【作者】 王刚；

【导师】 牛全福； 李克恭；

【作者基本信息】 兰州理工大学 ， 测绘工程（专业学位）， 2025， 硕士

【摘要】 植被动态变化是遥感监测的主要研究内容之一,归一化植被指数（Normalized Difference Vegetation Index,NDVI）是量化植被时空演变规律的关键指标,能够有效评估生态工程成效、揭示植被对气候变化的响应机制,可为区域生态保护政策制定提供重要的科学依据。黄河流域甘肃段作为我国西北地区重要的生态安全屏障区,在气候变化和人类活动的双重胁迫下,其生态系统结构和功能面临严峻挑战。因此,基于NDVI时序数据,系统解析该区域植被动态变化规律及其对气候变化和人类活动的响应机制,对于深入理解该区域环境演变过程和区域生态保护等具有重要科学意义。本研究以黄河流域甘肃段为研究对象,综合运用多尺度地理加权随机森林回归降尺度模型、BFAST（Breaks For Additive Season and Trend,BFAST）、多元回归残差分析法、改进型遥感生态指数（Improved Remote Sensing Ecological Index,IRSEI）、Sen-Mann Kendall趋势分析法以及地理探测器等模型方法,系统开展了2001-2022年植被覆盖动态变化、生态环境质量评价及其驱动机制研究。主要研究结论如下:（1）本研究通过耦合多尺度地理加权和随机森林回归模型,将MODIS NDVI数据进行降尺度,构建了黄河流域甘肃段2001-2022年30米分辨率的月尺度NDVI数据集。在精度评价中,通过选择典型区将降尺度后的NDVI数据与同期Landsat NDVI数据进行精度对比,得出两种数据的相关性较好,其决定系数R²为0.9082,均方根误差RMSE为0.0478。表明降尺度后的NDVI数据有效解决了单一数据源在时空连续性上的不足,可适用于黄河流域甘肃段植被变化动态监测。（2）研究区植被变化动态分析显示:年际变化呈显著上升趋势(增速为5.21×10^-3yr^-1),空间上呈现东部、东南部和西部高值、北部低值的分布格局;年内特征表现为7-8月NDVI峰值显著;季节特征表现为夏季NDVI均值最高（0.45）,冬季最低（0.14）;突变检测发现研究区有63.3%的像元存在突变,主要分布于黄土高原、青藏高原东南缘及兰州-白银都市圈,其中66%的突变发生在2011年后;突变类型以单调递增（含突变）（31%）和递增-递减（31%）转换为主,自然植被以单调递增（22.6%）和转换型（递增到递减（20.9%）、递减到递增（11%））为主,栽培植被则更多表现为单调递增（带正断点）（15.2%）;由多元回归残差分析得出人类活动贡献率（60.61%）显著高于气候变化（39.39%）。（3）由于传统RSEI指数在干旱半干旱区域的局限性,本研究针对研究区土壤侵蚀现状,创新性地引入土壤侵蚀因子（RUSLE）构建了IRSEI指数。研究表明,黄河流域甘肃段的IRSEI总体上呈"增长-下降-回升"的波动趋势,其中森林区稳定性较高而草原区波动显著。由空间自相关分析得出,IRSEI存在显著集聚特征,高质量区主要分布在东南部林区和甘南草原。由地理探测器单因子探测得出,降水和相对湿度对IRSEI变化的解释力最高（q>0.3）,而坡度与人口密度解释力最低（q<0.1）。交互探测得出,土地利用类型与降水的交互作用最为显著（q=0.60）,表明在水土流失严重的黄河流域甘肃段,合理的土地利用方式能显著增强降水资源的生态效益。更多还原

【Abstract】 Vegetation dynamic change is one of the main research areas in remote sensing monitoring.The Normalized Difference Vegetation Index(NDVI)is a key indicator for quantifying the spatiotemporal evolution patterns of vegetation.It can effectively assess the effectiveness of ecological projects,reveal the response mechanisms of vegetation to climate change,and provide important scientific foundations for the formulation of regional ecological protection policies.The Gansu section of the Yellow River Basin,as a critical ecological security barrier in China’s northwest region,faces severe challenges to its ecosystem structure and function under the dual pressures of climate change and human activities.Therefore,systematically analyzing the dynamic changes in vegetation and their response mechanisms to climate change and human activities based on NDVI time-series data holds significant scientific importance for deepening the understanding of environmental evolution processes and regional ecological protection in this area.This study focuses on the Gansu section of the Yellow River Basin.By comprehensively employing models and methods such as the multi-scale geographically weighted random forest regression downscaling model,BFAST(Breaks For Additive Season and Trend),multiple regression residual analysis,the Improved Remote Sensing Ecological Index(IRSEI),Sen-Mann Kendall trend analysis,and the geographical detector,the study systematically investigates the dynamic changes in vegetation coverage,ecological environment quality,and their driving mechanisms from 2001 to 2022.The main research conclusions are as follows:(1)This study coupled the multiscale geographically weighted regression model with random forest regression to downscale MODIS NDVI data,successfully constructing a monthly NDVI dataset at 30-meter resolution for the Gansu section of the Yellow River Basin from 2001to 2022.For accuracy validation,the downscaled NDVI data were compared with contemporaneous Landsat NDVI data in selected representative areas.The results demonstrated strong correlation between the two datasets,with a coefficient of determination(R~2)of 0.9082and root mean square error(RMSE)of 0.0478.These findings indicate that the downscaled NDVI data effectively addressed the spatiotemporal continuity limitations inherent in single data sources,making them suitable for dynamic vegetation monitoring in the Gansu section of the Yellow River Basin.(2)The analysis of vegetation change dynamics in the study area showed that the inter-annual variation showed a significant upward trend(with a growth rate of 5.21×10-3yr-1),with a spatial distribution pattern of high values in the east,southeast and west,and low values in the north;the intra-annual characteristics showed a significant peak of NDVI in July-August;the seasonal characteristics showed the highest mean NDVI in the summer(0.45),and the lowest in the winter(0.14);and the detection of mutations revealed that 63.3%of the pixels in the study area had mutations.Mutations were found in 63.3%of the pixels in the study area,mainly in the Loess Plateau,the southeast edge of the Tibetan Plateau,and the Lanzhou-Baiyin metropolitan area,with 66%of the mutations occurring after 2011;the mutation types were dominated by monotonically increasing(including mutations)(31%)and increasing-decreasing(31%)transitions,and the natural vegetation was dominated by monotonically increasing(22.6%)and transition types(increasing to decreasing(Natural vegetation was dominated by monotonically increasing(22.6%)and switching(increasing to decreasing(20.9%),decreasing to increasing(11%)),while cultivated vegetation showed more monotonically increasing(with positive breakpoints)(15.2%);from the analysis of residuals of the multiple regression,the contribution rate of anthropogenic activities(60.61%)was significantly higher than that of climate change(39.39%).(3)Due to the limitations of the traditional RSEI index in arid and semi-arid regi ons,this study innovatively introduced the soil erosion factor(RUSLE)to construct th e IRSEI index for the current situation of soil erosion in the study area.The study sh ows that the IRSEI in Gansu section of the Yellow River Basin generally shows a flu ctuation trend of“increasing-decreasing-rebounding”,in which the stability is higher in the forest area and the fluctuation is significant in the grassland area.From the spatial autocorrelation analysis,IRSEI has significant clustering characteristics,and the high quality areas are mainly distributed in the southeastern forest area and Gannan grassla nd.From the single-factor detection of the Geo Detector,precipitation and relative hu midity had the highest explanatory power(q>0.3)for IRSEI changes,while slope and population density had the lowest explanatory power(q<0.1).The interaction detection yielded that the interaction between land use type and precipitation was the most signi ficant(q=0.60),indicating that in the Gansu section of the Yellow River Basin,where soil erosion is serious,reasonable land use can significantly enhance the ecological be nefits of precipitation resources.更多还原