Capturing Temporal Dynamics in Large-Scale Canopy Tree Height Estimation

Abstract

With the rise in global greenhouse gas emissions, accurate large-scale tree canopy height maps are essential for understanding forest structure, estimating above-ground biomass, and monitoring ecological disruptions. To this end, we present a novel approach to generate large-scale, high-resolution canopy height maps over time. Our model accurately predicts canopy height over multiple years given Sentinel-1 composite and Sentinel~2 time series satellite data. Using GEDI LiDAR data as the ground truth for training the model, we present the first 10m resolution temporal canopy height map of the European continent for the period 2019-2022. As part of this product, we also offer a detailed canopy height map for 2020, providing more precise estimates than previous studies. Our pipeline and the resulting temporal height map are publicly available, enabling comprehensive large-scale monitoring of forests and, hence, facilitating future research and ecological analyses.

Figures (9)

• Figure 1: Comparison of six canopy height maps with precise measurements obtained via aerial laser scanning (ALS). The patches contain tall trees exceeding 30 m in height. Our model is the only one that can accurately estimate the height of such trees.
• Figure 2: Spatial distribution of 1,500 randomly selected validation locations across Europe. Each location covers a $2.56\,\mathrm{km} \times 2.56\,\mathrm{km}$ area in which estimated canopy heights are evaluated against GEDI measurements. Note that due to GEDI's flight path, no labels are available above $51.6^\circ\mathrm{N}$.
• Figure 3: Boxplots for each model showing the 2020 mean error in every $5\,\mathrm{m}$ bin between $10\,\mathrm{m}$ and $40\,\mathrm{m}$. Although liu2023, paulsestimating and langGlobalCanopyHeight2022 perform well on smaller trees, our models performs especially well for taller trees.
• Figure 4: Qualitative comparison of canopy height maps for the reference year 2020: liu2023, tolan2023, langGlobalCanopyHeight2022, paulsestimating, turubanova_europe (2020) and our model, 3D-Stack-MultiYear-L.
• Figure 5: Left: Scatterplots between 2020 GEDI labels and prediction for langGlobalCanopyHeight2022liu2023paulsestimatingtolan2023turubanova_europe and our model including $R^2$ for all labels and $R^2_7$ for labels exceeding $7\,\mathrm{m}$. Right: Histograms of GEDI labels and all maps. turubanova_europe and tolan2023 saturate at $28\,\mathrm{m}$, our model is the only one matching above $40\,\mathrm{m}$.