EMPIRIC

Map of 2 m air temperature over Puerto Rico at 2.5 km resolution, downscaled by EMPIRIC. The coast is warm and the central mountains are cool.

Typical global climate and weather forecast models have resolutions of 25 km or coarser, which fail to resolve the complex microclimates of islands and coastal regions. We work to build high-quality weather and climate data at the kilometer scale, focusing on site-specific hazards in concert with local stakeholders. Our focus is on understanding compound relationships between climate, weather, health, agriculture and infrastructure.

People

  • Photo of Prof. Christopher Horvat

    Prof. Christopher Horvat

    Assistant Professor, Brown University

    Sea ice floes, ice-albedo feedback, ecology under the ice, remote sensing, sub-mesoscale turbulence and its interaction with ice.

  • Photo of Dr. Paul Hall

    Dr. Paul Hall

    Senior Research Software Engineer, Brown University

    Mathematical modeling, scientific programming, geology, oceanography, data analysis.

  • Photo of Dr. Bingjie Zhao

    Dr. Bingjie Zhao

    Postdoctoral Researcher, Brown University

    Climate change, hydrology, machine learning, remote sensing.

  • Photo of Abrielle Mannino

    Abrielle Mannino

    Graduate Student, Brown University

    Tropical climate, ocean-atmosphere interactions, machine learning, co-modeling of extremes.

  • Konstantin Dichev

    Undergraduate Researcher, Brown University

    Machine learning.

  • Photo of Prof. Soupy

    Prof. Soupy

    Senior Researcher

    Rabbits, badgers, squirrels, running around in circles, nibbling.

Everyone in the group

Recent publications

All publications

Tools and code

  • EMPIRIC_DS

    Kilometre-scale downscaling. Generative transfer learning that maps coarse meteorological fields to kilometre scale, on timescales of hours to centuries. In development.

  • EMPIRIC_TC

    Tropical cyclone prediction. A Fourier Neural Operator emulator of the STORM tropical cyclone model. Produces site-specific cyclone hazard for the South Pacific in seconds.

  • OPT

    Optimal extreme statistics. The Optimal Path Threshold method determines the optimal characterization of extremes, with applications in human health, sea ice loss, evaporation, and agriculture.

  • EMPIRIC_SDS

    Spatial decision support. A web GIS platform showing site-specific hazard and uncertainty alongside health-facility data. Pre-alpha.

  • Risk outlooks for coastal regions

    Sub-seasonal forecasting of extremes. OPT-enabled forecasting of compound extreme climate events on sub-seasonal timescales: a short-term prediction and early warning system for extremes and their health impacts. Prototype.

  • TAC-to-IWXXM

    Future-proofing meteorological data. Converts aviation weather messages (METAR, SPECI, TAF, SIGMET) from traditional alphanumeric code to the WMO IWXXM standard, with validation. In production.

github.com/empiric2

Support

  • Climate Change AI, 2023 Innovation Grant: EMPIRIC_AI: AI-enabled Ensemble Projections of Cyclone Risk for Health Infrastructure in Pacific Island Countries and Territories
  • U.S. National Science Foundation, RISE 2528950: C2H2 EAGER: Site-specific Climate Risks to South Pacific Health and Health Infrastructure
  • CHAIRS-C, Brown University School of Public Health, Pilot Grant: Robust Associations between Multi-Type Compound Extreme Events and Older Adult Health
  • Brown University Office of Sustainability & Resiliency, 2025 Sustainability Seed Grant: Tracking and Forecasting Health Impacts of Compound Extreme Climate Impacts in the Providence Area