akalenda
commented
7 years ago From https://landsat.usgs.gov/how-are-thermal-infrared-sensor-tirs-thermal-bands-aboard-landsat-8-used :
Since the launch of Landsat 8 in 2013, thermal energy from outside the normal field of view (stray light) has affected the data collected in Bands 10 and 11 of the Thermal Infrared Sensor (TIRS). This stray light increases the reported temperature by up to four degrees Kelvin (K) in Band 10 and up to eight K in Band 11. The errors vary throughout the scene and depend upon radiance outside the instrument field of view, which users cannot correct in the Landsat Level-1 data product.
The amount of stray light in the scenes is estimated using scenes acquired before and after the target scene, as well as the edge pixels of the target scene. The stray light estimate is then subtracted from the target scene. Users are cautioned that the results after this correction may still not be considered precise enough for a split-window algorithm application and it is recommended that users refrain from using band 11 data in science studies.
For more information see: (Montanaro, M.; Gerace, A.; Lunsford, A.; Reuter, D. Stray Light Artifacts in Imagery from the Landsat 8 Thermal Infrared Sensor. Remote Sens. 2014, 6, 10435-10456)
More details about the stray light are documented in Appendix A of the Landsat 8 Data User Handbook.
The original fmask used a single TIRS band for brightness temperature, whereas Landsat8 provides two bands that together cover a wider spectrum. Currently the code just uses reflectance from the first of the TIRS bands.
A simple way to use both is to average the two and use that in BT (brightness temperature) calculations. But that is a naive approach. The two bands respond differently to different surfaces (clouds, water, land, vegetation, rocks, etc), and so we should be able to use them to improve our accuracy, which will require experimentation.