Streamflow Prediction at the Intersection of Physics and Machine Learning: A Case Study of Two Mediterranean‐Climate Watersheds

Abstract Accurate streamflow predictions are essential for water resources management. Recent studies have examined the use of hybrid models that integrate machine learning models with process‐based (PB) hydrologic models to improve streamflow predictions. Yet, there are many open questions regardin...

Full description

Saved in:
Bibliographic Details
Main Authors: S. Adera, D. Bellugi, A. Dhakal, L. Larsen
Format: Article
Language:English
Published: Wiley 2024-07-01
Series:Water Resources Research
Subjects:
Online Access:https://doi.org/10.1029/2023WR035790
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850115229265952768
author S. Adera
D. Bellugi
A. Dhakal
L. Larsen
author_facet S. Adera
D. Bellugi
A. Dhakal
L. Larsen
author_sort S. Adera
collection DOAJ
description Abstract Accurate streamflow predictions are essential for water resources management. Recent studies have examined the use of hybrid models that integrate machine learning models with process‐based (PB) hydrologic models to improve streamflow predictions. Yet, there are many open questions regarding optimal hybrid model construction, especially in Mediterranean‐climate watersheds that experience pronounced wet and dry seasons. In this study, we performed model benchmarking to (a) compare hybrid model performance to PB and machine learning models and (b) examine the sensitivity of hybrid model performance to PB model parameter calibration, structural complexity, and variable selection. Hybrid models were generated by post‐processing process‐based models using Long Short‐Term Memory neural networks. Models were benchmarked within two northern California watersheds that are managed for both municipal water supplies and aquatic habitat. Though model performance varied substantially by watershed and error metric, calibrated hybrid models frequently outperformed both the machine learning model (for 72% of watershed‐model‐metric combinations) and the calibrated process‐based models (for 79% of combinations). Furthermore, hybrid models were relatively insensitive to PB model calibration and structural complexity, but sensitive to PB model variable selection. Our results demonstrate that hybrid models can improve streamflow prediction in Mediterranean‐climate watersheds. Additionally, hybrid model insensitivity to PB model parameter calibration and structural complexity suggests that uncalibrated or less complex PB models could be used in hybrid models without any loss of streamflow prediction accuracy, improving model construction efficiency. Moreover, hybrid model sensitivity to the selection of PB model variables suggests a strategy for diagnosing poorly performing PB model components.
format Article
id doaj-art-687a57e8b14e4a3c8f352c4d96ad6619
institution OA Journals
issn 0043-1397
1944-7973
language English
publishDate 2024-07-01
publisher Wiley
record_format Article
series Water Resources Research
spelling doaj-art-687a57e8b14e4a3c8f352c4d96ad66192025-08-20T02:36:39ZengWileyWater Resources Research0043-13971944-79732024-07-01607n/an/a10.1029/2023WR035790Streamflow Prediction at the Intersection of Physics and Machine Learning: A Case Study of Two Mediterranean‐Climate WatershedsS. Adera0D. Bellugi1A. Dhakal2L. Larsen3University of California‐Berkeley Berkeley CA USAUniversity of California‐Berkeley Berkeley CA USASan Francisco Public Utilities Commission Sunol CA USAUniversity of California‐Berkeley Berkeley CA USAAbstract Accurate streamflow predictions are essential for water resources management. Recent studies have examined the use of hybrid models that integrate machine learning models with process‐based (PB) hydrologic models to improve streamflow predictions. Yet, there are many open questions regarding optimal hybrid model construction, especially in Mediterranean‐climate watersheds that experience pronounced wet and dry seasons. In this study, we performed model benchmarking to (a) compare hybrid model performance to PB and machine learning models and (b) examine the sensitivity of hybrid model performance to PB model parameter calibration, structural complexity, and variable selection. Hybrid models were generated by post‐processing process‐based models using Long Short‐Term Memory neural networks. Models were benchmarked within two northern California watersheds that are managed for both municipal water supplies and aquatic habitat. Though model performance varied substantially by watershed and error metric, calibrated hybrid models frequently outperformed both the machine learning model (for 72% of watershed‐model‐metric combinations) and the calibrated process‐based models (for 79% of combinations). Furthermore, hybrid models were relatively insensitive to PB model calibration and structural complexity, but sensitive to PB model variable selection. Our results demonstrate that hybrid models can improve streamflow prediction in Mediterranean‐climate watersheds. Additionally, hybrid model insensitivity to PB model parameter calibration and structural complexity suggests that uncalibrated or less complex PB models could be used in hybrid models without any loss of streamflow prediction accuracy, improving model construction efficiency. Moreover, hybrid model sensitivity to the selection of PB model variables suggests a strategy for diagnosing poorly performing PB model components.https://doi.org/10.1029/2023WR035790streamflowpost‐processingpredictionbenchmarkingphysics‐informed machine learninghybrid model
spellingShingle S. Adera
D. Bellugi
A. Dhakal
L. Larsen
Streamflow Prediction at the Intersection of Physics and Machine Learning: A Case Study of Two Mediterranean‐Climate Watersheds
Water Resources Research
streamflow
post‐processing
prediction
benchmarking
physics‐informed machine learning
hybrid model
title Streamflow Prediction at the Intersection of Physics and Machine Learning: A Case Study of Two Mediterranean‐Climate Watersheds
title_full Streamflow Prediction at the Intersection of Physics and Machine Learning: A Case Study of Two Mediterranean‐Climate Watersheds
title_fullStr Streamflow Prediction at the Intersection of Physics and Machine Learning: A Case Study of Two Mediterranean‐Climate Watersheds
title_full_unstemmed Streamflow Prediction at the Intersection of Physics and Machine Learning: A Case Study of Two Mediterranean‐Climate Watersheds
title_short Streamflow Prediction at the Intersection of Physics and Machine Learning: A Case Study of Two Mediterranean‐Climate Watersheds
title_sort streamflow prediction at the intersection of physics and machine learning a case study of two mediterranean climate watersheds
topic streamflow
post‐processing
prediction
benchmarking
physics‐informed machine learning
hybrid model
url https://doi.org/10.1029/2023WR035790
work_keys_str_mv AT sadera streamflowpredictionattheintersectionofphysicsandmachinelearningacasestudyoftwomediterraneanclimatewatersheds
AT dbellugi streamflowpredictionattheintersectionofphysicsandmachinelearningacasestudyoftwomediterraneanclimatewatersheds
AT adhakal streamflowpredictionattheintersectionofphysicsandmachinelearningacasestudyoftwomediterraneanclimatewatersheds
AT llarsen streamflowpredictionattheintersectionofphysicsandmachinelearningacasestudyoftwomediterraneanclimatewatersheds