loading page

On strictly enforced mass conservation constraints for modeling the rainfall-runoff process
  • +2
  • Jonathan Frame,
  • Frederik Kratzert,
  • Paul Ullrich,
  • Hoshin Gupta,
  • Grey Nearing
Jonathan Frame
University of Alabama

Corresponding Author:jmframe@crimson.ua.edu

Author Profile
Frederik Kratzert
Google
Author Profile
Paul Ullrich
University of California Davis
Author Profile
Hoshin Gupta
University of Arizona
Author Profile
Grey Nearing
Google
Author Profile

Abstract

It has been proposed that conservation laws might not be beneficial for accurate hydrological modeling due to errors in input (precipitation) and target (streamflow) data (particularly at the event time scale), and this might explain why deep learning models (which are not based on enforcing closure) can out-perform catchment-scale conceptual and process-based models at predicting streamflow. We test this hypothesis at the event and multi-year time scale using physics-informed (mass conserving) machine learning and find that: (1) enforcing closure in the rainfall-runoff mass balance does appear to harm the overall skill of hydrological models, (2) deep learning models learn to account for spatiotemporally variable biases in data (3) however this “closure” effect accounts for only a small fraction of the difference in predictive skill between deep learning and conceptual models.
15 Aug 2022Submitted to Hydrological Processes
17 Aug 2022Reviewer(s) Assigned
17 Aug 2022Submission Checks Completed
17 Aug 2022Assigned to Editor
18 Sep 2022Review(s) Completed, Editorial Evaluation Pending
20 Sep 2022Editorial Decision: Revise Major
15 Feb 20231st Revision Received
25 Feb 2023Submission Checks Completed
25 Feb 2023Assigned to Editor
25 Feb 2023Reviewer(s) Assigned
27 Feb 2023Review(s) Completed, Editorial Evaluation Pending
27 Feb 2023Editorial Decision: Accept