Uncertainties in the Projection of Dynamic Sea Level in CMIP6 and FGOALS-g3 Large Ensemble

Abstract

Abstract Decision-makers need reliable projections of future sea level change for risk assessment. Untangling the sources of uncertainty in sea level projections will help narrow the projection uncertainty. Here, we separate and quantify the contributions of internal variability, intermodel uncertainty, and scenario uncertainty to the ensemble spread of dynamic sea level (DSL) at both the basin and regional scales using Coupled Model Intercomparison Project phase 6 (CMIP6) and FGOALS-g3 large ensemble (LEN) data. For basin-mean DSL projections, intermodel uncertainty is the dominant contributor (>55%) in the near term (2021–40), midterm (2041–60), and long term (2081–2100) relative to the climatology of 1995–2014. Internal variability is of secondary importance in the near- and midterm until scenario uncertainty exceeds it in all basins except the Indian Ocean in the long term. For regional-scale DSL projections, internal variability is the dominant contributor (60%–100%) in the Pacific Ocean, Indian Ocean, and western boundary of the Atlantic Ocean, while intermodel uncertainty is more important in other regions in the near term. The contribution of internal variability (intermodel uncertainty) decreases (increases) in most regions from midterm to long term. Scenario uncertainty becomes important after emerging in the Southern, Pacific, and Atlantic Oceans. The signal-to-noise ratio (S/N) maps for regional DSL projections show that anthropogenic DSL signals can only emerge from a few regions. Assuming that model differences are eliminated, the perfect CMIP6 ensemble can capture more anthropogenic regional DSL signals in advance. These findings will help establish future constraints on DSL projections and further improve the next generation of climate models. Significance Statement Clarifying the sources of uncertainty in DSL projections is fundamental for further improving projections. Hence, this study aims to separate and quantify the three uncertainties in both basin and regional scale DSL projections using CMIP6 models and a 110-member large-ensemble experiment: internal variability, intermodel, and scenario uncertainty. We show that intermodel uncertainty is the dominant contributor at both the basin and regional scale, even though internal variability plays an important role in most regions of the Indian Ocean, Pacific Ocean, and Atlantic Ocean. Scenario uncertainty is negligible before it emerges at both the basin and regional scale in the long term. This provides the direction for constraining uncertainty in DSL projections.