Evaluation of Random Forest Regressor as Multiscale Regionalization Scheme for Streamflow Predictions Using Monte Carlo Method
Abstract
Machine learning (ML) algorithms are attractive alternatives to traditional hydrologic modeling for streamflow predictions in ungauged basins (PUB). However, physical processes affecting streamflow generation are inherently variable across multiple spatiotemporal scales and basin characteristics. Therefore, there is a need to investigate the robustness of ML algorithms to capture these variabilities effectively across all reaches in a large watershed. This study aims to evaluate the potential of Random Forest Regressor (RFR) in predicting streamflow across multiple spatial scales using the Monte Carlo method. RFR is implemented at different Hydrologic Unit Code (HUC) scales and the cumulative distribution function (CDF) of model performance is estimated using the Monte Carlo method for resampling training-testing dataset. Comparison of CDFs shows that RFR performance is significantly influenced by the variability in basin characteristics, such as dam density, percentage of urbanized area and averaged elevation. However, the RFR performance is not significantly affected across different spatial scales.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2021
- Bibcode:
- 2021AGUFM.H22G..09L