A0536
Title: Extract long-term trend from large-missing gap Atlantic meridional overturning circulation (AMOC) data
Authors: Yizi Cheng - University of Cincinnati (United States) [presenting]
Won Chang - University of Cincinnati (United States)
Roman Olson - Institute of Industrial Science (Japan)
Jongsoo Shin - Yonsei University (Korea, South)
Soon-il An - Yonsei University (Korea, South)
Abstract: Atlantic Meridional Overturning Circulation (AMOC) is a large-scale ocean circulation that transports cold and dense water in the deep North Atlantic to the equator and warm and salty water in the upper layers of the North Atlantic to the pole. AMOC plays an important role in heat and carbon transport, thus having considerable impacts on climate change, and in response, on natural and human systems. There is an assumption about persistent AMOC weakening in response to anthropogenic forcing. However, due to technical limitations, AMOC was barely measured before 2004, leading to the difficulty in understanding how AMOC has changed over the last century. The Denoising Variational Auto-encoder (DVAE) framework is applied to reconstruct the long-term trend of AMOC in the previous 140 years. A DAVE model is built based on an ensemble of CESM model runs, which can extract the long-term trend from noisy observations on AMOC strength and sea surface temperature data while filtering out the effects of natural variability and quantifying the estimation uncertainty. The trained model is then applied to real observations, the RAPID AMOC project, and NOAA Extended Reconstructed Sea Surface Temperature V5 data. The results show that there has been a persistent AMOC weakening over the last century, with a 0.997 probability of decreasing over 1SV.
