Improving Decadal Climate Prediction in a Warming World

Dramatic recent increases of sea surface temperature in the mid-latitude North Atlantic are strongly influenced by decadal variability, in part linked to changes in ocean circulation. Such changes may be predictable. Experiments with the HadCM3 climate model showed that the Atlantic meridional overturning circulation (AMOC) is potentially predictable up to a decade ahead. Such studies also support a tentative conclusion that decadal predictive skill is higher when the AMOC is strong. However, the spurious numerical mixing commonplace in z-coordinate ocean models may artificially limit predictive skill in climate models such as HadCM3. Such mixing is much reduced in climate models featuring alternative 'hybrid' ocean coordinates, such as the new CHIME climate model, developed at NOCS. The aim of the present project is to investigate predictive skill in CHIME, and compare it to models such as HadCM3. Guided by supervisors and NERC scientists engaged in Theme 1 of Oceans 2025, the student will undertake the following three modelling activities: (1) A control simulation of CHIME will be sampled for initial states, from which decadal ensembles will be carried out. Ensemble-mean predictions of AMOC intensity will be compared with evolution in the corresponding decade of the control simulation, to establish the extent of predictive skill in CHIME. So far as possible, experimental design will follow previous practice in studies with HadCM3, thus enabling direct comparison of predictability in the two different models. This exercise may be repeated for a CO2-forced simulation, in order to establish whether AMOC predictability will change in the future. Both control and CO2-forced CHIME simulations are being carried out as part of Oceans 2025. (2) The ocean component of CHIME (HYCOM) will be forced using surface data spanning several recent decades, to obtain a set of realistic initial states. The resulting hindcast will be evaluated alongside observations and hindcasts carried out with higher resolution ocean models. (3) Initializing CHIME with selected hindcast states from (2), further ensembles will then be carried out, to investigate the extent to which different past decadal periods were predictable, by comparing ensemble-mean predictions with corresponding periods in the hindcast. Predictability of the AMOC in CHIME will be compared with that in HadCM3 in the specific context of mixing. Specific investigations may also focus on mechanisms controlling decadal variability, such as the advection of persistent salinity anomalies across mid-latitudes, slowly propagating baroclinic planetary waves, and episodic export of sea ice and fresh water from the Arctic. While traditional approaches to climate prediction have not attempted to forecast decadal variability, new climate forecasting systems are now becoming operational. The DePreSys system, developed at the UK Met Office, comprises assimilation of observed ocean states into the HadCM3 climate model. While DePreSys predictions of surface air temperature are substantially improved at a global scale, little improvement is achieved over the North Atlantic. As decadal climate forecasts are now becoming critical for a wide range of policymaking decisions, there is an outstanding need for research that leads to improvements in the skill of these forecasts. In possibly extending the limit of Atlantic decadal predictability, the proposed research could prove of great value. In bringing together research groups at the National Oceanography Centre, Southampton, the University of Reading, and the UK Met Office, the project will help to strengthen national capability in decadal prediction.