关键词: 森林碳汇, 森林经营, 遥感, 时空分异, 机器学习

Abstract: In the context of global carbon neutrality, quantifying the carbon gains of ecological engineering and the carbon losses from urbanization is crucial for optimizing carbon sink strategies and achieving sustainable development. With the Wuhan metropolitan area as a representative urbanized region in China, we explored how ecological restoration projects and urbanization processes shape regional forest spatial patterns and quality, and consequently affect forest carbon sink function and its spatial-temporal patterns. Remote sensing, field plot data, and machine learning were integrated to construct a forest carbon stock model to examine the spatiotemporal differentiation patterns of forest carbon benefits from 2004 to 2023. The results showed that: (1) The average forest carbon density in the study area increased from 20.67±0.99 Mg C·hm^-2 to 27.43±1.17 Mg C·hm^-2, and the total carbon stock increased from 33.22±1.59 Tg C to 46.28±1.97 Tg C, showing an initial rapid and then stable increasing trend. (2) Carbon sink gain zones (accounting for 60.26% of the area, carbon gain contribution rate of 129.58%) and carbon source loss zones (16.11% of the area, contributing 30.91% to carbon loss) coexisted, while the carbon balance areas (23.63%) reflected a dynamic equilibrium between human disturbance and forest succession, with a minimal carbon gain contribution rate of 1.33%. (3) Afforestation areas (10.17%) and conserved forest areas (75.35%) were the major forest management types, dominating the growing carbon sink. Their carbon gain contribution rates were 34.97% and 77.94% respectively, demonstrating the significant carbon benefits of ecological restoration projects. In contrast, due to urbanization, the cutting and regeneration areas (8.45%) showed a polarized trend of significant increase and decrease in carbon density, with a carbon gain contribution rate of 3.78%, while forest conversion areas (6.03%) revealed the ecological costs of urbanization, contributing a carbon loss rate of 16.69%. By elucidating the antagonistic effects between ecological restoration projects and urbanization stress on forest carbon benefits, this study provides a methodological framework and decision-making support for regional carbon sink assessments and spatial optimization.

Key words: forest carbon sink, forest management, remote sensing, spatiotemporal differentiation, machine learning