Identification of Worldwide Optimal Pseudo-Invariant Calibration Sites for Post-Launch Radiometric Calibration of Earth Observation Satellite Sensors

The primary objective of this project was to identify extremely stable sites on the Earth's surface known as Pseudo-Invariant Calibration Sites (PICS). A recently developed technique for monitoring the long term stability of earth observing satellite sensors was based on using PICS for detecting trends in the radiometric response of these instruments. In a manner analogous to using a known reflectance or radiance source in a laboratory, this method relied on the stability of the Earth's surface over time. To perform this task, the Landsat 5 Thematic Mapper (TM) sensor was used to identify the most invariant locations or PICS on the Earth's surface by monitoring the temporal stability of carefully selected ground sites on Earth. Ground sites were selected to ensure minimal surface and atmosphere change that could affect the observed reflectance, thus enabling a means to monitor the radiometric stability of space instruments. PICS mainly consist of playa (dry lakebeds), salt flats and desert sand sites located in arid regions with low probability of cloud cover, spatial homogeneity, constant surface spectral reflectance and BRDF over short and long periods of time.;Potential PICS were evaluated and chosen for the study based upon their size, location, climate characteristics, and scene availability in the USGS data archive. A grid-based approach was used to determine and recommend the areas of each PICS that was considered most invariant. This approach relied on the PICS min-noise algorithm developed recently at SDSU, where the mean radiance of each grid was calculated for each scene and the grid with lowest temporal standard deviation of the mean was considered as most invariant. The Levene Test of equality of variance was used to optimize the size of worldwide PICS, and uncertainties using those optimal locations were calculated for comparison. A catalog of recommended sites was developed: seven in the Sahara Desert and one each in North America, South America, Australia, Middle East and Asia. Sahara sites exhibit uncertainties of 2% or less in VNIR and 2--3% in SWIR across most sites. Global sites exhibit uncertainties in the range of 1--5%. Finally, as a validation exercise, Landsat-7 Enhanced Thematic Mapper plus (ETM+) reflective band radiometric gain trends were estimated using the optimal locations of four worldwide PICS.