European Contribution Towards a Global Assessment of Agricultural Soil Organic Carbon Stocks

The chapter discusses a study that predicts the global organic carbon stocks for agricultural soils using European databases with geostatistical analysis and modeling. The overall statistical model consists of two submodels namely donor and donee modules. The donor module uses statistics to quantify the relationships between soil organic carbon (SOC) and environmental covariates. The covariates were selected based on their availability at global scale and their roles as major drivers that affect the carbon cycle in terrestrial ecosystems. Multiple linear regression was used in the donor module with the selected covariates and dense SOC measurements coming from LUCAS soil database (Toth et al., 2013a). The LUCAS soil database has more than 22,000 SOC measurements from European countries and a standardized sampling procedure was used routinely to collect samples of around 0.5 kg of topsoil (0-20 cm) each. The donor module reveals and quantifies the relationships between SOC mass concentration in soil and the predictors to be used in the donee model to extend the prediction at global scale using the same set of predictors. We used the WorldClim dataset (Hijmans et al., 2005), which is comprised of global climate data layers representing long-term conditions for the years from 1950 to 2000. The land cover data were extracted from the GlobCover 2009 (ESA and Universite' Catholique de Louvain, 2010) provided by the European Space Agency (ESA), the terrain parameters were derived from CGIAR-CSI SRTM 90 mDatabase (Jarvis et al., 2008), the soil layers obtained from Harmonized World Soil Database (FAO/ IIASA/ISRIC/ISSCAS/JRC, 2012), and the Normalized Difference Vegetation Index (NDVI) data were obtained from Copernicus Global Land Service Data Portal (Copernicus Global Land Service, 2015). The study yielded promising results which are broadly consistent with similar efforts predicting global agricultural SOC stocks. Our model fits the SOC data well (R-2 = 0.35) and preliminary results suggest a global agricultural SOC estimate of 100.34 Pg (Petagrams) in the first 20 cm. The study predicts the global agricultural SOC stocks using a geostatistical approach and the results are consistent with previous studies that used process-based SOC models.