时空无缝的SMAP植被光学厚度逐日产品（2015-2022）

植被光学厚度是衡量植被对微波辐射衰减作用的重要物理量，反映了植被的密度与生长状况，在生物多样性评估等研究中具有关键意义。使用多通道协作算法反演的SMAP植被光学深度数据集具有较长的时间序列，但由于卫星轨道扫描间隙限制，无法实现对全球陆地的完全覆盖。本数据集基于三维部分卷积神经网络模型，对多通道协同反演算法生成的MCCA SMAP植被光学厚度产品进行时空无缝填补。通过更新掩膜，该模型可以提取有效区域的特征，同时忽略无效区域，从而提高效率。通过无偏根均方误差，相关系数，偏差对重建的VOD进行了时空评估，结果表明，重构的VOD数据集具有更高的覆盖率、准确性和可靠性。此外，重构的VOD能更好地反映叶片水势的昼夜变化，有利于与干旱和植被生态系统有关的各种研究。

Vegetation Optical Depth (VOD) is an important physical quantity that quantifies the attenuation effect of vegetation on microwave radiation, reflecting vegetation density and growth status, and is of critical significance in research such as biodiversity assessment. The SMAP Vegetation Optical Depth dataset inverted via the multi-channel collaborative inversion algorithm has a long time series, yet it cannot achieve complete global land coverage due to limitations of satellite orbital scan gaps. This dataset adopts a 3D partial convolutional neural network model to perform spatiotemporal seamless gap filling on the MCCA SMAP Vegetation Optical Depth product generated by the aforementioned multi-channel collaborative inversion algorithm. By updating the mask, the model can extract features from valid regions while ignoring invalid ones, thereby improving computational efficiency. The reconstructed VOD was evaluated spatiotemporally using unbiased root mean square error (ubRMSE), correlation coefficient, and bias. The evaluation results demonstrate that the reconstructed VOD dataset exhibits higher coverage, accuracy, and reliability. Furthermore, the reconstructed VOD can better reflect the diurnal variation of leaf water potential, facilitating various studies related to drought and vegetation ecosystems.

搜集汇总

背景与挑战

背景概述

该数据集提供了2015年至2022年全球范围内的SMAP植被光学厚度逐日产品，空间分辨率为10km至100km，时间分辨率为日。通过三维部分卷积神经网络模型对原始数据进行时空无缝填补，提高了数据的覆盖率、准确性和可靠性，适用于生物多样性评估、干旱和植被生态系统研究。