Muhammad Talha AI for hydrology and Earth-system prediction
All research

Oak Ridge National Laboratory · 2026 – present

Ongoing

GNNs for Distributed Streamflow

Graph neural networks over the river network, pretrained on National Water Model simulations and fine-tuned on gauges.

  • Graph neural networks
  • River networks
  • Transfer learning
Code
Satellite view of the Tennessee Valley with TVA reservoirs winding between ridges
Tennessee Valley reservoirs near Knoxville, Landsat 8 OLI, April 10, 2022. NASA Earth Observatory.

This project predicts streamflow at many points along a river network at once. The graph is the river network itself: nodes are reaches and catchments, and edges follow upstream–downstream connectivity, so information propagates the way water does.

Approach

  • Pretraining on National Water Model (NWM) simulations, which cover every reach, then fine-tuning on observed gauge records, which are sparse but accurate.
  • Benchmarking against neural baselines and Tennessee Valley Authority (TVA) operational forecasts.

Related work