant Professor in Statistics at the UConn, I lead the research on Extreme Events, Risks, and Disasters that develops state-of-the-art models that help advance climate science and shape our awareness of future disaster chains. My expertise are on three key areas which I believe are the most pressing issues of our time: Extreme Events, Risks, and Disasters Modeling.
By thinking of the Earth as a complex system and by identifying the vulnerabilities of its components due to extreme events, we can build resilience and minimize losses of any kind. By combining space-time statistics, extreme statistics, Bayesian statistics, machine learning techniques, physics models, and other methods for massive datasets with complex dependence structures, powered by high-performance computing technologies, we can develop new models that can more faithfully render topographic, geologic, atmospheric, and biological details over small regions.
My research outputs include a high-resolution and high-dimensional map of space-time correlations of different climate variables anywhere on Earth and forecasts of climate events such as floods, droughts, and fires, from days to years in advance. Furthermore, my research in these three areas will address the concerns of the following institutions, among many others.
extreme and catastrophic events, risks, disasters, crash protection, antifragility, spatial and spatio-temporal statistics, high-dimensional and multivariate statistics, high performance computing, computational statistics, environmental data science, big data, machine learning, deep learning, artificial intelligence
PHD IN STATISTICS, King Abdullah University of Science and Technology (KAUST), Jeddah, Saudi Arabia (Jan 2017 – Jul 2021)
Thesis: Lagrangian Spatio-Temporal Covariance Functions for Multivariate Nonstationary Random Fields, Advisor: Marc G. GentonMS IN APPLIED MATHEMATICS, Ateneo de Manila University, Manila, Philippines (Aug 2015 – Jul 2016)
BS IN APPLIED MATHEMATICS, Ateneo de Manila University, Manila, Philippines (Jun 2011 – Mar 2015)
Assistant Professor in Statistics, Department of Statistics, University of Connecticut (December 2023 - present)
Postdoctoral Researcher, Department of Mathematics, University of Houston (August 2021 – July 2023)
Al-Kindi Statistics Research Student Award, King Abdullah University of Science and Technology (KAUST), Jeddah, Saudi Arabia (2021)
American Statistical Association Member
New England Statistical Society Member
Ateneo Innovation Center Research Fellow
Chair & Organizer, The 36th New England Statistics Symposium, Boston University, MA, USA (June 3-6, 2023) — Opportunities and Challenges in the Use of Statistics in Disaster Science
A Self-Organized Criticality Model of Extreme Events and Cascading Disasters
University of Missouri, USA
Learn More2025 Joint Statistical Meetings (JSM 2025)
Nashville, TN, USA
Learn MoreMulti-Hazard Bayesian Hierarchical Model for Damage Prediction
2025 Joint Statistical Meetings (JSM 2025)
Nashville, TN, USA
Learn MoreA Spatial Bayesian Causal Network for Cascading Disasters Modeling
8th International Conference on Econometrics and Statistics (EcoSta 2025)
Waseda University, Tokyo, Japan
Learn More65th ISI World Statistics Congress 2025 (ISI 2025)
The Hague, the Netherlands
Learn More2024 Joint Statistical Meetings (JSM 2024)
Portland, OR, USA
Learn More37th New England Statistics Symposium (NESS 2024)
University of Connecticut, Storrs, CT, USA
Learn MoreAirports. Power grids. Communication lines. One chain reaction brings them down. One model sees it coming.
Powered by Self-Organized Criticality
Use Case: Assess national infrastructure vulnerability to climate shocks and develop resilience strategies.
Use Case: Predict cascading airport closures and optimize flight routing during extreme weather events.
Use Case: Identify critical transmission nodes and prevent blackout cascades during peak demand or storms.
Use Case: Map supply chain vulnerabilities and develop backup routing strategies for critical goods.
Use Case: Quantify cascading failure risks for portfolio management and catastrophe bond pricing.
Use Case: Model cascading outbreaks across public health nodes and evaluate system-wide response fragility.
Use Case: Prepare coordinated response plans for multi-system failures during natural disasters.
Use Case: Evaluate how economic fragility spreads through interconnected systems and design upstream interventions to prevent cascade failures.
Use Case: Monitor how stress propagates across banking, payments, and capital markets; anticipate contagion; and coordinate systemic safeguards before cascade failures emerge.
Use Case: Pinpoint weak links, simulate cascade triggers, and coordinate preemptive actions across power, transport, and comms sectors—before critical systems go down.
Choose from aviation, power grid, or transportation networks
Adjust stress thresholds and failure propagation settings
Execute SOC analysis and monitor real-time progress
Review critical nodes, dangerous sequences, and risk assessments
Pinpoint infrastructure components most likely to trigger system-wide failures
Map how failures propagate through networks with conditional probabilities
Track system criticality levels and early warning indicators
Leverage self-organized criticality theory for accurate failure forecasting
Need integration with your existing monitoring systems? Custom network configurations? Real-time API access? Contact our team for enterprise-grade early warning solutions tailored to your infrastructure.
Contact for Enterprise Solutions
