标题: Dual pathways of forest landscape dynamics in China: Integrating ecological land evolution Index (ELEI) and machine learning to decipher fragmentation and restoration

作者: Xie, YF (Xie, Yifan); Liu, S (Liu, Sui); Pang, BW (Pang, Bowen); Wang, NN (Wang, Nannan); Xu, JW (Xu, Jiwei); Huang, H (Huang, He); Cui, Y (Cui, Yao); Yang, JM (Yang, Jiaming); Liu, YL (Liu, Yaolin); Liu, YF (Liu, Yanfang)

来源出版物: INTERNATIONAL JOURNAL OF APPLIED EARTH OBSERVATION AND GEOINFORMATION 卷: 142 文献号: 104724 DOI: 10.1016/j.jag.2025.104724 Published Date: 2025 AUG

摘要: Forest ecosystems underpin global ecological security, yet China's forest landscapes are subjected to dual pressures of restoration and degradation, with complex spatiotemporal dynamics that remain insufficiently understood. While existing studies focus on static landscape metrics, the nonlinear interactions between natural and anthropogenic drivers of forest fragmentation and connectivity remain elusive. This study addresses this gap by introducing the Ecological Land Evolution Index (ELEI) to quantify patch-scale forest dynamics, integrating multiscale geographically weighted regression (MGWR) with XGBoost-SHAP analysis to dissect driving mechanisms across China from 2000 to 2020. Key findings reveal distinct spatial patterns: Ecological Land Restoration (ELR) processes with high connectivity potential concentrated in northeastern China, the Qinba Mountains, and southwestern regions. "Connection-type" patches (comprising 48 % of ELR area) dominated low-elevation plains (e.g., Yunnan-Guizhou Plateau), while "infilling-type" restoration prevailed in hilly/mountainous areas, driven by shrubland-to-forest conversions. Ecological Land Degradation (ELD) driven by cropland expansion caused severe fragmentation in the Sichuan Basin and Yunnan-Guizhou Plateau."Enclosing-type" ELD dominated highrelief mountains, while "fragmentation-type" patches thrived in lowlands under agricultural/urban pressures. MGWR analysis identified slope, elevation, and the Grain for Green Program as key positive drivers of ELR connectivity, while urbanization and agricultural expansion emerged as dominant ELD fragmentation forces with pronounced spatial heterogeneity. XGBoost-SHAP further revealed nonlinear interactions: road density (>0.5 km/km(2)) exacerbating fragmentation in depopulating peri-urban zones, and climate-topography thresholds shaping restoration efficacy (e.g., precipitation > 1,300 mm). This national-scale assessment highlights the need for connectivity-oriented strategies: prioritize contiguous forest expansion in plains, promote infilling-type restoration in hills/mountains, and facilitate passive recovery in high-altitude regions. By quantifying patchscale connectivity changes and disentangling complex driver interactions, the study provides a scientific foundation for adaptive forest management, bridging ecological conservation and socioeconomic development.

作者关键词: Forest fragmentation; Landscape connectivity; Machine learning; Nonlinear relationship; Spatial heterogeneity

KeyWords Plus: CLIMATE-CHANGE; ECOSYSTEM SERVICES; HABITAT; CONNECTIVITY; BIODIVERSITY; IMPACTS; GRAPHS

地址: [Xie, Yifan; Liu, Sui; Pang, Bowen; Wang, Nannan; Xu, Jiwei; Huang, He; Cui, Yao; Yang, Jiaming; Liu, Yaolin; Liu, Yanfang] Wuhan Univ, Sch Resource & Environm Sci, 129 Luoyu Rd, Wuhan 430079, Peoples R China.

[Liu, Yaolin; Liu, Yanfang] Wuhan Univ, Collaborat Innovat Ctr Geospatial Technol, 129 Luoyu Rd, Wuhan 430079, Peoples R China.

[Liu, Yaolin] Duke Kunshan Univ, Kunshan 215316, Peoples R China.

通讯作者地址: Liu, YL (通讯作者)，Wuhan Univ, Sch Resource & Environm Sci, 129 Luoyu Rd, Wuhan 430079, Peoples R China.

电子邮件地址: yaolin610@yeah.net

影响因子：8.6