USE OF CONDITIONAL STOCHASTIC-MODELS TO GENERATE CLIMATE-CHANGE SCENARIOS

Stochastic models have been proposed as one technique for generating s cenarios of future climate change. One particular daily stochastic wea ther generator, termed Richardson's Model or WGEN, has received much a ttention. Because it is expressed in a conditional form convenient for simulation (e.g., temperature is modeled conditional on precipitation occurrence), some of its statistical characteristics are unclear. In the present paper, the theoretical statistical properties of a simplif ied version of Richardson's model are derived. These results establish that when its parameters are varied, certain unanticipated effects ca n be produced. For instance, modifying the probability of daily precip itation occurrence not only changes the mean of daily temperature, but its variance and autocorrelation as well. A prescription for how best to adjust these model parameters to obtain the desired climate change s is provided. Such precautions apply to conditional stochastic models more generally.