Interested in watching this video? You have two options:
This video is part of the SER Conference Library. If you want to learn more about this resource please see this guide.
You can purchase a pass for this video on our website.
Already purchased access to this video, or want to redeem credit for a new order? Just enter your order number or email below:
Sign in below to get unrestricted access:
Robin Chazdon, Hawthorne Beyer, Brooke Williams, Matthew Fagan, Marina Schmoeller, Bronson Griscom, Starry Sprenkle-Hyppolite, Nikola Alexandre, Renato Crouzeilles
Background and objectives: Assisted natural regeneration of forest is a proven cost-effective restoration strategy for climate change mitigation, biodiversity conservation, and livelihood enhancement. But the potential for natural regeneration varies spatially, emphasizing the need to identify and map areas estimated to have a high potential for assisted natural regeneration. As a collaborative project between Conservation International and the International Institute for Sustainability-Australia, we modelled the probability of natural regeneration at a high-resolution (30m) in tropical and subtropical forests (between ±25° latitude) globally. Methods: Using machine learning, specifically a LightGBM model using the platform driverless H2O, we generated predictions of natural regeneration potential (ANR potential) with high classification accuracies (c. 88 – 93%). We trained this model using sample deforested locations where natural regeneration did or did not occur between 2000 and 2012 and 41 biophysical and socioeconomic covariates that reflect processes known to influence land use, management, and forest regeneration processes. Results: Predictors of ANR potential varied across the three major regions we modelled. In the Neotropics, forest density and elevation were important predictors of ANR potential, whereas in the Afro-tropics, the most important predictor variables were distance to forest and biome. In the Indo-Malayan tropics, although biophysical variables such as forest density were also important, the human development index (a socio-economic variable) was the most important variable explaining the occurrence of natural regeneration. Conclusion: Our model and results provide important insights for supporting decisions on spatial planning of cost-effective forest restoration across the tropics and subtropics.
Conference Presentation, SER2021
Pre-approved for CECs under SER's CERP program