Aarhus University Seal

Amélie Marie Beucher

Mapping potential acid sulfate soils in Denmark using legacy data and LiDAR-based derivatives

Research output: Contribution to journal/Conference contribution in journal/Contribution to newspaperJournal articleResearchpeer-review

  • Amélie Beucher
  • Kabindra Adhikari, Department of Soil Science, University of Wisconsin-Madison, Madison, WI, United States
  • Henrik Breuning-Madsen, University of Copenhagen, Denmark
  • Mette Balslev Greve
  • P. Österholm, Åbo Akademi University, Finland
  • S. Fröjdö, Åbo Akademi University, Finland
  • N.H. Nielsen, Roskilde University, Denmark
  • Mogens Humlekrog Greve
Leaching large amounts of acidity and metals into recipient watercourses and estuaries, acid sulfate (a.s.) soils
constitute a substantial environmental issue worldwide. Mapping of these soils enables measures to be taken
to prevent pollution in high risk areas. In Denmark, legislation prohibits drainage of areas classified as potential
a.s. soilswithout prior permission fromenvironmental authorities. Themapping of these soils was first conducted
in the 1980’s.Wetlands, inwhich Danish potential a.s. soils mostly occur,were targeted and the soilswere surveyed
through conventional mapping. In this study, a probability map for potential a.s. soil occurrence was
constructed for thewetlands located in Jutland, Denmark (c. 6500 km2), using the digital soilmapping (DSM) approach.
Among the variety of available DSM techniques, artificial neural networks (ANNs) were selected. More
than 8000 existing soil observations and 16 environmental variables, including geology, landscape type, land
use and terrain parameters,were available as input datawithin themodeling. Predictionmodels based on various
network topologieswere assessed for different selections of soil observations and combinations of environmental
variables. The overall prediction accuracy based on a 30% hold-back validation data reached 70%. Furthermore,
the conventional map indicated 32% of the study area (c. 2100 km2) as having a high frequency for potential
a.s. soils while the digital map displayed about 46% (c. 3000 km2) as high probability areas for potential a.s.
soil occurrence. ANNs, thus, demonstrated promising predictive classification abilities for themapping of potential
a.s. soils on a large extent.
Original languageEnglish
JournalGeoderma
Volume308
Pages (from-to)363-372
Number of pages10
ISSN0016-7061
DOIs
Publication statusPublished - 2017

See relations at Aarhus University Citationformats

ID: 107938746