Knowledge Platform

Bayesian Belief Networks (BBN)

In this group we include methods mainly aimed at describing the structure of relationships among different agents that lead to an improved understanding of cause-effect interconnections, which support the representation of quantitative measures of uncertain values as well as evidence about the states of the variables.

General Purpose and Application

Bayesian Belief Networks (BBN) have proven to be a useful tool in ecosystem service modelling, particularly by integrating the biophysical and the socio-economic aspects in the analysis. The cause-effect analysis of different interaction among a variety of variables can be useful to simulate and make predictions to mainly support decision-making. Bayesian Networks in such case have demonstrated potential to be used to structure the problem and acquire knowledge from data/experience. Other modelling techniques which can be compared to BBN models in the context of ecosystem service modelling are dynamic models, fuzzy logic, artificial neural networks and decision trees. It has been applied in the case study area of Ferrara lowland (Italy) and Chlapowski landscape park (Poland).

Lessons learned

One of the main problems of comprehensive modelling of landscape services provision considering uncertainty is that empirical data is often unavailable. The experience carried out in the project confirmed the potential and limitation highlighted above.

In focusing on second order effects, chain modelling instruments could be a useful instrument, though the related literature is somehow still in its infancy. For example the application of BBN to evaluate the influence of landscape on the creation of second‐order effects for agritourism in the Ferrara lowland case shows how public goods and landscape are inputs for the generation of second‐order services, which is a step forward in feeding economic relationships into the BBN structure and shows explicitly different layers of interactions.

The BBN allowed quantifying the connection among ecosystem services and competitiveness and their relationships in a consistent framework. The study found that the BBN was helpful in analysing second‐order‐effects of landscape‐related public goods on several grounds. Its ability of combining empirical information with stakeholders’ information revealed to be an advantage in this case. In spite of the potential comprehensiveness of these models, applications where finally done with a high degree of simplification. In the experience of the Italian case study, one limitation of the method is time and space‐related. The BBN does not allow feedback loops among variables (static), but this weakness can be improve building dynamic systems that change over time, while with appropriate data the model can be spatially explicit.