Comparing the determinants of cropland abandonment in Albania and Romania using boosted regression trees

Daniel Müller, Pedro J. Leitão, Thomas Sikor

Research output: Contribution to journalArticle

150 Citations (Scopus)

Abstract

The collapse of socialist governance structures in Central and Eastern Europe led to the widespread abandonment of agricultural land. We estimated and compared the determinants of cropland abandonment in Albania and Romania during the postsocialist transitional period from 1990 to 2005. The data set included cropland abandonment derived from satellite image analysis, spatially continuous biogeophysical indicators, and socioeconomic surveys. Data were analyzed using boosted regression trees. Boosted regression trees can account for nonlinearities and interactions between variables and combine high predictive accuracy with appealing options to interpret the results. The results revealed important similarities between cropland abandonment in the countries and showed a strong correlation of abandonment with elevation and slope. Differences between cropland abandonment in Albania and Romania were apparent when the influence of topography was excluded. While physical accessibility tended to be more important in Albania, the density of cropland and input intensity were more decisive in Romania. The immediate time period following the collapse of socialism was dominated by extensive cropland abandonment in areas where agricultural production was no longer profitable. Gradual changes were observed in later stages of the transition period.
Original languageEnglish
Pages (from-to)66-77
Number of pages12
JournalAgricultural Systems
Volume117
DOIs
Publication statusPublished - 1 May 2013

Keywords

  • Land use change
  • Agricultural abandonment
  • Transition
  • Boosted regression trees
  • Eastern Europe

Cite this