Soil Organic Carbon estimation in Ferrara (Northern Italy) combining in situ geochemical analyses, hyperspectral remote sensing and neural networks

The critical zone (CZ) is the uppermost portion of the Earth’s crust, extremely vulnerable to human activities, and comprises plants, soil, air, water, rocks, and living organisms. In fact, anthropic processes such as agricultural practices involve the superficial CZ layer, exposing the soil organic matter to factors that lead to a variable depletion of the organic carbon pool. Obviously, organic carbon is a proxy of fertility and thus of soil health, therefore is a crucial element that must be monitored in the CZ. In this study, we investigated whether surface soil organic carbon (SOC) content could be measured using hyperspectral data provided by the Italian Space Agency PRISMA satellite. We collected 100 representative topsoil samples in an area of 30 × 30 Km2 in the Ferrara province (Northern Italy) and estimated their SOC content by elemental analysis. We matched the soil parameters to the spectra of the sampled areas that were measured by PRISMA on April 7th 2020 and used this data to train and test an Artificial Neural Network (ANN) for SOC estimation. Our research showed that this methodology can work also with few samples (100) in an area where the SOC concentration varies significantly (values from 0.7 to 9.3 wt%) over a spatial scale of just a few meters. The combination of geochemical analysis of in situ samples, hyperspectral remote sensing and neural network machine learning can thus be used in areas whose soils have a complex nature to reconstruct a detailed organic carbon distribution map. This can have a variety of practical applications, including monitoring the degradation of agricultural soils.