Derivation of Landsat 5 Tm Detector Relative Gain Models Using the Usgs Image Assessment System (ias)

Historical efforts to gather and utilize Landsat Thematic Mapper radiometric information from Earth images and the sensor internal calibration systems were typically manual in nature and hence limited in terms of data sampling and subsequent analysis. Multiple government agencies and private corporations operated the Landsat TM missions over their respective lifetimes resulting in a sparse sporadic collection of information in terms of operating conditions, failures, etc. These detrimental external factors pointed to the need for an automated analysis system to study Landsat data and improve product quality. The Image Assessment System (developed by NASA, USGS, SDSU, and other partners) has recently been expanded to incorporate the Landsat 4 & 5 TM instruments [1]. Ongoing processing via the IAS, and the public data production system (LPGS) has resulted in a database of calibration information exceeding 120,000 scenes that is being actively mined for historical information on the performance of the Landsat TM sensors. This paper will elaborate on the data mining effort and discuss in detail the derivation of L5 TM detector relative gain models, used to remove detector stripping. Many historical efforts have been made (typically using limited data) to model Landsat 5 TM relative gains. One such extensive and recent study performed without benefit of the IAS which can be used in a baseline comparative sense to compare / assess calibration improvements is: [2, 3].