Use of belief network modelling to assess the impact of buffer zones on water protection and biodiversity

摘要

Bayesian belief network (Fully Connected Belief Networks, FC BeNe) was used to estimate uncertainty in the functioning of established buffer zones in Finland. Four experts in the field of water protection and four in biodiversity were asked to assess the roles of 25 key variables involved in the functioning of buffer zones. The matrix comprised five levels of information: field properties, buffer zone properties, management measures, site data and concerns. The eight experts were asked to estimate probability distributions and causal relationships between the 25 variables, which included in particular erosion, particle bound P, dissolved P, and plant, insect and bird species diversity. Any of the three phases of the analysis; prior distributions, links and posterior distributions are end products which are usable as such.