Xu et al. (2026) Representing dynamic grassland density in the land surface model ORCHIDEE r9010

Identification

• Journal: Geoscientific model development
• Year: 2026
• Date: 2026-01-05
• Authors: Siqing Xu, Sebastiaan Luyssaert, Yves Balkanski, P. Ciais, N. Viovy, Liang Wan, Jean Sciare
• DOI: 10.5194/gmd-19-1-2026

Research Groups

• Laboratoire des Sciences du Climat et de l’Environnement (LSCE), CEA, CNRS, UVSQ, Université Paris-Saclay, Gif-sur-Yvette, France
• The Cyprus Institute, Climate and Atmosphere Research Center (CARE-C), Nicosia, Cyprus
• Amsterdam Institute for Life and Environment, Department of Ecological Science, Vrije Universiteit Amsterdam, Amsterdam, the Netherlands
• Research Institute of Agriculture and Life Sciences, Seoul National University, Seoul, Republic of Korea

Short Summary

This study introduces a dynamic grassland density approach in the ORCHIDEE land surface model, linking density to plant carbon status to address limitations of a fixed density assumption. The new approach significantly reduces unrealistic mortality events, improves Leaf Area Index (LAI) simulation, and better captures grassland responses to environmental variability, particularly in semi-arid regions.

Objective

• To simulate dynamic grassland density in the ORCHIDEE land surface model based on vegetation growth indicators (reserve and labile carbon content).
• To simulate the response of the bare soil fraction in grasslands to environmental changes, providing a foundation for predicting long-term spatiotemporal dust emissions.
• To enhance grassland survival, particularly in semi-arid regions, by reducing unrealistic mortality events.
• To better represent the grassland Leaf Area Index (LAI), a key variable for land-atmosphere processes like photosynthesis, transpiration, albedo, and the energy budget.

Study Configuration

• Spatial Scale: Global simulations at 2° × 2° resolution; regional field-based comparisons; satellite data regridded to 2° × 2°.
• Temporal Scale: 200-year spin-up; 51-year simulation period for mortality analysis; 17-year averaging period (2004–2020) for grassland density and mean annual LAI.

Methodology and Data

• Models used: ORCHIDEE (ORganizing Carbon and Hydrology In Dynamic EcosystEms) trunk version r9010, part of the IPSL-CM Earth System Model.
• Data sources:
  • Climate Forcing: CRU-JRA climate forcing data (2004–2020).
  • Land Cover: ESA CCI Land Cover dataset (2004) for Plant Functional Type (PFT) maps.
  • Field Observations: Regional field-based estimates of grassland density from five published case studies (France, Mongolian Plateau, USA, Senegal, Australia).
  • Satellite Observations (Fractional Vegetation Cover - FVC): Copernicus Land Monitoring Service FCOVER product (2004, regridded from ~0.003° to 2° × 2°).
  • Satellite Observations (Leaf Area Index - LAI):
    • MODIS LAI (2004–2020), 1 km resolution, 4 d temporal frequency (regridded to 2° × 2°).
    • Sentinel-2 LAI (2019), 10 m spatial resolution (aggregated to 2° × 2°).
  • Ancillary Data: Zomer et al. (2022) aridity index map; MCD12Q1 and GLC_FCS30D land cover products for LAI filtering.

Main Results

• The dynamic density approach simulates grassland density globally within a range of 0.05 to 1, with 56% of temperate C3, 66% of C4, and 33% of tropical C3 grasslands maintaining maximum density.
• Simulated grassland density showed consistency with field-based estimates across five regional case studies (e.g., 0.95 simulated vs. 0.91–0.99 observed in France; 0.63 simulated vs. 0.68 observed in USA).
• The dynamic density approach improved the spatial correlation (Pearson's r from 0.11 to 0.24) and reduced the Root Mean Square Error (RMSE from 0.26 to 0.22) for fractional vegetation cover (FVC) compared to satellite FCOVER.
• A positive correlation emerged between precipitation and simulated grassland density for all three grass PFTs, with peak probability density at 0.9–1 density and 0.5–4 mm d−1 precipitation.
• Mortality events were significantly reduced: 98% of temperate C3, 97% of C4, and 99% of tropical C3 grassland grid cells experienced reduced mortality over a 51-year simulation period.
• The aridity threshold for frequent mortality (≥5 events over 51 years) increased from 0.3 (fixed density) to 0.7 for C4 grasslands and 0.9 for temperate and tropical C3 grasslands with the dynamic approach.
• Globally, the dynamic density approach modestly improved LAI simulation, increasing Pearson's r from 0.51 to 0.56 and decreasing RMSE from 0.60 to 0.59 compared to MODIS LAI.
• Regionally, the coefficient of determination (R2) for mean annual LAI improved or remained unchanged, and RMSE decreased in three of four semi-arid regions (Australia, Central Asia, southern Africa). Seasonal LAI correlation (r) with MODIS improved in southern Africa from 0.77 to 0.93.
• For the majority of grid cells (84% temperate C3, 81% C4, 75% tropical C3), the reduction in LAI was less pronounced than the reduction in mortality events, indicating a beneficial trade-off.
• 97% of the remaining mortality events in the dynamic approach occurred in regions identified as unsuitable for grassland survival (hyper-arid, LAI < 0.1, or aridity > 0.83), suggesting potential errors in the PFT maps.

Contributions

• Introduces a novel, computationally efficient physiology-based approach to simulate dynamic grassland density in ORCHIDEE r9010, linking it directly to plant carbon status (reserve and labile carbon).
• Significantly improves the representation of bare soil fraction within grasslands and their dynamic response to environmental changes, laying a foundation for more accurate dust emission estimates and land-atmosphere feedback studies.
• Enhances the ecological realism of the model by substantially reducing unrealistic frequent mortality events in grasslands, particularly in resource-limited semi-arid regions.
• Improves the simulation of grassland Leaf Area Index (LAI) globally and regionally, which is a critical variable for accurately modeling land-atmosphere processes.
• Demonstrates that the observed positive correlation between precipitation and grassland density emerges as a natural property of the model's dynamics, rather than being explicitly prescribed.
• Eliminates the need for arbitrary bare soil fractions in PFT maps by allowing the bare soil fraction to emerge dynamically from the simulation, thereby improving the consistency and realism of land cover representation.

Funding

• National Research Agency (Agence Nationale de la Recherche) under the France 2030 program (reference ANR-22-EXTR-0009).
• European Union’s Horizon Europe research and innovation program under Grant Agreement No. 101071247 (Edu4Climate – European Higher Education Institutions Network for Climate and Atmospheric Sciences).
• HPC resources from GENCi-TGCC on grant 06328 (2023–2025).

Citation

@article{Xu2026Representing,
  author = {Xu, Siqing and Luyssaert, Sebastiaan and Balkanski, Yves and Ciais, P. and Viovy, N. and Wan, Liang and Sciare, Jean},
  title = {Representing dynamic grassland density in the land surface model ORCHIDEE r9010},
  journal = {Geoscientific model development},
  year = {2026},
  doi = {10.5194/gmd-19-1-2026},
  url = {https://doi.org/10.5194/gmd-19-1-2026}
}