Self-Organizing Gaussian-Based Downscaling of Climate Data for Simulation of Urban Drainage Systems

Abstract

Climate change and its impacts on hydrometeorological variables and surface runoff have been demonstrated in many investigations around the world. General circulation models (GCMs) are widely used in climate change studies; however, their applications are limited because of low resolution for regional investigations. Different downscaling models have been developed to overcome this shortcoming, but most of them cannot preserve the monthly characteristics of rainfall that are mostly important in water resources planning and management. In this study, a self-organizing Gaussian-based downscaling (SOGDS) scheme is proposed to preserve the monthly characteristics of rainfall variations. The Gaussian mixture distribution (GMD) was employed to determine the monthly rainfall class based on observed climatic predictors. The monthly characteristics of rainfall were estimated, including the number of dry days (days without rainfall) and the maximum number of wet and dry spells, by developing an artificial neural network (ANN) model. The output of the ANN model was applied to a probabilistic scheme to develop a daily rainfall time series. Finally, the generated daily rainfall time series was used to evaluate the performance of a drainage system in the northeastern part of the Tehran metropolitan area in Iran, to consider climate change impacts. The results indicated that the proposed model was capable of downscaling rainfall by preserving its statistical characteristics when compared to a statistical downscaling model (SDSM), which was used as an alternative model. The results indicated that climate change will significantly increase the flood volume/risk in the study region, which should be considered in the future planning of drainage systems in the study area.