H038-0013
Forecasting Streamflows in Mekong Delta Using Ensemble Learning Regression
Tuesday, 8 December 2020
Poster
Jinkyoo Choi1, Donghwan Kim2, Hyongki Lee1, Le Thuy Tien Du1, Chi-Hung Chang3, Du Duong Bui4 and Chinaporn Meechaiya5, (1)University of Houston, Department of Civil and Environmental Engineering, Houston, TX, United States, (2)National Center for Airborne Laser Mapping, Houston, TX, United States, (3)University of Houston, Houston, TX, United States, (4)National Center for Water Resources Planning and Investigation, Hanoi, Vietnam, (5)Asian Disaster Preparedness Center, Bangkok, Thailand
Abstract:
Low-lying deltas in Greater Mekong region are densely populated and extensively irrigated. In particular, the Mekong Delta (MD) is the largest delta in Vietnam that covers 39,000 km
2 with population of more than 18 million. It is also Vietnam’s largest rice bowl, making Vietnam as the 2
nd largest rice exporter in the world after Thailand. MD is the only area in the Mekong Basin where farmers can harvest up to seven rice crops every two years. In order to optimize crop yields, water supply to crops are regulated during the wet season while supplemented during the dry season (MRC). Hence, it is imperative to provide accurate and timely information of water availability to inhabitants in MD. However, it is difficult to measure or predict channel discharge downstream from Kratie due to significant movement of water between channels and floodplains and the flow reversal between the Tonle Sap Lake and the channels (MRC, 2005).
A recent innovative approach of estimating historical streamflows in Mekong Delta has been developed using Ensemble Learning Regression (termed ELQ) (Kim et al., RS 2019). ELQ combined base learners presenting hydrological variability over the Mekong mainstem and Tonle Sap Lake and successfully provided accurate discharge estimates validated with in-situ measurements. In this study, we aim at providing both short-term (1 – 16 days ahead) and seasonal (1 – 6 months ahead) forecasts of streamflow in MD, using forecasted discharges over Mekong mainstem (upstream of Kratie) generated from the Hydrological Predictions for the Environment (HYPE) model developed over the Greater Mekong region (Du et al., J Hydrol 2020) and forecasted water levels/inundation extents over the Tonle Sap Lake generated from the Multivariate El Niño/Southern Oscillation (ENSO) Index (MEI) (Chang et al., RSE 2020), combined by ELQ. The forecasted streamflow in MD by ELQ is expected to augment the current effort of “Asean Water Portal” (waterportal.vaci.org.vn) which is “to support water users, practitioners and researchers in understanding character and natural variability of water resources and operational short-term/long-term forecasting for sector-specific water planning in the Greater Mekong region”.
