Automated Land Category Data from Danish Historical Maps, Late 1800s

300 km² of land in Jutland was analyzed to produce machine-readable spatial datasets for five land categories from late-1800s Danish topographical maps. The project, led by Gregor Levin of Aarhus University, used object-based image analysis, GIS, color segmentation, and machine learning. An accuracy assessment indicates results for most categories are around 90% or higher.

Use Cases

• Analyzing historical land cover change based on the automated classification of heath, sand dune, wetland, forest, and water bodies.
• Training machine learning models for automated feature extraction from historical map imagery.
• Comparing land use dynamics from the late 1800s to the present day based on the described change assessment.
• Benchmarking automated geo-data production methods against traditional manual vectorisation for resource efficiency.

Strengths

• Accuracy assessment indicates results for most land categories are around 90% or higher.
• Covers a defined study area of approximately 300 km² in Jutland, Denmark.
• Methodology combines multiple automated techniques including machine learning, GIS, and image analysis.

Limitations

• Row count, file formats, and column-level documentation are unknown, limiting suitability assessment.
• The dataset's scale is limited to two pilot study areas.
• Last update date is unknown; freshness unverified.

Provenance

Source

Aarhus University

Collection Method

Automated geo-data production comprising object-based image analysis, vector GIS, colour segmentation, and machine learning processes.

Time Range

Late 1800s

Geography

Two study areas in Jutland, Denmark