Four plots comparing the accuracy of predicted latent heat and sensible heat fluxes with observations from flux towers.
Comparing the accuracy of predicted latent heat (evaporation) and sensible heat fluxes with observations from flux towers. NSE and KGE are two alternative measures of accuracy, with a value of 1 meaning a perfect representation of an observed time series. The plots show the cumulative frequency distributions of NSE and KGE. The more the curves are positioned to the right, the better the results. The results show that a deep learning method that exchanges information one-way with the underlying hydrological model (NN1W) does better than the original process-based parameterization (SA), but when the deep learning method is allowed to also use information from the hydrological model itself (NN2W) even better results are obtained. This shows the added value of synergistically using process-based modelling and data-science methods. Credit: Bennett and Nijssen [2021], Figure 3
Source: Water Resources Research

The past few years has seen a surge of papers applying machine learning and deep learning, a particular form of neural networks, to predicting hydrological variables. Although, predictions by deep learning methods are often more accurate than physically based models, they are usually restricted to single components of the hydrological cycle. Bennett and Nijssen [2021] use a component-based hydrological modeling framework to replace a physically based parameterization of turbulent heat fluxes with trained deep learning representations. Evaluation with observations shows that when more information is allowed to exchange between the physically based models and the deep learning methods, predictions are increasingly accurate.

Citation: Bennett, A., & Nijssen, B. [2021]. Deep learned process parameterizations provide better representations of turbulent heat fluxes in hydrologic models. Water Resources Research, 57, e2020WR029328. https://doi.org/10.1029/2020WR029328

—Marc F. P. Bierkens, Editor, Water Resources Research

Text © 2021. The authors. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.