277 Australian Meteorological and Oceanographic Journal 62 (2012) 277–285 Observational constraints on parameter estimates for a simple climate model (Manuscript received June 2012; revised August 2012) Roger W. Bodman 1 , David J. Karoly 1 , Susan Wijffels 2 and Ian G. Enting 3 1 School of Earth Sciences, The University of Melbourne, Melbourne, Australia 2 CSIRO Wealth from Oceans Flagship, Hobart, Australia 3 MASCOS, The University of Melbourne, Melbourne, Australia Introduction Simple climate models have been used in each of the IPCC assessment reports to estimate future global mean temperature changes from a range of emission scenarios. Such models continue to be widely used in integrated assessment models and by policymakers as they provide a flexible and economical alternative to AOGCMs (Atmosphere/Ocean General Circulation Models). A model such as MAGICC can, when tuned appropriately, closely emulate the global mean temperature projections of much more complex models (Meinshausen et al. 2011; Wigley and Raper 2001). The model can then provide insight into how key global processes interact to produce warming in the climate system. There are a number of sources of uncertainty that apply to MAGICC and its ability to project temperature changes, principally climate sensitivity, ocean effective vertical diffusivity, aerosol forcing and the carbon cycle, as well as the structural simplifications of the model (Meinshausen 2006; Wigley and Raper 2001). These parameter estimates can also have large impacts on the economic assessment of mitigation policies from integrated assessment models that are built on simple climate models, such as MAGICC (van Vuuren et al. 2011). Here we focus on constraining the model’s ocean diffusivity parameter and explore the implications that follow from revising its previous standard value. This investigation begins by recognising the significance of the ocean’s role in future global mean temperature changes and therefore in MAGICC, and it identifies the key climate system parameters that need to be estimated for the model to simulate 20 th century global mean temperature changes. Methods for estimating these parameters are then applied that are designed to isolate important individual parameters in the simple climate model using largely independent observational constraints. This helps to avoid the problem of how to deal with cross-correlations and weighting of different observations that arises in previous studies. As well, we address the mismatch between the observed and modelled ocean temperature change profile. Corresponding author address: Roger Bodman, Centre for Strategic Eco- nomic Studies, Victoria University, Melbourne, Victoria, 8001, Australia Email: roger.bodman@vu.edu.au Observations of ocean temperature and ocean heat content changes are investi- gated in order to better estimate the primary ocean parameters in a simple climate model, namely MAGICC (Model for the Assessment of Greenhouse gas Induced Climate Change). A re-examination of the simulated 20 th century ocean tempera- ture changes has been made possible by the release of new observational data, which indicates that complex climate models and MAGICC are mixing too much heat into the deeper ocean. Goodness-of-fit testing between the simulated and observed 1960–2008 world ocean temperature change leads to a revised set of parameters for the simple climate model that include a value for the ocean effec- tive vertical diffusivity that is nearly a quarter of that used in the IPCC Third and Fourth Assessment Report versions of the model. Testing an independent con- straint based on the ratio of changes in sea surface temperature to 700 m ocean heat content changes produces similar results. The lower ocean diffusivity affects the surface temperature results and alters the best estimate for the climate sensi- tivity parameter. The projected temperature changes for a high-growth emissions scenario show a larger increase in temperature by 2100 even with a reduced cli- mate sensitivity.