Search

Warming of the climate system accumulates mostly in the ocean and discrepancies in how this is modelled contribute to uncertainties in predicting sea level rise. In this study, regional temperature changes in an atmosphere–ocean general circulation model (HadCM3) are partitioned between excess (due to perturbed surface heat fluxes) and redistributed (arising from changing circulation and perturbations to mixing) components. In simulations with historical forcing, we firstly compare this excess–redistribution partitioning with the spice and heave decomposition, in which temperature anomalies enter the ocean interior either along isopycnals (spice) or across isopycnals (heave, without affecting the temperature-salinity curve). Secondly, heat and salinity budgets projected into thermohaline space naturally reveal the mechanisms behind temperature change by spice and heave linked with water mass generation or destruction. Excess warming enters the ocean as warming by heave in subtropical gyres whereas it mainly projects onto warming by spice in the Southern Ocean and the tropical Atlantic. In subtropical gyres, Ekman pumping generates excess warming as confirmed by Eulerian heat budgets. In contrast, isopycnal mixing partly drives warming and salinification by spice, as confirmed by budgets in thermohaline space, underlying the key role of salinity changes for the ocean warming signature. Our study suggests a method to detect excess warming using spice and heave calculated from observed repeat profiles of temperature and salinity.

Highlights: • Arctic liquid freshwater budget simulated in 14 CORE-II models is studied. • The models better represent the temporal variability than the mean state. • Multi-model mean (MMM) FW fluxes compare well with observations. • MMM FWC shows an upward trend in the recent years, with an underestimated rate. • FW flux interannual variability is more consistent where volume flux determines it. Abstract: The Arctic Ocean simulated in 14 global ocean-sea ice models in the framework of the Coordinated Ocean-ice Reference Experiments, phase II (CORE-II) is analyzed in this study. The focus is on the Arctic liquid freshwater (FW) sources and freshwater content (FWC). The models agree on the interannual variability of liquid FW transport at the gateways where the ocean volume transport determines the FW transport variability. The variation of liquid FWC is induced by both the surface FW flux (associated with sea ice production) and lateral liquid FW transport, which are in phase when averaged on decadal time scales. The liquid FWC shows an increase starting from the mid-1990s, caused by the reduction of both sea ice formation and liquid FW export, with the former being more significant in most of the models. The mean state of the FW budget is less consistently simulated than the temporal variability. The model ensemble means of liquid FW transport through the Arctic gateways compare well with observations. On average, the models have too high mean FWC, weaker upward trends of FWC in the recent decade than the observation, and low consistency in the temporal variation of FWC spatial distribution, which needs to be further explored for the purpose of model development.

Highlights: • Arctic sea ice extent and solid freshwater in 14 CORE-II models are inter-compared. • The models better represent the variability than the mean state. • The September ice extent trend is reasonably represented by the model ensemble mean. • The descending trend of ice thickness is underestimated compared to observations. • The models underestimate the reduction in solid freshwater content in recent years. Abstract: The Arctic Ocean simulated in fourteen global ocean-sea ice models in the framework of the Coordinated Ocean-ice Reference Experiments, phase II (CORE II) is analyzed. The focus is on the Arctic sea ice extent, the solid freshwater (FW) sources and solid freshwater content (FWC). Available observations are used for model evaluation. The variability of sea ice extent and solid FW budget is more consistently reproduced than their mean state in the models. The descending trend of September sea ice extent is well simulated in terms of the model ensemble mean. Models overestimating sea ice thickness tend to underestimate the descending trend of September sea ice extent. The models underestimate the observed sea ice thinning trend by a factor of two. When averaged on decadal time scales, the variation of Arctic solid FWC is contributed by those of both sea ice production and sea ice transport, which are out of phase in time. The solid FWC decreased in the recent decades, caused mainly by the reduction in sea ice thickness. The models did not simulate the acceleration of sea ice thickness decline, leading to an underestimation of solid FWC trend after 2000. The common model behavior, including the tendency to underestimate the trend of sea ice thickness and March sea ice extent, remains to be improved.