The estimation of biophysical variables is at the core of remote sensing science, allowing a close monitoring of crops and forests. Deriving temporally resolved and spatially explicit maps of parameters of interest has been the subject of intense research. However, deriving products from optical sensors is typically hampered by cloud contamination and the trade-off between spatial and temporal resolutions. In this work we rely on the HIghly Scalable Temporal Adaptive Reflectance Fusion Model (HISTARFM) algorithm to generate long gap-free time series of Landsat surface reflectance data by fusing MODIS and Landsat reflectances. An artificial neural network is trained on PROSAIL inversion to predict monthly biophysical variables at 30 m spatial resolution with associated, realistic uncertainty bars. We emphasize the need for a more thorough analysis of uncertainty, and propose a general and scalable approach to combine both epistemic and aleatoric uncertainties by exploiting Monte Carlo (MC) dropout techniques from the trained artificial network and the propagation of HISTARFM uncertainties through the model, respectively. A model recalibration was performed in order to provide reliable uncertainties. We provide new high resolution products of several key variables to quantify the terrestrial biosphere: Leaf Area Index (LAI), Fraction of Absorbed Photosynthetically Active Radiation (FAPAR), Canopy Water Content (CWC) and Fractional Vegetation Cover (FVC) are at 30 m Landsat spatial resolution and over large continental areas. Two study areas are considered: the large heterogeneous but moderately cloud covered contiguous United States, and the homogeneous but largely cloud covered Amazonia. The produced vegetation products largely agree with the test dataset (R = 0.90, RMSE = 0.80 m²/m² and ME = 0.12 m²/m² for LAI, and R = 0.98, RMSE = 0.07 and ME = 0.01 for FAPAR) providing low error and high accuracy. Additionally, the validation considers a thorough comparison with operational and largely validated medium resolution products, such as the Moderate-Resolution Imaging Spectroradiometer (MODIS) and Copernicus Global Land Service. Our products presented a good agreement and consistency with both MODIS (R = 0.84 and R = 0.85 for LAI and FAPAR, respectively) and Copernicus (R = 0.92 and R = 0.91 for LAI and FAPAR, respectively). To foster a wider adoption and reproducibility of the methodology we provide an application in GEE and source code at:https://github.com/IPL-UV/ee_BioNet/

生物物理变量的估计是遥感科学的核心，可以对农作物和森林进行密切监测。导出感兴趣参数的时间分辨和空间显式地图一直是深入研究的主题。然而，从光学传感器获取产品通常受到云污染以及空间和时间分辨率之间的权衡的阻碍。在这项工作中，我们依靠高度可扩展的时间自适应反射率融合模型 (HISTARFM) 算法通过融合 MODIS 和 Landsat 反射率来生成 Landsat 表面反射率数据的长无间隙时间序列。人工神经网络在 PROSAIL 反演上进行训练，以预测 30 m 空间分辨率的每月生物物理变量，并带有相关的、现实的不确定性条。我们强调需要对不确定性进行更彻底的分析，并提出一种通用且可扩展的方法，通过利用训练有素的人工网络中的蒙特卡洛 (MC) 辍学技术和通过模型传播 HISTARFM 不确定性来结合认知不确定性和任意不确定性，分别。为了提供可靠的不确定性，进行了模型重新校准。我们提供几个关键变量的新高分辨率产品来量化陆地生物圈：叶面积指数 (LAI)、吸收的光合有效辐射分数 (FAPAR)、冠层含水量 (CWC) 和植被覆盖率 (FVC) 为 30 m Landsat 空间分辨率和大片大陆地区。考虑了两个研究区域：美国本土大面积异质但云层覆盖适中的区域，和同质但大部分云层覆盖的亚马逊地区。生成的植被产品与测试数据集基本一致（R  = 0.90，RMSE = 0.80 m ² / m ²和 ME = 0.12 m ² / m ²用于 LAI，R  = 0.98，RMSE = 0.07 和 ME = 0.01 用于 FAPAR）提供低误差和高精度。此外，该验证还考虑与可操作且经过大量验证的中等分辨率产品进行彻底比较，例如中分辨率成像光谱仪 (MODIS) 和哥白尼全球陆地服务。我们的产品与 MODIS（ LAI 和 FAPAR 分别为R  = 0.84 和R = 0.85）和哥白尼（ R  = 0.92 和R = LAI 和 FAPAR 分别为 0.91）。为了促进该方法的更广泛采用和可重复性，我们在 GEE 和源代码中提供了一个应用程序：https://github.com/IPL-UV/ee_BioNet/