河口湿地受河流与海洋间的相互作用，自然环境错综复杂，资源丰富，是最具有生产力和活力的生态系统之一。盐沼植被作为河口湿地的重要部分，在提供栖息地支持、维护生物多样性、固碳等方面具有重要作用，是三角洲湿地健康的保护者和指示器。黄河三角洲是中国最年轻的湿地，受湿地开发、气候变化等影响，盐沼植被大面积退化，加之互花米草入侵物种扩张，侵占了原生盐沼植被生存空间。因此，研究黄河三角洲近三十年的典型盐沼植被分布格局及演变规律，分析影响盐沼植被分布格局变化的因素，对盐沼湿地的恢复以及河口生态健康的维护具有重要意义。
基于此，本文在分析典型盐沼植被光谱差异并获取特征波段的基础上，结合实地植被坐标提取与遥感分析构建典型盐沼植被分类模型，解决了滨海湿地植被分类准确度不高的技术问题，揭示近三十年黄河三角洲典型盐沼植被分布格局，并进一步揭示典型盐沼植被分布格局演变规律，具体研究内容与结果如下：
（1）典型盐沼植被地物光谱特征分析：通过地物光谱仪实地获取芦苇、互花米草、翅碱蓬、柽柳盐沼植被的光谱曲线，运用一阶导数变换、包络线去除变换并结合马氏距离识别植被可区分的特征波段，结果表明，植被原始光谱曲线变化趋势相似，难以进行种类的区分；一阶导数光谱虽无法直接区分各类植被，但根据红边参数及光谱特征参数可以识别植被的不同生长阶段；包络线去除可有效增加可区分特征波段，使其分布范围发生变化，由原始光谱曲线的集中于近红外和SWIR2特征波段转变为在可见光波段、近红外波段和短波红外SWIR1、SWIR2波段处均有分布。
（2）基于多源遥感数据的典型盐沼植被遥感反演模型构建：通过无人机与RTK获取植被分布坐标；获取1991-2022年期间可使用的Sentinel-1、Sentinel-2、Landsat系列图像，并利用2019年及2022年的Sentinel-1、Sentinel-2、Landsat系列图像提取光谱波段并计算特征指数，运用随机森林、支持向量机、梯度提升决策树三种算法进行反演，比较其性能参数；对Sentinel-1、Sentinel-2构建基于月际变化的反演模型，对Landsat系列构建基于季节变化的反演模型，生成近三十年黄河三角洲植被分布格局图。从整体分类结果上看，随机森林分类效果最优，支持向量机的精度最低，基于月际反演的随机森林算法总体精度为91.80%，Kappa系数为0.9024；随机森林模型的重要特征变量主要位于植被生长旺盛期（8月），而生长初期（5月）和生长衰退期（10月）次之，SAR-VH波段、NDBI、NDVI、EVI2指数的重要性最高。采取重要特征变量可有效提高遥感的分类结果。
（3）典型盐沼植被时空格局演变规律分析：基于反演的近三十年黄河三角洲植被分布格局，结合植被面积转移矩阵、景观格局指数分析方法、植被景观重心转移模型，从类型水平和整体水平两方面对植被景观格局进行了长期分析，探究湿地植被景观格局空间位置的变化及演变的规律。结果表明：现行黄河口保护区的湿地植被在空间上呈条带状交替分布，斑块间连通性降低、形状趋于简单化；芦苇群落是优势物种，总体面积平稳增长，景观重心变化不大；互花米草的景观重心逐渐趋向于河口附近，自2009年起扩张速度加快，优势度提高，在2017年后增长逐渐平缓；翅碱蓬破碎程度较高，景观重心自2009年呈现后退趋势；柽柳植被交错带的形状更为简单，抵抗人类活动影响的能力更差，受黄河改道影响，景观重心整体向东南方向的迁移。

Estuarine wetlands are influenced by the interaction between rivers and oceans, and have complex natural environments. They are one of the most productive and dynamic ecosystems, and salt marsh vegetation, as an important part of estuarine wetlands, plays an important role in providing habitat support, maintaining biodiversity, carbon sequestration, etc. It is the protector and indicator of the health of the wetland in the Yellow River Delta. Affected by wetland development, climate change and other factors, salt marsh vegetation has degraded on a large scale, and Spartina alterniflora, an invasive species, has expanded and occupied the living space of native salt marsh vegetation. Therefore, it is of great significance to study the distribution pattern and evolution law of typical salt marsh vegetation in the Yellow River Delta in the past 30 years, and to analyze the factors affecting the distribution pattern change of salt marsh vegetation for the restoration of salt marsh wetlands and the maintenance of estuarine ecological health.
Based on this, this paper analyzes the spectral differences of typical salt marsh vegetation and obtains characteristic bands, combines field vegetation coordinates extraction and remote sensing analysis to construct a classification model of typical salt marsh vegetation, solves the technical problem of low accuracy of vegetation classification in coastal wetland, reveals the distribution pattern of typical salt marsh vegetation in the Yellow River Delta in the past 30 years, and further reveals the evolution law of typical salt marsh vegetation distribution pattern. The specific research contents and results are as follows: 
(1) Spectral characteristics analysis of typical salt marsh vegetation objects: The spectral curves of Phragmites australis, Spartina alterniflora, Suaeda salsa and Tamarix chinensis salt marsh vegetation were obtained by field spectrometer. The first-order derivative transform and envelope removal transform were used to identify the distinguishable characteristic bands of vegetation based on Mahalanobis distance. The results show that the original spectral curves of vegetation change similarly and are difficult to distinguish; although the first-order derivative spectra cannot directly distinguish different types of vegetation, different growth stages of vegetation can be identified according to red edge parameters and spectral characteristic parameters; envelope removal can effectively increase distinguishable characteristic bands, making their distribution range change from original spectral curves concentrated in near-infrared and SWIR2 characteristic bands to visible light bands, near-infrared bands and short-wave infrared SWIR1 and SWIR2 bands.
 (2) Construction of remote sensing inversion model for typical salt marsh vegetation based on multi-source remote sensing data: The coordinates of vegetation distribution were obtained by UAV and RTK; Sentinel-1, Sentinel-2 and Landsat series images available from 1991 to 2022 were obtained. Spectral bands were extracted and characteristic indices were calculated. Three algorithms: gradient boosting decision tree, random forest, support vector machine were used for inversion. Their performance parameters were compared. A monthly variation-based inversion model was constructed for Sentinel-1 and Sentinel-2. A seasonal variation-based inversion model was constructed for Landsat series. The vegetation distribution pattern map of the Yellow River Delta in the past 30 years was generated. From the overall classification results, random forest is better than gradient boosting decision tree, while support vector machine has the lowest accuracy; The important feature variables of random forest model are mainly located in the period of vigorous growth of vegetation (August), followed by the initial stage (May) and decline stage (October) of growth. SAR_VH band, NDBI, NDVI and EVI2 indices have the highest importance. Taking important feature variables can effectively improve the classification results of remote sensing.
(3) Analysis of spatio-temporal pattern evolution of typical salt marsh vegetation: Based on the inversion of the Yellow River Delta vegetation distribution pattern map in the past 30 years, combined with vegetation area transfer matrix, landscape pattern index analysis method, vegetation landscape center of gravity transfer model, the long-term analysis of vegetation landscape pattern from the type level and the overall level, to explore the spatial position change and evolution of wetland vegetation landscape pattern. The results showed that the current wetland vegetation in the Yellow Estuary protected area was distributed alternately in strips, and the connectivity between patches decreased and the shape tended to be simplified. Phragmites australis community is the dominant species, the overall area increased steadily, but the landscape center of gravity changed little. The landscape center of gravity of Spartina alterniflora gradually tended to be near the estuary. Since 2009, the expansion speed was fast and the dominance degree was improved. After 2017, the growth rate was gradually gentle. The fragmentation degree of Suaeda salsa was higher, and the center of gravity of landscape showed a receding trend from 2009. The shape of Tamarix chinensis vegetation crisscross is simpler and its ability to resist human activities is worse. Influenced by the diversion of the Yellow River, the landscape center of gravity gradually moves to the southeast.