ABSTRACT Atlantic Meridional Overturning Circulation (AMOC) is an important component of the climate system that profoundly affects the climate of surrounding areas. AMOC is expected to slowdown due to anthropogenic climate change. The scientific community has long been at odds regarding how much AMOC has already slowed down, and whether there was any weakening at all. Key reasons for this conundrum are a limited direct observational record and the presence of substantial natural variability. The existing reconstruction methods using simple statistical approaches are not adequate to address these challenges. Moreover, they typically provide results without any uncertainty estimates. Here, we utilize a climate model ensemble experiment and a deep learning‐based approach to reconstruct the long‐term historical AMOC trend based on direct and indirect AMOC observations, while filtering out natural variability and providing adequate uncertainty quantification. Our analysis rules out the possibility of no decline with greater than 95% probability.
Cheng et al. (2026) studied this question.