Open AccessKnowledge graphs to support real‐time flood impact evaluation
Author(s) -
Johnson J. Michael,
Narock Tom,
SinghMohudpur Justin,
Fils Doug,
Clarke Keith C.,
Saksena Siddharth,
Shepherd Adam,
Arumugam Sankar,
Yeghiazarian Lilit
Publication year - 2022
Publication title -
ai magazine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.597
H-Index - 79
eISSN - 2371-9621
pISSN - 0738-4602
DOI - 10.1002/aaai.12035
Subject(s) - geospatial analysis , computer science , workflow , data science , variety (cybernetics) , open data , built environment , interoperability , plan (archaeology) , world wide web , engineering , database , artificial intelligence , geography , civil engineering , cartography , archaeology
A digital map of the built environment is useful for a range of economic, emergency response, and urban planning exercises such as helping find places in app driven interfaces, helping emergency managers know what locations might be impacted by a flood or fire, and helping city planners proactively identify vulnerabilities and plan for how a city is growing. Since its inception in 2004, OpenStreetMap (OSM) sets the benchmark for open geospatial data and has become a key player in the public, research, and corporate realms. Following the foundations laid by OSM, several open geospatial products describing the built environment have blossomed including the Microsoft USA building footprint layer and the OpenAddress project. Each of these products use different data collection methods ranging from public contributions to artificial intelligence, and if taken together, could provide a comprehensive description of the built environment. Yet, these projects are still siloed, and their variety makes integration and interoperability a major challenge. Here, we document an approach for merging data from these three major open building datasets and outline a workflow that is scalable to the continental United States (CONUS). We show how the results can be structured as a knowledge graph over which machine learning models are built. These models can help propagate and complete unknown quantities that can then be leveraged in disaster management.