Short term calibration of Landsat TM: recent findings and suggested techniques

Landsat Thematic Mapper (TM) data has traditionally been radiometrically calibrated on an individual scene basis. While this may be adequate for many applications, particularly single date applications, there are many cases where more advanced techniques are necessary. With emphasis today on global change, multi-temporal analysis is common place. Thus, forms of radiometric calibration are required that discriminate temporal change from changes in instrument response. To address this issue, a project is underway to evaluate TM calibration at several different time scales: short term, intermediate term, and lifetime. This paper presents some of the short term results. Short term is defined as within a single orbit. Continuous swaths of night data have been collected and the individual scenes, along with calibration files, stitched together. Analysis of the calibration files has revealed a number of interesting characteristics about detector gain and bias as well as instrument anomalies such as memory effect and scan correlated shift. Detector bias is a rapidly varying parameter that requires evaluation and correction as frequently as every scan. Detector gain varies significantly within an orbit in the primary focal plane in a predictable manner. Scan correlated shift has been accurately quantified and its effects may easily be removed from the data. Memory effect is in the process of being characterized and an algorithm is being developed to remove it. Line droop has been shown to simply be due to memory effect. Results from these investigations are being used in the development of radiometric correction algorithms for the Landsat 7 Image Assessment System