EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

A hybrid data-driven approach for modeling post-earthquake housing recovery

Lianyan Li, Alice Chang-Richards, Megan Boston, Ken Elwood

Earthquake Spectra · 2025-11 · 41 (4) · 3045-3073

DOI: 10.1177/87552930251345841 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Earthquake Spectra, online erschienen am 21.06.2025. Verfasst von Lianyan Li und 3 weiteren Autor:innen. Die Arbeit wurde bisher 7-mal zitiert (Crossref) und nennt 64 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0). Gefördert von: Chinese Government Scholarship.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Empirical data is fundamental for deriving accurate and high-resolution estimates of the post-earthquake building recovery process. To this end, a unique and extensive dataset comprising damage and recovery data from residential buildings affected by the 2010–2011 Canterbury Earthquake Sequence (CES) has been systematically compiled to introduce a novel hybrid data-driven approach for modeling post-earthquake recovery at both individual and community levels. At the individual building level, key temporal variables are modeled using their respective probability distribution functions (PDFs), which are iteratively refined through Bayesian inference with the incorporation of collected empirical data. Specifically, Markov Chain Monte Carlo (MCMC) and Empirical Bayes (EB) methods are employed to derive updated posterior distributions for these variables, offering a contextually grounded understanding of recovery dynamics within the New Zealand setting. At the community level, a suite of machine learning models, including Random Forest (RF), Gradient Boosting Machine (GBM), Decision Tree (DT), and Extreme Gradient Boosting (XGBoost), is developed to rapidly predict the recovery trajectories across residential building portfolios. The marginal posterior distributions derived from the MCMC and EB methods are directly applicable to future recovery modeling efforts, offering valuable insights into the temporal duration and inherent variability of recovery activities tailored to a region’s contextually specific socio-economic profiles and recovery practices. The findings further underscore the superior performance of the RF model in predicting community-level recovery trajectories, whereas the XGBoost model demonstrates excellence in estimating overall recovery timeframes. Consequently, an integrated approach utilizing both RF and XGBoost models is recommended to provide a comprehensive understanding of community-level recovery dynamics and overall recovery time in the aftermath of earthquakes.

Förderung

  • Chinese Government Scholarship

Zitieren

APA

Li, L., Chang-Richards, A., Boston, M., & Elwood, K. (2025). A hybrid data-driven approach for modeling post-earthquake housing recovery. Earthquake Spectra, 41(4), 3045-3073. https://doi.org/10.1177/87552930251345841

Vancouver

Li L, Chang-Richards A, Boston M, Elwood K. A hybrid data-driven approach for modeling post-earthquake housing recovery. Earthquake Spectra. 2025;41(4):3045-3073. doi:10.1177/87552930251345841

BibTeX
@article{Li2025,
  author = {Li, Lianyan and Chang-Richards, Alice and Boston, Megan and Elwood, Ken},
  title = {A hybrid data-driven approach for modeling post-earthquake housing recovery},
  journal = {Earthquake Spectra},
  year = {2025},
  volume = {41},
  number = {4},
  pages = {3045--3073},
  doi = {10.1177/87552930251345841}
}
Kostenlos starten →