#### Abstract

The Deep Convective Cloud (DCC) method is an efficient method to estimate the relative radiometric uncertainty of Earth Observation sensors. The method has been used to monitor the inter-satellite relative uncertainty of the Copernicus Sentinel-2 constellation.

The data selection process relies on thresholds on the NIR (B08) and cirrus (B10) bands on data acquired in the intertropical region. More than 1000 products for each satellite are processed each month. Histograms of the reflectance values are extracted for all spectral bands and each detector module. Once enough products are collected, a skewed- Gaussian is fit to the accumulated histogram. The second inflexion point of the distribution is used as a reflectance indicator. This indicator is preferred to the more standard choice of the mode of the distribution because it is statistically more robust. The reflectance indicators of the two satellite units can then be compared and their evolution in time monitored. Thanks to the large number of available data, it is possible to estimate the statistical uncertainty of the method by repeating the evaluation over the same time period. More precisely, the collected products are randomly assigned to 5 different batches and processed independently. The mean value and standard deviation of the 5 different measurements can then be computed to evaluate the sampling uncertainty. Similarly, results can be computed on different geographic zones to analyse the sensibility to local climatology.

The method has been used operationally to follow the radiometry of Sentinel-2 MSI instrument since January 2022. One particular objective of the analysis was to estimate the efficiency of the radiometric harmonization of the VIS-NIR channels implemented at the end of January 2022. The measured inter-satellite bias is smaller than 1.1% for bands B01 to B09 with 67% confidence, while a larger bias is observed on the SWIR B12 band.

Monitoring Sentinel-2 MSI Inter-calibration Using Deep Convective Clouds

The Deep Convective Cloud (DCC) method is an efficient method to estimate the relative radiometric uncertainty of Earth Observation sensors. The method has been used to monitor the inter-satellite relative uncertainty of the Copernicus Sentinel-2 constellation.

The data selection process relies on thresholds on the NIR (B08) and cirrus (B10) bands on data acquired in the intertropical region. More than 1000 products for each satellite are processed each month. Histograms of the reflectance values are extracted for all spectral bands and each detector module. Once enough products are collected, a skewed- Gaussian is fit to the accumulated histogram. The second inflexion point of the distribution is used as a reflectance indicator. This indicator is preferred to the more standard choice of the mode of the distribution because it is statistically more robust. The reflectance indicators of the two satellite units can then be compared and their evolution in time monitored. Thanks to the large number of available data, it is possible to estimate the statistical uncertainty of the method by repeating the evaluation over the same time period. More precisely, the collected products are randomly assigned to 5 different batches and processed independently. The mean value and standard deviation of the 5 different measurements can then be computed to evaluate the sampling uncertainty. Similarly, results can be computed on different geographic zones to analyse the sensibility to local climatology.

The method has been used operationally to follow the radiometry of Sentinel-2 MSI instrument since January 2022. One particular objective of the analysis was to estimate the efficiency of the radiometric harmonization of the VIS-NIR channels implemented at the end of January 2022. The measured inter-satellite bias is smaller than 1.1% for bands B01 to B09 with 67% confidence, while a larger bias is observed on the SWIR B12 band.