Understanding the role of the ocean in non-flux-adjusted perturbed physics ensembles

This project will address two fundamental scientific questions: 1. What determines the ocean's control over the rate of global warming to externally driven climate change: the atmosphere or the ocean? 2. If the atmospheric CO2 increase slows down the thermohaline circulation (THC), what are the chances of it recovering if CO2 levels are reduced? It will also address one practical question: 3. What influence do ocean resolution and the use of flux adjustment have on probabilistic climate forecasts using a perturbed physics approach? The project will provide answers to them by conducting two-phase ensemble climate change simulations using the coupled atmosphere-ocean general circulation model (AOGCM), HadCM3, which is well known for its intrinsic long-term stability. Many different models will be run with slightly different initial conditions and model physics, both of which contain large uncertainties. The results will be combined to make probabilistic forecasts of the ocean's heat uptake, which impacts the rate of global warming, and the THC. They will be compared with the results from our current climateprediction.net (CPDN) experiment, which uses a coarse-oceanic-resolution version of HadCM3 and flux adjustment. Flux adjustment refers to the addition of artificial heat and freshwater fluxes to keep the model from drifting to unrealistic states, and is widely used in climate change experiments. However, flux adjustment has no physical basis, so it is desirable not to use it. Analysis of the current CPDN ensemble experiments has identified significant difficulties in simulating the THC with flux-adjusted models. In the first, 'spinup', phase, the ocean GCM is run with time-constant forcing until each simulation is stable. Unlike the current CPDN experiment, in which we employ flux adjustment to force a stable base climate, this project will not use flux adjustment but instead fully exploit the vast computing resources provided by our distributed network to identify and select stable and realistic models. This requires a novel approach to spinup in which we use thousands of 50 to 100 year simulations to identify models with realistic base climates that are stable on century timescales. This is in contrast to the commonly adopted method of using multi-century simulations to spin up a single model, hoping to reach a realistic equilibrium. In the second, 'transient forcing', phase, we will run the models under historical and future forcing from 1920 to 2080 as in the current CPDN experiment. The outcome will be the key sensitivities of the ocean climate and climate change. It is not yet known, for example, if ocean mixing parametrisation or the representation of the hydrological cycle play a larger role in determining large-scale ocean circulation. We will also conduct complementary 'transient forcing' simulations in collaboration with non-NERC-funded researchers at the Hadley Centre to examine the reversibility of the change in the THC strength. Starting from a selection of spun-up models from the first phase, CO2 will be rapidly increased, then rapidly decreased. This transient change in CO2 forcing is expected to impact the THC intensity. The Hadley Centre will analyse the results to quantify the anthropogenic changes to the THC in the large ensemble, in particular whether the results are reversible with a return to the pre-industrial climate. Computing will be performed on thousands of PCs, whose idle computing time are donated by volunteers from the general public worldwide. Experiments will be performed using the now well-established CPDN distributed-computing infrastructure. A Microsoft-funded high-performing-computer cluster, which will be available at the Oxford e-Research Centre in the next year, will be used to test the models before their public distribution. However, computing support will be required to introduce a new model version, additional diagnostics, and new experimental design.