Joint PDFs for Australian climate in future decades and an idealized application to wheat crop yield

Abstract

The possible range of climate variables at a location for decades in the future is represented using a probability density function derived by summing distributions for both change forced by global warming and unforced decadal variability. Both components are estimated from CMIP3 climate model simulations. Standard deviations of decadal means in temperature and rainfall for locations in the Australian region are presented. The methods are extended to determine joint PDFs for rainfall and temperature, and these allow for the anticorrelations diagnosed for the pair over much of Australia. The distributions derived for the change in climate by 2070 in central New South Wales are then combined with a simple impact function for wheat, to illustrate a potential application of joint PDFs. The forced change in the crop in this idealized example has a distribution that ranges from small declines to substantial increases, largely driven by the effect of CO2 on plant growth. The anticorrelation broadens the range of impact, compared to the case where this is not allowed for. The distribution is also broadened in the decadal case, in which warmer and drier decades can produce declines in the crop. Increases are less likely in corresponding results for three southern and western wheat regions, where decreased rainfall is likely. This joint PDF approach can potentially allow efficient estimation of various impacts in multiple locations provided realistic impact functions are available.