Answers are grounded in this paper and cite their sources.
The NOAA Extended Reconstructed Sea Surface Temperature version 6 (ERSSTv6) provides a globally complete monthly SST dataset from 1850 to present, reconstructed from ship, buoy, and near-surface Argo observations using an artificial neural network (ANN) approach. ERSSTv6 incorporates enhanced quality control, large-scale spatial filtering, and an updated ice–SST proxy and includes an operationally feasible and physically consistent uncertainty framework that supports robust assessments of SST variability, trends, and ocean-related applications. We quantify total SST uncertainty as the combination of parametric and reconstruction uncertainties. Parametric uncertainty is estimated from a 1000-member ensemble generated by perturbing 30 internal parameters, including ANN coefficients, quality control, bias adjustments, and ice-SST proxy. ANN-related uncertainty is approximated using a random perturbation (RP) to ANN coefficients, whose magnitude is calibrated by validation errors. Reconstruction uncertainty is assessed with a 500-member ensemble applied to a space-time-complete reference dataset, representing algorithmic error under ideal observations. Results show that grid-scale uncertainty is largest in regions with strong SST variability and sparse data, including western boundary currents, the eastern equatorial Pacific, and ice-covered areas, with global mean uncertainty decreasing from 0.4°–0.5°C in the 1850s to ∼0.3°C after 1950. Uncertainty in global mean SST is smaller, ranging from 0.06°–0.08°C in the 1850s to ∼0.02°C in recent decades. Factor analysis indicates that ANN RP dominates the grid-scale uncertainty, while ship bias correction and ice–SST proxy dominate the uncertainty of global-mean SST. Comparisons show that both SST and its uncertainty in ERSSTv6 are broadly consistent with those other SST products.
No takes yet. Share an insight, caveat, or question.
Huang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: