Posts |

Congratulations Dr. Ziqi Yin! Successful PhD Defense on Greenland Ice Sheet Surface Melt

Sat, 16 May 2026 00:00:00 +0000

We are incredibly proud to announce that Ziqi Yin has successfully defended his PhD dissertation today in the Department of Atmospheric and Oceanic Sciences (ATOC) at the University of Colorado Boulder.

Dissertation title: Advancing Understanding of Greenland Ice Sheet Surface Melt Using Physics-Based and Machine Learning Models

Defense committee: Dr. Aneesh Subramanian (advisor, CU Boulder), Dr. Alexandra Jahn (CU Boulder), Dr. Rajashree Tri Datta (Delft University of Technology), Dr. Adam Herrington (National Center for Atmospheric Research), and Dr. Jianwu Wang (University of Maryland, Baltimore County).

Research Overview

Ziqi’s doctoral work marks a major step forward in polar climate research, bridging advanced physically-coupled Earth system modeling with cutting-edge data science and deep learning. His research focuses on improving our understanding of Greenland Ice Sheet (GrIS) mass loss — specifically surface melt, which has emerged as the dominant contributor to Greenland’s ice sheet degradation and subsequent global sea level rise in recent decades.

His dissertation consists of three core research thrusts:

1. Topographic Resolution Matters in Long-Term Projections (Published in the Journal of Advances in Modeling Earth Systems, JAMES)

Ziqi investigated how atmospheric grid resolution impacts century-scale ice sheet simulations. Using a variable-resolution (VR) grid featuring a ¼° regional refinement over the Arctic within the fully coupled CESM2.2–CISM2.1 framework, he demonstrated that conventional 1° coarse models flatten Greenland’s steep coastal topography. This flattening artifact artificially accelerates the positive melt–albedo feedback. By accurately resolving the terrain, his refined grid projected a multi-century sea level rise contribution (831 mm by year 350) that is 20–40% smaller than traditional models, indicating that coarse-resolution models may be overestimating future sea level rise.

2. Disentangling Melt Drivers via Causal Discovery (Under review at Geophysical Research Letters)

Moving beyond traditional correlation metrics, Ziqi applied the PCMCI+ causal discovery algorithm to isolate direct physical causes of summer melt anomalies in the ablation zone. His work successfully identified net shortwave radiation (the melt–albedo feedback) and turbulent heat fluxes (sensible and latent heat) as the dominant contemporaneous drivers of monthly summer melt anomalies. Under late-century high-warming scenarios (SSP3-7.0), he found that turbulent heat links become undirected, indicating a transition toward a more tightly synchronous and intensely coupled surface–atmosphere regime.

3. Accelerating Climate Projections with Graph Transformers (In preparation for The Cryosphere)

Ziqi developed a novel machine learning spatial emulator using a hybrid Graph Transformer architecture that combines local message passing with global self-attention. Trained on the 100-member CESM2 Large Ensemble (LENS2), the emulator faithfully replicates complex annual spatial melt fields under various climate conditions with an R² score above 0.99 and a root-mean-square error below 10%. When deployed across available CMIP6 models, it projects a surface melt increase of 89% (under low-emission SSP1-2.6) to 267% (under high-emission SSP5-8.5) by end of century. Under high-emission scenarios, Greenland’s northern basins are projected to experience the strongest acceleration in melting and eventually become the largest regional contributors to mass loss.

Impact

By combining the predictive power of variable-resolution climate physics, causal graphs, and spatial machine learning emulators, Ziqi’s work provides the scientific community with highly efficient, robust tools to track ice sheet–atmosphere interactions. These frameworks open new pathways for fast uncertainty quantification, risk assessments, and the identification of potential climate tipping points.

Please join us in extending our warmest congratulations to Dr. Ziqi Yin on this stellar milestone and wishing him all the best in his future scientific career!

Dr. Timothy Higgins Launches Operational AI-Based Ensemble Forecasting System for West Coast Atmospheric Rivers

Sat, 16 May 2026 00:00:00 +0000

Accurate probabilistic forecasting of extreme weather events is critical for water management and disaster preparedness along the U.S. West Coast. We are excited to share that postdoctoral research associate and climate fellow Dr. Timothy B. Higgins has successfully developed and operationalized a cutting-edge, diffusion-based ensemble forecasting system for Integrated Vapor Transport (IVT).

The model — the culmination of the final chapter of Dr. Higgins’ dissertation — is now officially live and operational at the Center for Western Weather and Water Extremes (CW3E).

Bridging the Gap in Probabilistic Forecasting

Traditional ensemble forecasting systems are computationally expensive, often limiting the number of scenarios meteorologists can simulate in real time. Dr. Higgins’ diffusion model fundamentally shifts this paradigm, demonstrating an ability to generate realistic 1,000-member ensembles within a highly manageable timeframe.

The model exhibits exceptional probabilistic forecasting skill, comparing favorably against premier global standards including the European Centre for Medium-Range Weather Forecasts Integrated Forecasting System (ECMWF IFS) and the Global Ensemble Forecast System (GEFS).

A Collaborative Effort

This work began during Dr. Higgins’ time in the Advanced Study Program (ASP) Graduate Visitor Program at the National Center for Atmospheric Research (NCAR), where he worked closely with William Chapman. The project was co-authored alongside Dr. Aneesh Subramanian (University of Colorado Boulder) and Dr. Luca Delle Monache (CW3E / Scripps Institution of Oceanography).

View the Live Model

While the formal manuscript is currently under peer review, the real-time model outputs are already publicly accessible:

Congratulations to Tim and the entire team on moving this impactful AI-driven climate research from theory into active operations!

Applications Open: FERS Summer School on AI and Machine Learning for Earth System Modeling and Prediction

Wed, 22 Apr 2026 00:00:00 +0000

Applications are now open for the FERS Summer School on AI and Machine Learning for Earth System Modeling and Prediction, co-directed by Prof. Aneesh Subramanian and Prof. Will Chapman (University of Colorado Boulder).

This intensive summer school, organized by the CMCC Foundation’s Future Earth Research School (FERS), brings together early-career researchers to explore the latest advances in machine learning and artificial intelligence for climate science and Earth system modeling.

Dates: 8 – 19 June 2026

Location: Bertinoro (FC), Italy

Application deadline: May 3, 2026

For full details on the curriculum, dates, location, eligibility, and how to apply, visit the course page:

Luke Howard Defends His PhD Thesis

Mon, 06 Apr 2026 00:00:00 +0000

Congratulations to Lucas (Luke) Howard on defending his PhD dissertation today!

Title: Advancing Earth System Data Assimilation and Prediction with Probabilistic Machine Learning

Date: Monday, April 6, 2026, 10:00 AM MT

Location: SEEC N128

Abstract

Probabilistic machine learning methods offer significant potential for advancing earth system prediction, but existing applications have largely been deterministic, limiting their utility in contexts where uncertainty quantification is essential. This dissertation advances probabilistic machine learning for earth system prediction across three distinct but complementary applications, each targeting a persistent bottleneck in operational forecasting or data assimilation.

In the first application, a probabilistic U-Net-based neural network is developed for subseasonal marine heatwave forecasting in the northern Indian Ocean and Arabian Sea. The model produces skillful probabilistic forecasts at lead times of up to 10 weeks, outperforming persistence and climatology benchmarks across a range of deterministic and probabilistic skill metrics and performing comparably to or better than the ECMWF S2S dynamical forecast. The results suggest that planetary waves and low-frequency ocean dynamics provide windows of predictability that a probabilistic machine learning approach can exploit.

In the second application, a convolutional neural network is used to augment an ensemble Kalman filter for the assimilation of high-resolution observations that would otherwise be discarded due to computational constraints. Demonstrated as a proof-of-concept on the Lorenz-96 system, the augmented method reduces analysis error by 37% compared to an ensemble Kalman filter operating on spatially thinned observations alone, and produces more accurate and reliable ensemble forecasts at lead times of up to 10 days.

In the third application, a probabilistic neural network emulator of the Community Radiative Transfer Model is developed for the GOES Advanced Baseline Imager. The emulator matches the accuracy of the full physics-based model at a fraction of the computational cost, while generating reliable uncertainty estimates that could improve observation error characterization in data assimilation systems. Explainable AI methods applied across the second and third applications confirm that the trained networks extract physically meaningful information, increasing confidence in their reliability on out-of-sample data.

We are proud of Luke’s contributions to the group and wish him all the best in his next chapter!