Macro-Scale Flow Modelling of the Mekong River with Spatial Variance

This dissertation, "Macro-scale Flow Modelling of the Mekong River With Spatial Variance" by Ying, Tian, 田英, was obtained from The University of Hong Kong (Pokfulam, Hong Kong) and is being sold pursuant to Creative Commons: Attribution 3.0 Hong Kong License. The content of this dissertation has not been altered in any way. We have altered the formatting in order to facilitate the ease of printing and reading of the dissertation. All rights not granted by the above license are retained by the author. Abstract: Abstract of thesis entitled Macro-scale flow modelling of the Mekong River with spatial variance submitted by Ying Tian for the degree of Doctor of Philosophy at The University of Hong Kong in June 2007 The Mekong River is one of the largest and most important international rivers in the world. River flow modelling and forecasting at high time resolution are important for mitigating flooding consequences. A parametric model using the Artificial Neural Network (ANN) approach and a conceptual rainfall-runoff model (CRR) using the Variable Infiltration Capacity (VIC) model have been applied to model the daily flow of the Mekong River. The VIC model was calibrated for a macro-scale study after selection of an optimization algorithm. The research results are expected to have some significance for water management for the Mekong River. Daily discharge at the gauging station Pakse was predicted based on the ANN for 1-day, 7-day and 14-day lead-times using the time-lagged daily discharges at Pakse itself and upstream gauging stations. It is shown that the reliability of the prediction accuracy decreases with increasing lead-time. Optimization algorithm comparisons were made between the Shuffled Complex Evolution (SCE-UA) and the Multi-Objective Shuffled Complex Evolution Metropolis (MOSCEM-UA) algorithms. Although not as well as the SCE-UA algorithm on test functions, both the single and multi-objective modes of the MOSCEM-UA algorithm performed well in locating the "true" optimum for the Sacramento Soil Moisture Accounting (SAC-SMA) model and the VIC model using synthetic data. It is further demonstrated that the optimization algorithm in multi-objective mode is more effective in locating the global optimum. Daily flow modelling of the Mekong River was carried out by coupling of the VIC model and linear reservoir routing at grid resolutions ranging from 22 to 0.1250.125. For spatial variance, it is discovered that 22 is too coarse, and model performances at 11 and 0.50.5 are satisfactory after calibration. Integrated consideration of model performance and computation loss and gain, 11 is an effective resolution. Investigation of calibrations on the distributed approach shows that the calibrations with information of interior basin stations improve the simulation performance. It is found that the parameter population optimized for the Upper Mekong is compatible for simulations in the Lower Mekong, and the parameter populations optimized at coarse grid resolutions are applicable to finer grid resolutions but not vice versa. With associated uncertainties, indirect parameter explanations for CRR models help to explain optimization results in terms of parameter changes and model performance improvements. For comparison between the two modelling approaches, the ANN performs better than the conceptual VIC model at Pakse for the Mekong River. Further, a Peak Flow Under- and Overestimation Treatment Algorithm was proposed to improve peak simulation performance, aimed at enabling water managers to be better prepared against related dangers. Embedded in the MOSCEM-UA algorithm, three case studies with increasing complexity show that it is effective for the simple synthetic model and can improve the performance of the SAC-SMA model and the VIC model using historical data. DOI: 10.5353/th_b3873555 Subjects: Streamflow - Mekong