Data-Driven Earthquake Multi-impact Modeling: A Comparison of Models

Hamish Patten; Max Anderson Loake; David Steinsaltz

doi:10.1007/s13753-024-00567-5

Volume 15 Issue 3

Jun. 2024

Turn off MathJax

Article Contents

Article Navigation > International Journal of Disaster Risk Science > 2024 > 15(3): 421-433

Hamish Patten, Max Anderson Loake, David Steinsaltz. Data-Driven Earthquake Multi-impact Modeling: A Comparison of Models[J]. International Journal of Disaster Risk Science, 2024, 15(3): 421-433. doi: 10.1007/s13753-024-00567-5

Citation:

Hamish Patten, Max Anderson Loake, David Steinsaltz. Data-Driven Earthquake Multi-impact Modeling: A Comparison of Models[J]. International Journal of Disaster Risk Science, 2024, 15(3): 421-433. doi: 10.1007/s13753-024-00567-5

Citation:

PDF

Data-Driven Earthquake Multi-impact Modeling: A Comparison of Models

doi: 10.1007/s13753-024-00567-5

1. Information Management, International Federation of Red Cross and Red Crescent Societies (IFRC), 1211, Geneva, Switzerland;
2. Department of Statistics, University of Oxford, Oxford, OX1 3LB, UK

Funds:

This research was funded by the Engineering & Physical Sciences Research Council (EPSRC) Impact Acceleration Account Award EP/R511742/1.

Accepted Date: 2024-05-29

Available Online: 2024-10-26

Publish Date: 2024-06-14

Abstract

Abstract

In this study, a broad range of supervised machine learning and parametric statistical, geospatial, and non-geospatial models were applied to model both aggregated observed impact estimate data and satellite image-derived geolocated building damage data for earthquakes, via regression- and classification-based models, respectively. For the aggregated observational data, models were ranked via predictive performance of mortality, population displacement, building damage, and building destruction for 375 observations across 161 earthquakes in 61 countries. For the satellite image-derived data, models were ranked via classification performance (damaged/unaffected) of 369,813 geolocated buildings for 26 earthquakes in 15 countries. Grouped k-fold, 3-repeat cross validation was used to ensure out-of-sample predictive performance. Feature importance of several variables used as proxies for vulnerability to disasters indicates covariate utility. The 2023 Türkiye-Syria earthquake event was used to explore model limitations for extreme events. However, applying the AdaBoost model on the 27,032 held-out buildings of the 2023 Türkiye-Syria earthquake event, predictions had an AUC of 0.93. Therefore, without any geospatial, building-specific, or direct satellite image information, this model accurately classified building damage, with significantly improved performance over satellite image trained models found in the literature.
- Disaster risk modeling,
- Earthquake impact models,
- Machine learning,
- Disaster statistics,
- Satellite image-derived building damage

FullText(HTML)

References(35)

References

[1]	Albulescu, A.C. 2023. Open source data-based solutions for identifying patterns of urban earthquake systemic vulnerability in high-seismicity areas. Remote Sensing 15(5). https://doi.org/10.3390/rs15051453.
[2]	CHRR (Center for Hazards and Risk Research, Columbia University) and CIESIN (Center for International Earth Science Information Network, Columbia University). 2005. Global earthquake hazard frequency and distribution. Palisades, New York: NASA Socioeconomic Data and Applications Center (SEDAC). https://doi.org/10.7927/H4765C7S.
[3]	CIESIN (Center for International Earth Science Information Network). 2018. Gridded population of the world, version 4.11 (GPWv4): Population count. https://doi.org/10.7927/H4JW8BX5.
[4]	Copernicus Emergency Management Service. n.d. Damage assessment. https://emergency.copernicus.eu/mapping/book/ export/html/138313. Accessed 29 May 2024.
[5]	Earle, P.S., D. Wald, K. Jaiswal, T. Allen, M. Hearne, K. Marano, A.J. Hotovec, and J. Fee. 2009. Prompt Assessment of Global Earthquakes for Response (PAGER): A system for rapidly determining the impact of earthquakes worldwide. US Geological Survey Open-File Report 1131(2009): Article 15.
[6]	Ehrlich, D., and G. Zeug. 2008. Assessing disaster risk of building stock: Methodology based on earth observation and geographical information systems. JRC Scientific and Technical Reports. Luxembourg: European Commission, Joint Research Centre, Institute for the Protection and Security of the Citizen.
[7]	Greenwell, B.M., B.C. Boehmke, and B. Gray. 2020. Variable importance plots—An introduction to the vip Package. The R Journal 12(1): 343-366.
[8]	Guha-Sapir, D. 2023. Emergency Events Database (EM-DAT)-CRED/OFDA International Disaster Database. https://public.emdat.be/. Accessed 12 May 2024.
[9]	He, C., Q. Huang, X. Bai, D.T. Robinson, P. Shi, Y. Dou, B. Zhao, and J. Yan et al. 2021. A global analysis of the relationship between urbanization and fatalities in earthquake-prone areas. International Journal of Disaster Risk Science 12(6): 805-820.
[10]	Heath, D.C., D.J. Wald, C.B. Worden, E.M. Thompson, and G.M. Smoczyk. 2020. A global hybrid VS 30 map with a topographic slope-based default and regional map insets. Earthquake Spectra 36(3): 1570-1584.
[11]	Hengl, T., G.B.M. Heuvelink, and D.G. Rossiter. 2007. About regression-kriging: From equations to case studies. Computers & Geosciences 33(10): 1301-1315.
[12]	Jaiswal, K., and D. Wald. 2010. An empirical model for global earthquake fatality estimation. Earthquake Spectra 26(4): 1017-1037.
[13]	LeCun, Y., Y. Bengio, and G. Hinton. 2015. Deep learning. Nature 521(7553): 436-444.
[14]	Lee, J., J.Z. Xu, K. Sohn, W. Lu, D. Berthelot, I. Gur, P. Khaitan, K.-W. Huang, et al. 2020. Assessing post-disaster damage from satellite imagery using semi-supervised learning techniques. https://doi.org/10.48550/arXiv.2011.14004. eprint: 2011.14004.
[15]	Li, X., P.M. Bürgi, W. Ma, H.Y. Noh, D.J. Wald, and S. Xu. 2023. DisasterNet: Causal Bayesian networks with normalizing flows for cascading hazards estimation from satellite imagery. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 6-10 August 2023, Long Beach, CA, USA, 4391-4403.
[16]	Loos, S., D. Lallemant, J. Baker, J. McCaughey, S.-H. Yun, N. Budhathoki, F. Khan, and R. Singh. 2020. G-DIF: A geospatial data integration framework to rapidly estimate post-earthquake damage. Earthquake Spectra 36(4): 1695-1718.
[17]	Miura, H., S. Midorikawa, and M. Matsuoka. 2016. Building damage assessment using high-resolution satellite SAR images of the 2010 Haiti earthquake. Earthquake Spectra 32(1): 591-610.
[18]	OpenStreetMap contributors. 2023. Planet dump retrieved from https://planet.osm.org. https://www.openstreetmap.org. Accessed 21 Apr 2024.
[19]	Paulik, R., N. Horspool, R. Woods, N. Griffiths, T. Beale, C. Magill, A. Wild, and B. Popovich et al. 2022. RiskScape: A flexible multi-hazard risk modelling engine. Natural Hazards 119: 1573-1840.
[20]	Rozelle, J.R. 2018. International adaptation of the HAZUS earthquake model using global exposure datasets. Ph.D. dissertation. University of Colorado Denver, Denver, CO, USA.
[21]	Shultz, S. 2017. Accuracy of HAZUS general building stock data. Natural Hazards Review. https://doi.org/10.1061/(ASCE)NH.1527-6996.0000258.
[22]	Silva, V., H. Crowley, M. Pagani, D. Monelli, and R. Pinho. 2014. Development of the OpenQuake engine, the Global Earthquake Model’s open-source software for seismic risk assessment. Natural Hazards 72: 1409-1427.
[23]	Silva, V., S. Akkar, J. Baker, P. Bazzurro, J.M. Castro, H. Crowley, M. Dolsek, and C. Galasso. 2019. Current challenges and future trends in analytical fragility and vulnerability modeling. Earthquake Spectra 35(4): 1927-1952.
[24]	Smits, J., and I. Permanyer. 2019. The subnational human development index database. Scientific Data 6(1): 1-15.
[25]	Svatonova, H. 2015. Aerial and satellite images in crisis management: Use and visual interpretation. In Proceedings of the International Conference on Military Technologies (ICMT) 2015, 19-21 May 2015, Brno, Czech Republic. https://doi.org/10.1109/MILTECHS.2015.7153705.
[26]	UNDRR (United Nations Office for Disaster Risk Reduction). 2023. DesInventar-Disaster Information Management System. https://www.desinventar.net/DesInventar/download.jsp. Accessed 29 May 2024.
[27]	University of California, Berkley. 2022. Global administrative areas version 4.1. https://www.gadm.org/. Accessed 13 Jan 2023.
[28]	Wald, D.J. 2005. ShakeMap manual: Technical manual, user’s guide, and software guide. U.S. Geological Survey Techniques and Methods. Reston, VA: U.S. Geological Survey.
[29]	Wald, D.J. 2008. Quantifying and qualifying USGS ShakeMap uncertainty. Reston, VA: US Geological Survey.
[30]	World Bank. 2022. Population estimates and projections. https://datacatalog.worldbank.org/search/dataset/0037655/ Population-Estimates-and-Projections. Accessed 14 Feb 2023.
[31]	Xia, Z., Z. Li, Y. Bai, J. Yu, and B. Adriano. 2022. Self-supervised learning for building damage assessment from large-scale xBD satellite imagery benchmark datasets. In Database and expert systems applications, ed. C. Strauss, A. Cuzzocrea, G. Kotsis, A.M. Tjoa, and I. Khalil, 373-386. Cham: Springer.
[32]	Xu, S., J. Dimasaka, D.J. Wald, and H.Y. Noh. 2022. Seismic multi-hazard and impact estimation via causal inference from satellite imagery. Nature Communications 13(1): Article 7793.
[33]	Xu, J.Z., W. Lu, Z. Li, P. Khaitan, and V. Zaytseva. 2019. Building damage detection in satellite imagery using convolutional neural networks. https://doi.org/10.48550/arXiv.1910.06444. eprint: 1910.06444.
[34]	Yepes-Estrada, C., V. Silva, T. Rossetto, D. D’Ayala, I. Ioannou, A. Meslem, and H. Crowley. 2016. The global earthquake model physical vulnerability database. Earthquake Spectra 32(4): 2567-2585.
[35]	Zhang, H., M. Wang, Y. Zhang, and G. Ma. 2022. TDA-Net: A novel transfer deep attention network for rapid response to building damage discovery. Remote Sensing 14(15): 3687.