Abstract
Radiometric cross-calibration of Earth Observation sensors is a crucial need in order to guarantee or quantify the consistency of measurements from a given sensor with measurements from other sensors. In this context, many applications are concerned about possible radiometric bias between sensors. Twenty desert sites, selected in a previous study, are revisited and their radiometric profiles are described for the visible to near infrared spectral domain. Therefore, acquisitions by various sensors over these 20 desert sites are collected into a dedicated database, SADE, defined to manage operational calibrations and the required SI-traceability. A cross-calibration method over desert sites is detailed. Starting from measurements by a reference sensor, surface reflectance are computed through an atmospheric correction step. A spectral interpolation is then performed to derive the surface reflectance for spectral bands of the sensor to calibrate. Then the atmospheric contribution is added to compute top-of-atmosphere reflectance. The comparison to reflectance acquired by the sensor to calibrate provide an estimation of the cross-calibration with the reference sensor. Results illustrate the efficiency of the method for various pairs of sensors based on archive for AQUA-MODIS, ENVISAT-MERIS, PARASOLPOLDER, and SPOT5-VEGETATION. One main limitation identified in the discussion is due to the spectral interpolation on the surface reflectance which can lead to a few percents bias depending on consistency between spectral bands from both sensor to calibrate and reference sensor. The improvement section identify in progress topics that will improve the method.
Cross-Calibration over Desert Sites: Description, Methodology and Operational Implementation
Radiometric cross-calibration of Earth Observation sensors is a crucial need in order to guarantee or quantify the consistency of measurements from a given sensor with measurements from other sensors. In this context, many applications are concerned about possible radiometric bias between sensors. Twenty desert sites, selected in a previous study, are revisited and their radiometric profiles are described for the visible to near infrared spectral domain. Therefore, acquisitions by various sensors over these 20 desert sites are collected into a dedicated database, SADE, defined to manage operational calibrations and the required SI-traceability. A cross-calibration method over desert sites is detailed. Starting from measurements by a reference sensor, surface reflectance are computed through an atmospheric correction step. A spectral interpolation is then performed to derive the surface reflectance for spectral bands of the sensor to calibrate. Then the atmospheric contribution is added to compute top-of-atmosphere reflectance. The comparison to reflectance acquired by the sensor to calibrate provide an estimation of the cross-calibration with the reference sensor. Results illustrate the efficiency of the method for various pairs of sensors based on archive for AQUA-MODIS, ENVISAT-MERIS, PARASOLPOLDER, and SPOT5-VEGETATION. One main limitation identified in the discussion is due to the spectral interpolation on the surface reflectance which can lead to a few percents bias depending on consistency between spectral bands from both sensor to calibrate and reference sensor. The improvement section identify in progress topics that will improve the method.