Forthcoming deployment of the Copernicus Sentinel-2 products geometric refinement

22 March 2021

We are pleased to announce the upcoming start of the geometric refinement for Copernicus Sentinel-2 products. The goal of this evolution is to improve the multi-temporal co-registration and the absolute geolocation accuracy of the products, by using a Global Reference Image (GRI) as source of Ground Control Points (GCPs) to constrain the geometric model of the images.

The refinement processing will be gradually deployed in operations during the coming months. In a first step to start in the next couple of weeks, the refinement processing will be activated for every new MSI observation that intersects (globally or partially) a subset of the Global Reference Image (GRI) covering most of continental Europe. The GRI orbital segments activated in this first phase are depicted in the following figures.

While every new Sentinel-2 observation intersecting the GRI will be a candidate for geometric refinement, the refinement algorithm itself may not be systematically applied depending on the image features impacting the ability to extract GCPs from the underlying GRI images (e.g. observations impaired by substantial cloud or sea cover, short observations over isolated islands, etc.).

The refining status of a product will be systematically reported in a specific metadata element (see the attribute flag "REFINED" or “NOT_REFINED" of the XML tag Image_Refining in the product Datastrip metadata).

Based on the preliminary validation performed, the geometric performance expected for geometrically refined Level-1C products is:

  • 8 m (at 95% circular-error confidence) absolute geolocation accuracy
  • 5 m (at 95% circular-error confidence) multi-temporal co-registration accuracy between product pairs sharing the same relative orbit

As far as Level-2A products are concerned, notwithstanding the geometric performance improvements directly inherited from that of the Level-1C, a correlated improvement of the temporal consistency of the surface reflectance values is expected.

The product geometry of non-refined products will remain entirely based on the raw sensor-to-target viewing model as presently applied.

The forthcoming deployment features a set of additional evolutions, as outlined below:

  • Use of the new 90 m Copernicus Digital Elevation Model (DEM), with corresponding improvement of the geolocation performance over mountainous areas in particular
  • Improvement of the Level-1C cloud detection algorithm over high-altitude terrains
  • Improvement of the Level-1C Technical Quality (TECQUA) mask accuracy
  • Additional On-Line Quality Control checks applicable to the geometrically refined products
  • Correction of a light halo observed along Level-2A image boundaries, either at the edge of the swath or at the end of data-strips
  • Improvement of the Level-2A cloud probability layer near the boundary of the swath
  • Improvement of the topographic correction thanks to the use of the Copernicus 90 m DEM

It is clarified that all above evolutions are independent of the geometric refinement, and will consequently be featured in every product regardless of whether the refinement processing was effectively applied.

The Processing Baseline identifier '03.00' will tag every new Level-1C or Level-2A product following operational deployment. The product format features minor updates with respect to that of the current baseline that have been captured in issue 14.6 of the Sentinel-2 Product Specifications Document. The corresponding set of XSD schemas is provided within the PDF document as an attachment.

You are invited to already familiarise yourselves with the updated format with the support of corresponding sample products made available for the purpose: Level-1C sample product and Level-2A sample product

A separate communication will follow shortly with the dates of the upcoming transition as soon as scheduled.

Navigation Menu

Share your stories

success stories

Calling on all interested users of Sentinel data, who would like to submit their results, turning their experiences into 'success stories'.

If you have a good story to tell, of how any of the Sentinel satellites are producing data that bring benefit to your work and/or to society, please contact the Sentinel Online Editor Malì Cecere at: with your proposals.