Benchmarking of land parameter upscaling methods

Scientific challenge

Satellite land-product validation requires reference measurements representative of areas much larger than those directly sampled by individual ground instruments. Upscaling ground-based observations to satellite-pixel scales therefore introduces assumptions about vegetation structure, spatial heterogeneity and the representativeness of measurement locations.

This issue is central to the Ground-Based Observations for Validation (GBOV) service, which supports the validation of Copernicus land products using ground reference measurements and derived land products. Upscaling uncertainties are particularly difficult to quantify when independent reference values at the satellite scale are unavailable.

S2GOS contribution

DTE-S2GOS provides a physically based framework in which ground measurements, satellite observations and reference land parameters are simulated from the same three-dimensional scene. This allows the effects of vegetation structure, illumination, atmospheric conditions and measurement geometry to be explored in a controlled setting.

The use case focuses on uncertainties in the upscaling of land parameters such as surface albedo, leaf area index (LAI), fraction of absorbed photosynthetically active radiation (FAPAR) and fraction of vegetation cover (fCover). The scenes will partially incorporate terrestrial laser scanning (TLS) data to represent terrain and vegetation structure in three dimensions.

The planned observation configurations include:

• Ground observations: a tower-mounted broadband shortwave radiometer (pyranometer) for albedo and a network of below-canopy photosynthetically active radiation (PAR) sensors.
• Satellite observations: synthetic Sentinel-2/MSI and Sentinel-3/OLCI measurements.
• Reference quantities: pixel-level top-of-canopy bidirectional reflectance factor (BRF), bihemispherical reflectance (BHR), FAPAR, LAI and fCover.

Demonstration

Two contrasting sites are considered: San Rossore near Pisa, Italy, and Patagonia National Park, Chile. Simulated ground and satellite observations will be combined to test the spatial representativeness of local measurements and compare ground-to-satellite upscaling approaches.

The technical design includes seasonal configurations, daily observation cycles and vegetation perturbations. FAPAR is planned to be simulated hourly under clear skies, shallow cumulus clouds and stratus cloud conditions, allowing the role of changing illumination and atmospheric conditions to be investigated.

An illustrative Patagonia configuration includes a HyperNet land tower equipped with a Hypstar radiometer. These simulations demonstrate the level of geometric detail available for modelling an actual ground sensor and its field of view.

Expected results

The use case aims to provide a quantitative assessment of uncertainties associated with upscaling ground observations by comparing simulated ground measurements, upscaled estimates and satellite-scale reference quantities.

In particular, the analysis will examine how vegetation heterogeneity, sensor placement, observation geometry, seasonal changes and atmospheric conditions influence the representativeness of ground-based measurements. The resulting framework is intended to support more robust validation of satellite-derived albedo, FAPAR, LAI and fCover.

The interpretation will also account for limitations in the 3D scene representation, including uncertainties in tree geometry and foliage properties derived from TLS and other input data. Specific upscaling algorithms and evaluation metrics will be finalised as part of the demonstration protocol. Land surface temperature and soil moisture are outside the scope of this phase.

Register for our newsletter to stay informed about DTE-S2GOS developments and use cases.