Uncovering the driving factors of trade-driven mangrove loss through interpretable machine learning

Understanding the driving factors of mangrove loss is complex, as loss in one country may be driven by economic activities in another through international trade. Moreover, the process through which anthropogenic activities transfer mangrove pressures across countries remains poorly understood. In this study, temporal trends in the global displacement of mangrove losses embodied in international trade were quantified, and significant driving factors were identified through machine learning techniques.We highlight how countries, including China and Malaysia, increasingly imported or exported mangrove losses during 2000–2016, contributing to a polarized global distribution of mangrove resource pressure. Key driving factors include agricultural value added, population size, industrial share of GDP, and protected area coverage.Moreover, interpretative machine learning facilitates decision-makers to reveal non-linear and interacting relationships for policy recommendations. For example, protected area percentage exhibit an inverted U-shaped relationship: it positively links with trade-driven loss between 9% and 54%, especially high gross capital formation countries. These findings reflect conservation paradox and provide state-of-the-art research in supporting ‘half-earth’ strategy that large-scale protected area can facilitate mangrove sustainability.