Vanderbecken (2026) WP4 - supplementary data - Using deep learning to assimilate sun-induced fluorescence satellite observations in the ISBA land surface model: Datasets

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Open MIND
• Year: 2026
• Date: 2026-02-17
• Authors: Pierre Vanderbecken
• DOI: 10.5281/zenodo.18668100

Research Groups

Not explicitly mentioned in the provided text.

Short Summary

This study utilizes a neural network and the Hyplant model to assimilate TROPOMI SIF and LAI data to improve the simulation of Leaf Area Index (LAI) and Gross Primary Production (GPP) within the Ebro basin.

Objective

• To evaluate the impact of assimilating TROPOMI Solar-Induced Fluorescence (SIF) and LAI (individually or combined) on the accuracy of LAI and GPP simulations.

Study Configuration

• Spatial Scale: Regional (Ebro basin) and Continental (Europe domain).
• Temporal Scale: June 2018 to June 2020 (with specific training/test periods from June 2018 to May 2019 and June 2019 to May 2020, respectively), including data for 2021.

Methodology and Data

• Models used: Hyplant model, Neural Network (for SIF estimation).
• Data sources:
  • TROPOMI (SIF and LAI).
  • ECOCLIMAP-II and ECOCLIMAP-SG (Plant Functional Type [PFT] fractions).
  • Hyplant SIFO2A measurement campaign (aircraft-based SIF measurements).

Main Results

• The study provides comparative simulation results for LAI and GPP based on four assimilation scenarios: TROPOMI SIF only, LAI only, both, or none.
• Performance metrics for the neural network (RMSE, NRMSE, and Pearson correlation) were computed for SIF retrievals across the European domain.
• Analysis increments for LAI were quantified for the year 2021 using a 20% observation error for TROPOMI SIF across 12 distinct PFTs.
• Time-series data were generated for specific coordinates (41.65°N, 1.05°E) comparing "analyse" and "open-loop" simulations for variables including LAI ($\text{m}^2\text{m}^{-2}$), GPP ($\text{mg CO}2\text{m}^{-2}\text{s}^{-1}$), daily cumulated evapotranspiration ($\text{kg H}2\text{O}\text{m}^{-2}\text{s}^{-1}$), and total soil moisture ($\text{m}^3\text{H}_2\text{O}\text{m}^{-3}$).

Contributions

• Development of a data assimilation framework integrating satellite-derived SIF (TROPOMI) via neural networks into the Hyplant model to refine vegetation productivity and structure estimates.
• Provision of a comprehensive dataset including PFT-specific LAI increments and high-resolution aircraft measurement data for validation.

Funding

Not explicitly mentioned in the provided text.

Citation

@article{Vanderbecken2026WP4,
  author = {Vanderbecken, Pierre},
  title = {WP4 - supplementary data - Using deep learning to assimilate sun-induced fluorescence satellite observations in the ISBA land surface model: Datasets},
  journal = {Open MIND},
  year = {2026},
  doi = {10.5281/zenodo.18668100},
  url = {https://doi.org/10.5281/zenodo.18668100}
}