Analysis of Capabilities of Bias-corrected Precipitation Simulation from Ensemble of Downscaled GCMs in Reconstruction of Historical Wet and Dry Spell Characteristics

Precipitation outputs from downscaled GCMs are often used as input data for hydrological models to study the impact of climate change on hydrological cycles and water resources. However, precipitation outputs from downscaled GCMs may bear systematic errors, such as incorrect estimates of seasonal cycles and precipitation extremes and hence cannot be directly used as input to hydrological models. In this context, bias correction methods are often applied to post-process precipitation outputs from downscaled GCMs to adjust the statistics or distributions of the model output to those of historical observations. Though bias corrected precipitation outputs from downscaled GCMs show great improvements over the raw precipitation outputs in terms of reconstruction of seasonal cycles, extreme precipitation, etc., they might misrepresent the wet and dry spell characteristics of precipitation, which play an important role in climate impact studies on surface water quality, agriculture, domestic water use, etc. Therefore, this study presents a detailed analysis of the wet and dry spell characteristics of bias corrected precipitation outputs in Singapore region from a set of dynamically downscaled GCMs, and compare them with those of the historical observations during the baseline period between 1980 and 2009. The cumulative distribution functions of wet and dry spell characteristics (duration and magnitude), and the contributions of wet or dry spells of certain durations or magnitudes to total duration or total accumulated depth for bias corrected downscaled GCM precipitation simulation and historical observations are analysed and compared. The results show that the bias corrected precipitation simulations from the downscaled GCMs tend to overestimate the duration of wet and dry spells, as well as the magnitude of wet spells, in the upper tail of the distributions. Besides, frequencies and contributions of extreme wet and dry spells are also overestimated. The results have implications for evaluation of climate change impacts on water resources management.