Research Unit #1: Multi-risk Analysis with Bayesian approach
DESCRIPTION OF THE RESEARCH ACTIVITY OF RU1

The activity of RU1 will be developed in two phases: (1) development of the model ByMuR (Bayesian Multi-Risk) and (2) assessment of single risks for the city of Naples and the related open-source software development.


PHASE 1: Development of the model ByMuR (Bayesian Multi-Risk)

The first step in developing the ByMuR model will be the definition in details of the general rules for assessing single risks. The general goal of this definition is to make homogeneous not only the final estimates, but also the quality and the way models, past data and the available information about the target area will be used for the achievement of the final estimates. The homogenization procedure will be done through two conceptual steps: Each one of the factors retrieved in the risk disaggregation (step A) will be related to the previous ones; in this way, a logical structure (tree) will be developed. This structure will represent the whole path to be followed in assessing the single risk. Each one of those factors will be intended as (logically) conditioned to the occurrence of previous factors. The three basic factors considered will be hazard, vulnerability and exposure. Each one of them will be further factorized taking into account their intrinsic characteristics and possible variations. The goals of this factorization are (i) to clarify in details where and how assumptions and choices will be taken, and, consequently, (ii) to allow identifying the steps in which there will be higher or lower knowledge/uncertainty, and, finally, (iii) to simplify the estimate of the probability of each factor, since conditional probabilities are generally simpler to be assessed.

As concerns the hazard estimations, we will be concentrating on the possibility that the various "causes" (sources of risk) leading to the hazardous phenomenon in the target area can be considered separately. As regards the vulnerability, we will be considering the direct and indirect effects as separated factors. As concerns the exposure, the temporal variations, in particular the diurnal and seasonal variations, will play the most important role.

Each one of those factors or sub-factors will be analyzed with a Bayesian approach (step B). This analysis will allow to exhaustively merge all the information, as well as its quality and quantity, available for the factor itself. All the probability distributions obtained will be combined together following the overall logic; such combination will allow combining the quantitative results as well as correctly propagating the uncertainties. At this stage, the conditions for the homogeneity of models and past data will be discussed, together with the possibility that the risk estimates can intrinsically include the interaction/amplification effects. If, for example, the catalog of seismic events used for the seismic hazard assessment includes also the events with a volcanic origin, it will necessary that also the models describing the earthquake occurrence will account for this possibility; in this case, the estimated seismic hazard will intrinsically include the amplification of the seismic hazard due to volcanic eruptions. This issue is particularly relevant for medium- and long-term assessments.

In this phase 1, all the researchers of RU1 will be involved. A fundamental contribution will be provided by the experience gained in probabilistic volcanic and tsunami hazard assessments. In both cases, clue aspects have been the factorization procedure and the Bayesian assess of the probability of each factor. Such studies represent the most important research activity carried out in the last years by the coordinator, Jacopo Selva, as well as of Anita Grezio and Warner Marzocchi. On the other hand, the technical and methodological contribution of Angela di Ruocco will be a key factor. Her experience gained within the European project Na.R.As. (Natural Risks Assessment), together with Warner Marzocchi, and the MEDRISK project (related to natural risks prevenction and reduction in the Eastern Mediterranean sea) gave her a general overview on multi-risk and related problems. Moreover, her experience in hydrogeologic risk assessment as well as her academic background will be key contributions to assess the issues related to the degree of damage.


PHASE 2: assessment of single risks for the city of Naples and open-source software development

The single risk assessments will be based on the results of single hazards provided by the RU2. The evaluation of vulnerability and exposure for the city of Naples willbe the specific task of RU1. The second step will be the comparison among risks and their hierarchization in several areas of the city of Naples. In parallel, another specific goal of RU1 will be the analysis and discussion of possible interaction/amplification effects in the framework of natural risks for Naples. All of those results will be collected and visualized in a open-source software. Such software will ensure a simple access and consultation to the results of the project. This software will also allow a detailed analysis of all the factors concurring to the risk estimates; the goal is to make simple the identification of the most uncertain factors in order to focus future studies in their direction.

Also in this phase 2, all the researchers of RU1 will be involved. As regards single risk assessments, the most important work will be collecting the available information and data relative to the degree of damage. At this stage, the contribution of Angela di Ruocco and the AMRA staff will be of fundamental importance. The experience on multi-risk analyses, gained by Angela di Ruocco in the European project Na.R.As. (Natural Risks Assessment) and on general applied statistics, gained by Jacopo Selva during its whole research activity, will help in the analysis and comparison of the results, given its novelty in literature. In this issue, the consultancy of Warner Marzocchi will be of particular relevance, given the long-term experience on statistics in geophysics. As regards the software development, it will be of great help the experience gained by Jacopo Selva in programming the free software BET_EF and BET_VH (http://www.bo.ingv.it/bet/), for eruption forecasting and volcanic hazard respectively. Also his collaboration in building databases both at national (DPC projects) and international (WOVOdat project) level for volcanological data.