【摘要】 植被覆盖度(FVC)是区域生态环境质量的重要参数，揭示长时序FVC时空分异特征及识别主导影响因子可为评估地表植被状况提供科学参考。以天津市为例，基于Google Earth Engine(GEE)云平台的1990—2020年Landsat遥感影像，结合像元二分模型计算FVC。运用一元线性回归、F检验进行FVC时空演变趋势分析，利用变异系数、Hurst指数分析FVC波动情况及未来变化趋势。在此基础上，应用随机森林(RF)模型的特征重要性排序、偏依赖图技术量化FVC与同期的气候因素(年均气温、年降水量)和人类活动因素(土地利用、人口密度)的非线性响应关系，并进一步分析多因素耦合对FVC空间分布的解释力度。结果表明：(1)在时间上，1990—2020年天津市FVC呈显著下降趋势，变化率为0.001 5(p<0.05),其中1990—2012年FVC显著降低，2012年后变化趋势不显著(p>0.05)。不同等级FVC时序差异明显，低、中低和中植被覆盖比例均呈现显著增加趋势(p<0.05),而中高和高植被覆盖比例一致表现为显著降低(p<0.05)。(2)在空间上，1990—2020年天津市FVC呈现由市内六区向外逐渐增高的空间分布格局，整体以高植被覆盖为主(49.68%)。研究区内32.62%的区域降低趋势显著，主要分布在环城四区以及滨海新区，27.42%的区域增加趋势显著，主要分布在远郊区县及市内六区。近30年内区域植被退化比重高于改善比重，且未来FVC变化趋势将以退化为主。(3)人类活动因素(57.27%)对FVC的影响程度远高于气候因素(42.73%),其中土地利用为影响FVC动态变化的主导因子。相比于单一气候因子，年均气温与年降水量的交互作用能够增强FVC空间分布的解释力度。研究结果有助于客观揭示天津市植被演变驱动因素，为制定植被保护政策提供参考依据。更多还原

【Abstract】 Fraction of vegetation coverage(FVC) is an important parameter for regional ecological environment quality. Revealing the temporal and spatial variation characteristic of long-term FVC and identifying the dominant influencing factor can provide a scientific reference for evaluating vegetation condition. Taking Tianjin as the research area, we extracted FVC by combining the pixel binary model based on Landsat images of the Google Earth Engine(GEE) during 1990—2020. The temporal and spatial variation trend of FVC was analyzed by linear regression and F test. Fluctuation change and future trend of FVC were analyzed by coefficient of variation and Hurst index. The order of feature importance and partial dependency plots of random forest(RF) were applied to quantify the non-linear response relationship between FVC and the same period of climate factors(mean annual temperature, annual precipitation) and human activity factors(land use, population density), then we further analyzed the spatial influencing magnitude of multi-factor coupling on FVC. The results showed that:(1) In terms of time, FVC presented the downward trend during 1990—2020 with a significant decrease of 0.001 5(p<0.05); FVC showed the significant decrease during 1990—2012(p<0.05); however, no significant change was observed after 2012(p>0.05); classifications of FVC displayed significant differences in time series; the proportions of low, medium-low and medium FVC showed significant increase(p<0.05), while the proportions of medium-high and high FVC showed significant decrease consistently(p<0.05);(2) spatially, FVC showed a spatial distribution pattern that gradually increased from six districts to outside and the types were mainly high(49.68%); FVC in about 32.62% of the research area showed a significant decrease trend which mainly distributed in four suburban districts and Binhai New Area; FVC in about 27.42% of the research area showed a significant increase trend which mainly distributed in outer suburban districts and six districts; In the past 30 years, the proportion of degraded area was larger than improved area and the future trend of FVC would be dominant degeneration;(3) the influence of human activities(57.27%) on FVC was much higher than the climate factors(42.73%), of which land use was the dominant factor; compared with the single climate factor, the interaction between temperature and precipitation could enhance the spatial influencing magnitude of FVC. This research can reveal the driving factors of FVC and provide a reference for policy-making on vegetation protection.更多还原