Land cover and its development is an object of study in applied landscape ecology and landscape conservation. Historical aerial photos are the basic data source for studies of long-term landscape development. All the data could be analysed by the use of geographical information systems, GIS. GIS IDRISI should be recommended for academic and nature conservation purposes, because of its powerful tools for spatial pattern (landscape structure) evaluation, user-friendliness and low price.

Case study is located in the borderland along the former "Iron Curtain". After the landscape was abandoned, the processes of "desynanthropisation" have started. In the Bohemian Forest Mts. in SW Bohemia, (i) patch size and shape were evaluated and (ii) spatial pattern was characterised with various indices: diversity, fragmentation, contagion, etc. Land cover development is presented in special triangular "landcovergram". The changes of indices are closely, but not simply related to the intensity of human influence (management). Two types of local particular differences are important: original situation ("the starting point") and both quality and quantity of postsynanthropic management. In the upper floodplains near Bohemian/Bavarian borderline, two main developments of landscape structure have gone: (i) the managed parcels were homogenised and in the same time (ii) wetlands were abondoned within their natural borders.

Key words: abandoned landscape, aerial photos, contagion, desynathropisation, fractal dimension, GIS IDRISI, Iron Curtain, land cover change, mosaic measures.

The Czech frontier countryside which has abondoned for decades is a challange for natural studies—starting with the species diversity and finishing with succession processes. Such a landscape is to be found in regions along the former Iron Curtain. Among cultural spruce forests are non-forest patches with ruined villages surrounded by meadows and pastures with remnants of a dense network of roads leading nowhere.

The relatively densily populated countryside with a fading glass industry was gradually depopulated after the World War II. Some villages perished completely, in others their development was substantially stifled. In five non-forest enclaves we have studied the post-war development of the landscape structure in the surroundings of past villages. Before the war over 2.000 people lived there. At present, approximately twenty permanent inhabitants do stay here. The facts about the landscape structure were gained from military aerial photographs from the years 1947 and 1987 (see Fig. 1). In 1947 the land was agricuturally arabled, after dereliction land-use changed to pasture, or in the case of small farms the land was overgrown with shrubs and fast-growing trees (Fig. 2). For the future analysis of land-cover/use change, three main classes of land cover have been interpreted: forest, grassland and intermediate.

Landscape structures were computerised, measured, and landcover indices calculated in geographical information system, GIS IDRISI (Eastman 1995). Contours of landcover segments were vectorised in ROOTS, exported to TOSCA module, then rasterised and processed in IDRISI. The spatial analysis in GIS IDRISI is similar to software for landscape measurement and evaluation, i.e. SPAN (Turner 1990) or HISA (Gustafson & Parker 1992).

The pattern of landscape mosaic is the basic information of landscape processes (Turner 1989, Wiens et al. 1993). We reduced the patch-corridor-matrix model (Forman 1995) to patch-matrix model, because the change of land cover during last 40 years didn’t affect streams, and new corridors were man-made (frontier railings and roads).

We calculated some landscape measures (see O’Neill 1988, Gustafson & Parker 1992, Forman 1995, Farina 1998) to monitor land cover change. The most important characters are patch number and size, and derived indices of patch shape and perimeter (LaGro 1991). Measures of spatial arrangement of patches were Shannon diversity and contagion, used as fragmentation index showing landscape dynamic.

In a large scale, the agriculturally managed lands were abandoned, thus letting the ground free for a secondary succession, particularly colonization of fields, meadows and pastures by pioneer woody plants. This process started from forest borders and hedges (see Holland et al. 1991), and also was developing on ruined buildings in the past villages. Also very distinctive has been the spreading of shrubby lines along the water courses, past hedges and past roads. This process is well ilustrated by the two photographs of the Úpor enclave. All those processes have reflected themselves in the landscape structure.

Most papers on landscape change are oriented to changes after settlement and/or increase of human activities (urbanisation, landfilling, fire suppression, etc., i.e. Iverson 1988, Jean & Bouchard 1991, Baker 1992), but there are only few works that are focused to opposite process —landscape change after dereliction and land abandonment (Pirnat 1993). Anthropically conditioned land cover changes in natural landscape mosaic display in landscape segments as follows:

• matrix stabilisation, porosity decreasement;
• decrease of grain size and density of network, in agriculture landscapes balk ploughing, consolidation of parcels and forest boundaries; in present we can see opposite processes;
• disappear and perforation of natural corridors, reduction of curvilinearity; in abandoned landscapes natural corridors fund new patches, and in long-term development this process leads to change of matrix;
• patches are rounded (consolidation of perimeter), in cultural landscapes predominate aggregated or regular pattern over random one;
• contrast of ecotones and their permeability increases;
• all segments are under repeated disturbance and fragmentation with the biomass movement and additional energy supplement.

• matrix destabilisation: proportion of land cover classes equalized (see direction to central position in landcovergram triangle, Fig. 4);
• patch number increased (Fig. 5), the patches are smaller;
• complexity (= fractal dimension) of undeveloped land cover classes increased (longer and more complicated boundaries and shapes);
• diversity of land cover increased;
• fragmentation rapidly (!) increased.

In general, the development of long-term landscape processes is simply reveal in the triangle landovergram (Fig. 4), where three main trends are in opposite directions—forestery, natural succession, and agriculture. The landscape matrix is becoming stabilized, the contrast between the ecotones deminish and their density is growing. Varied patches are added to the landscape and their form is getting more complex. The land cover diversity is growing, the species diversity gradually declining. The landscape not always is turning to its original stage due to stores of supplementary energy and global euthrophication. For this stage of landscape development, very rare in Europe, we suggest the term desynanthropisation.

We are grateful to S. Pecháčková for help with aerial photos interpretation, and to Grant Agency of Charles University, Prague, for the support of our project No. 77/PřF/95/B-BIO.

Fig. 2. Scenario of village development after the Iron Curtain establishment in fifties.

Fig. 5. Proportional distribution of patch size in 1947 and 1987.

Table 1 — The patches development and land cover (LC) change during 1947—1987

Explanations: complexity of land cover class was computed as fractal dimension index (Burrough 1981, Krummel et al. 1987), d=2*log P/log A, where P is perimeter and A is area; diversity H’=-å (A_i log A_i), where A_i is area of i class; fragmentation is explained by contagion (O’Neill et al. 1988), C=2n*log n + å _i å _j (P_ij*log P_ij), where P_ij is relative proportion of common borders of classes i and j, n is total class number.