A machine learning framework to assess global mangrove forestation potential under current and future climate...

Mangrove forestation is one of the most efficient forestry practices for carbon sequestration. This study developed a machine learning framework that integrated the random forest algorithm, SHapley Additive exPlanations (SHAP), and partial dependence plots (PDP) to assess global mangrove cover potential and its drivers, utilizing a suite of 48 environmental layers encompassing climatic, topographic, soil, and marine characteristics. Based on the mangrove cover potential, this study quantified the mangrove forestation potential under socioeconomic and ecological land-use constraints, as well as the carbon storage potential of forestation potential. The result showed that there is 156,682 km² of mangrove forestation potential under current climate conditions. When assessing the distribution of mangrove forestation potential across Marine Ecoregions of the World (MEOW) provinces, MEOW ecoregions, and countries, the greatest forestation capacity is observed in the Tropical Northwestern Atlantic, Amazonia, and Indonesia. SHAP and PDP results revealed that soil saturated water content and distance to sea are the key factors controlling mangrove cover potential. Under contrasting shared socioeconomic paths (SSP1-2.6 and SSP5-8.5), mangrove cover potential shows a general increase due to climate changes. However, under SSP5-8.5, sea-level rise alone could reduce the current forestation potential by 26,820 km². Furthermore, only 19,361 km² of the current forestation potential coincides with areas where future cover potential is projected to increase across both scenarios, indicating that the synergistic enhancement effect brought about by climate change on the forestation results is limited. From a national perspective, the five countries with the highest carbon storage potential from mangrove forestation are Indonesia, Brazil, Australia, Mexico, and the Philippines, with 1.016, 0.514, 0.409, 0.391, and 0.317 GtC, respectively. The global mangrove forestation potential map with clear spatial granularity provided in this study can offer important support for international-scale mangrove forestation.