Skip to Main Content
Remote sensing of long-term vegetation monitoring relies on the analysis of multisensor and multitemporal time-series measurements. Cross-sensor calibration is therefore important to prevent artifacts in the temporal signal due to inherent differences in sensors configurations. Variations in spectral response functions (SRFs) are among the major causes of differences in multisensor reflectances and products. In this paper, we report on the SRF comparability of the upcoming Sentinel-2 Multispectral Instrument (MSI) sensor with a number of operational sensors National Oceanic and Atmospheric Administration (NOAA)/ Advanced Very High Resolution Radiometer (AVHRR9), Landsat 7 Enhanced Thematic Mapper Plus (ETM+), Satellite Pour l'Observation de la Terre VEGETATION1 (VGT1), Moderate Resolution Imaging Spectroradiometer (MODIS), and Medium Resolution Imaging Spectrometer (MERIS) relevant for vegetation monitoring. SRF cross-sensor calibration methods for the conversion of red and near infrared (NIR) reflectances and Normalized Difference Vegetation Index (NDVI) values of operational sensors in reference to the Sentinel-2 MSI sensor were evaluated. Calibration data sets obtained using the soil-leaf-canopy radiative transfer model; a state-of-the-art airborne imaging spectrometer Airborne Prism Experiment (APEX); and univariate and multivariate regression models were considered for SRF cross-sensor calibration. For AVHRR9 and VGT1, reflectances in the red spectral region differed more than 30% from Sentinel-2 reflectances. These differences translated in NDVI deviations of up to 10%. The developed SRF cross-sensor calibration method reduced the differences by factors up to 6, 3, and 7 for red, NIR, and NDVI values, respectively. All but AVHRR9 have been found to be cross-calibrated to within 5% differences for reflectances and NDVI values. The present work is considered as part of a broader harmonization effort aimed at preparing for the integration of Sentinel-2 MSI data- with existing historical data records and product time series.