Evaluating Transformer and LSTM Frameworks for Prediction in Ungauged Basins

A study evaluating encoder-only Transformer and LSTM frameworks for streamflow prediction in ungauged basins found that the LSTM outperformed the Transformer across both upstream-only and combined configurations. Incorporating downstream information boosted median Nash-Sutcliffe efficiency (NNSE) by more than 60% for all models. The results indicate that recurrent memory is better aligned with upstream reconstruction tasks, while downstream hydrologic context significantly improves prediction skill across architectures.

arXiv:2606.02791v1 Announce Type: new 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. This study evaluates whether an encoder-only Transformer provides an advantage over an LSTM for upstream streamflow inference under limited hydrologic information, using retrospective simulations from the NOAA National Water Model NWM . Across both upstream-only and combined configurations, the LSTM showed stronger overall performance than the Transformer model across the two configurations. Incorporating downstream information further boosted performance for all models, increasing median NNSE by more than 60%. Rather than treating this as a leaderboard-style comparison, we interpret the experiments as a test of architectural inductive bias for hydrologic sequence inference. The results indicate that recurrent memory remains better aligned with this upstream reconstruction task than an encoder-only Transformer, while downstream hydrologic context provides a strong auxiliary constraint that substantially improves prediction skill across architectures