Can composite indices explain multidimensionality of tree risk assessment? A case study in an historical monumental complex

Urban trees are a source of ecosystem services and contribute to increase quality of life for many communities and their residents. At the same time they could also determine disservices, such as damages to structures and risk to human safety due to critical combination of tree defects and environmental factors. Although knowledge about tree structure and host-pathogen interactions has grown, the assessment of risk has not been fully explored yet. In particular, the multidimensionality of tree risk assessment as well as the concern about weighting techniques in the construction of indices to measure it need to be addressed. The aim of this paper is to develop a composite tree risk index (R) useful for the management of amenity trees, with a two-fold perspective. First, the multidimensionality issue is explored by considering three dimensions of the tree risk: the Hazard, that represents the likelihood of failure of the tree, the Contact factor, i.e. the nature and the value of the target, and the Damage factor, i.e. the potential for injury or damage. Each dimension is assessed as a function of a set of observed variables related to dendrometric attributes, health status, proximity to buildings and artifacts, tree esthetic value, derived both from field data and spatial analysis performed under GIS environment. Second, in the aggregation of variables into specific-dimension indicators we verify if and to what extent various positive and normative weighting techniques influence the construction of composite index. The R index, obtained as the product of the specific-dimension indicators, was computed for 130 trees located within the formal garden and the park of Villa Lante in Bagnaia, Viterbo (Central Italy), a well-known Renaissance monumental complex, where the needs of cultural heritage conservation and human safety have to be balanced with recreational uses. Sensitivity analyses were carried out in order to test the robustness of R with respect to the considered techniques. © 2015 Elsevier GmbH.