Evaluating Transformer and LSTM Frameworks for Prediction in Ungauged Basins
This study evaluates the effectiveness of Transformer and LSTM frameworks for predicting streamflow in ungauged basins. The findings indicate that LSTM models outperform Transformer models in upstream streamflow inference, particularly when incorporating downstream information. The research highlights the importance of recurrent memory in hydrologic sequence inference tasks.
- ▪The study compares the performance of Transformer and LSTM models for streamflow prediction in ungauged basins.
- ▪LSTM models demonstrated stronger overall performance than Transformer models across various configurations.
- ▪Incorporating downstream hydrologic information significantly improved prediction accuracy for all models.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2606.02791 (cs) [Submitted on 1 Jun 2026] Title:Evaluating Transformer and LSTM Frameworks for Prediction in Ungauged Basins Authors:Taye Akinrele, James Halgren, Noorbakhsh Amiri Golilarz, Sudip Mittal, Shahram Rahimi View a PDF of the paper titled Evaluating Transformer and LSTM Frameworks for Prediction in Ungauged Basins, by Taye Akinrele and 4 other authors View PDF HTML (experimental) Abstract:Watershed networks exhibit convergent topologies in which multiple tributaries merge into downstream channels,integrating diverse upstream hydrological processes. In ungauged basins, the absence of direct observations increases uncertainty and limits the ability to anticipate extreme events.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
