Harmonization of categorical maps by alignment processes and thematic consistency analysis

Abstract

This paper describes an approach for harmonizing historical vector categorical maps with related modern maps. The approach aims at the correction of geometric distortions and semantic disagreements using alignment processes and analysis of thematic coherence. The harmonized version of the map produced by this approach can already be overlaid with other maps, what was unfeasible with the original map. The positional errors of the old map are reduced by two consecutive geometric adjustments, which use transformations usually available in most GIS software. The thematic consistency between the old and the modern map is achieved by harmonizing their classification systems and by the inclusion of specific contents missing in the early map, but represented in the modern map (e.g. small rivers). This approach was tested in the geometric and thematic harmonization of the Portuguese Land Cover/Land Use (LCLU) map for 1990 (COS90). In this test, the 1995 orthorectified aerial images and the 1995 LCLU map (COS95) were used as reference sources of higher positional accuracy, to align the COS90 map. COS90 was firstly adjusted with the 1995 aerial images by an NTV2 grid transformation, developed by the authors. Then, for reduction of the local distortions, the map resulting from the first transformation was aligned with the COS95 by a rubber-sheeting linear interpolation transformation. This geometric harmonization enabled a decrease of the Root Mean Square Error of COS90 from 204 meters to 13 meters. The thematic harmonization of COS90 enabled its comparison with modern related maps, and the integration of 201 river sections, that were missing because the specifications used in the production of the original map did not allow their representation.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 2 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.