Quantifying Uncertainties in the Modelled Estimates of Extreme Precipitation Events at Upper Thames River Basin

Abstract

Assessment of climate change impact on hydrology at watershed scale incorporates downscaling of global scale climatic variables into local scale hydrologic variables and evaluation of future hydrologic extremes. The climatological inputs obtained from several global climate models suffer the limitations due to incomplete knowledge arising from the inherent physical, chemical processes and the parameterization of the model structure. Downscaled output from a single AOGCM with a single emission scenario represents only one of all possible future climate realizations; averaging outputs from multiple AOGCMs might underestimate the extent of future changes in the intensity and frequency of climatological variables. These available methods, thus cannot be representative of the full extent of climate change. Present research, therefore addresses two major questions: (i) should climate research adopt equal weights from AOGCM outputs to generate future climate?; and (ii) what is the probability of the future extreme events to be more severe? This paper explores the methods available for quantifying uncertainties from the AOGCM outputs and provides an extensive investigation of the nonparametric kernel estimator based on choice of bandwidths for investigating the severity of extreme precipitation events over the next century. The Sheather-Jones plug-in kernel estimate appears to be a major improvement over the parametric methods with known distribution. Results indicate increased probabilities for higher intensities and frequencies of events. The applied methodology is flexible and can be adapted to any uncertainty estimation studies with unknown densities. The presented research is expected to broaden our existing knowledge on the nature of the extreme precipitation events and the propagation and quantification of uncertainties arising from the global climate models and emission scenarios.