Affordable Access

Access to the full text

Modelling relationships between socioeconomy, landscape and water flows in Mediterranean agroecosystems: a case study in Adra catchment (Spain) using Bayesian networks

  • Ropero, Rosa F.1
  • Rumí, Rafael2
  • Aguilera, Pedro A.1
  • 1 University of Almería, Informatics and Environmental Research Group, Department of Biology and Geology, Almería, Spain , Almería (Spain)
  • 2 University of Almería, Department of Mathematics, Almería, Spain , Almería (Spain)
Published Article
Environmental and Ecological Statistics
Springer US
Publication Date
Mar 05, 2019
DOI: 10.1007/s10651-019-00419-2
Springer Nature


In Mediterranean areas, the co-evolution between social and natural systems has given rise to heterogeneous and complex systems of interactions called agroecosystems, in which strong relationships between socioeconomy, landscape and water flows have been identified. In this context, water resources management is a prominent area of research, particularly in semi-arid conditions, where a special set of challenges requires novel tools to deal with uncertainty, multiple sources of information and expert knowledge. In this paper, Bayesian Networks are proposed as a means to model the relationships between socioeconomy, landscape and water flows in a Mediterranean agroecosystem, studying its behaviour under two scenarios of change in land use trends: maintenance of traditional Mediterranean agriculture, and agricultural intensification through the development of greenhouses. Results show that an increase in the area of traditional agriculture would lead to better control of runoff and increased primary productivity, measured as green water flows. By contrast, agricultural intensification of the territory would provoke an increase in evaporation and water losses. Due to the versatility of Bayesian networks, results can be expressed not only as probabilities, but also using other metrics that can be computed from them. Accordingly, Sensitivity Analysis to Evidence, Sensitivity Analysis to Parameters and the Kullback–Leibler divergence were carried out. Bayesian Networks have demonstrated their ability to deal with uncertainty inherent to natural systems, combining expert knowledge, data from regional datasets and Geographical Information Systems, and automatic training algorithms giving robust and proper results.

Report this publication


Seen <100 times