Forecasting Flood Levels and Areal Inundation in Downstream Mekong Using Remotely Sensed Data and Modeling
Inhabitants in the Mekong River Basin (MRB) floodplain, mainly inside Cambodia and Vietnam, have been extensively relying on resources from the river as their major food and income source by developing floodplain agriculture and freshwater fishery. However, despite the potential impact of increasing extents of climate and anthropogenic changes on the MR hydrology, which can particularly influence the livelihoods of people in the downstream MRB floodplain, there is no publicly and routinely issued water level forecast inside the Mekong Delta (MD), neither inundation extent forecast for the whole MRB floodplain. This may be because of (1) the concerns of heavy computational burden and limited accuracy of conventional approaches due to the complex hydraulic conditions and flat terrain in the region, and (2) less effective data exchange between countries due to geopolitical barriers. This dissertation introduces a question: How can we build skillful, computationally efficient, and sustainable water level and inundation extent forecasting systems for the MRB, specifically for the downstream floodplain? To answer the question, we have proposed: Chapter 3 - A freely accessible, computationally efficient daily water level forecasting system for the MR, which particularly addresses the challenges in the MD region; Chapter 4 - A satellite imagery-based inundation extent forecasting framework, called Forecasting Inundation Extents using Rotated empirical orthogonal function analysis (FIER), with the Tonle Sap Lake Floodplain as a test bed. It allows quick and continuous estimation of inundation extents of any time with available hydrological data and also addresses the concerns of heavy computational burden and extreme overestimation issues in the conventional inundation forecasting approaches. In Chapter 5, we further implemented FIER to the whole MRB floodplain, where conventional inundation extent forecasting approaches are quite challenging to be applied. The FIER pseudo-forecasted inundation extents were then applied to spatially predict flood hazard levels and rice paddies at risk which can serve as references for local stakeholders to do more efficient decision making for better flood damage containment. The systems we developed utilize remote sensing data and are based on computationally efficient methods and can be easily implemented on cloud-based platform with enhanced scalability and accessibility.