Our lab aim to address the critical challenges posed by climate extremes in a warming world. We focus on understanding and mitigating climate risks, with particular emphasis on the impacts of tropical cyclones. By integrating advanced physics-based models and cutting-edge data-mining techniques, we seek to enhance resilience and develop more effective adaptation strategies.

Tropical Cyclones in a Warming Climate

Tropical cyclones, or hurricanes, present growing threats to human life, infrastructure, and economic stability, with U.S. damages averaging over $28 billion annually. In a warming climate, the frequency, intensity, and complexity of these storms are increasing, driven by changes in storm patterns and sea-level rise. Our research employs high-resolution, physics-based simulations to model individual and compound hazards—such as storm surge, inland rainfall flooding, and their dangerous interactions—in coastal urban areas. These simulations, grounded in synthetic hurricane scenarios, enable a detailed evaluation of how climate change alters the spatial extent and severity of hurricane-induced flooding under both present and future conditions.

To better understand and quantify these evolving risks, we integrate data-driven machine learning approaches—including Conditional Random Fields and deep learning—with traditional hydrodynamic models. This hybrid framework allows us to assess the economic impacts of hurricane hazards on infrastructure and residential areas, projecting that compound flooding events like Hurricane Sandy could occur five times more frequently by the end of the century. Our findings emphasize the need for adaptive strategies that incorporate both engineered and nature-based defenses. These insights support urban planners and policymakers in designing resilient cities capable of withstanding the escalating impacts of tropical cyclones in a rapidly changing climate.

Compound Extremes and Multidimensional Risk

Climate change, driven by human activities, is intensifying the frequency, severity, and spatial reach of extreme weather and climate events. Particularly concerning are compound extremes—events where multiple climate stressors, such as heatwaves and droughts, occur simultaneously or sequentially. These events have disproportionately large impacts on human systems, ecosystems, and infrastructure. Our research develops advanced Bayesian, time-evolving frameworks to quantify and analyze the multidimensional nature of these risks in a nonstationary, warming climate.

These frameworks allow us to examine how climate change alters the joint probability and co-occurrence of extreme conditions across time and space. For example, we find that the probability of simultaneous warm and dry years has doubled globally since the 1961–1990 baseline, with especially sharp increases in critical agricultural regions. In the U.S., the risk of compound droughts—where meteorological and hydrological droughts coincide—has risen by up to 20%, particularly in the western states. These compound events often display strong temporal memory, meaning past events increase the likelihood of future ones.

By integrating statistical models such as the bivariate GARCH model with climate data, and by focusing on causal attribution and co-dependence across extremes, our methodology informs water resource management, agriculture, and energy planning. Importantly, we show that ambitious climate mitigation policies, like those under the Paris Agreement, can significantly reduce the likelihood of compound events occurring across multiple regions. Our multidimensional risk modeling thus provides a critical tool for building resilience to cascading climate hazards in a rapidly changing world.

Advanced Machine Learning and Data Mining

Anthropogenic climate change is driving more severe and frequent climate and oceanic extremes, increasing the urgency for improved predictive tools. To address this challenge, our research focuses on developing advanced data-mining and machine learning techniques that enhance the predictability of hydro-climate extremes.

Specifically, we pioneer methods such as supervised dimensionality reduction integrated with machine learning to improve statistical downscaling and forecasting in a warming climate. These innovations aim to strengthen early warning systems and support more informed decision-making in the face of escalating climate risks.