The challenge of the CSIRO decadal climate forecasting project is to deliver multi- year to decadal forecasts that will enable Australian Industries and Regulators to better deal with climate variability and climate extremes. Internationally, considerable research investment is occurring with the aim of developing useful multi-year to decadal climate forecasts. This endeavor has been identified by the World Climate Research Program (WCRP) as a grand challenge in near term climate prediction with the anticipation that useful climate forecasts are attainable. As with all research, this project carries some risk, however, the rewards to the Australian community of delivering useful climate forecasts at the multi-year to decadal timescales carries potentially huge benefits for industry and our ability to manage our natural assets. To be successful will require building a climate forecasting system underpinned by identifying and developing a deep understanding of the climatic processes that can be skillfully predicted.
An important philosophy of the project is to demonstrate how the understanding gained through the observations and process studies translates into improved decadal climate forecasts. To this end, climate forecasts will be continually assessed by generating hindcasts over the recent decades and verified against the observed climate. In this way we can quantify the skill of the CAFE system and document how the forecast skill improves with time. To assess the forecast skill we will use a range of key climate indices (e.g. NINO3.4) and predictants (e.g. rainfall extremes). We will need to develop novel ensemble based probabilistic climate forecast metrics similar to those currently employed in numerical weather prediction. Probabilistic assessments will be used to characterise the forecast skill, to help identify where to direct our resources, and to evaluate the progress towards our ultimate goal of reliable and useful forecasts. Such assessments will also help direct the development of the climate model (i.e. identification of model bias), improve data assimilation by assessing the value of different observational data streams, and to better constrain ensemble initialisation through systematic characterisation of forecast error growth.
To ensure success there will need to be integration of the three activities of the project. Further the project must engage other research providers as well as internal and external clients. In our engagement we must effectively communicate the success and challenges of delivering useful multiyear climate forecasts.
To balance the need to deliver climate forecasts with the requirement to advance our fundamental understanding of the climate system to ensure we achieve the goal of delivering useful climate forecasts the project has 3 key activities: