<p>Seismic signals generated by landslides offer valuable insights to understand their dynamics. However, comprehensively analyzing these signals to directly retrieve volumes and runout distances from raw seismic data poses challenges. Leveraging recent advances in seismology and machine learning, we present a novel approach to estimate the propagation properties of seismogenic landslides. Using a dataset comprising seismic recordings and corresponding landslide characteristics, we trained gradient boosting machine learning models to predict volumes and runout distances. Our models achieved a median error of 42% and an <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="10346_2025_2496_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="24" /> </InlineMediaObject> <EquationSource Format="TEX">$\varvec{R}^{\varvec{2}}$</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mrow> <mi mathvariant="bold-italic">R</mi> </mrow> <mrow> <mn mathvariant="bold">2</mn> </mrow> </msup> </math></EquationSource> </InlineEquation> of 0.84 for volume estimation. Runout distance prediction yielded a median error of 19%, albeit with a lower <InlineEquation ID="IEq2"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="10346_2025_2496_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="24" /> </InlineMediaObject> <EquationSource Format="TEX">$\varvec{R}^{\varvec{2}}$</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mrow> <mi mathvariant="bold-italic">R</mi> </mrow> <mrow> <mn mathvariant="bold">2</mn> </mrow> </msup> </math></EquationSource> </InlineEquation>. Our approach does not require inversion, modeling, or simulations of the seismic sources, offering a practical and efficient method for estimating landslide volumes directly from their seismic signals. Further refinement holds promise for enhanced landslide risk assessment, potentially enabling real-time monitoring and mitigation efforts.</p>

Seismogenic landslides volume and runout estimation with machine learning

期刊论文
发表日期：

2025年4月16日

查看全文

Clément Hibert,
Charlotte Groult,
Jean-Philippe Malet

摘要

Seismic signals generated by landslides offer valuable insights to understand their dynamics. However, comprehensively analyzing these signals to directly retrieve volumes and runout distances from raw seismic data poses challenges. Leveraging recent advances in seismology and machine learning, we present a novel approach to estimate the propagation properties of seismogenic landslides. Using a dataset comprising seismic recordings and corresponding landslide characteristics, we trained gradient boosting machine learning models to predict volumes and runout distances. Our models achieved a median error of 42% and an $\varvec{R}^{\varvec{2}}$ $R^{2}$ of 0.84 for volume estimation. Runout distance prediction yielded a median error of 19%, albeit with a lower $\varvec{R}^{\varvec{2}}$ $R^{2}$ . Our approach does not require inversion, modeling, or simulations of the seismic sources, offering a practical and efficient method for estimating landslide volumes directly from their seismic signals. Further refinement holds promise for enhanced landslide risk assessment, potentially enabling real-time monitoring and mitigation efforts.