Colliander et al. (2025) A review of forward modelling and retrieval approaches for forest soil moisture and vegetation optical depth using L-band radiometry

Identification

• Journal: Remote Sensing of Environment
• Year: 2025
• Date: 2025-12-15
• Authors: Andreas Colliander, Mike Schwank, Yiwen Zhou, Mehmet Kurum, Cristina Vittucci, Leung Tsang, Alex Roy, Aaron Berg
• DOI: 10.1016/j.rse.2025.115158

Research Groups

• Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, USA
• Swiss Federal Institute for Forest, Snow and Landscape Research WSL, Birmensdorf, Switzerland
• GAMMA Remote Sensing Research and Consulting AG, Gümligen, Switzerland
• University of Georgia, Athens, GA, USA
• University of Rome Tor Vergata, Rome, Italy
• University of Michigan, Ann Arbor, MI, USA
• Université du Québec à Trois-Rivières, Québec, Canada
• University of Guelph, Guelph, ON, Canada
• Finnish Meteorological Institute, Helsinki, Finland

Short Summary

This review evaluates current L-band radiometry retrieval approaches for forest soil moisture (SM) and vegetation optical depth (L-VOD), highlighting persistent systematic uncertainties and the lack of adequate validation data for forest ecosystems. It emphasizes the need for new algorithms and enhanced validation to fully leverage L-band radiometry for forest monitoring.

Objective

• To review the current state of L-band radiometry-based soil moisture (SM) and vegetation optical depth (L-VOD) retrieval methodologies and their validation for forest environments.
• To identify challenges and future directions for advancing scientifically robust SM, above-ground biomass (AGB), and vegetation water column (VWC) retrievals in forests.

Study Configuration

• Spatial Scale: Global, regional, and local scales, including satellite footprints (e.g., 40 km, 9 km), airborne campaigns, and ground-based observations. Downscaled products to 1 km resolution are also discussed.
• Temporal Scale: Decadal, interannual, multiyear, seasonal, daily, and diurnal variations.

Methodology and Data

• Models used:
  • Microwave Emission Models (MEMs): τ-ω (tau-omega) model, Two-Stream Microwave Emission Model (2S-MEM), L-band Microwave Emission of the Biosphere (L-MEB), Tor Vergata model, Microwave Emission Model for Layered Vegetation (MEMLV).
  • Radiative Transfer (RT) Theory: Discrete Scatterers model, Successive Orders of Scattering (SoS), Albedo Expansion.
  • Multiple Scattering Theory (MST-FL): Foldy-Lax equations, Fast Hybrid Method (FHM).
  • Dielectric Mixing Models: Topp et al. (1980), Mironov et al. (2004, 2009, 2010, 2013, 2019).
  • Roughness Models: HQN model, Shi’s fast-model, Air-to-Soil (A2S) transition model.
  • Snow Models: MEMLS, HUT-nlayers, DMRT-QMS, DMRT-ML, WALOMIS.
  • Retrieval Algorithms: SMOS Level 2/3, SMOS-IC, SMAP Single Channel Algorithm (SCA), SMAP Dual Channel Algorithm (DCA), Modified DCA (MDCA), Regularized DCA (RDCA), Multitemporal DCA (MTDCA), SMAP-IB, Soil Moisture Index (SMI).
  • Machine Learning: Random Forest (RF), Shallow Neural Networks, Deep Learning (Convolutional Neural Networks, Long Short-Term Memory models, ConvLSTM), Hybrid ML, Physics-informed ML.
• Data sources:
  • Satellite Radiometry: SMOS (L-band), SMAP (L-band), AMSR2, FY-3, SSMI, Windsat, AMSR-E (C/X-band).
  • Satellite Lidar: GEDI, ICESat-2.
  • Satellite Optical: MODIS NDVI.
  • Satellite Radar: Sentinel-1, ASCAT.
  • Airborne Campaigns: ESTAR radiometer, PALS, SMAPVEX (2012, 2015, 2016, 2019-2022).
  • Ground-based Observations: ELBARA II L-band radiometer, dielectric probes, dendrometers, sap flow, Global Navigation Satellite System Transmissometry (GNSS-T), terrestrial laser scanners, in situ soil moisture networks (International Soil Moisture Network (ISMN), Soil Climate Analysis Network (SCAN), National Ecological Observatory Network (NEON), Forest Soil Moisture Experiments (FOSMEX), Boreal Ecosystem Research and Monitoring Sites (BERMS), Hydrological Open Air Laboratory (HOAL), BIONTE, Coweeta, Harvard Forest, Millbrook, Sodankylä).
  • Reanalysis/Models: European Centre For Medium-Range Weather Forecasts (ECMWF), ERA-5 Land, Goddard Earth Observing System Model, Version 5 (GMAO GEOS-5).
  • Ancillary Data: Land cover, soil texture, topography, Leaf Area Index (LAI), canopy height, biomass, forest fraction, tree density, clumpiness, phenology, rainfall data.

Main Results

• L-band brightness temperature (TB) demonstrates sensitivity to forest soil moisture (SM) and L-band vegetation optical depth (L-VOD), but current retrieval algorithms often perform suboptimally in forests due to complex canopy structure, heterogeneous organic soils, and parameterization issues.
• Key challenges in forest SM retrieval include significant spatial heterogeneity of SM and vegetation, parameterization of ground roughness and volume scattering, and the effects of organic cover layers, snow, and freeze-thaw cycles.
• L-VOD shows strong spatial correlations (Pearson correlation coefficient R up to 0.94) with forest structural properties like canopy height and above-ground biomass (AGB), often outperforming higher-frequency VODs and optical indices.
• L-VOD is a valuable tool for mapping forest biomass and monitoring slow processes (e.g., deforestation), but its interpretation requires caution at temperatures below 0 °C due to changes in sap water dielectric properties.
• Machine learning approaches, including Random Forest and deep learning models, show promise in improving the spatial resolution and accuracy of SM products in forests, often outperforming traditional methods.
• Validation of forest SM and L-VOD products is severely limited by a lack of suitable ground truth and reference data, particularly for L-VOD, and underdeveloped validation techniques. Sparse in situ networks often introduce representativeness errors.
• New retrieval algorithms are needed to explicitly account for unique forest characteristics, and validation efforts require significant quantitative and qualitative enhancement.

Contributions

• Provides a comprehensive review of L-band radiometry-based soil moisture (SM) and vegetation optical depth (L-VOD) retrieval methodologies, specifically focusing on forest environments.
• Highlights the unique challenges and complexities of applying L-band remote sensing to forests, which are often generalized in broader reviews.
• Evaluates the performance of existing operational retrieval algorithms (e.g., SMOS, SMAP) in forested regions and identifies persistent systematic uncertainties.
• Discusses advanced microwave emission models, such as Multiple Scattering Theory of Foldy-Lax (MST-FL) and Microwave Emission Model for Layered Vegetation (MEMLV), and their potential to improve retrievals by better representing complex forest structures.
• Summarizes the current state of ground truth data availability and validation techniques for forest SM and L-VOD, emphasizing critical gaps and the need for expanded validation infrastructure.
• Offers a clear roadmap for future research directions, including algorithm development, enhanced validation strategies, and improved integration with ancillary datasets (e.g., lidar, optical, meteorological data).

Funding

• National Aeronautics and Space Administration (NASA) through a contract with Jet Propulsion Laboratory, California Institute of Technology.
• SMAP Science Team (NASA).
• Global Navigation Satellite System Research (NASA, reference code: 80NSSC25K7954).
• Remote Sensing Theory programs (NASA).
• European Space Agency (ESA) ESTEC (reference code: 4000137990/22/NL/IA).

Citation

@article{Colliander2025review,
  author = {Colliander, Andreas and Schwank, Mike and Zhou, Yiwen and Kurum, Mehmet and Vittucci, Cristina and Tsang, Leung and Roy, Alex and Berg, Aaron},
  title = {A review of forward modelling and retrieval approaches for forest soil moisture and vegetation optical depth using L-band radiometry},
  journal = {Remote Sensing of Environment},
  year = {2025},
  doi = {10.1016/j.rse.2025.115158},
  url = {https://doi.org/10.1016/j.rse.2025.115158}
}