Zhou et al. (2026) Synergistic retrievals of leaf area index and leaf chlorophyll content in deciduous broadleaf forests from Sentinel-2 and Landsat

Identification

• Journal: Remote Sensing of Environment
• Year: 2026
• Date: 2026-03-25
• Authors: Haoqiang Zhou, Mingzhu Xu, Jing M. Chen, Xingchang Wang, Rong Shang, R.Z. Wang, Yulin Yan, Jiao Wang
• DOI: 10.1016/j.rse.2026.115382

Research Groups

• Key Laboratory of Humid Subtropical Eco-Geographical Process of Ministry of Education, School of Geographical Sciences, Fujian Normal University, Fuzhou, China
• Department of Geography and Program in Planning, University of Toronto, Toronto, ON, Canada
• School of Ecology, Key Laboratory of Sustainable Forest Ecosystem Management − Ministry of Education, Northeast Forestry University, Harbin, China
• College of JunCao Science and Ecology (College of Carbon Neutrality), Fujian Agriculture and Forestry University, Fuzhou, China

Short Summary

This study systematically evaluates synergistic Leaf Area Index (LAI) and Leaf Chlorophyll Content (LCC) retrievals for deciduous broadleaf forests from Sentinel-2 and Landsat data. It identifies limitations in canopy structural representation as a primary driver of mutual error compensation and demonstrates that integrated parameterization strategies significantly improve retrieval accuracy and seasonal dynamics.

Objective

• To evaluate how the choice of a radiative transfer (RT) model and key parameterization strategies influence the accuracy of synergistic LAI and LCC retrievals and their characteristic error-compensation effects.
• To benchmark the performance of the optimal retrieval scheme identified in this study against existing operational algorithms (SL2P and SL2P-CCRS).
• To compare the performance of LAI and LCC retrievals derived from Sentinel-2 and Landsat, and to examine the influence of the retrieval algorithm.
• To investigate the impact of leaf spectral parameterization on the accuracy of retrieved seasonal LAI dynamics.

Study Configuration

• Spatial Scale: Field measurements from 9 elementary sampling units (ESUs) of 20 m × 30 m at Maoershan site (China) and 213 ESUs of 40 m × 40 m across 22 NEON deciduous broadleaf forest (DBF) sites (North America). Satellite reflectance extracted from 30 m × 30 m areas.
• Temporal Scale: LAI time series from 2016–2018 (Maoershan) and 2014–2022 (NEON). LCC measurements from 2018 (Maoershan, seasonal) and 2017–2022 (NEON, peak season). Satellite imagery from Sentinel-2 (September 2015 onwards) and Landsat-7/8 (2013–2022), selected within ±7 days of field sampling.

Methodology and Data

• Models used:
  • Radiative Transfer (RT) Models: PROSAIL (1D model, 4SAIL + PROSPECT-D), 5-Scale (geometric-optical RT model, 4-Scale + PROSPECT-D).
  • Soil Reflectance Model: Brightness-Shape-Moisture (BSM).
  • Parameterization Strategies: Pure Vegetation Fraction (FV), Canopy Non-Photosynthetic Component Fraction (FS), Prior-Knowledge (PK) constraint based on LAI-vegetation index relationship.
  • Comparison Algorithms: SL2P (PROSAIL-based neural network), SL2P-CCRS (4SAIL2-based neural network).
  • Inversion Method: Look-Up Table (LUT) method using Root Mean Square Relative Error (RMSRE) cost function.
• Data sources:
  • Satellite: Sentinel-2 L1C Top-Of-Atmosphere (TOA) reflectance (10 m resolution), Landsat-7/8 Level-2 Collection 2 Tier 1 Top-Of-Canopy (TOC) reflectance (30 m resolution).
  • Ground measurements:
    • Maoershan site: Synchronous LAI (litterfall collection, leaf seasonality modeling) and LCC (spectrophotometric analysis).
    • NEON sites: LAI (Digital Hemispherical Photographs, DHPs), LCC (Plant Foliar Traits product).
  • Ancillary data: MODIS Level-3 atmosphere products, NCEP/NCAR reanalysis, TOMS/OMI ozone dataset for atmospheric correction.
  • Prior knowledge: Empirical relationship between Fractional Vegetation Cover (FVC) and Plant Area Index (PAI) derived from NEON data.

Main Results

• Limitations in canopy structural representation of RT models (e.g., 1D PROSAIL) are a primary driver of mutual error compensation between LAI and LCC retrievals.
• Integrated parameterization (FV + FS + PK for PROSAIL; FS + PK for 5-Scale) enables more balanced and accurate synergistic retrieval, improving LAI and LCC accuracy by 24.0–46.3% and 12.0–45.4%, respectively, across sites and models. Simpler strategies tend to improve one parameter at the expense of the other.
• Both enhanced models (PROSAIL (FV + FS + PK) and 5-Scale (FS + PK)) show higher retrieval accuracy for LAI and LCC than the operational SL2P algorithm.
• Relative to SL2P-CCRS, 5-Scale (FS + PK) achieves comparable LAI accuracy but more accurate LCC, particularly at low and high LCC extremes, and better captures seasonal dynamics. SL2P-CCRS exhibited a practical lower bound for LCC estimates near 20 μg cm−2.
• The integrated parameterizations reduce discrepancies between Sentinel-2 and Landsat-7/8 retrievals, supporting the feasibility of combining these sensors for continuous time series.
• An advanced leaf-spectra parameterization, accounting for all canopy materials (green and non-photosynthetic components), is essential for capturing seasonal LAI dynamics, especially during autumn senescence. The 5-Scale (FS) algorithm showed a mean difference of -3 ± 9 days for the middle of senescence (MOS) compared to observed, significantly better than fixed-leaf spectral (-24 ± 16 days) and baseline (-11 ± 9 days) algorithms.

Contributions

• Systematically evaluated the influence of RT model choice (1D vs. geometric-optical) and parameterization strategies (FV, FS, PK) on synergistic LAI and LCC retrieval accuracy and error compensation in deciduous broadleaf forests.
• Identified canopy structural representation as a critical factor driving LAI-LCC mutual error compensation.
• Developed and validated enhanced physically based algorithms (e.g., 5-Scale (FS + PK)) that achieve more balanced and accurate synergistic LAI and LCC retrievals, outperforming existing operational algorithms (SL2P, SL2P-CCRS) in terms of LCC range and seasonal dynamics.
• Demonstrated the importance of explicitly accounting for non-photosynthetic components and incorporating prior knowledge to develop robust retrieval algorithms and improve cross-sensor consistency between Sentinel-2 and Landsat.
• Highlighted the necessity of dynamic leaf spectral parameterization for accurately capturing seasonal LAI dynamics, particularly during autumn senescence.
• Provides a robust framework for monitoring forest ecosystem functioning using decametric-resolution satellite observations, with implications for future algorithm development and operational product generation.

Funding

• National Natural Science Foundation of China (42201360, U23A2002, and 42471356)
• Natural Science Foundation of Fujian Province (2025J01646)
• U.S. National Science Foundation (through the NEON Program)

Citation

@article{Zhou2026Synergistic,
  author = {Zhou, Haoqiang and Xu, Mingzhu and Chen, Jing M. and Wang, Xingchang and Shang, Rong and Wang, R.Z. and Yan, Yulin and Wang, Jiao},
  title = {Synergistic retrievals of leaf area index and leaf chlorophyll content in deciduous broadleaf forests from Sentinel-2 and Landsat},
  journal = {Remote Sensing of Environment},
  year = {2026},
  doi = {10.1016/j.rse.2026.115382},
  url = {https://doi.org/10.1016/j.rse.2026.115382}
}