Jhayron S. Pérez-Carrasquilla

PhD Candidate in Atmospheric & Oceanic Science at UMD | Earth System Predictability, Variability & Machine Learning

I develop and apply machine-learning methods to better understand and predict Earth system variability, with interests spanning subseasonal predictability, large-scale circulation, tropical variability, and extreme events.

About

I’m a PhD candidate in Atmospheric and Oceanic Science at the University of Maryland, where I develop and apply machine-learning methods to better understand and predict Earth system variability. My research focuses on predictability across timescales—from subseasonal forecasting to longer-term climate variability and change—with interests in large-scale circulation, tropical variability, and extreme events, while emphasizing process-oriented understanding.

Across academic, public-sector, and applied research settings, I’ve worked on problems spanning subseasonal prediction, air quality, extreme events, renewable-energy forecasting, and climate risk. My work combines Earth system models, reanalyses, observations, and data-driven methods—including tree-based models, deep learning, explainable AI, and causal discovery—to study both sources of predictability and uncertainty.

I contribute to international efforts at the intersection of climate science and AI, including the WCRP Fresh Eyes on CMIP initiative and the AMS Committee on AI Applications to Environmental Science. I was also an ASP Graduate Visitor at US NSF NCAR, where I studied large-scale circulation changes and associated extremes. Broadly, I’m interested in using scientific insight and machine learning to tackle challenging questions of prediction and uncertainty in a complex and changing Earth system.

I earned a bachelor’s degree in engineering and a master’s degree in water resources from the Universidad Nacional de Colombia. Outside research, I enjoy sports, music, movies, and reading.

Research

Predictability across timescales

I study the sources and limits of atmospheric predictability, especially for North American weather regimes at subseasonal-to-seasonal lead times. This work uses machine learning, Earth system reanalyses, and coupled model output to diagnose when and why useful forecast information emerges.

Large-scale circulation and extremes

I analyze how large-scale circulation patterns vary across observations and models, how they shape extreme heat and other impacts, and how their behavior may change in a warming climate.

Causal and interpretable machine learning

I use explainable AI and causal discovery methods to connect predictive models with physical interpretation, including work on dynamical precursors of extreme events from gridded Earth system data.

Applied forecasting experience

Before and alongside my PhD, I worked on applied forecasting problems involving air quality, streamflow, meteorological variables, renewable-energy resources, and hurricane risk products for public and private-sector decision-making.

Subseasonal prediction Weather regimes Extreme events Machine learning Causal discovery Large ensembles Reanalysis Satellite and radar data

Publications

Selected experience

Graduate Research Assistant
University of Maryland, Department of Atmospheric and Oceanic Science | 2022–Present
  • Develop machine-learning methods to diagnose sources of subseasonal predictability in North American weather regimes.
  • Analyze weather regime variability and long-term changes using reanalysis data, climate model output, and large ensembles.
ASP Graduate Visitor
US NSF National Center for Atmospheric Research | Boulder, CO | 2024
  • Studied changes in North American weather regimes and associated surface extremes using Earth system model ensembles.
Research Scientist and Software Developer
Corporación Clima S.A.S. | Medellín, Colombia | 2020–2022
  • Developed machine-learning forecasts for meteorological, hydrological, energy-sector, and insurance-sector applications.
  • Built automated hurricane risk reports using National Weather Service model output.
Research Scientist
SIATA Early Warning System of the Aburrá Valley | Medellín, Colombia | 2018–2021
  • Developed machine-learning forecasts of air quality and analyzed drivers of convection and poor air quality.
  • Led atmospheric sampling campaigns and communicated scientific results with public-sector and community stakeholders.

Awards and honors

Connect

Contact: jspecar@gmail.com; jhayron@umd.edu