Editor’s Note: This article was contributed by KnowWhereGraph.
In the hours and days following a catastrophic hurricane’s landfall, humanitarian responders face numerous challenges. Sifting through disjointed and complex data sets should not be one of them.
This is one of the basic premises underlying KnowWhereGraph, a location-enabled knowledge graph with billions of interconnected facts about humans and the environment for quickly answering questions, including who is most vulnerable? What are the critical health risks facing the population? Where are the nearest medical facilities and shelters that can accommodate them? Who has expertise relevant to this situation and place? And how can the emergency response be tailored to respect the historical and cultural sensitivities of affected populations?
Direct Relief has partnered with a consortium of researchers in academia, private-sector industry, and the U.S. federal government to develop and demonstrate KnowWhereGraph’s capabilities for supporting public health and humanitarian aid. Funded by the National Science Foundation, the project aims to provide integrated, cleaned, and trustworthy environmental and social data to rapidly promote situational awareness during and following a disaster.
lAYERS OF RESPONSE
Disasters are often unpredictable and complex events, which challenge Direct Relief to understand and respond to many different issues simultaneously and immediately. From the physical state of hospitals and health clinics to the prevalence of diabetes and other chronic medical conditions within disaster-affected populations, diverse data from many sources can inform a strategic approach to humanitarian response.
“There are so many different variables we need to be mindful of when planning and executing our humanitarian response, including understanding the demographics and health of a population, the socio-economic and cultural factors in play, and the geography of the impacted area. Sometimes we need to reach out to local experts as well, but first, we must figure out who those people are. KnowWhereGraph can deliver all that information in seconds. It enhances our situational awareness,” said Anna Lopez-Carr, Direct Relief’s Monitoring and Evaluation Specialist and one of the collaborators on the project.
For example, data can help to predict where power outages may threaten individuals’ home supplies of insulin that require refrigeration to remain effective and safe. Upon identifying these areas, Direct Relief can muster the appropriate cold-chain resources to deliver insulin or other medications likely to be in short supply.
People and data drive Direct Relief’s response. Yet often, simply finding experts with relevant knowledge in a particular disaster type, health specialty or geographic region to provide meaningful interpretation of key data poses its own unique challenge. The KnowWhereGraph project uses machine learning and semantic web techniques applied to a vast body of published literature to identify medical professionals and scholars whose insight may help Direct Relief to respond more efficiently and effectively.
To be immediately useful, data must be reliable and easy to discover, access, visualize and integrate with other data to give a multi-faceted picture of the situation at hand. KnowWhereGraph provides these capabilities as a geospatial knowledge graph—a database that stores and represents data about places, times, people, events and concepts, and—critically—the relationships between them. Sitting on top of the knowledge graph is the GeoGraphVis tool, an online interface that provides access to the data through an interactive, map-based visualization.
Greater Than the sum of its parts
In isolation, the destruction of 2020’s Hurricane Laura would have made headlines, but local Covid-19 impacts complicated the response further. Unlike a traditional geographic information system (GIS), built on the concept of separate data layers, KnowWhereGraph is designed to integrate data in a way that highlights when the effect of a compound disaster may be greater than the sum of its parts. This was the case after the massive Thomas Fire, which ignited in Southern California in December 2017. A few weeks later, heavy rain fell on areas recently charred and denuded by the fire, causing catastrophic debris flows that resulted in multiple casualties and extensive property damage. KnowWhereGraph makes it easier to identify patterns in events that strike at different times but overlapping places, potentially helping responders to better prepare for otherwise unexpected consequences.
In contrast to a data portal, KnowWhereGraph creates smart data, not software that rapidly goes out of date. Similarly, while portals split data across geographic units such as states or counties, KnowWhereGraph delivers data across all sorts of geographic entities as disasters do not respect human-defined boundaries.
“This is such an exciting opportunity to apply novel graph technology to help better understand and mitigate impacts of natural disasters, and to support developing a more resilient community and society, ” said Wenwen Li, project co-PI and Professor of the School of Geographical Sciences and Urban Planning at Arizona State University.
“It has been a great experience and collaboration to develop a useful tool to help people whose safety and lives are endangered by disasters, and I hope our GeoGraphVis tool could also increase public awareness in terms of the importance of disaster preparation and relief,” said Sizhe Wang, PhD student and lead developer of GeoGraphVis at the Cyberinfrastructure and Computational Intelligence lab of Arizona State University.
The KnowWhereGraph project is supported by the National Science Foundation under the Convergence Accelerator program’s Open Knowledge Network initiative. Direct Relief is collaborating with a team that includes the Center for Spatial Studies and the National Center for Ecological Analysis and Synthesis at the University of California, Santa Barbara; Kansas State University; Michigan State University; Arizona State University; University of Southern California; Esri; Oliver Wyman; Hydronos Labs; the US Geological Survey; and the US Department of Agriculture. The GraphDB triple store (database) powering the graph was kindly provided by Ontotext. Finally, KnowWhereGraph has recently partnered with the Scalable Precision Medicine Open Knowledge Engine (SPOKE) and the Urban Flooding Open Knowledge Network (UFOKN) to provide cross-walks among these graphs.