About

My research focuses on managing climate risk, particularly in how climate- and human-system uncertainties and human responses to environmental hazards influence the efficacy of climate risk management strategies. The overarching goal of this research is to help design robust approaches to reducing the impacts of climate change. I work on both mitigation (the reduction of greenhouse gas emissions) and adaptation, particularly focusing on flood risk and energy-system reliability. I am also deeply committed to open science and maximizing research transparency and reproducibility, which are essential for building credibility in climate risk projections and associated risk-management interventions.

Research topics

• Artificial Intelligence
• Computer Science
• Ecology
• Engineering
• Sociology
• Social Science
• Geography
• Management science
• Atmospheric sciences
• Mathematics
• Climatology
• Environmental science
• Econometrics
• Statistics
• Risk analysis (engineering)
• Chemistry
• Business
• Data science
• Engineering ethics
• Simulation
• Economics
• Knowledge management
• Meteorology

Selected publications

• Varying sources of uncertainty in risk-relevant hazard projections

  2026-03-14

  articleOpen access1st author

  A growing number of societal actors rely on high-resolution meteorological information to understand a changing landscape of physical hazards. Within this context, accounting for uncertainty is crucial to quantify and manage risks, but can be challenging given the potential for various sources of uncertainty to manifest differently across use-cases. Here, we combine three state-of-the-art downscaled ensembles to characterize how different uncertainties affect projections of several temperature- and precipitation-based risk metrics across the contiguous United States. We focus on long-term trends of aggregate indices as well as the intensity of rare events with 10- to 100-year return periods. By leveraging new downscaled initial condition ensembles, we characterize the role of internal variability at local scales and estimate its importance relative to other sources of uncertainty. Our results demonstrate systematic differences in patterns of uncertainty between average and extreme indices, across recurrence intervals, and between temperature- and precipitation-derived variables. We show that temperature metrics are more sensitive to the choice of radiative forcing scenario and Earth system model, while internal variability is often dominant for precipitation-based metrics. Additionally, we find that the statistical uncertainty from extreme value distribution fitting can often exceed the uncertainties related to Earth system modeling, particularly at recurrence intervals of 50 years or longer. Our results can provide guidance for researchers and practitioners conducting physical hazard risk assessment.

  PublisherDOI

• Output data: Varying sources of uncertainty in risk-relevant hazard projections across the United States

  Zenodo (CERN European Organization for Nuclear Research) · 2026-05-17

  datasetOpen accessSenior author

  Abstract Physical climate risk assessment requires understanding how different sources of uncertainty affect hazard projections. However, the relative importance of these uncertainties can differ across end-uses. Here, we combine three state-of-the-art downscaled climate model ensembles to characterize how different uncertainties affect projections of several temperature- and precipitation-based risk metrics across the contiguous United States. We focus on long-term trends of aggregate indices as well as the intensity of rare events with 10- to 100-year return periods. By including new downscaled initial condition ensembles, we characterize the role and relative importance of internal variability at local scales. Our results demonstrate systematic differences in patterns of uncertainty between average and extreme indices, across recurrence intervals, and between temperature- and precipitation-derived metrics. We show that temperature metrics are more sensitive to the choice of emissions scenario and Earth system model, while internal variability can be dominant for precipitation-based metrics. Additionally, we find that the statistical uncertainty from extreme value distribution fitting can often exceed climate-related factors, particularly at recurrence intervals of 50 years or longer. These results highlight the challenge of providing general guidance for climate impacts assessment and the need to consider a wide variety of potential uncertainties when quantifying climate risk. Journal reference Currently undergoing peer review: https://doi.org/10.22541/essoar.15003332/v1 Data description - `ensemble_summary` contains the ensemble means and upper/lower quantiles for each downscaling method and SSP combination; the trend metrics are grouped into a single netcdf file while the return level outputs are separated by variable. - `trends` contains the individual trend estimates for each member of the meta-ensemble; there are separate files for each variable and for each sub-ensemble (i.e., downscaling method). - `GEV` contains the GEV outputs, including the GEV parameter estimates and calculated associated return levels; there separate files for the GEV parameters and return levels, and for each variable and sub-ensemble (i.e., downscaling method). - `uncertainty_results` contains our main uncertainty decomposition results; the trend metrics are grouped into a single netcdf file while the return level outputs are separated by variable. Notes: - NetCDF files are compressed via zlib so may take longer than expected to open, especially the trend and GEV estimates that contain outputs for all individual ensemble members - To ensure maximal compatibility, strings coordinates (e.g. for SSPs or ESMs) are encoded as fixed-length byte strings (NetCDF4 char arrays); python users can use `ds[coord].astype(str)` to convert after loading. - All trend and GEV calculations were performed on the native grids of the downscaled outputs, then regridded to the LOCA2 grid using a nearest neighbors algorithm. All uncertainty results are valid over CONUS only. Contact Additional details can be found in the preprint (https://doi.org/10.22541/essoar.15003332/v1) or corresponding GitHub repository (https://github.com/david0811/conus_comparison_lafferty-etal-2026). Email: dcl257@cornell.edu

  PublisherDOI

• Neglecting Human Response Leads to Biased Distributional Flood Risk Outcomes

  2026-03-14

  articleOpen accessSenior author

  Flood-risk assessments increasingly consider how flood risk is distributed across populations. However, future flood risk is subject to a number of uncertainties related to flood hazard, exposure, vulnerability, and human response, which are often not fully considered in such assessments. These uncertainties can be amplified by the finer scales required for distributional analyses. To better understand which uncertainties are relevant for distributional impacts, we perform a large-scale uncertainty characterization experiment using a calibrated agent-based model over the course of a multi-decadal simulation. We find that failing to account for key uncertainties, particularly related to flood damage estimation and human response, results in major biases in future flood losses and recovery. Furthermore, the relative importance of these uncertain factors vary depending on the population of interest. For example, we find that behavioral risk factors towards flooding are the most influential in shaping high-income population recovery, but factors related to housing preference and affordability are the most influential in shaping low-income recovery. Our results highlight the need to systematically account for multiple sources of uncertainty to better understand the distribution of flood risks.

  PublisherDOI

• Unrefined national building inventories can mislead risk assessments and decisions

  2026-03-13

  articleOpen access

  Flood-risk assessments increasingly rely on large-scale building inventories that offer fine spatial detail but limited and uneven quality assurance. As a result, exposure is often treated as a static, “ready-to-use” input, even though small errors in where assets are located or how they are characterized can propagate into loss estimates. Despite the centrality of exposure for understanding changing risk under climate and socio-economic change, the implications of adopting exposure data without refinement remain poorly quantified. Here, we test how exposure data quality influences flood-loss estimates and decision-relevant metrics by comparing damages derived from a widely used national building inventory to estimates produced with high-quality, feature-rich local building data across an ensemble of flood scenarios. We find that adopting an unrefined building inventory can systematically distort decision-relevant damage metrics. For example, roughly one-fifth of areas are misclassified with respect to a funding priority status metric used in the U.S. Simple, transferable exposure refinements—particularly corrections to building locations—substantially reduce these errors, yielding near-complete agreement with rankings based on high-quality local data. Our findings demonstrate that credible assessments of flood risk require explicit attention to the spatio-temporal reliability of exposure inputs, not only improved hazard characterization or vulnerability functions. We provide actionable guidance for diagnosing exposure errors and implementing practical corrections.

  PublisherDOI

• New Level-3 datasets demonstrate that robust field-level groundwater use accounting is possible at continental scales: what is next for transboundary aquifer governance?

  2026-03-14

  articleOpen access

  According to the UN Valuing Water Initiative, accurate measurement of water resources is critical to valuation, decision making, and governance. Rotary drilling and submersible pump technologies, which proliferated in the 1950s-1970s, facilitated rapid and widespread development of global groundwater reserves. Currently, 25% of global water use is from groundwater, trending towards 100% in arid regions, and 75% of this groundwater is used for agriculture. Because it is difficult to measure, regulation of groundwater use is sparse and underenforced, and international treaties governing use of groundwater are virtually non-existent. Given the promise of spatial ubiquity in satellite observations, the hydrologic remote sensing research community has made tremendous progress in the measurement of hydrologic fluxes that are aliased by sparse in-situ networks. Two promising data derivatives—mainly energy-balance actual evapotranspiration derived from radiometric surface temperature and surface displacement associated with groundwater extraction from interferometric synthetic aperture radar—have fundamentally changed our understanding of how anthropogenic groundwater use in particular is modifying the hydrosphere, and enable estimates of relative groundwater extraction rates at the well/farm scale. Due to the high computational costs and technical complexity associated with processing these datasets, particularly at the spatial scales required for national to transboundary water accounting, their use in operational water management has been limited.Two operational datasets published in the United States this year allow for both large and small-scale accounting of agricultural groundwater use: the OpenET project DisALEXI dataset, and the OPERA project’s Sentinel 1 interferometric LOS displacement dataset. We demonstrate how independent error sources in these two datasets assist with uncertainty characterization, and benchmark their performance against GRACE observations of total water storage flux. We demonstrate how they allow us to estimate both regional (aquifer to nation-level) and local (field and well-level) groundwater use. We focus our analysis on 34 aquifers that span the United States/Mexico border, where a deepening water crisis is playing out in the absence of international agreements on transboundary aquifer use. We use this case study to demonstrate how investment in production of Level-3 datasets from entire satellite archives can enable international collaboration on natural resource development.

  PublisherDOI

• Bayesian hypergraph inference from scarce and noisy dynamical observations

  arXiv (Cornell University) · 2026-05-05

  preprintOpen access

  Inferring higher-order interaction structure from observations of dynamics is a central challenge in complex systems, particularly when data are scarce, noisy, or concentrated in lower-dimensional regions of state space. We develop Bayes-THIS, a Bayesian extension of Taylor-based Hypergraph Inference using SINDy (THIS), which reconstructs hypergraph structure from time-series data by identifying sparse Taylor coefficients associated with pairwise and higher-order interactions. By replacing fixed-threshold sparse regression with sparse Bayesian regression using automatic relevance determination, Bayes-THIS explicitly models residual variance and applies adaptive, term-wise coefficient shrinkage, improving robustness in data-limited, high-noise, and ill-conditioned regimes. The resulting Gaussian posterior also enables an uncertainty-aware inference workflow: a posterior predictive check assesses whether the data contain sufficient higher-order signal to reliably support inference beyond a pairwise model, and credible-interval pruning selects hyperedges whose inferred coefficients are statistically distinguishable from zero. Finally, we characterize a fundamental limitation of the Taylor-based inference framework: when higher-order interactions concentrate on nodes that lack lower-order connections, the Taylor expansion systematically inflates lower-order coefficient estimates, producing spurious edges indistinguishable from genuine lower-order interactions. This structural non-identifiability cannot be resolved by either THIS or Bayes-THIS.

  PublisherDOI

• Identification of pressure points in modern power systems using transfer entropy

  Cell Reports Sustainability · 2026-03-26

  articleOpen accessSenior author

  Transfer entropy reveals how grid stress propagates to power shortages Reliability failures arise from system-wide interactions, not isolated bottlenecks Stress pathways vary across weather and technology scenarios and are hard to predict

  PublisherOA PDFDOI

• Output data: Varying sources of uncertainty in risk-relevant hazard projections across the United States

  Zenodo (CERN European Organization for Nuclear Research) · 2026-05-17

  datasetOpen accessSenior author

  Abstract Physical climate risk assessment requires understanding how different sources of uncertainty affect hazard projections. However, the relative importance of these uncertainties can differ across end-uses. Here, we combine three state-of-the-art downscaled climate model ensembles to characterize how different uncertainties affect projections of several temperature- and precipitation-based risk metrics across the contiguous United States. We focus on long-term trends of aggregate indices as well as the intensity of rare events with 10- to 100-year return periods. By including new downscaled initial condition ensembles, we characterize the role and relative importance of internal variability at local scales. Our results demonstrate systematic differences in patterns of uncertainty between average and extreme indices, across recurrence intervals, and between temperature- and precipitation-derived metrics. We show that temperature metrics are more sensitive to the choice of emissions scenario and Earth system model, while internal variability can be dominant for precipitation-based metrics. Additionally, we find that the statistical uncertainty from extreme value distribution fitting can often exceed climate-related factors, particularly at recurrence intervals of 50 years or longer. These results highlight the challenge of providing general guidance for climate impacts assessment and the need to consider a wide variety of potential uncertainties when quantifying climate risk. Journal reference Currently undergoing peer review: https://doi.org/10.22541/essoar.15003332/v1 Data description - `ensemble_summary` contains the ensemble means and upper/lower quantiles for each downscaling method and SSP combination; the trend metrics are grouped into a single netcdf file while the return level outputs are separated by variable. - `trends` contains the individual trend estimates for each member of the meta-ensemble; there are separate files for each variable and for each sub-ensemble (i.e., downscaling method). - `GEV` contains the GEV outputs, including the GEV parameter estimates and calculated associated return levels; there separate files for the GEV parameters and return levels, and for each variable and sub-ensemble (i.e., downscaling method). - `uncertainty_results` contains our main uncertainty decomposition results; the trend metrics are grouped into a single netcdf file while the return level outputs are separated by variable. Notes: - NetCDF files are compressed via zlib so may take longer than expected to open, especially the trend and GEV estimates that contain outputs for all individual ensemble members - To ensure maximal compatibility, strings coordinates (e.g. for SSPs or ESMs) are encoded as fixed-length byte strings (NetCDF4 char arrays); python users can use `ds[coord].astype(str)` to convert after loading. - All trend and GEV calculations were performed on the native grids of the downscaled outputs, then regridded to the LOCA2 grid using a nearest neighbors algorithm. All uncertainty results are valid over CONUS only. Contact Additional details can be found in the preprint (https://doi.org/10.22541/essoar.15003332/v1) or corresponding GitHub repository (https://github.com/david0811/conus_comparison_lafferty-etal-2026). Email: dcl257@cornell.edu

  PublisherDOI

• Unlocking the benefits of transparent and reusable science for climate risk management

  Proceedings of the National Academy of Sciences · 2026-01-14

  articleOpen access

  People around the world seek climate risk information to guide their decisions. For instance, projections about future flood risk inform where households choose to live, how lenders manage credit risks, and which communities receive federal funding. Yet data limitations and fundamental validation challenges raise important concerns about the reliability of such projections. The principles of transparency and reusability help address these concerns by enabling scrutiny of assumptions and methods, development of foundational data and tools, and consistent application of evaluation standards. While there is ongoing debate about how much transparency commercial climate risk services should provide, many expect noncommercial actors to lead the way on operationalizing transparency and reusability to fulfill their knowledge-building role in the climate risk ecosystem. However, despite prominent success stories, we find a substantial gap between principles and practice: Only four percent of the most-cited peer-reviewed climate risk studies in recent years fully share their data and code although this is a widely accepted minimum standard for transparency. We highlight low-cost measures that noncommercial researchers can take now to improve transparency and reusability. We also emphasize that transformative progress requires substantial investment, cross-sector collaboration, and careful consideration of tradeoffs, data rights, and multiple perspectives on equity. We hope this perspective accelerates both immediate actions and longer-term conversations to improve the ability of science to effectively support timely, evidence-based, and sound climate risk management.

  PublisherDOI

• Human Behavior and Hazard Uncertainties Shape Distributional Flood Risk Outcomes

  SSRN Electronic Journal · 2026-01-01

  preprintOpen accessSenior author
  PublisherDOI

Frequent coauthors

• Klaus Keller

  Dartmouth College

  65 shared

• Jennifer Morris

  Moscow Institute of Thermal Technology

  17 shared

• Tony E. Wong

  Rochester Institute of Technology

  16 shared

• Jonathan Lamontagne

  Tufts University

  16 shared

• Wei Peng

  National Marine Environmental Forecasting Center

  14 shared

• Jim Yoon

  Pacific Northwest National Laboratory

  13 shared

• Heng Wan

  10 shared

• Julianne D. Quinn

  University of Virginia

  9 shared

Labs

• Srikrishnan GroupPI

  1-2 sentence research focus

Education

• Ph.D., Energy & Mineral Engineering

  Penn State

  2018