Detection of long-term landscape changes and trajectories in a Pannonian sand region: comparing land-cover and habitat-based approaches at two spatial scales

A key driver of biodiversity loss is human landscape transformation. Change detection and trajectory analysis are frequently applied methods for studying landscape change. We studied to what degree habitat-specific change detection and trajectory analysis provide different information on landscape change compared to the analysis with land-cover statistics. Our research was carried out at two spatial scales (regional, 1800 km², 360 random points; local, 23 km², polygon-based maps) in the Kiskunság, Hungary. Spatio-temporal databases were prepared using historical maps, aerial photos and satellite images from 1783, 1883, 1954, and 2009. Local expert knowledge of landscape history and recent vegetation was used during the historical reconstructions. We found large differences at both scales between land-cover based and habitat-specific analyses. Habitat-specific change detection revealed that grassland loss was not continuous in the different habitats, as land-cover based analysis implied. Ploughing affected open sand grasslands and sand steppes differently in the periods studied. It was only apparent from the habitat-specific analyses that from the grasslands only mesotrophic and <i>Molinia</i> meadows were relatively constant, up until the 1950s. The gradual increase in forest area revealed by land-cover CHD analyses was split into natural and anthropogenic processes by habitat-specific analyses. Habitat specific trajectory analysis also revealed ecologically important historical differences between habitats. Afforestation affected especially the open sand grasslands, whereas wetland habitats were relatively stable. The most important trajectory was the one in which closed sand steppes were ploughed during the 19^th century, and remained arable fields until present. Fifty percent of the regional trajectories of 18^th century open sand grasslands terminated in tree plantations at present, though 82% of the current open sand grasslands of the local site can be regarded as ancient. We concluded that dividing land-cover categories into finer habitat categories offered an opportunity for a more precise historical analysis of key habitats, and could reveal important ecological processes that cannot be reconstructed with land-cover based analyses. It also highlighted habitat-specific processes making natural and social drivers better interpretable. Information on the diversity of habitat-histories may serve as a basis for spatially more explicit conservation management.