Optimizing high-resolution climate data with artificial intelligence for enhanced hydrological applications in Thailand | |
| Author | Shakif, Shanjidul Hassan |
| Call Number | AIT Thesis no.WM-26-16 |
| Subject(s) | Artificial intelligence--Environmental aspects--Thailand Climate change mitigation--Technological innovations--Thailand Hydrology--Mathematical models |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering in Water Engineering and Management |
| Publisher | Asian Institute of Technology |
| Abstract | Reliable precipitation information is essential for hydrological modeling, flood forecasting, and water resources planning in Thailand, where monsoon rainfall is highly variable in space and time. However, satellite, reanalysis, and numerical weather prediction products often contain systematic bias, false rainfall detections, and degraded performance during intense wet-season events. This study developed an artificial intelligence-based framework to improve both historical precipitation estimation and medium-range precipitation forecasting for hydrological applications. The research was conducted through three connected objectives. First, a two-stage Double Machine Learning Thailand Precipitation Product (DML-TPP) was developed by combining an XGBoost rain/no-rain classifier with an LSTM rainfall-intensity regressor. The product merged six precipitation inputs, CHIRPS, ERA5-Land, GSMaP, IMERG, PERSIANN-CCS, and PERSIANN-PDIR, with HII gauge observations over Thailand. On the independent 2022-2024 test period, DML-TPP achieved the strongest rainfall event detection performance, with the highest CSI and lowest FAR among all evaluated products, while maintaining competitive continuous-error performance and a balanced bias profile. Second, the same DML concept was extended to bias-correct medium-range NWP precipitation forecasts from GFS, IFS, AIFS, and ICON. The DML-corrected forecast provided the best overall 10-day performance, achieving the lowest RMSE, the highest correlation, and the only positive macro-averaged NSE among the full-horizon products. However, ICON remained highly competitive during its shorter available lead window. Third, the hydrological value of the improved precipitation product was tested using CWatM across major Thai river basins. DML-TPP substantially reduced discharge overestimation compared with raw precipitation products and produced competitive performance for terrestrial water storage, evapotranspiration, and surface soil moisture. Overall, the study demonstrates that AI-based precipitation correction can produce more balanced and hydrologically useful rainfall inputs for Thailand. Its main value lies not only in improving numerical accuracy, but in reducing false rainfall signals and producing more stable forecasting and water-management applications. |
| Year | 2026 |
| Type | Thesis |
| School | Faculty of Civil and Environmental Engineering (2026) |
| Department | Other Field of Studies (No Department) |
| Academic Program/FoS | Water Engineering and Management (WEM) |
| Chairperson(s) | Natthachet Tangdamrongsub; |
| Examination Committee(s) | Hassan, Nurul Husna Binti;Mongkol Ekpanyapong;Shanmugam, Mohana Sundaram; |
| Scholarship Donor(s) | Thai Pipe Scholarship;AIT Scholarship; |
| Degree | Thesis (M. Eng.) - Asian Institute of Technology, 2026 |
 | |