The role of digital traces in assessing recovery after disasters in urban areas

Student thesis: Doctoral ThesisDoctor of Philosophy (PhD)

Abstract

After a natural disaster people are disrupted from their normal day to day life. Whether it be displacement from disasters or governments enforcing a national lockdowns due to COVID-19, we hope that normal behaviour resumes after a period of `downtime'. Unfortunately defining such 'downtime' is difficult as there are many aspects of life that return to normal at different rates. Because of this, knowing what data to collect is critical - surveys have been used in the past, but they are expensive, time consuming and can have small sample sizes. As a result, research has turned measuring recovery indirectly with methods that are more readily available, cheaper and available at a higher temporal resolutions than surveys. In this work we argue that the study of businesses are a proxy for recovery of a community after an event. This is because communities are built upon businesses operating as normal. Here we show that interactions made by businesses on Facebook can be used to measure recovery after disaster in real time, as well as showing that features local to each business can be used to predict recovery over time. Measuring recovery with social media is cheap, has a minute by minute time resolution, and is achievable by anyone in the world with an internet connection. Most alternative strategies to downtime measurement do not share all three of these traits. Showing that different areas recover differently based on features local to each business informs researchers that study models of mobility under different lockdown scenarios, as mobility is not homogeneous over a region. Knowing then why we predict the differences in recovery to different types of `places of interest' (POIs) gives governments and non governmental organisations (NGOs) the ability to better target restrictions to specific locations to reduce the spread of COVID-19, without needing to close all locations.
Date of Award9 May 2023
Original languageEnglish
Awarding Institution
  • University of Bristol
SupervisorFilippo Simini (Supervisor) & Nikolai W F Bode (Supervisor)

Cite this

'