Use of Pseudo-Invariant Sites for Long-Term Sensor Calibration

Over the past several years, pseudo-invariant sites for trending the radiometric gain of multispectral satellite imaging systems has proven to be very useful and accurate. For example, the lifetime calibration of the Landsat 5 Thematic Mapper (TM) was recently updated based in large part upon this approach. This effort reduced the error in the gain estimate from roughly 13 percent to less than 5 percent and integrated a variety of absolute calibration estimates for the sensor. Typically, pseudo-invariant sites are located in arid regions where very little temporal and spatial change occurs over long periods of time and significant extent. Sites in the Sahara desert are especially attractive. However, drawbacks of this approach are that the sites are relatively unacceptable and, for some instruments, only limited numbers of data sets are available. As a result, it would be helpful to use sites that are observed often, such as locations in North America, so that the technique could be applied to more sensors both historically and in the future. The drawback to these sites is that they are typically of much smaller extent and, therefore, are more difficult to use from a geometric perspective. Recent work has shown that small pseudo-invariant sites, if carefully chosen, can also be used for accurate trending of satellite instruments. This paper will illustrate an approach that suggests the calibration site commonly known as Sonoran Desert can be used to observe the characteristic calibration curve of Landsat 5 TM. Because of the significant number of observations of this site, and other similar sites in North America, there is good potential to extend this technique to additional multispectral instruments for enhanced long term monitoring of radiometric gain.