Modelling farmland abandonment: A study combining GIS and data mining techniques

This paper studies the driving factors in the process of farmland abandonment. The text is organised in three sections: (i) the concept of abandonment is reviewed by identifying the most important drivers in the scientific literature; (ii) a case study in a Spanish Mediterranean area is presented; and (iii) a new approach is proposed that combines geographical information systems (GIS) and knowledge discovery in databases (KDD) to define the most important variables for studying the abandonment process in this area. The variables considered in this study can be grouped into environmental, socio-economic, and those related to farming practices. Bearing in mind the potentially large number of variables, a few questions had to be answered at the outset: (1) which of the variables are really driving factors in the abandonment process of farming plots?; (2) how important is each of these variables by itself and in combination with others?; and (3) is it possible to consider all variables in the same analysis? We demonstrate the capability of data mining techniques to select the most important features for creating useful scenarios. Finally, two experiments demonstrate that the applied KDD method can be useful for selecting the best variable combinations in studies on farmland abandonment.

► The concept of agricultural land abandonment and its driving factors is reviewed.
► We used data mining and GIS to asses a variable selection process.
► Well-known geographical relationships for a particular study area were confirmed.
► A few previously non-considered relationships were discovered.
► This methodology will help in designing better scenarios for farmland abandonment.