土壤水分和植被是地表與大氣相互作用的關鍵參數。作為土壤蒸發和冠層蒸騰的總和，可以使用地表熱通量模型和地表溫度數據來估算蒸散。通常採用熱紅外（TIR）的衛星圖像來檢索遙感技術中的地表溫度數據。 Landsat-8 OLI和TIR傳感器已廣泛用於提供30米空間分辨率的地表溫度數據。然而，溫度，濕度，風速和土壤濕度等環境因素的快速變化會影響蒸散的動態。因此，蒸發蒸騰的估算除了高空間分辨率之外還需要高時間分辨率，用於日常，晝夜或甚至每小時分析。另一方面，Landsat-8的時間分辨率為16天。因此，更高空間分辨率的衛星傳感器具有較低的時間分辨率是一個挑戰，反之亦然。先前的研究通過開發空間和時間自適應反射融合模型（STARFM）解決了這種限制。此外，由於一些其他限制，STARFM算法被修改和擴展為增強的空間和時間自適應反射融合模型（ESTARFM）。 STARFM和ESTARFM通常用於融合可見圖像。對於相同的衛星傳感器，與可見圖像相比，TIR圖像的空間分辨率通常更粗糙。因此，對TIR圖像採用時空圖像融合方法成為一項挑戰。在這項研究中，通過融合Himawari-8和Landsat-8的TIR圖像，改進了最初的STARFM以合成具有10分鐘空間分辨率的Landsat等新TIR圖像。我們的目的是比較STARFM，ESTARFM和STARFM在融合Himawari-8和Landsat-8 TIR波段時的性能。最後，改進的STARFM顯示了具有30米空間分辨率的每小時地表溫度產品的有希望的結果，其適用於估計具有高空間和時間分辨率的實際蒸發蒸騰。;Soil moisture and vegetation are key parameters in land-atmosphere interactions. As the sum of soil evaporation and canopy transpiration, Evapotranspiration (ET) can be estimated using land surface heat flux models and land surface temperature (LST) data. Thermal infrared (TIR) satellite images are generally employed to retrieve LST data in remote sensing. Landsat-8 operational land imager (OLI) and thermal infrared sensors (TIRS) can provide LST data with 30 m spatial resolution that have been widely used. However, rapid changes in environmental factors such as temperature, humidity, wind speed and soil moisture will affect to dynamic of ET. Therefore, ET estimation needs high temporal resolution beside high spatial resolution for daily, diurnal or even hourly analysis. On the other hand, temporal resolution of Landsat-8 is 16 days. Thus, it has been a challenge that higher spatial resolution satellite sensors have lower temporal resolution, and vice versa. Previous study has solved such kind of limitation by developing a spatial and temporal adaptive reflectance fusion model (STARFM). In addition, due to some other limitations, the STARFM algorithm was modified and extended as enhanced spatial and temporal adaptive reflectance fusion model (ESTARFM). Both STARFM and ESTARFM are commonly used for fusing visible images. For the same satellite sensors, TIR images are generally coarser in spatial resolution compared to visible images. Therefore, it becomes a challenge to employ the spatiotemporal image fusion methods for TIR images. In this study, the original STARFM was improved to synthesize Landsat-like TIR images with 10 minutes spatial resolution by fusing TIR images of Himawari-8 advanced Himawari imager (AHI) and Landsat-8 TIRS. We aim to compare the performances of STARFM, ESTARFM and an improvement of STARFM in blending Himawari-8 and Landsat-8 TIR bands. Finally, the improved STARFM shows the promising result of hourly LST products with 30 m spatial resolution that are applicable in estimating actual ET with high spatial and temporal resolutions.
