Constructing a soil class map of Denmark based on the FAO legend using digital techniques

Abstract

Soil mapping in Denmark has a long history and a series of soil maps based on conventional mapping approaches have been produced. In this study, a national soil map of Denmark was constructed based on the FAO–Unesco Revised Legend 1990 using digital soil mapping techniques, existing soil profile observations and environmental data. This map was developed using soil-landscape models generated with a decision tree-based digital soil mapping technique. As input variables in the model, more than 1170 soil profile data and 17 environmental variables including geology, land use, landscape type, area of wetlands, digital elevation model and its derivatives were compiled. The predicted map showed that Podzols and Luvisols were the most frequent soil groups, covering almost two-thirds of the area of Denmark. Geographically, Podzols occupied a major portion of western Denmark, where the soils have developed on sandy parent material, whereas eastern Denmark mostly contained Luvisols developed on loamy basal till. The occurrence of the predicted soil groups was assigned using several variables, of the most important was clay content in the topsoil and subsoil, elevation, geology and landscape type. The overall prediction accuracy based on a 20% hold-back validation data was 60%, but increased to 76% when prediction accuracy of similar soil groups was considered. Podzoluvisols and Alisols were among the weakly predicted groups (<48% prediction confidence), whereas Podzols and Luvisols had the highest accuracy of prediction (>70%). Overall, the average prediction uncertainty was less than 34%. Compared to the existing conventional soil map, the new map showed promising predictions. Validation of the predicted map with different techniques (point validation, prediction confidence analysis, and map-to-map comparison) confirmed that the output is reliable and can be used in various soil and environmental studies without major difficulties. This study also verified the importance of GlobalSoilMap products and a priori pedological information that improved prediction performance and quality of the new FAO soil map of Denmark.