Investigation of urban transformation parameters with causal graphs

Urban transformation refers to the whole of actions put to continuously improve physical, social, economic, and environmental conditions of an urban area experiencing deterioration and collapse, via systematic and integrated approaches. Due to the losses induced by earthquakes in the 21st century and the current structure stocks having been deteriorated or affected by earthquakes in our country, transformation of urban structures requires a proactive approach. In order to take the necessary measures against the life and property loss, determining risky areas by prioritization is of critical importance. Within the context of urban transformation, the necessity of evaluating a large number of parameters and challenges in determination of weights negatively affect the detection process. For this purpose, a causal graphs-based model is presented in this dissertation, to investigate for examining the parameters used in detection of individual masonry structures. The number of parameters is reduced by selection based on the features of parameters used in risk detection via logistic regression analysis during the preliminary stages of the model. Relations between the selected parameters are determined via path analysis, and causality structure representing cause and effect relations between two parameters, and showing either direct or indirect effect of the independent variable on the dependent variable is developed and transferred on graphs. In establishment of causal relation model, both the statistical analysis results and theory are evaluated together. While the correct classification rate is 89% in the analyses conducted using only direct effect coefficients of the chosen parameters, this rate is increased to 96% once the direct effects are also taken into consideration with the developed causal graph model. With this model, by optimizing the weeks-long detection process necessitating field work, data acquisition from the structure, and simulation steps, gathering information on the risk situation of the structure, either Risky or Risk-free, is provided. The results reveal that the approach proposed in this dissertation performs risky structure predictions with high accuracy, can be used for pre-assessment, and model's effectiveness and classification performance are acceptable according to engineering and scientific literature. By contributing to the decision process with the pre-assessment model developed, it is aimed to reduce the risk assessment period and cost, as well as the damages that may take place on the structure. The approach developed form the analyses performed considering structures located at four different citys performs risk assessment via the parameters reduced through the developed software. The main disadvantage of rapid and pre-assessment techniques is that they are limited to one region. In this regard, the characteristic and updateability of the used dataset are highly important in development of the methodology. The method is calibrated via new data entry and updating of the effect weights of newly added regions. Updating of the database and thus the weights between parameters are important in reducing limitations of the methodology.