Advances in spatio-temporal modeling of ocean heat content with Argo floats
Date:
Abstract: Estimating the global ocean heat content (OHC) with reliable uncertainties is critical for understanding the evolution of Earth’s climate, as most excess energy accumulated in the climate system by the Earth Energy Imbalance is stored in the ocean. Argo floats provide a rich 20-year dataset of temperature measurements across the global ocean, but its large size and nonstationarity present computational and statistical challenges. We first present our latest work on addressing these challenges using a mapping and uncertainty quantification pipeline based on conditionally simulating from a locally stationary bivariate Gaussian process model. This bivariate approach allows both space-time dependence and vertical dependence across pressure levels to be incorporated in the uncertainties. Using this reproducible pipeline, we present 2004-2022 estimates and uncertainties for the OHC trend and related quantities on global and regional scales. In addition, we propose further computational improvements using recent developments in neural inference. In the second part of this talk, we will show preliminary results of applying neural likelihood and conditional simulation to the ocean heat content modeling framework.