We briefly describe the research work done in the last two decades regarding numerical modelling of climate. First, we describe the characteristics of global circulation models, which allow simulating the future evolution of climate according to different forcing scenarios related to human activities. These emission scenarios are the main source of uncertainty to model climate change; thus, a set of likely scenarios is usually considered in climate change studies. We also describe the effect of different sources of uncertainty in the final results; for instance, model errors due to approximate parametrizations, etc. These uncertainties put hard constraint to this problem and force to analyze it within the framework of probability, using new “ensemble prediction techniques” to quantify the existing uncertainty, and require huge computational resources. Therefore, the spatial resolution of climate simulation is constrained to 200-500 km, providing only global average information. However, in the recent years, an increasing research activity is focussing on the generation of regional climate change scenarios, applying different techniques to project global into local information.
