Sistema predictivo multiescala de la degradación del frente urbano edificado

Evidence from progressive deterioration of the urban building stock is well known. This brings about a large social and economic impact due to maintenance costs and associated personal hazards. This thesis explores ways to tackle these phenomena by means of predictive methodologies from a big urban areas point of view. To this end, we propose a predictive system about the deterioration of urban building areas. In this thesis we contribute to the existing theories of durability and life cycle analysis by introducing survival analysis prediction techniques in the durability analysis of the urban building stock. The dissertation is divided into three parts. In Part I, we present a literature review regarding the theoretical and legislative framework. After that, we address the following issues: a) classification of the observed phenomena, b) architectonical typologies emerging from the study, c) their location in the urban network, d) their relationship with climatological phenomena and pollution and finally, e) identification of the causes of deterioration and related structural damages, creating a classification according to its extent and its severity. Part II covers the methodology of inspections, the statistical analysis and the implementation of a SIG application. The template file for facade inspections has been revised by experts and it has been optimized by using multicriteria methods. We study as well the reliability of the inspection teams. In Chapter 5 we describe the survival analysis techniques employed in the nonparametric modeling for interval-censored data. The Turnbull estimator has been used as an alternative to parametric analysis due to the structure of the information available. This is a novel approach compared to the state of the art in prediction modeling. The inspection methodology and survival analysis have been integrated in a SIG management framework. The statistical application in QGIS allows for an exploration of the urban building stock analyzing the landscape, the morphology and the damages using the R libraries. The plugin, programmed in Python, is a relevant contribution as a tool for decision-making, as it allows making inferences about the deterioration process of the building stock. Then we apply the proposed prediction methodology to the analysis of a case study. The city that has been chosen for the study is L?Hospitalet de Llobregat. In part III we describe the inspections conducted in order to select the neighborhoods that allow for maximum data coverage. The research has avoided parametric assumptions and it has been focused on the topics that are more interesting from the point of view of the data collected in terms of orientation, territorial impact, percentage of damages in the representative parts of the facade and percentages of different types of damages. These descriptive statistics of the sample guide the sample selection in the following stages of the survival analysis study. Specifically, we propose and describe non-parametric techniques for the determination of the significant variables in the process of facade deterioration. Results from the case study show how to draw relevant information in order to conduct an effective prevention management. We conclude showing the validity of the predictive system for the prevention of the building stock deterioration. The final chapter also suggests advice for the improvement of inspections, such as ICT and inspector?s training, and for incorporating dynamic data about the landscape exploiting the QGIS tool and the statistical analysis of the FAD plugin. The PhD candidate postulates that such initiatives require a supporting infrastructure from institutions and he advocates the creation of a commission to implement predictive systems that improve the management of the urban building stock.