Development and Use of the Stochastic Climate Change Scenario Generator

The volume of GCM simulations available for climate change impact studies continually increases. This allows for better representation of uncertainties (between GCMs, between emission scenarios, between parameterizations, etc.), but, simultaneously, the volume of available GCM output data has become so large such that it poses a strong requirement for more effective organization of climate change impact analyses. In implementing the multi-model information for a given impact analysis, only scenarios from a subset of all available GCMs are mostly employed. Less frequently, the impact analysis is based on scenarios processed from all of the GCMs. However, this is not applicable in cases where an ensemble of GCM simulations is too large (for example, when dealing with the perturbed-physics ensemble available from climateprediction.net project). In such cases, one may use scenario emulators/generators, which may produce a large set of climate change scenarios representing the whole multivariate probability distribution function of the scenarios. In the first part of the presentation, the underlying model of the scenario generator is introduced. The generator is based on a multivariate parametric model whose parameters are derived from a set of GCM based scenarios (no limit on the size of the set, the model may also be calibrated with a very large perturbed-physics ensemble). Once calibrated, the generator may produce an arbitrarily large set of climate change scenarios. These scenarios consist of changes in monthly means and variabilities, and are easily linked with the stochastic weather generator M&Rfi, which produces daily weather series to be used as an input to the impact models. In the second part, the weather series produced by the weather generator linked to the scenario generator are used to make a probabilistic assessment of the climate change impacts on four soil climate parameters: (i) the length of vegetation window (number of days with suitable soil moisture and temperature conditions); (ii) soil hydric regime (following the USDA soil classification and enhanced Newhall classification scheme); (iii) number of days during year when the soil profile is completely dry; and (iv) mean annual soil temperature at 50 cm. The projection is made with use of the SoilClim model fed by the synthetic weather series for two periods (2050 and 2100) and for a set of European and U.S. stations. The climate change impacts are assessed in terms of changes in the probability distribution functions of the four soil climate parameters based on multi-year simulations for a number of climate change scenarios. The impacts obtained using the scenario generator are compared with the impacts obtained with a “classical” approach, which consists of pooling the results obtained with a set of single GCM based climate change scenarios (using the same set of GCMs coming from the IPCC-AR4 database, which was used to calibrate the scenario generator). Acknowledgements: The present study is supported by the AMVIS-KONTAKT project (ME 844) and the GAAV Grant Agency (project IAA300420806).