The atmospheric blockage could bring record rainfall to Hawaii. This phenomenon was also linked to the record heatwave and floods that affected Europe in 2023. Credit: Jason Miller via Unsplash (right) and Henrique Ferreira (left)
Researchers at the University of Hawaii at Mānoa are understanding the possible future effects of a weather phenomenon called atmospheric deprivation. A blocking event occurs when a large high-pressure system comes to rest and changes the direction of the jet stream, creating weather patterns that can be associated with record flooding and heat waves.
In a new study, University of Manoa atmospheric scientist Christina Calamperidou uses deep learning models, a type of artificial intelligence that uses algorithms that learn from large amounts of data, to estimate the frequency of interception events over the past 1,000 years. and how it was influenced. Climate change may influence future atmospheric interception phenomena.
“This study aimed to extract paleoclimate signals from the paleoclimate record using a deep learning model that infers the atmospheric cutoff frequency from surface temperature,” Karaperidou said. “This is a unique study and the first attempt to reconstruct a long record of blocking frequencies based on a complex and unknown relationship with surface temperature. Machine learning methods are extremely powerful for such tasks. .”
Training deep learning models
Article continues below ad Article continues below ad
Karamperidou developed a specialized deep learning model and trained it using historical data and a large ensemble of climate model simulations. The model is now able to estimate the frequency of anomaly interception events in seasonal temperature reconstructions over the past 1000 years. These past temperature reconstructions are relatively well constrained by an extensive network of temperature-sensitive tree-ring records during the growing season.
“This approach shows that deep learning models are a powerful tool to overcome the long-standing problem of extracting paleoclimate from paleoclimate,” said Karaperidou. “This approach can also be used for the instrumental period of climate history starting in the 18th century, when regular meteorological measurements were taken, since reliable data for identifying blocks have been available since the 1940s, or perhaps This is because there is only a satellite era (after the satellite era).
Frequency of future blocking events
There is still no scientific consensus on how climate change will change the frequency of blocking events. These strong and persistent mid-latitude highs could have a significant impact on Hawaii, which is experiencing flooding due to a persistent block in the North Pacific, and the summer block could bring extreme heat. It could also have a significant impact around the world, including in the Pacific Northwest and Europe, where there is a strong climate. wave.
Article continues below ad
Therefore, understanding changes in the frequency of these events is critical for Hawaii, especially as it relates to other factors that influence climate, such as El Niño and long-term patterns of sea surface temperatures in the tropical Pacific Ocean. This study allowed Karaperidou to link the frequency of blocking in mid-high and high latitudes to climate change in the tropical Pacific in the long context of the past millennium. This is essential for validating climate models and blocking uncertainties in future climate projections.
open research and transparency
Karamperidou collaborated with two UH Mānoa students to create a unique web interface for exploring deep learning models and resulting reconstructions. He said sharing results and methods in this way is important for open research best practices and transparency, especially as the applications of machine learning and artificial intelligence rapidly expand into many aspects of daily life. I emphasized one thing. The web interface is hosted on Jetstream-2, an NSF-supported cloud computing system. Its regional partners include UH Information Technology Services – Cyberinfrastructor and Hawaii Data Science Institute.
In the future, Karamperidou plans to explore various features and architectural enhancements of the deep learning model to expand its application to climate phenomena and variables that are directly related to socio-economic impacts.
