Natural forest re-growth reflects a decline in traditional agricultural practices that can be observed worldwide. Over the last few decades, natural forest re-growth has replaced much of the agricultural land in the Swiss mountains. This is a region where forms of traditional cultivation have preserved unique landscapes and habitats of high ecological value. This study aimed to characterise the locations in the Swiss mountains where agricultural land has been abandoned and overgrown by trees and bushes. Therefore, multivariate statistical models based on geo-physical and socio-economic variables were developed. Land-use change data were taken from two nationwide land-use surveys carried out in the 1980s and 1990s. In order to obtain reliable models, neighbourhood effects and the group structure in our data were accounted for. For the latter a robust estimation technique known as cluster-adjustment was used.
Results show that forest re-growth is largely restricted to former alpine pastures, land with grass and scrub vegetation and agricultural land with groups of trees at mid to high altitudes, steep slopes, stony ground and a low temperature sum. Some relationships were not as expected, e.g. many of the new forest areas were found to be relatively close to roads. A new finding from this study was that forest re-growth is largely restricted to regions with immigration, higher proportions of part-time farms as opposed to full-time farms and high farm abandonment rates. By accounting for neighbourhood effects, the model fit was improved. The considerable residual deviance of the models was interpreted as the result of undetected local characteristics, such as poor water availability, small-scaled topographic peculiarities (e.g. small trenches, stonewalls, soil damages by cattle) and the individual's motivation to abandon or maintain cultivation. The conclusion made was that general policy measures for the whole mountain area are not suitable for the prevention of land abandonment and forest re-growth, and that policy measures must pay more attention to local characteristics and needs.