植被覆盖度（Fractional Vegetation Cover，FVC）通常定义为绿色植被在地面的垂直投影面积占统计区总面积的百分比，是研究地球系统中水圈、大气圈、生物圈以及圈层间相互作用的一个重要植被参数。植被动态生长信息是不同植物种类及植物种群内在规律的表征。已有研究通过动态植被模型引入植被动态生长信息并进一步结合辐射传输模型研发了地表均质区高分辨率植被覆盖度估算方法，有效地提高了植被覆盖度估算精度。该方法假设一个低空间分辨率FVC像元对应区域内的植被生长状况相同，进而由时间序列低分辨率植被覆盖度产品每个像元构建一个动态植被模型，其空间对应的高空间分辨率像元共享此动态植被模型。然而，该方法仅适用于地表均质区的植被覆盖度估算，因为在地表异质区大多数低空间分辨率像元不是纯净像元，而是由各种不同地表组分构成的混合像元。因此，在地表异质区这种动态植被模型构建方法会引入较大的植被覆盖度估算误差。此外，该方法计算效率较低，而且需要覆盖整个植被生长季的遥感数据进行动态植被模型构建，进而导致无法实现近实时植被覆盖度估算。因此，本研究旨在发展结合植被动态生长信息的地表异质区高空间分辨率、高效率、近实时植被覆盖度估算方法，以实现区域尺度高空间分辨率植被覆盖度的高精度估算。 本研究首先研发了地表异质区高空间分辨率植被覆盖度估算方法以适应地表异质区大多数低空间分辨率FVC像元为混合像元的情况。基于低空间分辨率GLASS FVC像元是其空间对应的地表各种地物组分FVC值的线性组合的假设，该方法使用一个与低空间分辨率GLASS FVC像元大小相同的移动窗口，利用一个GLASS FVC像元和移动窗口内所有Landsat 8 OLI像元的NDVI为每个高空间分辨率Landsat 8 OLI像元提取高分辨率FVC先验信息，用于为每个Landsat 8 OLI像元构建各自独立的动态植被模型；然后利用动态贝叶斯网络框架结合动态植被模型和植被辐射传输模型估算Landsat空间尺度的植被覆盖度最优值。基于地面实测数据验证，发现本文发展的植被覆盖度估算方法精度 (R2 = 0.7757, RMSE = 0.0881) 优于地表均质区植被覆盖度估算方法(R2 = 0.7038, RMSE = 0.1125)以及常用的查找表法(R2 =0.7457, RMSE = 0.1249)。然而，这种方法仍然存在计算效率较低，仅适用于包含完整植被生长周期的历史数据分析的局限性。 本研究进一步发展了一种高效率地从低空间分辨率植被覆盖度产品中提取高空间分辨率植被动态生长信息、近实时植被覆盖度高效估算方法。该方法基于植被在某一时刻的生长状态与之前时刻的生长情况紧密联系的假设，结合气候因素、前后时刻的时间间隔大小及其他外部环境因素，建立能够随时间推移，定量FVC状态转移的变化幅度，动态更新模型估算结果并紧密结合至FVC估算方法中的高效率动态植被模型，高效地从低空间分辨率植被覆盖度产品中提取植被动态生长信息。高效率动态植被模型构建无需拟合模型参数，无需包含完整生长周期的植被覆盖度历史数据建立，每当有有效的新观测时便可加入至植被覆盖度估算过程中，能够实现植被覆盖度的高效、近实时估算。验证结果表明，本文提出的方法精度理想（R2=0.889, RMSE=0.0917），和本文第一部分发展的结合韦尔斯特Logistic动态植被模型的植被覆盖度估算方法精度（R2=0.884, RMSE=0.0913）相当。 总之，本研究发展的植被覆盖度估算方法能够在复杂地表情况下，利用植被动态生长信息，高效、近实时的估算高空间分辨率植被覆盖度，精度较高，可靠性强，有潜力应用于更大区域尺度的高空间分辨率植被覆盖度数据集高效、可靠的建立。

Fractional vegetation cover (FVC) is defined as the fraction of green vegetation of the total statistical area in the nadir view. It is an important parameter for studying atmosphere, pedosphere, hydrosphere, biosphere and their interactions in the Earth system. Vegetation dynamic growth information characterizes the inherent laws of different vegetation types and their populations. In previous studies, a high spatial resolution FVC estimation method for homogeneous areas based on the dynamic vegetation model and the radiation transfer model is proposed, and the accuracy of FVC estimation has been improved effectively. In their study, a low spatial resolution FVC pixel was assumed to have the same vegetation growth status in the corresponding region. Then a dynamic vegetation model was constructed from each pixel of the low spatial resolution FVC time series. The corresponding high spatial resolution pixels shared the dynamic vegetation model. However, this method is only suitable for estimating FVC in homogeneous areas, because most low spatial resolution pixels in heterogeneous areas are no longer pure pixels, but mixed pixels composed of various surface components. Therefore, the method of dynamic vegetation model construction in heterogeneous areas will bring large uncertainties. In addition, this method has low computational efficiency, and requires remote sensing data covering the whole vegetation growing season to construct a dynamic vegetation model, which leads to the difficulty of near real-time FVC estimation. Therefore, the purpose of this study is to develop a high spatial resolution, high efficiency, near real-time FVC estimation method in heterogeneous areas combining dynamic vegetation growth information, so as to achieve high-precision FVC estimates with high spatial resolution at regional scale. Firstly, a high spatial resolution FVC estimation method was developed to deal with the status that most of the low spatial resolution FVC pixels in heterogeneous areas were mixed pixels. Based on the assumption that the FVC value of a low spatial resolution GLASS FVC pixel is a linear combination of FVC values of the corresponding surface endmembers, this method first uses a moving window with the same size as the low spatial resolution GLASS FVC pixels, and uses a low spatial resolution GLASS FVC pixel and NDVI of all Landsat 8 OLI pixels in the moving window to extract prior vegetation growth information and construct an independent dynamic vegetation model for each Landsat 8 OLI pixel. Then, the dynamic Bayesian network framework is used to estimate optimal FVC of Landsat spatial scale combing dynamic vegetation model and vegetation radiation transfer model. The accuracy of the FVC estimation method developed in this study (R2 = 0.7757, RMSE = 0.0881) is better than that of the previous methods (R2 = 0.7038, RMSE = 0.1125) and the commonly used look-up table method (R2 = 0.7457, RMSE = 0.1249). However, this method still has some limitations, such as low computational efficiency, and it can only be applied to historical data analysis including complete vegetation growth cycle. This study further developed a time-efficient near real-time FVC estimation method based on extracting dynamic vegetation growth information with high spatial resolution from low-resolution FVC products. Based on the assumption that the vegetation growth state at a certain time is closely related to the growth state at the previous time, combined with climatic factors, time intervals and other external environmental factors, this method can quantitatively determine the change extent of FVC state transition over time, dynamically update the estimation results and integrate them into the algorithm without fitting model parameters. High-efficiency dynamic vegetation model can quickly extract dynamic vegetation characteristics from low-resolution vegetation coverage products. High-efficiency dynamic vegetation model need not be established by historical FVC data including complete growth cycle, and can be added to the process of vegetation coverage estimation whenever there are effective new observations. Validation results indicate that the performance of the proposed method is satisfactory (R2=0.889, RMSE=0.0917) and comparable to previous FVC estimation method in heterogeneous areas incorporating the dynamic vegetation growth model represented by modified Verhulst Logistic equation (R2=0.884, RMSE=0.0913). Therefore, the proposed method can estimate high-resolution FVC efficiently and in near real-time using dynamic vegetation growth information under complex surface conditions, with high accuracy and reliability, and has potential to be applied to the efficient and reliable acquisition of high spatial resolution FVC data at a larger regional scale.