Accelerated Assessment of Critical Infrastructure in Aiding Recovery Eforts During Natural and Human-made Disaster Gautam Thakur, Kelly Sims, Chantelle Rittmaier, Joseph Bentley, Debraj De, Junchuan Fan Tao Liu, Rachel Palumbo, Jesse McGaha, Phil Nugent Bryan Eaton, Jordan Burdette, Tyler Sheldon, Kevin Sparks {thakurg,simskm,rittmaiercm,bentleyjd,ded1,fanj}@ornl.gov,taoliu@mtu.edu {palumborl,mcgahajr,nugentpj,burdetteja,sheldondt,sparkska}@ornl.gov Human Dynamics Section, Geospatial Science and Human Security Division, Oak Ridge National Laboratory, USA ABSTRACT Relief and recovery from disasters (both natural and human-made) require a coordinated approach across several federal and state gov- ernment agencies. In order to achieve optimal resource allocation and deployment of frst responders, accurate and timely assessment of the impact and extent of destruction are the cornerstones to any recovery efort. Ideally, this knowledge should be gathered and shared within the frst 0-24 hours (termed as łAcute Phasež by the U.S. CDC guideline) for informed decision-making. But achieving this poses signifcant challenges for the data collection and data har- monization processes, particularly when voluminous data are being generated from diverse and distributed sources during the disaster responses. To this end, this work developed a scalable and efcient workfow to dynamically collect and harmonize crowd-sourced geographic multi-modal data, and then assess critical infrastructure (CI) damaged during disaster events. We demonstrate the applica- tion of our framework with two real-world experiences in addressing post-disaster recovery eforts - for the Bahamas (Natural - due to Hurricane Dorian, 2019) and Beirut (Human-made - due to explo- sion caused by the ammonium nitrate stored in a warehouse, 2020). We have illustrated that a coordinated efort is needed for planning as well as for execution to achieve informed decision making. CCS CONCEPTS · Information systems → Geographic information systems; Decision support systems; · Computer systems organization → Data fow architectures; · Computing methodologies → Machine learning. KEYWORDS Spatial data mining and knowledge discovery, geographic informa- tion retrieval, disaster response, geographic information system, assessment of critical infrastructure, damage assessment, data cura- tion and management, data reliability and quality, machine learning SIGSPATIAL ’21, November 2ś5, 2021, Beijing, China © 2021 Copyright held by the owner/author(s). ACM ISBN 978-1-4503-8664-7/21/11. https://doi.org/10.1145/3474717.3483947 ACM Reference Format: Gautam Thakur, Kelly Sims, Chantelle Rittmaier, Joseph Bentley, Debraj De, Junchuan Fan, Tao Liu, Rachel Palumbo, Jesse McGaha, Phil Nugent, and Bryan Eaton, Jordan Burdette, Tyler Sheldon, Kevin Sparks. 2021. Ac- celerated Assessment of Critical Infrastructure in Aiding Recovery Eforts During Natural and Human-made Disaster. In 29th International Confer- ence on Advances in Geographic Information Systems (SIGSPATIAL ’21), November 2ś5, 2021, Beijing, China. ACM, New York, NY, USA, 12 pages. https://doi.org/10.1145/3474717.3483947 1 INTRODUCTION Accelerated assessment of damages to Critical Infrastructure or CI (such as schools, hospitals, roads and bridges and other transporta- tion components, energy, food, telecommunication, etc.) is corner- stone to the post-disaster planning and mitigation eforts. Such planning eforts include creation of temporary shelters, identifca- tion of evacuation zones, and optimal placement of frst-responders, among the others. A coordinated approach is needed that includes multitude of activities: a) Top-down or high-level damage estima- tion from remote sensing data, b) human in loop validation and verifcation of CIs that fall in the path of disaster from the previ- ous step, c) Bottoms-up or higher-order feature extraction from spatially-explicit ground-level imagery (also serves as a second line of evidence), and d) information extraction along with information quality assurance from crowd-sourced data (social media, news, blogs, etc.). Besides, robust scalable algorithms and data-intensive computing infrastructure are essential for creating a comprehensive picture of the immediate event aftermath situation, and delivering it within 24 hours of post-disaster (called the łAcute Phasež by the U.S. CDC guideline [2]), for maximal efectiveness. Unfortunately, current approaches and ERM (emergency re- sponse management) pipeline in practice are limited in their ca- pacity to utilize the full values of multitude of analytics, modeling, algorithmic and computing solutions [8, 9, 18, 20]. For example, the work in [20] has found through conversations and collaborations with frst responders that there is limited use of optimized dispatch and informed decision-making after the incident occurrence. This is happening in real-world practice for two reasons [20]: (i) response to incidents follow a greedy strategy and geographically closest approach, for instantaneous dispatch efort after a report is received (despite the fact that optimizing dispatch can minimize response times in the long run); (ii) it has been hard to judge the severity of an incident from a service call. In 2017 hurricane season, after- action report [3] by FEMA (the Federal Emergency Management Agency) it is stated that ł.. FEMA and supporting federal agencies This work is licensed under a Creative Commons Attribution- NonCommercial-ShareAlike International 4.0 License