Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 2/16/2001
Publication Date: N/A
Citation: N/A Interpretive Summary: The Landsat satellites have supported a series of Thematic Mapper (TM) sensors that could soon span a period of 32 years of continuous earth observation. This long-term set of multi-spectral images provides an ideal research tool for monitoring land cover and land use change and the effects of anthropogenic activity. To best exploit these images for temporal studies, the images need to be normalized to a surface parameter that is unaffected by sensor signal drift and changing atmospheric conditions. An approach has been developed to convert the TM signal to values of surface reflectance which indicate changes in soil and plant conditions over time. This approach can be used operationally for future and archived images because it is based only on general atmospheric theory, minimal one-time ground measurements, and information derived from the image itself. This normalization approach has proven accurate for a 10-year time series of Landsat images for a rangeland site in southeast Arizona. With this simpl and accurate means to convert Landsat TM signal to surface reflectance, it will be possible to study regional land changes and make better resource management decisions to best utilize and sustain our limited agricultural resources.
Technical Abstract: The recent launch of Landsat-7 ETM+ extends the uninterrupted stream of TM and ETM+ images to a potential span of 32 years. This exceptional image set will allow long-term studies of natural resources, but will require an operational method for converting image digital number (dn) to the temporally-comparable surface reflectance factor. A refinement to the empirical line (EL) approach for reflectance factor retrieval (RFR) from the Landsat-5 and -7 TM and ETM+ has been proposed. The refined empirical line (REL) approach requires only one within-scene calibration target, minimal field measurements of that target, and a reasonable estimate of dn for zero reflectance using a radiative transfer model or value provided by this analysis. This study showed that the REL approach worked wellfor a 10-year Landsat-5 TM and Landsat-7 ETM+ image set in Arizona and was retrieved with an estimated accuracy of 0.01. A quantitative approach was proposed to determine the suitability of a within-scene target for the REL approach, and based on historical measurements, a variety of targets met the size and brightness requirements for the REL approach. This operational approach for RFR should encourage long-term investigation of natural resources to answer critical questions regarding resource management and effects of climate changes.