Landscape aesthetic modelling using Bayesian networks: Conceptual framework and participatory indicator weighting

Abstract Landscape aesthetics provides humans with health and social benefits contributing to overall well-being, thus representing a cultural ecosystem service. Landscape biophysical and social attributes create information that is interpreted as either beauty or blight by the mind of the beholder. The ARtificial Intelligence for Ecosystem Services (ARIES) modelling platform is quite suited to the landscape aesthetics paradigm, since it is characterized by both a strong focus on the spatial connectivity between ecosystems and beneficiaries and by the employment of Bayesian networks to quantify and communicate uncertainty. A conceptual framework based on landscape aesthetic abstraction levels was proposed to build these Bayesian networks, progressively linking tangible indicators to abstract dimensions and concepts. As input to ARIES, a simple and rapid participatory methodology was designed to weight indicators according to stakeholder preferences, from which values the probabilities were derived for use in canonical probabilistic models. The participatory indicator identification methodology generated both abstract and concrete terms, suggesting that the process should be supervised to obtain clear and tangible indicators. A sensitivity analysis revealed that individual visual blight indicators had more profound impacts on landscape aesthetic while the effect of beauty indicators was more subtle and balanced. Although the methodology may require a relatively large number of participants to derive probabilities, the procedure was not overly challenging for the participants. This methodology has the potential to be implemented widely, in various contexts and for different periods, accounting for alternative spatiotemporal variations and land cover contexts.