Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Optimizing runoff simulation in three mid-high latitude catchments by integrating terrestrial ecosystem modelling, hybrid machine learning, and causal inference

Zhou, Hao LU orcid ; Tang, Jing LU orcid ; Olin, Stefan LU orcid ; Guo, Renkui LU orcid and Miller, Paul A. LU orcid (2026) In Journal of Hydrology: Regional Studies 63.
Abstract

Abstract Study regionThree mid-high latitude catchments include the Krycklan (Boreal Sweden), the Danube (Central Europe) and the Mississippi (USA).Study focusWe develop a hybrid eco-hydrological framework that couples the process-based terrestrial ecosystem model LPJ-GUESS (Lund-Potsdam-Jena General Ecosystem Simulator) with five machine-learning (ML) algorithms through two hybrid model structures. Monthly runoff is simulated for each catchment and benchmarked against both stand-alone LPJ-GUESS and isolated ML models. Model attribution is carried out with the interpretable ML algorithm SHapley Additive exPlanations (SHAP), while causal forests estimate average treatment effects (ATEs) of key drivers on prediction bias, thereby bridging... (More)

Abstract Study regionThree mid-high latitude catchments include the Krycklan (Boreal Sweden), the Danube (Central Europe) and the Mississippi (USA).Study focusWe develop a hybrid eco-hydrological framework that couples the process-based terrestrial ecosystem model LPJ-GUESS (Lund-Potsdam-Jena General Ecosystem Simulator) with five machine-learning (ML) algorithms through two hybrid model structures. Monthly runoff is simulated for each catchment and benchmarked against both stand-alone LPJ-GUESS and isolated ML models. Model attribution is carried out with the interpretable ML algorithm SHapley Additive exPlanations (SHAP), while causal forests estimate average treatment effects (ATEs) of key drivers on prediction bias, thereby bridging correlation and causation.New hydrological insights for the regionHybrid models raise Nash–Sutcliffe efficiency by 0.4–1.9 relative to original LPJ-GUESS and reduce peak-timing phase bias, while demanding only modest extra computation. SHAP consistently ranks incoming radiation, temperature and precipitation as the leading factors on runoff, but shows that LPJ-GUESS underweights the radiation effect by 11–12 % in Krycklan and Mississippi catchments. Causal-forest inference analysis confirms a strong radiation-driven bias (ATE: –0.13, –0.27 and –0.59 s.d. for Krycklan, Danube and Mississippi catchment, respectively) even after controlling for other drivers. The results demonstrate missing energy-balance processes as an important source of model error and provide quantitative guidance for future model development in cold-region catchments.

(Less)
Please use this url to cite or link to this publication:
author
; ; ; and
organization
publishing date
type
Contribution to journal
publication status
published
subject
keywords
Causal forest inference, Hybrid modelling framework, LPJ-GUESS, Runoff modelling, SHAP analysis
in
Journal of Hydrology: Regional Studies
volume
63
article number
103085
publisher
Elsevier
external identifiers
  • scopus:105034570010
ISSN
2214-5818
DOI
10.1016/j.ejrh.2025.103085
language
English
LU publication?
yes
id
b9a22f6b-77b1-4fbe-88a7-10b7b0184467
date added to LUP
2026-06-12 11:58:35
date last changed
2026-06-12 11:58:57
@article{b9a22f6b-77b1-4fbe-88a7-10b7b0184467,
  abstract     = {{<p>Abstract Study regionThree mid-high latitude catchments include the Krycklan (Boreal Sweden), the Danube (Central Europe) and the Mississippi (USA).Study focusWe develop a hybrid eco-hydrological framework that couples the process-based terrestrial ecosystem model LPJ-GUESS (Lund-Potsdam-Jena General Ecosystem Simulator) with five machine-learning (ML) algorithms through two hybrid model structures. Monthly runoff is simulated for each catchment and benchmarked against both stand-alone LPJ-GUESS and isolated ML models. Model attribution is carried out with the interpretable ML algorithm SHapley Additive exPlanations (SHAP), while causal forests estimate average treatment effects (ATEs) of key drivers on prediction bias, thereby bridging correlation and causation.New hydrological insights for the regionHybrid models raise Nash–Sutcliffe efficiency by 0.4–1.9 relative to original LPJ-GUESS and reduce peak-timing phase bias, while demanding only modest extra computation. SHAP consistently ranks incoming radiation, temperature and precipitation as the leading factors on runoff, but shows that LPJ-GUESS underweights the radiation effect by 11–12 % in Krycklan and Mississippi catchments. Causal-forest inference analysis confirms a strong radiation-driven bias (ATE: –0.13, –0.27 and –0.59 s.d. for Krycklan, Danube and Mississippi catchment, respectively) even after controlling for other drivers. The results demonstrate missing energy-balance processes as an important source of model error and provide quantitative guidance for future model development in cold-region catchments.</p>}},
  author       = {{Zhou, Hao and Tang, Jing and Olin, Stefan and Guo, Renkui and Miller, Paul A.}},
  issn         = {{2214-5818}},
  keywords     = {{Causal forest inference; Hybrid modelling framework; LPJ-GUESS; Runoff modelling; SHAP analysis}},
  language     = {{eng}},
  publisher    = {{Elsevier}},
  series       = {{Journal of Hydrology: Regional Studies}},
  title        = {{Optimizing runoff simulation in three mid-high latitude catchments by integrating terrestrial ecosystem modelling, hybrid machine learning, and causal inference}},
  url          = {{http://dx.doi.org/10.1016/j.ejrh.2025.103085}},
  doi          = {{10.1016/j.ejrh.2025.103085}},
  volume       = {{63}},
  year         = {{2026}},
}