Abstract:
|
The accurate quantification of how much heat the oceans are absorbing is crucial for estimating the anticipated global temperature increase due to anthropogenic emissions. The Argo program, consisting of floats that measure vertical temperature profiles at a dense resolution over the global ocean, has provided a wealth of data from which to estimate ocean heat content. However, statistically modeling the uncertainty in such estimates remains challenging due to the need for a globally valid covariance model that can capture complex nonstationarity in the temperature field. Here, we develop a hierarchical Bayesian Gaussian process model that exhaustively models non-stationarity while remaining computationally feasible for large spatial datasets. Our model produces credible intervals for ocean heat uptake that, unlike previous estimates, directly quantify uncertainty due to parameter estimation. Additionally, we develop a statistical framework for testing the possibility that ocean heat content estimates are biased by preferential sampling arising from the dependence between the Argo observation locations (which are determined by density currents) and the underlying heat field.
|