Abstracts International Symposium 'Developing Countries facing Global Warming: a Post-Kyoto Assessment' pages:64-65
International Symposium 'Developing Countries facing Global Warming: a Post-Kyoto Assessment' location:Brussels date:12 - 13 June 2009
Most climate models do not accurately reproduce present climate situation with similar statistics. This means that direct application of projected climate variables to hydrological models for impact analysis is implausible. In order to reach an adequate impact analysis of climate change, the support climate models should provide perturbation factors (PFs) or climate change signals that are consistent and account for the different aggregation dependency. This study uses the quantile perturbation analysis to explore the variation of PFs of rainfall intensities with respect to different aggregations, the model resolution and frequency, for three future IPCC emission scenarios (A2, A1B and B1) from eighteen Global Climate Models (GCMs). This approach ensures that the perturbations in the observed events are consistent with similarly ranked events in the climate model series. The GCM scenario quantiles are ranked from high to low values from current (1971-1990) conditions and compared to ranked data of GCM simulations for future (2049-2060 and 2081-2100) conditions. These ratios (PFs) were graphically analyzed for consistency, uncertainty, and dependency on frequency (return period) for the different rainfall aggregations.
The PFs, derived from three GCMs (CGCM3.1t63, CM2.1U.H2 and CM2.0) outputs, are highly inconsistent with those derived from the other models for all the different rainfall aggregations considered and for both the 2050s and 2090s. Uncertainty in PFs, which depicts uncertainty in a given model, increases with increase in model temporal resolution. For most models, the PFs are above one and depend on the frequency for both the 2050s and 2090s and all three emission scenarios. For daily and monthly aggregations, the distributions of PFs tend to cluster around one for most models. The variation in PFs is higher for the lower quantiles for annual aggregation; but increases in the wet months than in the dry months for monthly aggregation. At daily aggregation, there are many outliers in the PFs for higher ranked quantiles for most of the GCMs. There is no evidence to suggest that the PFs are spatially dependent although CM4.1 model has shown significant variation from Katonga to Ruizi catchments. Models such as MK3.5, MK3.0 and ECHAM5, with higher spatial resolutions, tend to have PFs close to or lower than one. Future rainfall intensities will be frequently higher than the present rainfall events in Katonga and Ruizi catchments for both the 2050s and 2090s. Although PFs can easily be derived from annual and monthly aggregations, the challenge of deriving climate change signals from daily climate model series, especially for dry days, for application to observed series, is still big. The resulting uncertainty is carried to hydrological models; leading to uncertain impact analysis. However, perturbation analysis is important in assessing consistent models and uncertainty in climate projections as given by the different GCMs.