Extensive forest restoration is crucial for achieving sustainable development goals and mitigating climate change. Tree planting is a common but costly method, often resulting in less biodiverse forests compared to natural regeneration. Global initiatives like the Bonn Challenge aim for large-scale forest restoration, emphasizing the importance of tropical regions due to their high biodiversity and the significant areas already deforested. However, effective large-scale forest restoration requires cost-effective methods. While tree planting is widely used, natural regeneration is a far more cost-effective and often more successful approach in terms of biodiversity and long-term success. The lack of knowledge regarding where natural regeneration is most likely to succeed hinders the broader application of this approach. This study addresses this gap by providing a high-resolution map of the potential for natural regeneration across the global tropics, leveraging robust pantropical data on forest regrowth.

Previous studies on mapping forest restoration potential have relied on coarse-scale data, expert opinions, and assumptions about potential forest cover, leading to underestimates. This study builds upon recent pantropical remote sensing analyses that identified areas of natural regrowth, using this data to create a more accurate assessment of the potential for natural regeneration. The study acknowledges other studies estimating the potential for natural forest regrowth, but highlights its use of actual data on natural forest regrowth after deforestation for a more precise and spatially explicit assessment.

The study used machine learning methods to model the potential for natural regeneration across the global tropics. A sample of 5.4 million hectares of previously identified natural regrowth was used to train the model. The model incorporated a range of predictor variables at local and landscape scales known to influence tropical forest regrowth. These variables included distance to nearest tree cover, local forest density, land cover, soil metrics, bioclimatic variables, slope, net primary productivity, burned area, distance to water, population density, gross domestic product, human development index, road density, distance to urban areas, and protected area status. Spatially explicit values of these variables were used to predict the potential for natural regeneration in 2015 and 2030. The resulting continuous potential values (0-1) were translated into area-based values by multiplying the potential by the area of each pixel. A sensitivity analysis was conducted, considering cells with greater than 50% potential. The model's accuracy was assessed through validation, achieving an accuracy of 87.9%. A comparison was made between models incorporating both socioeconomic and biophysical variables and models with only biophysical variables, resulting in the selection of the biophysical-only model for spatial predictions due to its stability and higher spatial resolution.

The study estimates that biophysical conditions support natural regeneration in tropical forests over 215 million hectares globally until 2030. This area is distributed across the neotropics (98 Mha), Indomalayan tropics (90 Mha), and Afrotropics (25.5 Mha). Five countries—Brazil, Indonesia, China, Mexico, and Colombia—account for 52% of this global potential. The potential for natural regeneration was positively associated with local forest density and negatively associated with distance to existing forest, particularly pronounced in the neotropics. Soil organic carbon content was also a positive predictor. It is estimated that forest regeneration over this 215 million hectares could sequester 23.4 Gt of carbon (aboveground biomass) over 30 years, representing a significant contribution to global carbon sequestration. This estimate does not include belowground biomass, which could increase the total to 28.6–30.0 Gt of C. This would significantly increase current global carbon sequestration potential in primary and secondary tropical and subtropical forests. The study's high-resolution maps reveal fine-scale spatial variation in regeneration potential, highlighting the importance of spatially explicit targeting for restoration efforts.

The findings demonstrate substantial opportunities for cost-effective forest restoration through assisted natural regeneration, particularly in the Neotropics and Indomalayan tropics. This approach offers considerable carbon sequestration potential and biodiversity benefits, significantly mitigating current pantropical forest carbon losses and potentially increasing current global carbon sequestration by up to 14.3% annually. The study's high-resolution maps allow for precise targeting of restoration efforts, maximizing impact and efficiency. The identification of key countries (Brazil, Indonesia, China, Mexico, and Colombia) with high regeneration potential emphasizes the need for tailored national-level restoration strategies, leveraging both policy and market-based mechanisms. The study acknowledges limitations such as potential underestimation due to the positive feedback nature of natural regeneration and the exclusion of certain ecosystems like savannas. It also notes the influence of future climate change impacts on regeneration processes.

Natural forest regeneration presents a significant opportunity for large-scale, cost-effective forest restoration. The study highlights the substantial carbon sequestration potential and biodiversity co-benefits of this approach. The high-resolution maps provided serve as a valuable tool for informing restoration planning and policy at local, national, and global levels. Future research should focus on refining the model to incorporate factors such as climate change impacts and socioeconomic considerations to further enhance the accuracy and utility of predictions for guiding effective and equitable restoration strategies.

The study's estimates of regeneration potential might underestimate the long-term potential due to the positive feedback loop of regeneration—as forests regrow, they extend the area suitable for future regeneration. The analysis also focuses on dense natural forest ecosystems and excludes non-forested or sparsely forested areas such as savannas. Additionally, the impact of climate change on future regeneration success is not explicitly modeled. Finally, while the model is high-resolution, the input data varied in resolution, which needs to be considered when interpreting the results.