3D-SAR Tomography and Machine Learning for High-Resolution Tree Height Estimation

TL;DR

The paper tackles canopy height estimation for forest biomass by leveraging TomoSAR tomocubes and LiDAR ground truth, addressing the need for accurate, scalable forest height data. It compares tabular AutoML (AutoGluon) and 3D CNN approaches (compact U-Net variants) using the TomoSense dataset that includes L- and P-band tomographic SAR data with monostatic and bistatic configurations. Key findings include a best absolute MAE of $2.82$ m for L-monostatic union channels and a best normalized MAE of $1.02$ for P-band when accounting for vertical resolution, demonstrating the value of tomography and lightweight CNNs in data-constrained settings. The study highlights the importance of robust geospatial validation, characterizes how band/polarization choices affect accuracy, and provides publicly releasable model configurations to advance tree-height and biomass modelling for global carbon stock assessments.

Abstract

Accurately estimating forest biomass is crucial for global carbon cycle modelling and climate change mitigation. Tree height, a key factor in biomass calculations, can be measured using Synthetic Aperture Radar (SAR) technology. This study applies machine learning to extract forest height data from two SAR products: Single Look Complex (SLC) images and tomographic cubes, in preparation for the ESA Biomass Satellite mission. We use the TomoSense dataset, containing SAR and LiDAR data from Germany's Eifel National Park, to develop and evaluate height estimation models. Our approach includes classical methods, deep learning with a 3D U-Net, and Bayesian-optimized techniques. By testing various SAR frequencies and polarimetries, we establish a baseline for future height and biomass modelling. Best-performing models predict forest height to be within 2.82m mean absolute error for canopies around 30m, advancing our ability to measure global carbon stocks and support climate action.

Figures (8)

• Figure 1: Geographic splits of training/validation/test data. L-r: Swathe, square, quadrant.
• Figure 2: Height Estimation for union polarisation channels
• Figure 3: Forest canopy height reconstruction for L-monostatic, HH+HV+VV polarisation.
• Figure 4: 3D LiDAR scan of case study area
• Figure 5: Model architecture for Models 2 and 3. Block options are given in Figure \ref{['fig:encoder-block']} and \ref{['fig:resid_block']}