Main content area

Evaluating the potential of Bayesian networks for desertification assessment in arid areas of Iran

Boali, Abdolhossein, Bashari, Hossein, Jafari, Reza
Land degradation & development 2019 v.30 no.4 pp. 371-390
Bayesian theory, adaptive management, decision support systems, desertification, ecosystems, groundwater, land use, models, soil, soil quality, uncertainty, wind erosion, Iran
Spatiotemporal complexities increase the uncertainty of the results of various desertification assessment models. We integrated the Mediterranean desertification and land‐use (MEDALUS) model with Bayesian networks (BNs) to develop a novel method for desertification assessment in central Iran. We first derived the current desertification status of the study area from empirical observations of actual, current desertification conditions and mapped the obtained status based on the criteria of the MEDALUS model. We then utilized BNs to produce three separate causal models for the soil, groundwater, and wind erosion indicators of the MEDALUS model. We then integrated all the related indicators and indices into a desertification assessment BNs model. After the reevaluation and sensitivity analysis of the model, we compared its results with those of the MEDALUS model. The obtained results highlighted soil condition and wind erosion as the main determinants of desertification in the study area. Considering the ability of BNs to accommodate uncertainty in the assessment process, these models can assist relevant authorities in making informed decisions. They can also be applied as decision support tools for adaptive management in fragile ecosystems of arid areas.