A Mass Conservation Relaxed (MCR) LSTM Model for Streamflow Simulation Across CONUS

Abstract The recent development of the physics‐aware Mass‐Conserving Long Short‐Term Memory network (MC‐LSTM) provides an alternative to other data‐driven Deep Learning (DL) models in hydrology. Mass‐Conserving Long Short‐Term Memory incorporates mass conservation directly into the LSTM architecture...

Full description

Saved in:
Bibliographic Details
Main Authors: Yihan Wang, Lujun Zhang, N. Benjamin Erichson, Tiantian Yang
Format: Article
Language:English
Published: Wiley 2025-08-01
Series:Water Resources Research
Subjects:
Online Access:https://doi.org/10.1029/2024WR039131
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846173982852644864
author Yihan Wang
Lujun Zhang
N. Benjamin Erichson
Tiantian Yang
author_facet Yihan Wang
Lujun Zhang
N. Benjamin Erichson
Tiantian Yang
author_sort Yihan Wang
collection DOAJ
description Abstract The recent development of the physics‐aware Mass‐Conserving Long Short‐Term Memory network (MC‐LSTM) provides an alternative to other data‐driven Deep Learning (DL) models in hydrology. Mass‐Conserving Long Short‐Term Memory incorporates mass conservation directly into the LSTM architecture. Despite the theoretical advancements, studies have reported a surprisingly limited performance of the MC‐LSTM in streamflow simulation. We hypothesize that such a limitation is due to the unrealistic mass conservation scheme in MC‐LSTM, which overlooks unobserved incoming water fluxes beyond precipitation. As an attempt to verify this hypothesis, we propose a Mass Conservation Relaxed LSTM (MCR‐LSTM), which incorporates a bi‐directional mass relaxation (MR) component to account for potential incoming water fluxes beyond precipitation. We train and test the proposed MCR‐LSTM model across 531 watersheds in the contiguous United States (CONUS) against three baseline models: the Sacramento Soil Moisture Accounting, LSTM, and MC‐LSTM. Our results show that MCR‐LSTM outperforms MC‐LSTM despite its underperformance compared to LSTM. Specifically, MCR‐LSTM's advantage over MC‐LSTM is mainly seen in the Plains and Western U.S., where the newly incorporated MR component better simulates water loss and suggests the likely existence of additional incoming water fluxes beyond precipitation, respectively. The novelty and contribution of this study are twofold: firstly, it introduces an alternative physics‐aware DL tool (i.e., MCR‐LSTM) in hydrology with higher accuracy in specific regions compared to MC‐LSTM. Secondly, it provides a diagnosis of regions where strict, precipitation‐based mass conservation constraints may be unrealistic in streamflow simulation.
format Article
id doaj-art-071863f35a3b492baf3d102ec385b66b
institution Kabale University
issn 0043-1397
1944-7973
language English
publishDate 2025-08-01
publisher Wiley
record_format Article
series Water Resources Research
spelling doaj-art-071863f35a3b492baf3d102ec385b66b2025-08-26T12:02:54ZengWileyWater Resources Research0043-13971944-79732025-08-01618n/an/a10.1029/2024WR039131A Mass Conservation Relaxed (MCR) LSTM Model for Streamflow Simulation Across CONUSYihan Wang0Lujun Zhang1N. Benjamin Erichson2Tiantian Yang3School of Civil Engineering and Environmental Science University of Oklahoma Norman OK USASchool of Civil Engineering and Environmental Science University of Oklahoma Norman OK USALawrence Berkeley National Laboratory Berkeley CA USASchool of Civil Engineering and Environmental Science University of Oklahoma Norman OK USAAbstract The recent development of the physics‐aware Mass‐Conserving Long Short‐Term Memory network (MC‐LSTM) provides an alternative to other data‐driven Deep Learning (DL) models in hydrology. Mass‐Conserving Long Short‐Term Memory incorporates mass conservation directly into the LSTM architecture. Despite the theoretical advancements, studies have reported a surprisingly limited performance of the MC‐LSTM in streamflow simulation. We hypothesize that such a limitation is due to the unrealistic mass conservation scheme in MC‐LSTM, which overlooks unobserved incoming water fluxes beyond precipitation. As an attempt to verify this hypothesis, we propose a Mass Conservation Relaxed LSTM (MCR‐LSTM), which incorporates a bi‐directional mass relaxation (MR) component to account for potential incoming water fluxes beyond precipitation. We train and test the proposed MCR‐LSTM model across 531 watersheds in the contiguous United States (CONUS) against three baseline models: the Sacramento Soil Moisture Accounting, LSTM, and MC‐LSTM. Our results show that MCR‐LSTM outperforms MC‐LSTM despite its underperformance compared to LSTM. Specifically, MCR‐LSTM's advantage over MC‐LSTM is mainly seen in the Plains and Western U.S., where the newly incorporated MR component better simulates water loss and suggests the likely existence of additional incoming water fluxes beyond precipitation, respectively. The novelty and contribution of this study are twofold: firstly, it introduces an alternative physics‐aware DL tool (i.e., MCR‐LSTM) in hydrology with higher accuracy in specific regions compared to MC‐LSTM. Secondly, it provides a diagnosis of regions where strict, precipitation‐based mass conservation constraints may be unrealistic in streamflow simulation.https://doi.org/10.1029/2024WR039131deep learningstreamflow simulationinterpretable AI
spellingShingle Yihan Wang
Lujun Zhang
N. Benjamin Erichson
Tiantian Yang
A Mass Conservation Relaxed (MCR) LSTM Model for Streamflow Simulation Across CONUS
Water Resources Research
deep learning
streamflow simulation
interpretable AI
title A Mass Conservation Relaxed (MCR) LSTM Model for Streamflow Simulation Across CONUS
title_full A Mass Conservation Relaxed (MCR) LSTM Model for Streamflow Simulation Across CONUS
title_fullStr A Mass Conservation Relaxed (MCR) LSTM Model for Streamflow Simulation Across CONUS
title_full_unstemmed A Mass Conservation Relaxed (MCR) LSTM Model for Streamflow Simulation Across CONUS
title_short A Mass Conservation Relaxed (MCR) LSTM Model for Streamflow Simulation Across CONUS
title_sort mass conservation relaxed mcr lstm model for streamflow simulation across conus
topic deep learning
streamflow simulation
interpretable AI
url https://doi.org/10.1029/2024WR039131
work_keys_str_mv AT yihanwang amassconservationrelaxedmcrlstmmodelforstreamflowsimulationacrossconus
AT lujunzhang amassconservationrelaxedmcrlstmmodelforstreamflowsimulationacrossconus
AT nbenjaminerichson amassconservationrelaxedmcrlstmmodelforstreamflowsimulationacrossconus
AT tiantianyang amassconservationrelaxedmcrlstmmodelforstreamflowsimulationacrossconus
AT yihanwang massconservationrelaxedmcrlstmmodelforstreamflowsimulationacrossconus
AT lujunzhang massconservationrelaxedmcrlstmmodelforstreamflowsimulationacrossconus
AT nbenjaminerichson massconservationrelaxedmcrlstmmodelforstreamflowsimulationacrossconus
AT tiantianyang massconservationrelaxedmcrlstmmodelforstreamflowsimulationacrossconus