Jhayron S. Pérez-Carrasquilla

I’ll discuss our recent papers:

An Earth-System-Oriented View of the S2S Predictability of Weather Regimes (AIES)
doi.org/10.1175/AIES...

Observed and Modeled Amplification of the Frequency, Duration, and Extreme Heat Impacts of the Pacific Trough Regime (Earth's Future)
doi.org/10.1029/2025...

And to my talk on S2S predictability and long-term changes of the Pacific Trough weather regime:

- 13B.6 Predictability and Long-Term Changes of the Pacific Trough Regime and Associated Extreme Heat Impacts
(Thursday, 9:45–10:00 am)
ams.confex.com/ams/106ANNUA...

Excited to see you there!

#AMS2026 If you’re in Houston this week, I’d like to invite you to attend a session I’m co-chairing with my advisor, Maria Molina:

- 9B Artificial Intelligence for Actionable Insights and Applications in Climate Science
(Wednesday, 8:30–10:00 am)
ams.confex.com/ams/106ANNUA...

Screenshot of part of our paper "Increasing Frequency and Persistence of the Summertime Greenland High Regime Not Captured by a Seasonal Prediction Model Very Large Ensemble" in GRL

📣 New paper with Lorenzo Polvani published in @agu.org GRL!

Increasing Frequency and Persistence of the Summertime Greenland High Regime Not Captured by a Seasonal Prediction Model Very Large Ensemble

Open access: doi.org/10.1029/2025...

@earthscista.bsky.social @lamont.columbia.edu

Composite-mean normalised 500 hPa geopotential height anomalies for the four North American weather regimes: Pacific Trough, Pacific Ridge, Greenland High and Alaskan Ridge. The remaining 14% of days are classified as No Regime.

Updated time series of daily year-round North American weather regimes for 1 Jan 1979 to 31 Dec 2025 just published on Zenodo: doi.org/10.5281/zeno...

I plan to keep this dataset updated annually. Happy regime-ing!

Controllable Probabilistic Forecasting with Stochastic Decomposition Layers AI weather prediction ensembles with latent noise injection and optimized with the continuous ranked probability score (CRPS) have produced both accurate and well-calibrated predictions with far less ...

Even after the news last week, we are still doing great research here at NCAR. My group has released a new paper, led by John Schreck, on arxiv about an efficient method for generating calibrated AI weather ensembles with reproducible scale specific perturbations. arxiv.org/abs/2512.18815

Some festive vortices to spin you into the new year!

visualpde.com/fluids/vorti...

Physical drivers of the November 2023 heatwave in Rio de Janeiro Abstract. Because extreme heat has not historically been a major hazard for the city of Rio de Janeiro, the November 2023 Heatwave magnitude and timing were staggering. Here we conduct a case study of...

Our paper exploring the drivers of the November 2023 “Taylor Swift” heatwave in Rio de Janeiro is officially out in WCD! Take a read to learn more about the causes of this record-breaking heatwave and Rio’s emerging risk for extreme heat:

wcd.copernicus.org/articles/6/1...

I came to the US over 10 years ago for a postdoc at NCAR. It changed my life and supercharged my ability to do science. NCAR's unique community nature makes it a global node in atmospheric and climate research. It's really not possible to overstate its value for science and society.

Hudson Bay is finally starting to freeze over in earnest.

2025 will likely go down as the 2nd-longest ice-free period for the Western Hudson Bay polar bear population, continuing the long-term trend (now >1 month longer than in the 1980s).

The goal of the course is to discuss key strategies and pitfalls in scientific ML applications. We will cover topics like data preprocessing, imbalanced datasets, choosing the appropriate ML architecture, uncertainty quantification, explainability and interpretability, and recent generative AI.

Intermediate Machine Learning in Python for Environmental Science Problems 2026 The AMS is a global community committed to advancing weather, water, and climate science and service.

Happy to share this short course at #AMS2026 that I'll be helping with:

www.ametsoc.org/ams/educatio...

Excited to see you there!!

The WCRP Earth System Modelling and Observations (ESMO) Working Group on Subseasonal to Interdecadal Predictions (WGSIP) is launching a new webinar series in late 2025.

29 October 2025, 22:00 UTC — Theme: Predictability limits

More info: https://loom.ly/F-Vfj6o

#WCRP #ESMO #WGSIP

Diffenbaugh, N. S. (2025) Committed acceleration of climate stresses in the coming decades. Environmental Research: Climate, in press. doi.org/10.1088/2752...

Scientists Must Join Forces to Solve Forecasting’s Predictability Desert - Eos To strengthen societal resilience to worsening natural hazards, siloed Earth system science communities must collaborate to understand conditions that favor skillful subseasonal-to-seasonal forecasts.

Scientists Must Join Forces to Solve Forecasting’s Predictability Desert eos.org/opinions/sci...

Learning predictable and informative dynamical drivers of extreme precipitation using variational autoencoders Abstract. Large-scale atmospheric dynamics modulate the occurrence of extreme precipitation events and provide sources of predictability of these events on timescales ranging from days to decades. In ...

Excited to share that our new paper is out in Weather and Climate Dynamics and was selected as a highlight publication🎉

We introduce a machine-learning approach to identify physically interpretable drivers of regional precipitation extremes. Have a look here: wcd.copernicus.org/articles/6/9...

Its worth noting that when the first modern climate models were published in 1970 it was hardly clear that there was a warming trend; if anything there had been flat or slightly cooling global temperatures for the past three decades:

In a UN speech today, President Trump said that "all of these [climate] predictions were wrong".

Back in 2019 I led a research effort to digitize old climate model projections and assess how well they did. Turns out they got future warming pretty spot on!

🚨 Recruiting Two PhD/MS Students 🚨

I am looking to bring on at least two GRAs (M.S. or Ph.D. Level) beginning Spring or Fall 2026 to join our CHAOS research group. Research projects will be related to artificial intelligence and machine learning applications for extreme temperatures and rainfall

Tropical deforestation is associated with considerable heat-related mortality - Nature Climate Change The authors assess the impacts of tropical deforestation and its subsequent local warming on human heat-related mortality. They estimate that deforestation-related warming (+0.27 °C) is associated wit...

🌳 Deforestation and deadly heat

A new study links tropical deforestation to ~28,000 heat-related deaths each year.

Local warming from forest loss now accounts for over a third of climate heat mortality in affected regions.

🔗 www.nature.com/articles/s41...

#SciComm #ClimateCrisis 🧪

Thanks, Prasad!!

Thanks!!

Our paper aims to provide an ML-based view of mid-latitude S2S predictability that could help open new scientific pathways and increase our ability to handle weather-related risks.

Check it out here! doi.org/10.1175/AIES...

Leveraging ML, we show how predictability sources vary across seasons and regimes. Data-driven models confirmed previously found sources of predictability (like ENSO and MJO) but also highlighted new opportunities for improved predictions, like stratosphere-troposphere and land surface interactions.

🚨I’m excited to share that the final version of my first PhD paper is already published! With my amazing advisor, Maria Molina, we characterized the predictability of North American weather regimes.

Katie Dagon @ncar-ucar.bsky.social presents our next Lecture in #Climate #Data Science!

📅 THURS, 7/24/25
🕧 12:30p ET * please note time-shift this week *
📍 @columbiaseas.bsky.social Innovation Hub/Zoom
💻 RSVP: www.eventbrite.com/e/summer-202...

@climate.columbia.edu #LEAPEducation #CESM #AI

Please consider submitting an abstract to the Weather Regime session that I am co-chairing at the AMS 106th Annual Meeting. 👍🏼

New Article: "When and where soil dryness matters to ecosystem photosynthesis" rdcu.be/evfGV

A causality-guided explainable AI framework shows soil moisture dominates vapour pressure deficit in shaping global photosynthesis during water-limited conditions.

Space‐Time Causal Discovery in Earth System Science: A Local Stencil Learning Approach We introduce Causal Space-Time Stencil Learning (CaStLe) for learning local causal dynamical structure underlying space-time data CaStLe enables previously infeasible analyses of grid-cell-level ...

Published in JGR:MLC! We introduce CaStLe (Causal Space-Time Stencil Learning), a method for grid-level space-time causal discovery that scales efficiently in high-dimensional Earth system data. It enables causal analysis of grid-level processes like eruption plumes.

#EarthScience #CausalDiscovery

👏👏👏👏