Carreiras et al. (2026) Determinants of L-band backscatter in dry tropical ecosystems: Implications for biomass mapping

Identification

• Journal: Remote Sensing of Environment
• Year: 2026
• Date: 2026-01-08
• Authors: João M.B. Carreiras, Thomas Higginbottom, John L. Godlee, S. Harrison, Lorena Benitez, Penelope J. Mograbi, Aurora Levesley, Karina Melgaço, David T. Milodowski, Georgia Pickavance, Geoff Wells, Edmar Almeida de Oliveira, Luzmila Arroyo, Sam Bowers, Roel J. W. Brienen, Domingos Cardoso, Antônio Alberto Jorge Farias Castro, Ezequiel Chavez, Ítalo A.C. Coutinho, Tomas Domingues, Fernando Elias, Mário Marcos Espírito Santo, Ted R. Feldpausch, David Galbraith, Emanuel Gloor, Francisco M.P. Gonçalves, Tatenda Gotore, Francoise Yoko Ishida, Timothy J. Killeen, Yadvinder Malhi, Beatriz S. Marimon, Ben Hur Marimon-junior, Desirée M. Ramos, Simone Matias Reis, Ian McNicol, Edward T.A. Mitchard, Peter W. Moonlight, Paulo S. Morandi, Patricia Morellato, Anderson Muchawona, Jonathan Ilunga Muledi, Alejandro Murakami, Mylor Ngoy Shutcha, Paula Nieto-Quintano, Alexander Parada-Gutierrez, Nayane Cristina Candida dos Santos Prestes, Luciano Paganucci de Queiroz, Priscyla M.S. Rodrigues, Jhonathan Oliveira Silva, Rubens M. Santos, Tiina Särkinen, Domingos Fortunato P.F. Silva, Tony C. de Sousa Oliveira, Marc Steininger, José João Tchamba, Elmar Veenendaal, Débora C. Zuanny, Tim R. Baker, Kyle G. Dexter, Gabriele C. Hegerl, R. Toby Pennington, Oliver L. Phillips, Mathew Williams, Shaun Quegan, Casey M. Ryan
• DOI: 10.1016/j.rse.2025.115213

Research Groups

• University of Sheffield (United Kingdom)
• National Centre for Earth Observation (United Kingdom)
• University of Edinburgh (United Kingdom)
• University of Leeds (United Kingdom)
• University of Exeter (United Kingdom)
• University of Oxford (United Kingdom)
• Socio-Ecological Observatory for Studying African Woodlands (SEOSAW) (South Africa)
• ForestPlots.net (global collaboration, with significant contributions from South American institutions like Universidade do Estado de Mato Grosso, Museo de Historia Natural Noel Kempff Mercado, and UNESP - São Paulo State University)
• Charles Darwin University and Terrestrial Ecosystem Research Network (TERN) (Australia)
• Airbus Defence and Space (United Kingdom)
• Numerous other universities and research institutions across Africa, Australia, Europe, and South America.

Short Summary

This study empirically investigates the determinants of L-band Synthetic Aperture Radar (SAR) backscatter in dry tropical ecosystems across three continents, revealing that stem density is a stronger driver of volume scattering than mean stem biomass, and demonstrating the potential of full-polarimetric SAR to retrieve both for improved aboveground biomass density (AGBD) mapping.

Objective

• To identify the dominant scattering mechanisms driving the L-band SAR signal in dry tropical savannas and woodlands.
• To determine the relative importance of vegetation structure (stem density, mean stem biomass) and soil characteristics (soil water content, soil sand content) in controlling the L-band SAR signal from woody vegetation in the dry tropics.

Study Configuration

• Spatial Scale:
  • Ground Observations: 221 permanent plots (171 in Africa, 6 in Australia, 44 in South America). Most plots are 1 hectare, with some 0.5 hectare plots in the Brazilian Caatinga.
  • SAR Data: 25 meter ground resolution.
  • Soil Moisture: Approximately 10 kilometer spatial resolution (ERA5-Land).
  • Soil Texture: 250 meter grid spacing (ISRIC SoilGrids).
• Temporal Scale:
  • SAR Data: ALOS-1 PALSAR-1 (2006–2011) and ALOS-2 PALSAR-2 (2014-present), totaling 59 acquisition dates.
  • Ground Observations: Plot measurements represent a period of two years on either side of the census year; linear interpolation used for multiple censuses.

Methodology and Data

• Models used:
  • Polarimetric Decomposition: van Zyl (1993), Cloude-Pottier (1997), and Freeman-Durden (1998) decompositions applied to fully-polarimetric L-band SAR data.
  • Multivariate Analysis: Theory-informed Structural Equation Modelling (SEM) to analyze SAR-vegetation-soil relationships.
  • Visualization: Generalized Additive Models (GAMs) to visualize relationships between scattering mechanisms and ground observations.
• Data sources:
  • Satellite: ALOS-1 PALSAR-1 and ALOS-2 PALSAR-2 (L-band, fully-polarimetric Single-Look Complex (SLC) data).
  • Ground Observations: Tree census data from permanent plots across Africa (SEOSAW network), Australia (TERN-Landscapes and Joint Remote Sensing Research Program), and South America (ForestPlots.net). Data includes Diameter at Breast Height (DBH), height, and wood density for stems > 10 cm DBH. Aboveground biomass density (AGBD), stem density (N), and mean stem biomass (b) were derived.
  • Reanalysis: ECMWF ERA5-Land product for soil volumetric water content (0–7 cm layer).
  • Gridded Soil Dataset: ISRIC SoilGrids for soil sand content (0–5 cm layer).

Main Results

• Dominant Scattering Mechanisms: Volume scattering is the dominant mechanism in wooded areas, even at low AGBD, mixed with surface scattering and weaker double bounce.
• AGBD Sensitivity:
  • Van Zyl decomposition: Odd-bounce scattering shows a significant positive trend with AGBD up to 52 tonnes per hectare (t ha−1), then saturates. Cross-pol and even-bounce scattering show similar trends but with lower magnitudes.
  • Freeman-Durden decomposition: Volume scattering shows a significant positive trend with AGBD up to 58 t ha−1. Double bounce shows a positive trend up to 22 t ha−1. Surface scattering shows no significant relationship with AGBD.
• Determinants of Scattering Mechanisms (SEM):
  • Volume Scattering: Most sensitive to stem density (0.32 standard deviations (SDs) increase for each doubling of stem density), followed by soil water content (0.23 SD increase for each SD increase), and mean stem biomass (0.14 SD increase for each doubling).
  • Double Bounce Scattering: Strongly affected by stem density (0.25 SD increase for each doubling) and moderately negatively affected by soil sand content (−0.18 SD).
  • Surface Scattering: Only a small positive effect from stem density (0.08 SD).
• Implications for AGBD Mapping: Radar intensity is about twice as sensitive to stem density as to mean stem biomass. AGBD maps based solely on radar intensity may underestimate changes from stem growth and overestimate changes from stem recruitment or loss. Full-polarimetric data allows for the retrieval of both stem density and mean stem biomass, providing a more comprehensive description of forest structure.

Contributions

• Provides the first large-scale, multi-site empirical understanding of L-band SAR scattering processes in dry tropical ecosystems, addressing a critical knowledge gap for accurate AGBD estimation.
• Empirically demonstrates that stem density is a stronger determinant of L-band volume scattering than mean stem biomass, challenging existing AGBD mapping approaches that rely solely on radar intensity.
• Introduces a structural equation modeling framework that, when combined with full-polarimetric SAR data, enables the inversion of scattering mechanisms to retrieve both stem density and mean stem biomass, offering a more complete characterization of forest structure beyond AGBD alone.
• Highlights the critical importance of current and future quad-polarimetric SAR missions (e.g., ALOS-4 PALSAR-3, BIOMASS, ROSE-L) for advancing large-scale vegetation structure and carbon cycle monitoring in dry tropical regions.

Funding

• UKRI NERC grants: SEOSAW (NE/P008755/1), SECO (NE/T01279X/1; NE/W001691/1; NE/N011570/1).
• São Paulo Research Foundation (FAPESP): grant #2015/50488-5, fellowships #2017/17380-1, #2021/10639-5.
• Brazilian National Council for Scientific and Technological Development (CNPq): grants 312589/2022-0, 306563/2022-3, 308623/2021-5.
• FAPEMIG: APQ-03020-22.
• NERC National Centre for Earth Observation (NCEO).
• ForestPlots.net Research Project #198.
• JAXA (ALOS-1 PALSAR-1 and ALOS-2 PALSAR-2 data access: ER3A2N020, ER3A2N035).
• European Space Agency (ALOS-1 PALSAR-1 data access).
• Alaska Satellite Facility (ALOS-1 PALSAR-1 data access).
• Copernicus Climate Change Service (ERA5-Land data access).
• ISRIC-World Soil Information (SoilGrids data access).
• U.S. Geological Survey (Landsat 5 Thematic Mapper data).

Citation

@article{Carreiras2026Determinants,
  author = {Carreiras, João M.B. and Higginbottom, Thomas and Godlee, John L. and Harrison, S. and Benitez, Lorena and Mograbi, Penelope J. and Levesley, Aurora and Melgaço, Karina and Milodowski, David T. and Pickavance, Georgia and Wells, Geoff and Oliveira, Edmar Almeida de and Arroyo, Luzmila and Bowers, Sam and Brienen, Roel J. W. and Cardoso, Domingos and Castro, Antônio Alberto Jorge Farias and Chavez, Ezequiel and Coutinho, Ítalo A.C. and Domingues, Tomas and Elias, Fernando and Santo, Mário Marcos Espírito and Feldpausch, Ted R. and Galbraith, David and Gloor, Emanuel and Gonçalves, Francisco M.P. and Gotore, Tatenda and Ishida, Francoise Yoko and Killeen, Timothy J. and Malhi, Yadvinder and Marimon, Beatriz S. and Marimon-junior, Ben Hur and Ramos, Desirée M. and Reis, Simone Matias and McNicol, Ian and Mitchard, Edward T.A. and Moonlight, Peter W. and Morandi, Paulo S. and Morellato, Patricia and Muchawona, Anderson and Muledi, Jonathan Ilunga and Murakami, Alejandro and Shutcha, Mylor Ngoy and Nieto-Quintano, Paula and Parada-Gutierrez, Alexander and Prestes, Nayane Cristina Candida dos Santos and Queiroz, Luciano Paganucci de and Rodrigues, Priscyla M.S. and Silva, Jhonathan Oliveira and Santos, Rubens M. and Särkinen, Tiina and Silva, Domingos Fortunato P.F. and Oliveira, Tony C. de Sousa and Steininger, Marc and Tchamba, José João and Veenendaal, Elmar and Zuanny, Débora C. and Baker, Tim R. and Dexter, Kyle G. and Hegerl, Gabriele C. and Pennington, R. Toby and Phillips, Oliver L. and Sitch, Stephen and Williams, Mathew and Quegan, Shaun and Ryan, Casey M.},
  title = {Determinants of L-band backscatter in dry tropical ecosystems: Implications for biomass mapping},
  journal = {Remote Sensing of Environment},
  year = {2026},
  doi = {10.1016/j.rse.2025.115213},
  url = {https://doi.org/10.1016/j.rse.2025.115213}
}