Application of long short-term memory model in predicting hydrologic extremes under climate change and land use change scenarios in the Lancang-Mekong river basin | |
| Author | Tupaz, Kimberly Torrico |
| Call Number | AIT Thesis no.WM-22-12 |
| Subject(s) | Hydrology--Mekong River Basin Climatic changes--Mekong River Basin |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Water Engineering and Management |
| Publisher | Asian Institute of Technology |
| Abstract | With evidence such as warming of the climate system and intensified hydrologic events, climate change continues to be an important subject in the scientific community. It is said that food security and sustenance would be compromised by climate-related extremes for regions with agriculture and fisheries as dominant industries, such as the Lancang-Mekong River Basin. Recent news has shown the worsening drought and flood conditions in the area with the lowest drop of water level in 2019, attributed to both upstream dam politics and climate change. Hence, the assessment of how future extreme hydrologic conditions change is vital. Modern research on deep learning techniques for time series prediction has been made and for this study, Long Short-Term Memory was selected to forecast future flows under climate change SSP2-4.5 and SSP 5-8.5 scenarios. Prior to future projections, historical assessment of hydrologic extremes on selected stations was conducted using Indicators of Hydrologic Alteration (IHA) which demonstrated increasing trend of minimum flows in the lower Mekong region, validated using Mann-Kendall trend test. Bias correction using empirical quantile mapping was done on rainfall and average temperature variables of the 3 GCMs (EC-Earth 3, EC-Earth 3 Veg, and Nor-ESM 2MM to account for biases between observed (i.e., APHRODITE) and simulated values. Approximately, the largest average annual temperature increase is 0.23˚C/yr and 0.36˚C/yr under SSP2-4.5 and SSP5-8.5, respectively at Stung Treng station. The resulting median of the 3 GCMs were used as inputs to the LSTM model along with observed flow and land use data. Predicted annual average flows exhibit increasing and decreasing trends for the lower stations (i.e., Pakse and Stung Treng) and upper station (i.e., Chiang Saen), respectively. Further analysis shows increasing minimum flows but decreasing maximum flows. |
| Year | 2022 |
| Type | Thesis |
| School | School of Engineering and Technology (SET) |
| Department | Department of Civil and Infrastucture Engineering (DCIE) |
| Academic Program/FoS | Water Engineering and Management (WM) |
| Chairperson(s) | Shrestha,Sangam |
| Examination Committee(s) | Shanmugam, Mohana Sundaram;Ho Huu Loc |
| Scholarship Donor(s) | Luang Prabang Hydroelectric Power/Deedoke Hydroelectric Power Projects;Asian Institute of Technology Scholarships |
| Degree | Thesis (M. Sc.) - Asian Institute of Technology, 2022 |