A Spatio-Temporal Model for Mean, Anomaly, and Trend Fields of North Atlantic Sea Surface Temperature

Abstract

We consider the problem of fitting a statistical model to 30 years of sea surface temperature records collected over a large portion of the Northern Atlantic. The observations were collected sparsely in space and time with different levels of accuracy. The purpose of the model is to produce an atlas of oceanic properties, including climatological mean fields, estimates of historical trends, and a spatio-temporal reconstruction of the anomalies, i.e., the transient deviations from the climatological mean. These products are of interest to climate change and climate variability research, numerical modeling, and remote sensing analyses. Our model improves upon the current tools used by oceanographers in that it constructs instantaneous temperature fields before averaging them into the climatology, thus giving equal weight to all years in the time frame, regardless of the temporal distribution of data. It also accounts for nonisotropic and nonstationary space and time dependencies, owing to its use of discrete process convolutions. Particular attention is given to the handling of massive datasets such as the one under study. This is achieved by considering compact support kernels that allow an efficient parallelization of the Markov chain Monte Carlo method used in the estimation of the model parameters. Resulting monthly climatologies are compared with those of the World Ocean Atlas 2001, version 2. Different water masses appear better separated in our climatology, and a close link emerges between the kernels' shape and the dominating patterns of ocean currents. The subpolar and the temperate North Atlantic display opposite trends, with the former mainly cooling over the years and the latter mainly warming, especially in the Gulf Stream region. Long-term changes in annual cycles are also detected. As in any hierarchical Bayesian model, parameter estimates come with credibility intervals, which are useful to compare results with other approaches and detect areas where sampling campaigns are needed the most.