Abstract

During the commissioning phase of Sentinel-3B, the satellite was placed in close formation with Sentinel-3A for several months. This configuration provides a unique opportunity to compare measurements from the two satellites, opening new perspectives for inter-calibration. We will briefly present an overview of activities performed using tandem data and describe in more details two applications for Sentinel-3 optical instruments.

A first application is the estimation of inter-satellite calibration biases. We describe the methodology used to intercompare the multispectral OLCI A and B instruments, using re-gridding, conversion to reflectance and spectral adjustment of the Rayleigh signal. Statistics are then computed for the different classes of scene. Clouds are particularly interesting targets because of their abundance and white reflectance spectrum. Thanks to this method, it has been possible to estimated inter-calibration biases with an uncertainty lower than 0.5% across the full instrument field-of-view. Biases have been shown to be temporally stable during the tandem period.

The efficiency of this inter-calibration has been assessed by aligning OLCI-A on OLCI-B with a custom reprocessing. We also demonstrate the positive impact on inter-comparison of Level 2 land and ocean products.

A second type of application concerns the validation of per-pixel uncertainties. For this goal, differences between SLSTR A and B measurements (L1 or L2) are compared to ex-ante uncertainties provided by models. More precisely, the independent components of the uncertainties are used to normalize the inter-satellite differences. This normalized difference is expected to behave like a normal distribution with a standard deviation of 1. Although the agreement is relatively satisfactory for data with the highest quality level, some significant variations have been observed for lower quality indices. This information can help improve processing algorithms and/or uncertainty estimates.

Share

COinS
 
Sep 21st, 11:05 AM

New Perspectives for Inter-Calibration using Sentinel-3 Tandem Data

During the commissioning phase of Sentinel-3B, the satellite was placed in close formation with Sentinel-3A for several months. This configuration provides a unique opportunity to compare measurements from the two satellites, opening new perspectives for inter-calibration. We will briefly present an overview of activities performed using tandem data and describe in more details two applications for Sentinel-3 optical instruments.

A first application is the estimation of inter-satellite calibration biases. We describe the methodology used to intercompare the multispectral OLCI A and B instruments, using re-gridding, conversion to reflectance and spectral adjustment of the Rayleigh signal. Statistics are then computed for the different classes of scene. Clouds are particularly interesting targets because of their abundance and white reflectance spectrum. Thanks to this method, it has been possible to estimated inter-calibration biases with an uncertainty lower than 0.5% across the full instrument field-of-view. Biases have been shown to be temporally stable during the tandem period.

The efficiency of this inter-calibration has been assessed by aligning OLCI-A on OLCI-B with a custom reprocessing. We also demonstrate the positive impact on inter-comparison of Level 2 land and ocean products.

A second type of application concerns the validation of per-pixel uncertainties. For this goal, differences between SLSTR A and B measurements (L1 or L2) are compared to ex-ante uncertainties provided by models. More precisely, the independent components of the uncertainties are used to normalize the inter-satellite differences. This normalized difference is expected to behave like a normal distribution with a standard deviation of 1. Although the agreement is relatively satisfactory for data with the highest quality level, some significant variations have been observed for lower quality indices. This information can help improve processing algorithms and/or uncertainty estimates.