Machine Learning Applications in Hydrology

Machine Learning Applications in Hydrology


JS Rice, JM Vose, and RE Emanuel



Project Goal

The goal of this project was to use machine learning models for two applications in hydrology. First, we used a hierarchical machine learning ensemble model to predict streamflow. Then, we assessed model sensitivies to learn how changes in the percentage of different types of land cover in specific places in the watershed influenced (i.e., increase or decrease) streamflow. Second, we trained and tested a deep neural network (DNN) to convert downscaled, gridded General Circulation Model (GCM) hydroclimatic fluxes to watershed-scale runoff for over 2,700 watersheds across the conterminous United States. We also compared DNN performance to several other emperical grid-to-watershed-scale conversion methods.

Project Links


Thanks to the National Science Foundation and the US Forest Service for supporting this research.

sheila m. saia, PhD
Postdoctoral Researcher


. Improved Accuracy of Watershed-Scale General Circulation Model Runoff Using Deep Neural Networks. JAMES (in review) EarthArxiv (preprint), 2019.

Preprint Project