Clarifying tropical secondary forest estimates: results from a growing reference dataset on post-disturbance regrowth

The global forest sink, particularly the tropical forest sink, plays a critical role in the global carbon cycle (Bueno et al., 2020; Pan et al., 2024). Carbon stored in forests is equivalent to almost half the carbon emitted from fossil fuels between 1990 and 2019; however, persistent patterns of deforestation and degradation threaten this carbon sink and its vital role in climate regulation (Pan et al., 2024). Abandoned or left to recover, post-disturbance landscapes can lead to naturally regenerating forests that contribute to significant carbon sequestration, biodiversity benefits, and ecosystem services (Chazdon et al., 2016). Therefore, it is crucial to understand when and where regenerating forests occur to highlight their environmental contributions and promote forest restoration (de Jong et al., 2025). Advancements in RS technologies have enabled a proliferation of RS-based land cover datasets with the potential to estimate secondary forest age and extent (Xu et al., 2024; Baker et al., 2025). However, many have not been explicitly validated for the age or extent of post-disturbance forest regrowth. First comparisons of datasets on secondary forest extent reveal substantial differences at the pixel scale, calling for the need for a reference dataset.

Creating a high-quality reference dataset is essential to ensure the reliability of remote sensing-based maps (Baker et al., 2025) and deliver meaningful results to policymakers (Xu et al., 2024; Tyukavina et al., 2025). Here, we propose a robust methodology and reference dataset to validate different secondary forest estimates and support a broader analysis of tropical forest regrowth dynamics. Using a stratified sampling design based on Potapov et al. (2022) and Zhang et al. (2023)'s land cover products, samples were extracted from areas of agreement and disagreement on forest gain, forest loss, stable forest, and stable non-forest regions. Forest trajectories, drivers of disturbance and regrowth, and years of disturbance and regrowth were interpreted using Landsat, Planet, and Google Earth Pro imagery from 2000-2020. Focusing on tropical biomes outside the Amazon Basin which are often overlooked, weighted area estimates from a preliminary 500-sample reference dataset in South America reveal 87.97Mha (5.12% ± 1.00%) of deforestation, 21.70Mha (1.24% ± 0.25%) of secondary forests, and 82.34Mha (4.72% ± 0.92%) of degraded forests. Although the preliminary sample does not indicate systematic over- or under-estimation of secondary forest age, results reveal high commission errors in the extent of naturally regenerating secondary forests outside the Amazon Basin. Furthermore, high uncertainty in the interpretation of tropical dry forest samples highlights the challenges in identifying secondary forests outside the humid tropics and emphasizes the need for more research outside of the Amazon. 

This dataset has the potential to expand across the pan-tropics to harmonize essential information on regenerating forests and guide urgent action needed to protect the forest carbon sink. Learning from the preliminary study in South America, we emphasize the importance of data quality and draw attention to the uncertainty of large-scale secondary forest products.