2011 Reports

A framework for the simulation of regional decadal variability for agricultural and other applications

Greene, Arthur M.; Goddard, Lisa M.; Hansen, James W.

Climate prediction on decadal time scales is currently an active area of research, and reliable model-based forecasts of regional "near-term" climate change have yet to be demonstrated. In the absence of such forecasts, synthetic data sequences that capture the statistical properties of observed near-term climate variability have potential value. Incorporation of a climate change component in such sequences can help define risk estimates for a range of climatic stresses, including those lying beyond what has been experienced in the past. Properly conditioned simulations can be used to drive agricultural, hydrological or other application models, enabling resilience testing of adaptation or decision systems. The use of statistically-based methods enables the efficient generation of large ensembles of synthetic sequences and consequently, the creation of well-defined probabilistic risk estimates. In this report we examine some procedures for the generation of synthetic climate sequences that incorporate both the statistics of observed variability and expectations regarding future regional climate change. Model fitting and simulation are considered in the framework of classical time series analysis, with methodology conditioned by requirements particular to the decadal climate problem. A method of downscaling annualized simulations to the daily time step, while preserving subannual statistical properties, is presented and other possible methods discussed. Deployment in the applications setting, the details of which may vary considerably, depending on regional climate characteristics, available data and the design of follow-on models, is considered and elements of a case study presented.