Uncertainty analysis of Langat river flow projections using impact-based multi-model ensemble approaches

• Main Author: Galavi, Hadi
• Format: Thesis
• Language: English
• Published: 2015
• Online Access: http://psasir.upm.edu.my/id/eprint/58119/1/FK%202015%2088IR.pdf; http://psasir.upm.edu.my/id/eprint/58119/

One of the major manifestations of the climate change impacts in the 21st century in a water catchment is the precipitation—frequency and intensity—pattern alteration that may result in water scarcity. It is important therefore to define the basin-scale hydrologic features under changing/variable climate for sustainable management of water resources. Spatial changes of precipitation frequency and intensity because of climate change may influence the streamflows frequency and magnitude causing intensified floods and droughts and the associated substantial local and regional impacts on the economy. Assessment of climate change hydrological impacts deals with uncertainties resulting from the application of General Circulation Models (GCM), Greenhouse Gasses Emission Scenarios (ES), downscaling methods, and hydrological models, each with their inherent uncertainty. Uncertainty assessment of the climate change impacts on streamflow of the Hulu Langat Basin is the main objective of this study. To this end, the Soil and Water Assessment Tool (SWAT) is used to model the hydrological system of the catchment. It is calibrated based on the historical streamflow data of the catchment. An ensemble of 19 GCMs under two emission scenarios (ES) is used to provide a wide range of possible future climate scenarios. Next, bias-corrected GCM’s precipitation and temperature data were used to run the SWAT model for both the current and future climate. Uncertainty in obtained streamflow scenarios was analyzed with focus on hydrological model parameters, emission scenarios,and GCM uncertainties. This research has modified the existing uncertainty model of Reliability Ensemble Averaging (REA) to be applicable at impact level of climate studies; and a probabilistic ensemble approach that is referred to as Bootstrapped Ensemble Uncertainty Modeling (BEUM) was proposed for uncertainty modeling. In the baseline climate simulations, hydrologic model parameters uncertainty was found to be larger than the emission scenario uncertainty, while GCMs were the largest source of uncertainty. However,parameter uncertainty was the smallest source in future climate periods, while GCMs and emission scenarios were the larger sources with projections of 130% and 51% relative change in annual streamflow, respectively. The projected temporal pattern of monthly streamflow for 2070-2099 under emission scenario of RCP8.5 was found to be different from observed pattern, where the usual first peak flow of the year in April is changed to May and the lowest flow rate happens in February instead of July and August. The temporal change in uncertainty sources may have to be taken into cognizance when implementing water resources projects in the future. Based on the REA method, an approximately 3.5 and 2.9 m3/s increase in mean monthly streamflow during the 2016-2045 period respectively under the emission scenarios of RCP4.5 and RCP8.5, are anticipated. The modification applied to the REA method accommodated the inclusion of hydrological model parameter uncertainty into the total uncertainty assessment. The modified REA method was able to embrace a more reliable prediction interval compared to the original REA. In addition, a full coverage of prediction intervals was possible in the proposed BEUM method, although it proved to be computationally expensive in comparison with the REA method.