| description abstract | A statistical downscaling methodology was implemented to generate daily time series of temperature and rainfall for point locations within a catchment, based on the output from general circulation models. The rainfall scenarios were constructed by a two-stage process. First, for a single station, a conditional first-order Markov chain was used to generate wet and dry day successions. Then, the multisite scenarios were constructed by sampling from a benchmark file containing a daily time series of multiple-site observations, classified by season, circulation weather type, and whether the day is wet or dry at the reference station. The temperature scenarios were constructed using deterministic transfer functions initialized by free atmosphere variables. The relationship between the temperature and rainfall scenarios is established in two ways. First, sea level pressure fields define the circulation weather types underpinning the rainfall scenarios and are used to construct predictor variables in the temperature scenarios. Second, separate temperature transfer functions are developed for wet and dry days. The methods were evaluated in two Mediterranean catchments. The rainfall scenarios were always too dry, despite the application of Monte Carlo techniques in an attempt to overcome the problem. The temperature scenarios were generally too cool. The scenarios were used to explore the occurrence of extreme events, and the changes predicted in response to climate change, taking the example of temperature. The nonlinear relationship between changes in the mean and changes at the extremes was clearly demonstrated. | |
