The Yonsei Aerosol Retrieval (YAER) algorithm for the Geostationary Ocean Color Imager (GOCI) retrieves aerosol optical properties only over dark surfaces, so it is important to mask pixels with bright surfaces. The Advanced Himawari Imager (AHI) is equipped with three shortwave-infrared and nine infrared channels, which is advantageous for bright-pixel masking. In addition, multiple visible and near-infrared channels provide a great advantage in aerosol property retrieval from the AHI and GOCI. By applying the YAER algorithm to 10 min AHI or 1 h GOCI data at 6 km×6 km resolution, diurnal variations and aerosol transport can be observed, which has not previously been possible from low-Earth-orbit satellites. This study attempted to estimate the optimal aerosol optical depth (AOD) for East Asia by data fusion, taking into account satellite retrieval uncertainty. The data fusion involved two steps: (1) analysis of error characteristics of each retrieved result with respect to the ground-based Aerosol Robotic Network (AERONET), as well as bias correction based on normalized difference vegetation indexes, and (2) compilation of the fused product using ensemble-mean and maximum-likelihood estimation (MLE) methods. Fused results show a better statistics in terms of fraction within the expected error, correlation coefficient, root-mean-square error (RMSE), and median bias error than the retrieved result for each product. If the RMSE and mean AOD bias values used for MLE fusion are correct, the MLE fused products show better accuracy, but the ensemble-mean products can still be useful as MLE.

中文翻译：

---

在 2016 年 KORUS-AQ 和 2018 年 EMeRGe 活动期间整合 GOCI 和 AHI Yonsei 气溶胶光学深度产品

地球静止海洋彩色成像仪 (GOCI) 的延世气溶胶检索 (YAER) 算法仅检索暗表面上的气溶胶光学特性，因此用亮表面屏蔽像素非常重要。Advanced Himawari Imager (AHI) 配备三个短波红外和九个红外通道，有利于亮像素掩蔽。此外，多个可见光和近红外通道在从 AHI 和 GOCI 中检索气溶胶特性方面具有很大优势。通过将 YAER 算法应用于6 km×6 km 的10 min AHI 或 1 h GOCI 数据可以观察到分辨率、昼夜变化和气溶胶传输，这是以前低地球轨道卫星无法实现的。本研究试图通过数据融合估计东亚的最佳气溶胶光学深度（AOD），同时考虑到卫星反演的不确定性。数据融合包括两个步骤：（1）分析每个检索结果相对于地基气溶胶机器人网络（AERONET）的误差特征，以及基于归一化差异植被指数的偏差校正，以及（2）汇编融合产品使用集成均值和最大似然估计（MLE）方法。融合结果在预期误差内的分数、相关系数、均方根误差 (RMSE)、和每个产品的检索结果的中值偏差误差。如果用于 MLE 融合的 RMSE 和平均 AOD 偏差值是正确的，则 MLE 融合产品显示出更好的准确性，但集成均值产品仍然可以用作 MLE。