Catchment Regionalization Through The Eyes Of Hydrolstm
Presenter: Luis De la Fuente1
Co-Author(s): Hoshin Gupta
Advisor(s): Dr. Laura E. Condon
1Department of Hydrology and Atmospheric Sciences, University of Arizona
Regionalization is an issue that hydrologists have been working to solve for decades. It is used for example when we transfer parameters from one calibrated model to another, and we search for similarity between gauged to ungauged catchments. However, a unified method of regionalization that can successfully be applied to transfer parameters, and be able to show similarity between different regions while considering the differences in meteorological forcing, catchment attributes, and hydrological responses is still an open research question. Machine learning (ML) has shown promising results in the generalization of its results at temporal and spatial scales for streamflow prediction. That is an indication that ML models have learned some useful relationships of regionalization that we would like to extract. In this study, we explore how the HydroLSTM representation (modification of the traditional Long Short-Term Memory) can learn meaningful relationships between meteorological forcing and catchment attributes. At the same time, these relationships are able to generate different hydrological responses across the US. These results show that regionalization could be finally tackled in the near future by using machine learning models.