Otryakhin et al. (2025) Comparison of simulations from a state-of-the-art dynamic global vegetation model (LPJ-GUESS) driven by low- and high-resolution climate data

Identification

• Journal: Geoscientific model development
• Year: 2025
• Date: 2025-11-27
• Authors: Dmitry Otryakhin, David Martín Belda, Almut Arneth
• DOI: 10.5194/gmd-18-9101-2025

Research Groups

• Institute of Meteorology and Climate Research Atmospheric Environmental Research IMK-IFU, Karlsruhe Institute of Technology (KIT), Germany

Short Summary

This study explores the differences in vegetation outcomes from the dynamic global vegetation model LPJ-GUESS when driven by high- (0.05°) versus low-resolution (0.5°) climate data across Europe. It reveals significant systematic discrepancies, particularly in mountainous regions where high-resolution simulations show substantially smaller carbon pools (e.g., total carbon 37 %–39 % smaller), and quantifies the impact of under-represented orographic climate variability and shoreline features on regional predictions.

Objective

• To explore and quantify the systematic differences in vegetation outcomes between LPJ-GUESS simulations conducted at high (0.05°) and low (0.5°) spatial resolutions, specifically focusing on the impact of under-represented orographic climate variability.
• To identify and decompose the sources of bias in low-resolution LPJ-GUESS outputs across the European Union, including climate-response bias and shoreline-representation bias.

Study Configuration

• Spatial Scale:
  • Domain: Continental European Union plus Norway, Iceland, Switzerland, the United Kingdom, non-EU Balkan states, Moldova, Belarus, and parts of Ukraine, Russia, Morocco, Algeria, and Tunisia (λ ∈[26.75°W, 35.25°E]; φ ∈[34.75°N, 71.75°N]).
  • Resolutions: 0.5° (low resolution) and 0.05° (high resolution, upscaled from 0.0083(3)°).
  • Study region for ensemble experiments: Alps (high elevation variability).
  • Control region for ensemble experiments: Between Dinaric Alps and Carpathian Mountains (low elevation variability).
• Temporal Scale:
  • Climate data availability: 1850–2100 (historical, SSP1-2.6, SSP3-7.0, SSP5-8.5 scenarios).
  • Ensemble experiments: Historical period 1850–2014.
  • Europe-wide simulations: Averaged over 2010–2014.

Methodology and Data

• Models used:
  • Dynamic Global Vegetation Model (DGVM): LPJ-GUESS (European Applications branch, v4.0, r9710), including the SIMFIRE-BLAZE fire submodel.
  • Downscaling algorithm: CHELSA V2.1 (semi-mechanistic, terrain-informed algorithm for temperature, precipitation, and downwelling shortwave radiation).
  • Statistical test: Bootstrap two-sample heterogenic location test.
• Data sources:
  • Low-resolution climate forcing: ISIMIP3b (0.5°).
  • High-resolution climate forcing: CHELSA-downscaled ISIMIP3b (0.0083(3)° then upscaled to 0.05°).
  • 3D data for CHELSA: CMIP6 ensemble (MPI-ESM1-2-HR).
  • Static elevation data: Global Multi-resolution Terrain Elevation Data 2010 (GMTED2010) at 30 arcsec resolution.
  • Soil properties: Digitized Soil Map of the World.
  • Nitrogen deposition: Historical ISIMIP nitrogen deposition data (for Europe-wide simulations); constant pre-industrial rate of 2 kgN ha−1 yr−1 (for ensemble experiments).

Main Results

• In mountainous regions (Alps), LPJ-GUESS simulations show statistically significant differences between high- and low-resolution outputs for all considered ecosystem variables. High-resolution simulations consistently predict smaller carbon pools (e.g., total carbon pool 37 %–39 % smaller) and lower productivity (e.g., GPP -23.4 %, NPP -25.5 %) compared to low-resolution simulations.
• In a control region with low elevation variability, differences between resolutions were substantially smaller, with some variables showing no statistically significant differences.
• Europe-wide simulations reveal a total bias in low-resolution aggregated ecosystem variables ranging from -7.3 % (Net Ecosystem Productivity, NEP) to 6.6 % (Litter carbon, Clit).
• This total bias is decomposed into two main components:
  • Climate-response bias: Arising from the non-linear response of the model to orographic climate variability, ranging from -3.8 % (NEP) to 2.9 % (Clit, Soil carbon, Csoil).
  • Shoreline-representation bias: Due to the coarse grid's inability to accurately represent coastlines and inland water bodies, ranging from -3.5 % (NEP) to 4.1 % (Runoff, Roff).
• The shoreline-representation bias was found to be larger in magnitude than the climate-response bias for all variables in the Europe-wide domain.
• The non-conservative properties of the CHELSA downscaling (for temperature and radiation) introduce a bias comparable to altitude-related differences in low-variability regions but are less impactful in mountainous areas.
• Fire disturbance contributes to ensemble mean differences in the control region but has a negligible influence compared to orography-induced climate differences in the study region.

Contributions

• Development and public release of a novel high-resolution (0.05°) climate dataset for Europe (1850–2100) derived from ISIMIP3b data using the CHELSA downscaling algorithm, optimized for ecosystem modeling applications.
• Comprehensive quantification and attribution of systematic biases in LPJ-GUESS simulations caused by coarse spatial resolution, specifically highlighting the critical role of orographic climate variability and shoreline representation.
• Demonstration that the impact of spatial resolution on model outputs is highly region-dependent, with effects up to ~46 % in mountainous areas compared to ~3 % at the European scale.
• Provision of a robust statistical methodology (bootstrap two-sample heterogenic location test) for comparing model outputs with different variances, applicable to other model intercomparison studies.
• Elucidation of the underlying non-linear mechanisms (e.g., growing season duration, photosynthetic rates, soil water dynamics, radiation distribution) through which climate drivers and orography influence vegetation dynamics, suggesting broader applicability to other Dynamic Global Vegetation Models.

Funding

• The article processing charges for this open-access publication were covered by the Karlsruhe Institute of Technology (KIT).

Citation

@article{Otryakhin2025Comparison,
  author = {Otryakhin, Dmitry and Belda, David Martín and Arneth, Almut},
  title = {Comparison of simulations from a state-of-the-art dynamic global vegetation model (LPJ-GUESS) driven by low- and high-resolution climate data},
  journal = {Geoscientific model development},
  year = {2025},
  doi = {10.5194/gmd-18-9101-2025},
  url = {https://doi.org/10.5194/gmd-18-9101-2025}
}