Soil organic carbon prediction using satellite imagery by Klara Dvorakova

The concern about the role of soils in the global carbon budget and the effects of soil organic carbon (SOC) decline on soil quality has been incorporated in international treaties. Initiatives such as ‘4 per mille Soils for Food Security and Climate’ or ‘The Farm to Fork’ strategy are being implemented in the context of Paris Agreement to mitigate climate change. Implementation of such initiatives requires a robust, scientifically-sound measurement, reporting and verification (MRV) system to track that the policy goals are being met and that claimed increases in soil carbon (C) stocks in croplands are real. However, difficulties arise when implementing soil monitoring systems, as cost-effective and accurate SOC maps of croplands that cover large areas (regional/national), with high spatial (~20-50m) and temporal (5 years repeatability) resolution are needed.

This PhD thesis assesses the potential of remote sensing products to provide a tool for a rapid, repeatable, cost-effective and accurate monitoring for SOC content in croplands. The process of deriving final soil product from remote sensing imagery is however hindered by a number of factors, such as the condition of the soil surface, which needs to be bare. Therefore, the first objective of the thesis was to assess the effects of disturbing factors such as soil moisture, crop residues and soil crust on SOC prediction from Sentinel-2 satellite imagery. Secondly, this thesis aimed to test the potential of the created SOC maps as a tool for detecting areas under conservation agriculture.

This resulted in an original work as:

- We have defined a methodology that allows to obtain pure soil pixels from Sentinel-2 imagery in Belgium, and

- We have used Sentinel-2 derived SOC map and uncertainty map to evaluate the difference in SOC content of conservation agriculture fields from conventional agriculture fields, which has, to our knowledge, not been done before.