Development of data integration strategies to improve interdisciplinarity in hazards research

Abstract

Natural hazards are inherently interdisciplinary problems that pose risk to human life, property and prosperity. To provide holistic and actionable solutions in the face of these hazards, a more integrated approach to hazards research is needed. The current state of the hazards and disaster research continues to work more in disciplinary silos with progress being made around the world. This progress is ongoing, and this dissertation contributes to the investigation of these cross- and trans-disciplinary spaces in the context of natural hazards research and how they can be further fused and progressed, with a specific focus on data integration and modeling techniques that inform the complex problem of outmigration characterization following a hazard event. For contextualization, this dissertation first presents prior attempts at data integration. With the commonly echoed best practice of data integration from the earliest stages of data creation, a set of tools are developed for more interdisciplinary data collection in geographically large field studies. These tools are then implemented for the creation of a multi-community dataset tracking damage and recovery following the December 2021 Midwest Tornado Outbreak. This data can be utilized in the training and parameterization of long-term post-event models such as outmigration prediction. Modeling techniques for using the knowledge and data acquired in this field study are explored to arrive at actionable and predictive data for enhanced interdisciplinary hazards research. These modeling techniques include the combination of top-down and bottom-up approaches, linear multi-regression modeling, agent-based modeling, and hindcasting. Some or all these techniques are used to first develop a sheltering model to determine the viability of community tornado shelters during an event similar to that seen in the field study, and establish the knowledge needed to undertake the more complex outmigration model. The datasets and modeling techniques created and acquired are then leveraged to develop a top-down and bottom-up outmigration model after a hazard event that predicts rate of gross outmigration, gross inmigration, net migration, and demographic change following a hazard event. With this set of tools and resources, this dissertation aims to tangibly propel the task of interdisciplinarity in disaster and natural hazards research with the set of tools and resources provided here culminating in the development of a model for predicting long-term population flow following an event.