Unveiling critical morphological contributions in Forest vegetation carbon storage: An MSPA and explainable machine learning analysis of Jinhua City, China

Forests act as critical carbon sink ecosystems, with their carbon sequestration capacity influenced by multiple factors. However, the long-term relationship between forest spatial morphological patterns and carbon storage remains insufficiently explored. This study focuses on Jinhua City, Zhejiang Province, using multi-source data from 2002 to 2020, including morphological spatial pattern analysis (MSPA) metrics, climatic factors, land use types, and socioeconomic data. By employing MSPA and an explainable machine learning (ML) framework, we investigated the relationships between forest carbon storage and key influencing factors. The results indicate that: (1) Incorporating MSPA factors significantly enhances the predictive accuracy of vegetation carbon storage models. (2) NDVI, MSPA factors, and SSD (sunshine duration) are the most critical determinants of carbon storage levels, exhibiting pronounced nonlinear relationships with forest vegetation carbon storage. Specifically, NDVI, D_CORE (density of core), and SSD show the most significant positive contributions, whereas D_ISLET (density of islet), D_BRANCH (density of branch), and D_LOOP (density of loop) exhibit relatively lower and negative correlations. (3) Certain key influencing factors display threshold effects and optimal intervals. In Jinhua City, the significantly higher carbon sequestration benefits are associated with NDVI values ranging from 0.63 to 0.73, D_CORE between 63% and 89%, and SSD of 1482 h, providing actionable guidance for spatial planning. This study provides new insights into forest carbon management in Jinhua, suggesting that optimizing landscape ecological spatial patterns should be prioritized in ecological conservation efforts. Additionally, differentiated strategies should be developed for distinct regions to support sustainable forest management in alignment with China's dual carbon goals. • Relationship between morphological spatial pattern and carbon storage was analyzed. • MSPA factors significantly improve forest carbon storage prediction accuracy. • NDVI, MSPA factors, and SSD are key nonlinear determinants of carbon storage. • Indicators show nonlinear thresholds, guiding tailored spatial planning strategies.