Improvement of the Sub-pixel Weighted Algorithm for Retrieving Pixel Surface Emissivity and its Application

As well known, surface emissivity is one of the key parameters in the retrieval of surface temperature. The algorithms and the measurements about it are always the focus in the field of thermal remote sensing study. Among these methods for acquiring emissivity on pixel scale, day/night algorithm is used widely, for example, MODIS LST product; on the other hand, sub-pixel weighted method is an operational algorithm. Because the emissivities of the soil vary with different water contents, it is necessary to adopt the concept of relative thermal inertia to account for this effect. In addition, other influencing factors, such as type of soil, structure of soil and vegetation cover can also lead to different emissivity. In order to optimize this algorithm further, we did the experiments using an automatic field observation system to retrieve the component emissivity of mixed ground objects in November, 2005 developed by our group. In the experiment, the observed objects were composed of four sub-pixel components which have different combinations of soil content, soil type and vegetation cover. Then, the revised algorithm and the day/night algorithm to MODIS data are compared. Similar results were found in the two experimental sites. Since the day/night method requires day and night remote sensing data in a same day, it is difficult to be applied to TM and NOAA-AVHRR data, while the new sub-pixel weighted method will be a good choice.