How Tropical Conditions Influence Snowfall in Western Antarctica
A study from the University of Utah reveals that tropical climate patterns like ENSO modulate the amount of snow that falls annually on Western Antarctica. By combining ice cores and machine learning models, researchers clarify a long-suspected link between the tropics and vast ice sheets.
Each year, the amount of snow covering Western Antarctica fluctuates dramatically, and a new study shows these variations are driven by tropical climate conditions. Researchers from the University of Utah have identified a teleconnection mechanism that links temperature and precipitation anomalies in low latitudes to snow deposits over 3000 kilometers south.
What the study reveals: precise results and key figures
By analyzing ice cores taken along the western coast of Antarctica, scientists have identified layers of snow whose thickness varies according to the phases of the Southern Oscillation (ENSO). Years marked by intense tropical El Niño episodes consistently coincide with decreased snowfall, while La Niña periods favor increased snowfall. This correlation, observed over several decades, underscores the importance of tropical conditions as a driver of annual variations in Antarctic ice cover.
The authors emphasize that tropical variability explains a significant portion of the uncertainty in global climate models' predictions for the Antarctic region. In other words, without correctly integrating tropical signals, projections of ice mass and sea level remain largely imprecise. The work also highlights the role of high-altitude atmospheric currents that transport moisture from tropical oceans to polar latitudes, amplifying or reducing snow production depending on the phase of ENSO.
How: research method and data used
To establish this link, the team combined satellite data from the Copernicus mission with ice core archives dating back over 150 years. Density and stable isotope measurements allowed for the reconstruction of precipitation histories, while predictive models such as the ECMWF system and GraphCast network were employed to simulate atmospheric flows. A predictive model based on machine learning was trained using these datasets, integrating both in-situ observations and global atmospheric data.
The training process used deep neural networks capable of identifying nonlinear patterns between tropical anomalies and polar responses. In parallel, the Pangu-Weather model provided fine temporal resolution, reducing prediction uncertainty through its global attention architecture. Researchers also exploited the ECMWF's historical atmospheric data series to validate simulations and adjust machine learning algorithm parameters.
This hybrid approach, combining ice cores, satellite data, and artificial intelligence, offers new insights into the complex interplay between tropical climate patterns and Antarctic snowfall.
