Unified multi-task learning for hydrological processes using a shared transformer framework
Bekir Z Demiray, Ibrahim Demir · 2026
Read Unified multi-task learning for hydrological processes using a shared transformer framework by Bekir Z Demiray, Ibrahim Demir (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/PMC13507253.txt.
Description
Bekir Z Demiray and Ibrahim Demir present, in 2026 work, one shared transformer model for hydrology. It handles two jobs at once: forecasting river flow a day ahead and sharpening rainfall records to finer time steps. Most studies train separate models per variable, wasting shared structure between rain and runoff. Their design uses one encoder with a dedicated decoder per task.
Joint training improved flow forecast accuracy versus training the same model alone, while rainfall reconstruction stayed comparable. Outputs remained stable and physically sensible across tests. A unified, process-aware architecture could simplify forecasting centers that now maintain many single-purpose models. Flood warning and reservoir planning both depend on exactly these paired predictions.
Catalog record
| CREATOR | Bekir Z Demiray, Ibrahim Demir |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 61913 |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |