Assessing the spatial patterns of land abandonment and forest re-growth using landscape integrated models: case study from the northern part of the Carpathians

Many landscape changes are gradual, rather than abrupt changes in land cover categories. For example, forest re-growth which is connected with a decline in agricultural practices is common worldwide (Rudel et al. 2000; MacDonald et al. 2000), and has been widespread across Europe since the Middle Ages due to socio-economic and political changes (Mather 2001). Forest re-growth occurred when crop- and pasture- land is infrequently used, often in areas with poor soils, steep slopes, and limited accessibility. The most common cause of this process is ‘marginalisation’ of agricultural land (MacDonald et al. 2000).

Such graduate processes nevertheless can have major environmental consequences. For example, land abandonment and forest re-growth result in both positive (e.g., stabilisation of soils, carbon sequestration or increase of biodiversity) and negative effects (e.g., loss of traditional cultivation forms, long-term loss of species rich habitats or higher probability of wild-fires) (Gellrich et al. 2007). The question is though how to measure the spatial patterns of such gradual changes at the landscape scale, since the environmental effects of these changes certainly suggest that accurate detection and monitoring is warranted?

Most of quantitative methods which are used to quantify landscape patterns are based on discrete landscape representations, that can capture land cover well. However, categorization obscures the internal heterogeneity within a given land cover class, and can lead to losses of important ecological information. An alternative approach is to use landscape continuous models. Such gradient-based representations may be able to more accurately represent continuous spatial heterogeneity, which could lead to a better understanding of pattern-process relationships. Ultimately though, the integration of these two different approaches into one coherent model would probably provide the best representation of landscape structure (McGarigal et al. 2009, Dragut 2010).

The aim of our study is to develop an approach which allows to integrate categorical- and gradient-based landscape representations. We test our model by assessing a quintessential gradual process, i.e., land abandonment and forest re-grown and its spatial patterns in the northern part of the Carpathians where forest re-growth has been widespread in the last decades (Kuemmerle et al. 2008).

In our assessment, we used a near-annual, peak-growing-season time series of Landsat TM and ETM+ for 1984-2012 to generate annual maps of land abandonment and forest re-growth. Land abandonment was defined from a remote sensing perspective as open areas which have not been used for at least three years for crops, hay cutting or livestock grazing. Then, we analysed land abandonment and forest re-growth spatial patterns using a landscape integrated model (combination of categorical- and gradient-based landscape models) which was constructed based on adaptation and combination of existing methods like morphological image analysis and texture analysis.

Our preliminary results showed that the integrated approach allowed to detect changes in landscape pattern caused by land abandonment and forest-re-grown with high accuracy, in particularly in areas with relatively small-scale, gradual changes which are common in the northern part of the Carpathians.

Acknowledgments

Research supported by the National Science Centre, program SONATA, project no 2011/03/D/ST10/05568.