1 AIT Asian Institute of Technology

Evaluation of process-based and data-driven approaches for flood hazard assessement in the Upper Mun River Basin, Thailand

AuthorShrestha, Pooja
Call NumberAIT Thesis no.WM-23-03
Subject(s)Hydraulic models--Thailand--Upper Mun River Basin
Hydraulic engineering--Data processing
Neural networks (Computer science)--Thailand--Upper Mun River Basin--Databases
Flood forecasting--Thailand--Upper Mun River Basin
NoteA thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering in Water Engineering and Management
PublisherAsian Institute of Technology
AbstractFlood modelling is a key tool in flood risk management and selection of the proper flood modelling approach is important for the efficient utilization of time and resources, without compromising the output accuracy. This study evaluates the performance of two flood modelling approaches: a hydraulic modelling software, and supervised machine learning model. The evaluation has been done in terms of model accuracy, predictive capability, and characteristics of the models. The flood level and extent area of two of the extreme flood events (2011 and 2017 A.D.) in the Mun river basin in northeast Thailand were simulated using two types of models HECRAS and ANN. 2D- HEC-RAS model was developed using digital elevation model as the terrain model and flow hydrograph and stage hydrograph as the boundary conditions while ANN was trained using the past flood data, hydrometeorological data and basin properties to predict the flood level and flood extent area. Both HEC-RAS and ANN model exhibited a tendency to underpredict water levels in the river during flood. In terms of flood extent area, ANN-simulated flood inundation area was 19% lower than the observed extent area, while HEC-RAS’s result was closer to the observed area. The predictive capability of the models was quantified through the estimation of correctly forecasted flooded/non-flooded points. ANN demonstrated 8% higher accuracy than that of HEC-RAS in predicting both inundated and dry areas. ANN has been used for the first time in the Mun River basin for flood hazard assessment, and the results justify its efficiency and suitability for further use in this area. However, it is important to note that the choice of model depends on the specific requirements, data availability, and expertise of the stakeholders involved. The selection of either model is not a mutually exclusive choice, both models can be combined to utilize their respective strengths and achieve a more comprehensive understanding of the complex nature of flood.
Year2023
TypeThesis
SchoolSchool of Engineering and Technology (SET)
DepartmentDepartment of Civil and Infrastucture Engineering (DCIE)
Academic Program/FoSWater Engineering and Management (WM)
Chairperson(s)Babel, Mukand Singh
Examination Committee(s)Shrestha, Sangam;Shanmugam, Mohana Sundaram;Ho Huu Loc
Scholarship Donor(s)Asian Institute of Technology Scholarships
DegreeThesis (M. Eng.) - Asian Institute of Technology, 2023


Usage Metrics
View Detail0
Read PDF0
Download PDF0