Addressing semantics and historical data heterogeneities in cross-temporal landscape analyses

The process of recreating historical land cover dynamics, needed to understand trends and transient states in ecosystems, includes difficulties such as the sensitivity of applied spatial analyses to heterogeneities in historical material. This paper compares the handling of quantitatively dominating categories in two matrix analyses of land cover change within a Swedish boreal landscape (1725-1859). The focus is on how inconsistencies between historical maps can be handled without violating the inherent semantic potential. The study shows that analyses of land cover support different indications of change depending on the treatment of dominating categories. The type of landscape and research questions in focus should therefore be part of choosing matrix method and classification scheme. The observed patterns need to be evaluated against drivers of change and semantic plasticity in classification schemes to separate ecological change from semantic confusion. This paper recommends aggregated classification schemes with maintained original relationships between categories in comprehensive analyses. However, no pathway is persistent over time and categories should be allowed to disappear and new to appear. Analysis of historical dynamics with extended transition matrixes is recommended to account for the dynamics of small categories in relation to dominating categories within a landscape. (c) 2010 Elsevier B.V. All rights reserved.