Fan et al. (2025) Considering parameter seasonal variation to enhance process-based ecosystem model performance, evidence from the SWH model

Identification

• Journal: Ecological Indicators
• Year: 2025
• Date: 2025-11-29
• Authors: Zeng Fan, Yi Cheng, Jingzhe Wang, Weirong Zhang, Xiaoliang Ma, Qilin Zhu, Yi‐Fei Lu, Kun Zhao, Chuan Jin, Zhongmin Hu
• DOI: 10.1016/j.ecolind.2025.114480

Research Groups

• Hainan Baoting Tropical Rainforest Ecosystem Observation and Research Station, School of Ecology, Hainan University, Haikou, China
• School of Soil and Water Conservation, Beijing Forestry University, Beijing, China
• School of Artificial Intelligence, Shenzhen Polytechnic University, Shenzhen, China
• Key Laboratory of Agro-Forestry Environmental Processes and Ecological Regulation of Hainan Province, School of Environmental Science and Engineering, Hainan University, Haikou, China
• School of Breeding and Multiplication (Sanya Institute of Breeding and Multiplication), Hainan University, Sanya, China

Short Summary

This study demonstrates that incorporating seasonal variation into empirical parameters significantly enhances the performance of the SWH evapotranspiration (ET) partitioning model. A novel Monte Carlo-based calibration scheme with adaptive time windows achieved a 95% success rate and substantially improved R² values compared to traditional methods, approaching the accuracy of Extended Kalman Filtering.

Objective

• To investigate if key empirical parameters in the SWH model reflect ecological traits across sites.
• To identify the mechanisms driving the spatiotemporal variation of these parameters.
• To determine if proposed calibration schemes accounting for seasonal parameter variability can improve the SWH model's performance in simulating ET dynamics.

Study Configuration

• Spatial Scale: Global, utilizing data from 148 FLUXNET sites spanning all vegetated continents and principal terrestrial biomes. A subset of 20 long-term observation sites (with >10 years of continuous data) was used for advanced calibration schemes.
• Temporal Scale: Daily, seasonal, and annual. Parameter calibration was performed on daily-scale ET simulations, with schemes designed to capture intra-annual (seasonal) parameter variations over multiple years of observation.

Methodology and Data

• Models used:
  • SWH model (Shuttleworth-Wallace dual-source model enhanced with Ball-Berry stomatal conductance and Lin-Sun soil surface resistance models).
  • Monte Carlo-based parameter calibration (two-stage scheme).
  • Improved Monte Carlo-based parameter calibration with fixed and adaptive time window grouping strategies.
  • Extended Kalman Filtering (EKF) for dynamic parameter estimation.
• Data sources:
  • Global open-access eddy covariance flux datasets: FLUXNET2015, La Thuile synthesis, ChinaFLUX, and OzFlux (totaling 264 initial sites, refined to 148 after quality control).
  • MODIS LAI (Leaf Area Index) time series.
  • Meteorological and environmental variables (e.g., precipitation, soil moisture, temperature, vapor pressure deficit).

Main Results

• Key SWH model parameters (Prss, Prsc, PEc) effectively characterize ET states, showing distinct variations across ecosystem types and strong correlations with climatic factors (e.g., Prsc positively correlated with mean annual temperature and vapor pressure deficit, negatively with soil water content; PEc positively correlated with mean annual precipitation and LAI).
• These parameters exhibit consistent seasonal patterns, with Prsc typically following a U-shaped trajectory (elevated at year start/end, mid-year decline) and Prss generally showing an inverted U-shaped pattern (peaking mid-year).
• The improved Monte Carlo-based calibration scheme, incorporating adaptive time windows, achieved a 95% success rate in outperforming traditional constant-parameter strategies, with R² improvements of 10–30%.
• At 20 long-term observation sites, the performance of the improved Monte Carlo scheme approached that of the Extended Kalman Filtering (EKF), yielding R² values between 0.83 and 0.98.
• The optimal time window length for parameter calibration varied significantly among ecosystem types, aligning with their specific seasonal characteristics (e.g., smaller windows for grasslands/croplands, larger for deciduous broadleaf forests).

Contributions

• Provides compelling evidence for the necessity of considering seasonal parameter variability in process-based ecosystem models to accurately simulate ET dynamics.
• Develops and validates a practical and theoretically sound Monte Carlo-based parameter calibration scheme that dynamically adjusts parameters using adaptive time windows, significantly enhancing model performance.
• Systematically quantifies the spatiotemporal variations of key SWH model parameters and elucidates their driving environmental and vegetation mechanisms.
• Offers a robust framework that can be applied to improve parameterization and predictive capacity in other ecosystem process models, moving beyond static parameter assumptions.

Funding

• National Natural Science Foundation of China (Grant No. U23A2002; 62472216)
• National Key R&D Program of China (Grant No. 2020YFA0608100)
• Hainan Provincial Natural Science Foundation of China (Grant No. 425QN239; 423RC432)
• U.S.-China Carbon Consortium (USCCC)

Citation

@article{Fan2025Considering,
  author = {Fan, Zeng and Cheng, Yi and Zha, Tianshan and Wang, Jingzhe and Zhang, Weirong and Ma, Xiaoliang and Zhu, Qilin and Lu, Yi‐Fei and Zhao, Kun and Jin, Chuan and Hu, Zhongmin},
  title = {Considering parameter seasonal variation to enhance process-based ecosystem model performance, evidence from the SWH model},
  journal = {Ecological Indicators},
  year = {2025},
  doi = {10.1016/j.ecolind.2025.114480},
  url = {https://doi.org/10.1016/j.ecolind.2025.114480}
}