The extreme weather phenomena witnessed last summer, including extreme rainfall and heatwaves, were attributed to the changes in short-term precipitation patterns within 10 days, resulting from global warming.

An international research team, led by Professor Ham Yoo-geun (Faculty of Earth Systems and Environmental Sciences at CNU) and consisting of Dr. Kim Jung-hwan from CNU, Professor Min Seung-gi from Pohang University of Science and Technology, and international researchers from the USA and Germany developed a deep learning model that quantifies the relationship between the intensity of global warming and daily precipitation patterns globally. They applied this model to satellite precipitation observation data from 1980 to 2020.

The findings reveal that since 2015, over 50% of the total days, saw daily precipitation patterns clearly change beyond natural variability due to the artificial influence of global warming.

Additionally, regionally, the eastern United States, East Asia including Korea, the Amazon Rainforest region, and the subtropical Eastern Pacific region were found to be experiencing the greatest changes.

This study uses deep learning techniques to identify for the first time that global warming has already significantly changed global daily precipitation characteristics. The research team published a related research paper in the online edition of the world-renowned academic journal Nature on August 31. It also was be published in the print edition one month later.

The research team explained that the AI-based method offers a more accurate assessment than conventional methods, which primarily rely on long-term trends in monthly or annual average precipitation. The research team noted that a key indicator of global warming's effect is the increasing number of days with extreme rainfall, as well as a higher frequency of rainless days, which signify the most apparent impact of global warming.

The researchers also explained that the deep learning approach developed this time is effective in identifying these nonlinear responses to changes in rainfall patterns, such as enhanced daily precipitation variability, and it can overcome the limitations of existing research.

The research team emphasizes that convolution-based deep learning techniques provide a more detailed understanding of global warming's impact on daily rainfall patterns and could contribute to improving climate prediction models.

This study was conducted as part of the “Korea Carbon Project: Observation-Based Greenhouse Gas Spatial Information Construction Technology Development Project” implemented by the Ministry of Environment to support carbon neutrality.