A weather-type conditioned multi-site stochastic rainfall model for the generation of scenarios of climatic variability and change

Further developments of a stochastic rainfall model conditioned by weather types for the water resource region of Yorkshire, UK, are presented. The model is extended to multi-site and a new technique is developed to allow the reproduction of historical monthly rainfall cross-correlation statistics. Monte-Carlo simulation and sampling techniques are combined to preserve monthly historical rainfall cross-correlation between two sub-regional Neyman-Scott Rectangular Pulses (NSRP) rainfall models. These are conditioned seasonally with a semi-Markov weather generator and used to generate multiple long synthetic series for climate impact assessment in Yorkshire, encompassing an area of some 15,000 km2. An example application of the model in constructing a climate change scenario for 2021-2050 is detailed. Current UK climate change scenarios show change in both airflow patterns and rainfall properties. In climate scenario development it is therefore desirable to be able to change the frequency of weather state occurrence as well as the mean and variance statistics of rainfall. This methodology allows both the impact of variation in the frequency or persistence of weather states and changes in internal weather state properties such as increased intensity or proportion of dry days for example to be investigated. This methodology of simulating potential atmospheric circulation changes may provide a valuable tool for the future management of water resource systems and many other hydrological impact applications.