Providing application-specific climate projections datasets: CSIRO’s Climate Futures Framework

Abstract

CSIRO has been developing climate projections since the 1980s and released national projections for Australia in 1988, 1992, 1996, 2001 and 2007. As these were designed communicate nation- and state-wide changes to a range of users, the projections focused on individual climate variables, such as temperature, with a range of uncertainty derived from multiple climate models. Over this time, CSIRO has also been involved in many impact assessments with collaborators and clients. These usually require projections tailored for specific applications. Such applications require internally consistent combinations of climate variables such as those based on output from individual models. However, many end-users of projections find the complexity (including the range of uncertainty) difficult to deal with. There are projections from up to 24 global climate models (GCMs), up to six emission scenarios, and around a dozen climate variables. Given limited resources, end-users often seek a narrower range of projections for their applications. CSIRO Marine and Atmospheric Research has developed a novel approach to deal with this issue in a way that is tailored to the specific needs of applications. The Climate Futures Framework allows climate scientists, in close consultation with end-users, to provide simplified projections while still addressing uncertainty. This approach also facilitates the selection of a sub-set of climate models from which internally consistent data can be obtained for multiple variables. The Climate Futures Framework classifies the GCM projected changes into pre-defined categories (called climate futures) defined by two climate variables - commonly the change in annual mean surface temperature and annual average rainfall. Each climate future is assigned a relative likelihood based on the number of climate models that fall within that category. For example, if 12 of 24 models fall into the Warmer - Drier climate future, it is given a relative likelihood of 50%. In most cases, this results in a small number of climate futures, each with a relative likelihood. It is then possible to further simplify the projections using a risk assessment approach (where risk = consequence x likelihood). In consultation with the decision makers, key climate futures that directly influence the decision-making are identified for further assessment. Any other climate futures that will not affect decision-making can be ignored for the purposes of the particular impact assessment. Once the relevant climate futures and key variables have been identified, multi-variate ranking is used to determine the most representative model(s) from each climate future. This subset of models can then be used to generate internally consistent datasets. CSIRO has been using the Climate Futures Framework to provide climate change projections and associated datasets to clients within and outside Australia since 2009. Although focussed research into client attitudes to the approach is yet to be undertaken, initial feedback suggests it is a useful aid to understanding projected changes and their associated uncertainty, while simplifying the selection of a small number of models for use in risk assessment. The Framework also offers other benefits, including improved communication of projections to a wide audience. This paper will focus on its use in providing application-specific datasets.