Zhang et al. (2026) Integrating GLASS LAI into the SWAT Model for Improved Hydrological Simulation in Semi-Arid Regions

Identification

• Journal: Agronomy
• Year: 2026
• Date: 2026-03-18
• Authors: Xun Zhang, Yu Jiang, Ting Yan, Kun Xie, P. F. Li, Jiping NIU, Kexin Li, Genxu Wang
• DOI: 10.3390/agronomy16060639

Research Groups

• College of Water Resources and Architectural Engineering, Northwest A&F University, Yangling, China
• Key Laboratory of Arid Agricultural Soil and Water Engineering of Ministry of Education, Northwest A&F University, Yangling, China
• Xinjiang Research Institute of Agriculture in Arid Areas, Urumqi, China
• Shaanxi Belt and Road Joint Laboratory of Dryland Biological Resources and Green Smart Agriculture, Northwest A&F University, Yangling, China
• National Key Laboratory of Water Disaster Prevention, Nanjing Hydraulic Research Institute, Nanjing, China
• Research Center for Climate Change, Ministry of Water Resources, Nanjing, China

Short Summary

This study modified the SWAT model by integrating high-resolution Global Land Surface Satellite (GLASS) Leaf Area Index (LAI) data to improve hydrological simulations in the semi-arid Wuding River Basin, significantly enhancing runoff and evapotranspiration accuracy by correcting unrealistic vegetation dynamics.

Objective

• To improve the accuracy of hydrological simulations (runoff and evapotranspiration) in semi-arid regions by integrating high-resolution GLASS LAI data into the SWAT model's source code, thereby correcting its inherent vegetation growth module deficiencies and revealing ecohydrological coupling mechanisms.

Study Configuration

• Spatial Scale: Wuding River Basin, Loess Plateau, China (drainage area: 29,234 km²), delineated into 23 subbasins and 309 Hydrologic Response Units (HRUs). Digital Elevation Model (DEM) resolution: 30 m; Soil data resolution: 1 km; Land use data resolution: 30 m; GLASS LAI data resolution: 250 m (resampled to 30 m).
• Temporal Scale:
  • Warm-up period: 2008
  • Calibration period: 2009–2013
  • Validation period: 2014–2016
  • Total simulation period: 2009–2016 (with meteorological data from 2004–2018)

Methodology and Data

• Models used:
  • Soil and Water Assessment Tool (SWAT) model (SWAT_rev664 version), with source code modifications.
  • Simplified Environmental Policy Impact Climate (EPIC) model (default in SWAT, replaced by GLASS LAI).
  • Fortran-based Dynamic Dimension Search (DDS) algorithm for parameter calibration.
  • SPAW software (version 6.02.70) for calculating soil properties.
• Data sources:
  • Satellite/Remote Sensing:
    • Global Land Surface Satellite (GLASS) Leaf Area Index (LAI) product (daily, 250 m resolution, 2009–2016).
    • Advanced Spaceborne Thermal Emission and Reflection (ASTER) Global Digital Elevation Model (GDEM) (30 m resolution).
    • Chinese Land Cover Dataset (CLCD) (Landsat-based, 30 m annual land cover).
    • MODIS LAI (MCD15A2H) for control/comparison.
    • Multi-source fused Evapotranspiration (ET) dataset (2000–2020) of the Yellow River Basin (0.1° resolution) as reference.
  • Observation/Reanalysis:
    • Daily streamflow observation data (2009–2016) from Baijiachuan hydrological station.
    • China Surface Daily Climate Dataset (National Meteorological Science Data Center): daily precipitation, maximum/minimum temperature, wind speed, relative humidity, and sunshine duration (2004–2018) from seven meteorological stations.
  • Other:
    • Harmonized World Soil Database (HWSD) (1 km spatial resolution).

Main Results

• Runoff Simulation Improvement: The improved model significantly enhanced runoff simulation accuracy. For the calibration period, the coefficient of determination (R²) improved from 0.52 to 0.71, and the Nash-Sutcliffe Efficiency (NSE) improved from 0.52 to 0.7. For the validation period, R² improved from 0.21 to 0.58, and NSE improved from 0.2 to 0.51. The model corrected the underestimation of runoff peaks during extreme precipitation events (e.g., 2013 flood season peak runoff improved from ~150 m³/s to closer to observed ~350 m³/s) and optimized simulations during receding water and dry seasons, correcting the overestimation of base flow.
• LAI Correction: The improved model corrected the unrealistic default LAI peak values (from >5.0 to a physically reasonable range of 1.5–3.0 for major land use types). GLASS LAI data revealed complex, multi-peaked vegetation growth patterns and a longer growing season in semi-arid regions, which the default heat-driven model failed to capture.
• Evapotranspiration (ET) Changes: The multi-year average ET simulated by the improved model increased from 251.7 mm to 341.8 mm. Annual ET for different land use types showed increases (e.g., agricultural land median ET increased by 12.2%, pasture by nearly 20%). The ET pattern shifted from a single peak to a double peak, correcting early-season ET underestimation and summer ET overestimation. The improved model's ET showed a higher R² (0.92) with actual ET compared to the original model (R² = 0.75).
• Water Balance: The multi-year average blue water (total water production) in the watershed decreased from 161.8 mm to 84.1 mm, while the multi-year average percentage of blue water to ET decreased from 63.8% to 24.6%. This indicates a shift towards increased green water (evapotranspiration) due to more realistic vegetation dynamics.
• Phenology: Analysis of GLASS LAI revealed a significant south-east to north-west spatial gradient in vegetation phenology. A basin-wide trend of earlier Start of Season (SOS) (−0.5 days/year) and later End of Season (EOS) (+0.8 days/year) was observed, indicating an overall longer growing season.
• Model Robustness: The improved model demonstrated robust performance under extreme drought conditions (e.g., 2010), correcting the default SWAT model's severe overestimation of dry season base flow (Q95 value improved from 21.4 m³/s to 3.9 m³/s, closer to the measured 3.81 m³/s).

Contributions

• Developed a physically constrained SWAT model for semi-arid regions by integrating high-resolution GLASS LAI data at the source code level, explicitly replacing the empirical heat unit-based growth module.
• Quantitatively assessed how the improved vegetation dynamics altered the tradeoff between blue water (runoff) and green water (evapotranspiration), providing new insights into "vegetation–water–soil" interactions in ecologically fragile semi-arid regions.
• Demonstrated significant improvement in runoff simulation accuracy and correction of systematic biases in ET and LAI dynamics, enhancing the structural robustness of SWAT for water management.
• Implemented a Python-based automated processing workflow for mapping remote sensing LAI data to HRUs and an improved Fortran-DDS algorithm for efficient model calibration.

Funding

• National Key Research and Development Program of China (No. 2023YFC3206504)
• Key Research and Development Program of Shaanxi (Grant Nos. 2024SF-YBXM-533, 2023-YBNY-273, 2023-YBSF-380)
• Shaanxi Province Water Conservancy Science and Technology Project (Grant No. 2024slkj-10)
• Research Project of Shaanxi Laboratory for Arid Region Agriculture (No. 2024-22)
• National Science Foundation of China (Grant No. 52579045)

Citation

@article{Zhang2026Integrating,
  author = {Zhang, Xun and Jiang, Yu and Yan, Ting and Xie, Kun and Li, P. F. and NIU, Jiping and Li, Kexin and Wang, Genxu},
  title = {Integrating GLASS LAI into the SWAT Model for Improved Hydrological Simulation in Semi-Arid Regions},
  journal = {Agronomy},
  year = {2026},
  doi = {10.3390/agronomy16060639},
  url = {https://doi.org/10.3390/agronomy16060639}
}