Synthetic satellite observations for ACIX-IV

Scientific challenge

The Atmospheric Correction Inter-comparison eXercise (ACIX) compares the performance of atmospheric correction processors, primarily through their surface reflectance products. Aerosol optical depth (AOD) and total-column water vapour are also relevant outputs. The fourth phase, ACIX-IV, offers an opportunity to evaluate processors against controlled, physically consistent synthetic observations.

With real satellite images, reference surface reflectance and atmospheric properties cannot be measured accurately for every pixel at the satellite’s spatial resolution. This complicates the attribution of differences between processors. A synthetic benchmark can instead provide a known scene and a consistent set of reference quantities, while retaining realistic three-dimensional radiative effects, including adjacency effects.

S2GOS contribution

DTE-S2GOS is designed to generate synthetic top-of-atmosphere (TOA) radiance for both multispectral Sentinel-2/MSI (approximately 10 m) and hyperspectral CHIME (approximately 30 m). The observations are intended to be distributed in an orthorectified, Level-1C-like form, with instrument-related characteristics and radiometric noise represented in the simulation.

The same 3D radiative transfer calculations provide pixel-level reference surface hemispherical-directional reflectance factor (HDRF). Reference atmospheric fields, including AOD and total-column water vapour, are derived from the atmospheric scene configuration informed by the Copernicus Atmosphere Monitoring Service (CAMS).

An important feature of the ACIX-IV workflow is the separation of data access: participants receive the synthetic satellite observations and associated public metadata, while the reference HDRF and atmospheric quantities are retained for the benchmark evaluation. This prevents the reference values from influencing the processors being tested.

Demonstration

The planned exercise covers selected regions of interest spanning different land covers, including forest, cropland, urban and arid environments. Winter and summer configurations are foreseen, with atmospheric variations and cloud conditions, including cumulus clouds, considered in the simulation design.

Each synthetic image is intended to include appropriate ancillary information, such as geographic coordinates, solar and viewing geometry, terrain elevation, cloud or cloud-shadow masks where applicable, spectral band definitions, solar irradiance and radiometric-noise information.

Synthetic Sentinel-2 MSI RGB image over the Furioso region in December — Synthetic Sentinel-2/MSI RGB quicklook over the Furioso region in December, simulated with Eradiate.

Synthetic Sentinel-2 MSI RGB image over the Furioso region in June — Corresponding synthetic Sentinel-2/MSI RGB quicklook in June, illustrating seasonal changes in the simulated scene.

Expected results

ACIX-IV participants will process the synthetic TOA observations using their atmospheric correction algorithms. The retrieved surface reflectance will subsequently be compared with the simulated pixel-by-pixel HDRF reference, allowing a quantitative assessment across land-cover types, seasons and atmospheric conditions.

The exercise is expected to clarify the sensitivity of atmospheric correction algorithms to surface heterogeneity, atmospheric variability and three-dimensional radiative effects. It will also demonstrate how physically based synthetic images can support the development and validation of retrieval algorithms for future missions, including CHIME. Detailed comparison metrics remain subject to the ACIX-IV evaluation protocol.