Stimulating novel Technologies from Earth Remote Observation to Predict European Soil carbon

General introduction

The use of sattelites to estimate soil organic carbon contents of croplands across Europe was the goal of the STEROPES project. Soil experts along with researchers in digital image analysis have explored the possibilities of estimating soil organic carbon contents of cropland across Europe using the European Space Agency's Sentinel-2 satellites. A twofold conclusion was made: 1) current Sentinel-2-based prediction models are not yet accurate enough to detect the relatively small changes in slow soil carbon accumulation. 2) However, the Sentinel-2 images can give an idea of soil variance and thus show potential for establishing cost-effective sampling strategies.
Traditional monitoring of soil organic carbon is costly as it involves classical sampling and laboratory analysis. Here the researchers explored whether remote soil sensing techniques could offer a cheaper yet valid alternative to generate estimates with relatively high spatial and temporal resolution.

Research approach

First, models were created to estimate soil organic carbon contents based on a time series of reflectance/absorption spectra of cropland soils across Europe. A second aspect was to create correction procedures to deal with specific disturbance factors, e.g., cloudiness, soil moisture content, crop residue, soil texture and salinity. The final component explored the use of satellite data as a covariate for a more cost-efficient selection of sampling points in (more conventional) monitoring initiatives. The aim there was to provide a cheaper alternative to carbon monitoring and high-resolution dynamic soil mapping.

Relevance/Valorisation

Remote soil sensing techniques such as the freely available multispectral sensors from the Sentinel-2 satellites offer an alternative method with the potential to estimate soil parameters with relatively high spatial (10mx10m) and temporal (5 days) resolution and at a far lower cost than spectral and spatial models. However, current Sentinel-2-based prediction models are not yet accurate enough to detect relatively small changes in soil carbon, even after correcting for various confounding factors. Accurate detection of soil organic carbon based on the satellite data is not yet a reality: a wide variety of confounding factors must first be addressed. The images do provide information about the variance in the soil, thus they are useful when creating an efficient sampling strategy for determining soil organic carbon contents. 

Funding provider(s)
EU Horizon2020