| 基于XGBoost-SHAP模型的闽三角城市群生态环境质量时空演变及驱动因素分析 |
| 摘要点击 300 全文点击 25 投稿时间：2025-08-04 修订日期：2025-09-30 |
| 查看HTML全文
查看全文 查看/发表评论 下载PDF阅读器 |
| 中文关键词 生态环境质量 遥感生态指数（RSEI） 时空演变 XGBoost-SHAP模型 谷歌地球引擎（GEE） 闽三角城市群 |
| 英文关键词 ecological environment quality remote sensing ecological index （RSEI） spatiotemporal evolution XGBoost-SHAP model Google Earth Engine（GEE） Min Delta Urban Agglomeration |
| DOI 10.13227/j.hjkx.202508029 |
|
| 中文摘要 |
| 探究闽三角城市群生态环境质量的时空演变规律及其驱动因素，对于保障区域生态安全及高质量发展具有重要意义. 基于谷歌地球引擎平台的2000~2024年Landsat遥感数据集，采用主成分分析法构建了遥感生态指数（RSEI），结合变异系数分析了闽三角城市群RSEI的时空演变规律，并运用可解释极端梯度提升（XGBoost-SHAP）模型揭示了自然和人类活动因子对生态环境质量的关键驱动因子、非线性效应及交互效应. 结果表明：①时间尺度上，闽三角城市群RSEI呈先下降后上升的趋势，多年均值为0.62，最低值和最高值分别出现在2010年（0.53）和2024年（0.68），生态环境质量整体良好；②空间尺度上，研究区RSEI呈现“西北高、东南低”的分布格局，生态环境质量变好区域面积大于变差区域，两者占比相差4.51%；③研究区RSEI变异系数范围为0.58%~84.63%，整体处于稳定状态，但仍有9.10%的区域表现出不稳定状态；④生态环境质量变化的主导因素具有时间异质性：2000~2010年期间，该变化主要受植被覆盖度、年均气温和年降水量等自然因素驱动；而2015~2024年期间，则转变为植被覆盖度、夜间灯光强度和人口分布等自然和人类活动因素协同主导. 此外，自然与人类活动因素间的交互作用呈现高度复杂性，表现出阈值依赖性和效应方向可逆性. 研究结果可为闽三角城市群的生态文明建设和可持续发展提供科学参考. |
| 英文摘要 |
| Investigating the spatiotemporal evolution patterns of ecological environment quality and its driving factors in the Min Delta Urban Agglomeration is essential for ensuring regional ecological security and promoting high-quality development. Utilizing Landsat remote sensing datasets from 2000 to 2024 on the Google Earth Engine platform， this study calculated the remote sensing ecological index （RSEI） through principal component analysis. The spatiotemporal evolution patterns of RSEI in the Min Delta Urban Agglomeration were analyzed in conjunction with the coefficient of variation. Furthermore， the interpretable Extreme Gradient Boosting （XGBoost-SHAP） model was employed to identify key drivers， nonlinear effects， and interaction effects of both natural and human activity factors on ecological environment quality. The results showed that： ① Temporally， the RSEI of the Min Delta Urban Agglomeration exhibited an initial decline followed by an upward trend. The multi-year average RSEI was 0.62， with the lowest and highest values occurring in 2010 （0.53） and 2024 （0.68）， respectively， indicating generally good ecological quality overall. ② Spatially， the RSEI in the study area showed a distribution pattern of “higher in the northwest and lower in the southeast”. The area with improved ecological quality was larger than that with degraded quality， with a difference of 4.51% in their respective proportions. ③ The coefficient of variation of RSEI in the study area ranged from 0.58% to 84.63%， with a mean value of 11.97%. The overall ecological environment was stable， but significant spatial heterogeneity existed. Specifically， 54.86% of the area exhibited low RSEI volatility， indicating stability， primarily distributed in the western and northern regions with high vegetation coverage. In contrast， 9.10% of the area was shown to be unstable， mainly located in the eastern coastal regions. ④ The dominant factors driving changes in ecological environment quality demonstrated temporal heterogeneity： From 2000 to 2010， changes were primarily driven by natural factors such as fractional vegetation coverage （FVC）， annual average temperature （AT）， and annual precipitation. In contrast， from 2015 to 2024， changes shifted to being co-driven by both natural factors （FVC） and human activity factors （nighttime light intensity （NTL）， population distribution）. Moreover， the interactions between natural and human activity factors were highly complex， exhibiting distinct threshold dependency and reversibility of effect directions. For instance， the most pronounced interaction was observed between FVC and AT in 2000. Specifically， when FVC was below 50%， an increase in AT had a positive interactive effect； however， beyond this threshold， the influence of FVC on RSEI changes gradually shifted to a negative effect as AT increased. By contrast， in 2024， the most significant interaction occurred between FVC and NTL. When FVC was less than 45%， increasing NTL transitioned from a negative to a positive effect on RSEI dynamics； conversely， when FVC exceeded 45%， the interactive effect shifted from positive to negative. These findings provide a scientific reference for ecological civilization construction and sustainable development in the Min Delta Urban Agglomeration. |
