Research Purpose:
The National Centers for Environmental Information (NCEI) calculates the cost of extreme weather events as over $1b with a total of $1.16t from 2005 to 2019. The Midwest is regularly affected by tornadoes, hail storms, and flooding and the frequency and intensity are increasing as the climate changes. Record flooding of the Missouri River in 2019 cost $10.8b in damage. Aside from the financial costs of infrastructure damage, there are incalculable human costs鈥攄eaths, injuries, trauma, and PTSD suffered by whole communities. The US government鈥檚 Fourth National Climate Assessment estimated that the consequences of climate change could reduce the U.S. GDP by 10% by 2100.
The topics of disaster resilience, prediction, and management are both timely and relevant in light of the increased severity and frequency of weather events that are arising as the global climate changes. Many sources of data are available from different agencies鈥� archives as well as from personal smartphones and from new sensor platforms, terrestrial, aerial, and spaceborne. These data describe the geography, weather, infrastructure, socio-economic demographics, imagery, and behavioral patterns associated with disasters. At the same time elements of Artificial Intelligence, such as machine learning, neural networks, cellular automata, evolutionary algorithms, and image recognition can improve prediction and facilitate spatial and temporal visualization of resilience, vulnerability, risks, and hazards. The combination of the former big data and the machine learning algorithms can aid in improving prediction and in the assessment of community resilience to future disasters. Such assessments can be used for policy formulation and strategic decision-making at all levels of government.
While government departments and agencies have been able to enumerate the costs of damage to infrastructure, the ability to understand, describe and visualize human social effects is an ongoing challenge. Furthermore, while governments can improve physical infrastructure, they have been less successful at improving the resilience of human communities and the natural environment in the face of future disasters. The interactions between infrastructure, economics, and societal factors are less well understood and more work is needed to fill this gap.
Toward Transdisciplinary Solutions:
The DRAC research group aims to combine big data and machine learning to analyze the interactions between infrastructural, economic, and social elements of communities, with a view to visualizing it both spatially and temporally. This visualization will be an important component of a web-based and app-based dashboard. The dashboard will also incorporate stochastic elements for simulated disaster scenarios and parameters for altering elements of resilience. Our focus will be less on individual components and more on their interactions. No single discipline and its associated knowledge-base, understanding, and perspective will be sufficient to fully comprehend human community resilience. The collaboration between workers from multiple disciplines in this center will facilitate progress towards understanding the many Infrastructural-Economic-Sociological (IES) interactions.
Combination of disciplines in the initial DRAC team
An additional outcome of this collaboration will be expanding the team's collaborative circle of competence from individual disciplinary perspectives towards a Transdisciplinary Perspective necessary for investigating multiple complexities of collection of problems under the umbrella of Disaster Prediction and Resilience. It is our expectation that the DRAC team members, faculty and students in this collaborative project, will gaze anew upon their specific disciplines; from the limitations of their own discipline toward collaborative transdisciplinary perspectives.
Our general plan is to first identify the key data input required for machine learning and then to search for the available data sources in large databases. From the databases, the key data can be extracted and transformed into forms suitable for analysis (machine learning and testing). Most of the data will be geo-coded and time-stamped for visualization in a purpose-built dashboard. The dashboard, either as a web page or smartphone app, will be used for prediction, what-if scenarios, and assessment of community resilience. Layers within the GIS can be colour-coded for numerical ranges or different categories.