Scientific challenge

Bright desert Pseudo-Invariant Calibration Sites (PICS) play an important role in monitoring the radiometric stability and consistency of Earth observation instruments. Their relatively stable surface reflectance makes them suitable for long-term calibration and validation activities.

Conventional PICS-based approaches generally average satellite observations over a selected area to estimate a mean top-of-atmosphere (TOA) radiance or bidirectional reflectance factor (BRF). These spatial averages are then compared with reference values, often calculated using one-dimensional (1D) radiative transfer models. While effective for monitoring mean radiometric behaviour, averaging reduces the information available about spatial variability and individual detector pixels.

The challenge is to establish whether realistic, spatially resolved simulations can support pixel-by-pixel radiometric validation, while accounting for the uncertainties associated with both synthetic and actual observations.

S2GOS contribution

DTE-S2GOS proposes an alternative approach in which each simulated satellite pixel is retained rather than averaged over the calibration site. The Synthetic Scene Generator produces a detailed three-dimensional representation of the terrain and surface properties, while the Observation Simulator uses the Eradiate 3D radiative transfer model to simulate the radiometric signal observed by satellite and ground-based instruments.

The simulations are designed to account for:

• Surface heterogeneity and topography, represented using high-spatial-resolution data (approximately 1 m or better where available) and a digital elevation model.
• Local infrastructure, including the GHNA and GONA measurement masts and their surrounding fences, which can influence the measured radiometric signal.
• Observation conditions, including solar and viewing geometry, atmospheric properties and instrument spectral response.
• Uncertainty propagation, from the relevant simulation inputs through to the simulated observations.

The resulting synthetic observations can be compared with actual satellite measurements at the individual-pixel level, with uncertainty estimates accompanying the comparison.

Demonstration

The demonstration focuses on the Gobabeb region in Namibia, a well-characterised bright desert environment. It uses information from the Gobabeb HYPERNETS Namibia station (GHNA), equipped with a HYPSTAR radiometer, and the nearby Gobabeb calibration site (GONA), which is part of the RadCalNet network. The two stations are approximately 650 m apart.

The planned simulation area extends up to approximately 1.5 km around GHNA station. The demonstration will combine detailed surface characterisation with simulations of Sentinel-2/MSI and Sentinel-3/OLCI observations, as well as the ground-based HYPSTAR measurements at GHNA.

For comparisons with satellite acquisitions, simulations will reproduce the acquisition conditions as closely as possible, including illumination and viewing geometry and atmospheric conditions. Appropriate spectral and spatial instrument characteristics will be considered to make the simulated and measured quantities comparable.

Expected results

This use case aims to establish a framework for pixel-by-pixel radiometric assessment at bright desert calibration sites, extending conventional approaches based on spatially averaged observations.

The expected outcomes include:

• A detailed comparison between simulated and actual satellite observations, preserving spatial variability within the calibration area.
• Pixel-level radiometric differences and uncertainty estimates, allowing the significance of discrepancies to be assessed.
• An evaluation of the accuracy achievable with S2GOS when realistic surface structure, observation geometry and atmospheric conditions are represented.
• An assessment of the potential for characterising individual detector-pixel behaviour, which cannot be investigated using only spatially averaged calibration values.

The comparison methodology will account for uncertainties in both the simulated and observed quantities, as well as uncertainty introduced by differences in their spatial, spectral or temporal representation. The results are intended to inform future calibration and validation activities, including those supported by RadCalNet, HYPERNETS and ESA’s vicarious calibration services.

Register for our Newsletter to stay informed about DTE-S2GOS development and use cases.