A map of the top-soil organic carbon content of Europe generated by a generalized additive model

摘要

There is an increasing demand for up-to-date soil organic carbon (OC) data for global environmental andclimatic modelling. The aim of this study was to create a map of topsoil OC content at the European scaleAQ4 by applying digital soil mapping techniques to the first European harmonized geo-referenced topsoil (0–20 cm)database, which arises from the Land use/Cover Area Frame statistical Survey (LUCAS). A map of the associateduncertainty was also produced to support careful use of the predicted OC contents. A generalized additive model(GAM) was fitted on 85% of the dataset (R2 =0.29), using OC content as dependent variable; a backward stepwiseapproach selected slope, land cover, temperature, net primary productivity, latitude and longitude as suitablecovariates. The validation of the model (performed on 15% of the data-set) gave an overall R2 of 0.27 and an R2of 0.21 for mineral soils and 0.06 for organic soils. Organic C content in most organic soils was under-predicted,probably because of the imposed unimodal distribution of our model, whose mean is tilted towards the prevalentmineral soils. This was also confirmed by the poor prediction in Scandinavia (where organic soils are morefrequent), which gave an R2 of 0.09, whilst the prediction performance (R2) in non-Scandinavian countries was0.28. The map of predicted OC content had the smallest values in Mediterranean countries and in croplands acrossEurope, whereas largest OC contents were predicted in wetlands, woodlands and mountainous areas. The mapof the predictions’ standard error had large uncertainty in northern latitudes, wetlands, moors and heathlands,whereas small uncertainty was mostly found in croplands. The map produced gives the most updated generalpicture of topsoil OC content at the European Union scale.