大氣氣膠光學厚度(Aerosol Optical Depth, AOD)是空氣品質的重要指標之一，且具明顯的時間與空間分布變化之特性。對於大氣氣膠的觀測，除地面定點的測站量測外，衛星的觀測可獲取大範圍的AOD資訊，然而受限於衛星軌道與空間觀測特性，利用單一衛星將無法獲得完整的高時空解析之觀測。為了得到高時空解析的衛星觀測影像，前期研究改進Spatial-Temporal Adaptive Reflectance Fusion Mode (STARFM)方法，結合高時間解析的Himawari-8影像與高空間解析的Landsat-8影像之優勢，將大氣效應納入融合過程，以獲得高時空解析的大氣層頂反射率融合影像。再用Simultaneous Radiation Solution(SRS)反演獲取高時空解析大氣氣膠光學厚度資訊。為有效應用於實際觀測作業，本研究將建構自動化影像融合過程，並進行長期驗證，以確認此方法之可行性。前期反演方法經個案測試呈現，在太陽天頂角較大時段(早晨跟黃昏)，AOD會有高估的現象，因此本研究假設兩張參考時間點影像調整到相似的亮度，並建構地表雙向反射查找表以加入雙向地表反射率(bidirectional reflectance distribution function, BRDF)模型，檢驗AOD反演誤差之改進。研究結果顯示，調整後之參考影像亮度與加入雙向地表反射率之修正，可有效地改進太陽天頂角較大時段AOD之反演誤差，以2019/4/23個案為例，高時空解析融合影像所反眼AOD最大相對誤差從86%降到44%，最小相對誤差也從16%降到0.17%，獲得相當的改進成效。因為本研究在不同地物時都用相同的異向參數，未來可以改善地物分類，然後根據地物設定不同的異向參數。希望未來能應用在反演氣膠種類跟監測短周期的污染事件。;Aerosol Optical Depth (AOD) is a pivotal indicator of air quality which highly varies in spatial and temporal. Due to the characteristics of satellite orbit and spatial observation, acquiring AOD in high spatiotemporal resolution by a single satellite is not viable. Therefore, this study proposed an automatic image fusion technique that employs a modified Spatial-Temporal Adaptive Reflectance Fusion Model (STARFM) algorithm to synthesize Landsat and Himawari imagery for high spatiotemporal AOD retrieval. Bidirectional reflectance distribution function (BRDF) was incorporated in the model for adjusting the brightness of two reference time points, which could alleviate the overestimated AOD values retrieved in the morning and the sunset (i.e., period of large solar zenith angle). An additional look-up table storing fine surface reflectivity was built for assisting this procedure. The results showed a significant reduction of maximum and minimum relative error to 44% and 0.17%, respectively. This automatic model presented a promised method for short-term air quality monitor. Future studies could be improved by more meticulous land cover data and associated anisotropy parameters.
