Synergy of Satellite- and Ground-Based Aerosol Optical Depth Measurements Using an Ensemble Kalman Filter Approach
Synergy of Satellite- and Ground-Based Aerosol Optical Depth Measurements Using an Ensemble Kalman Filter Approach
复制标题
使用集成卡尔曼滤波器方法进行卫星和地面气溶胶光学深度测量的协同作用
DOI：
10.1029/2019jd031884
复制
发表时间：
2020
影响因子：
4.4
通讯作者：
Nakajima Teruyuki
中科院分区：
文献类型：
--
作者：
Li Jing;Kahn Ralph A.;Wei Jing;Carlson Barbara E.;Lacis Andrew A.;Li Zhanqing;Li Xichen;Dubovik Oleg;Nakajima Teruyuki
卫星和地面遥感是两种广泛使用的测量气溶胶特性的技术。然而，这两种方法都不是完美的，因为卫星检索受到各种不确定因素的影响，地面观测的空间覆盖范围有限。在这项研究中，着眼于提高大尺度气溶胶信息的估计，我们开发了一种基于集合卡尔曼滤波（EnKF）的数据协同技术，有效地将这两种类型的测量联合收割机，并产生一个月平均气溶胶光学厚度（AOD）产品与全球覆盖和提高精度。我们首先使用11个月平均的AOD数据集构建了一个474成员的集合，以代表AOD场的变化。然后选择中分辨率成像光谱仪AOD反演作为背景场，使用EnKF同化来自135个气溶胶机器人网络站点的地基测量值。与卫星资料相比，组合场的偏差和均方根误差大大减小，相关系数大大提高。此外，交叉验证表明，在地面观测未被同化的位置，均方根误差和偏差的减少以及相关性的增加仍然可以达到~ 20%。气溶胶光学厚度的空间代表性大或站点密度高的位置通常是变化最大的地方。这项研究表明，EnKF技术有效地将地面站点获得的信息扩展到更大的区域，为结合不同类型的测量信息提供了途径，从而更好地估计气溶胶特性及其时空变异性。
Satellite‐ and ground‐based remote sensing are two widely used techniques to measure aerosol properties. However, neither is perfect in that satellite retrievals suffer from various sources of uncertainties, and ground observations have limited spatial coverage. In this study, focusing on improving estimates of aerosol information on large scale, we develop a data synergy technique based on the ensemble Kalman filter (EnKF) to effectively combine these two types of measurements and yield a monthly mean aerosol optical depth (AOD) product with global coverage and improved accuracy. We first construct a 474‐member ensemble using 11 monthly mean AOD data sets to represent the variability of the AOD field. Then Moderate Resolution Imaging Spectroradiometer AOD retrievals are selected as the background field into which ground‐based measurements from 135 Aerosol Robotic Network sites are assimilated using the EnKF. Compared with satellite data, the bias and root‐mean‐square errors of the combined field are greatly reduced, and correlation coefficients are greatly improved. Moreover, cross validation shows that at locations where surface observations were not assimilated, the reduction in root‐mean‐square error and bias and the increase in correlation can still reach ~20%. Locations where the spatial representativeness of AOD is large or the site density is high are where the greatest changes are typically found. This study shows that the EnKF technique effectively extends the information obtained at surface sites to a larger area, paving the way for combining information from different types of measurements to yield better estimates of aerosol properties as well as their space‐time variability.