Scholarlink

<p>Landslide susceptibility maps (LSMs) are vital for hazard mitigation and planning, but their direct comparison remains challenging due to differing methodologies, map classes, and lack of accessible digital data. This study presents a novel approach for quantitative pixel-level comparison of published LSMs—often only available as images—using the PIA software. Fourteen LSMs from eight studies in the Alborz Mountains, Iran, were digitized, standardized to four susceptibility classes, and compared using three quality metrics: area under the curve (AUC), slope of the validation curve, and spatial bias towards mapped landslide cluster areas. Results show that 12 out of 14 LSMs achieve AUC values above 0.5, indicating moderate to good predictive performance, yet spatial agreement is low: only 0.1% of the area has full agreement across all LSMs, and statistically significant modal agreement (at least 8 out of 14 maps concur, <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="11069_2025_7487_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="62" /> </InlineMediaObject> <EquationSource Format="TEX">\(p&lt;0.05\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>p</mi> <mo>&lt;</mo> <mn>0.05</mn> </mrow> </math></EquationSource> </InlineEquation>) is present in just 55% of the region. The greatest inconsistencies occur in the “medium” and “very high” susceptibility classes. Feature importance analysis identifies map resolution and parameter choice as main drivers of LSM performance, exceeding method or algorithm. All performance metrics derived from PIA remain valid only in relative rather than absolute terms due to inevitable processing errors. Importantly, when multiple LSMs are available, decision-making should preferentially use the statistically significant mode map to ensure robust, consensus-based hazard assessment. This standardized framework enables reproducible, transparent comparisons, and can help guide both future LSM development and operational risk management.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Assessing Consistency and Inconsistency in Landslide Susceptibility Mapping: Quality Criteria and Pixel-Level Analysis in a Case Study from the Alborz Mountains, North of Tehran, Iran

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月21日

查看全文
  • Thomas M. Kreuzer,
  • Martina Wilde,
  • Christian Büdel,
  • Peter Priesmeier,
  • Alexander Fekete,
  • Birgit Terhorst

摘要

Landslide susceptibility maps (LSMs) are vital for hazard mitigation and planning, but their direct comparison remains challenging due to differing methodologies, map classes, and lack of accessible digital data. This study presents a novel approach for quantitative pixel-level comparison of published LSMs—often only available as images—using the PIA software. Fourteen LSMs from eight studies in the Alborz Mountains, Iran, were digitized, standardized to four susceptibility classes, and compared using three quality metrics: area under the curve (AUC), slope of the validation curve, and spatial bias towards mapped landslide cluster areas. Results show that 12 out of 14 LSMs achieve AUC values above 0.5, indicating moderate to good predictive performance, yet spatial agreement is low: only 0.1% of the area has full agreement across all LSMs, and statistically significant modal agreement (at least 8 out of 14 maps concur, \(p<0.05\) p < 0.05 ) is present in just 55% of the region. The greatest inconsistencies occur in the “medium” and “very high” susceptibility classes. Feature importance analysis identifies map resolution and parameter choice as main drivers of LSM performance, exceeding method or algorithm. All performance metrics derived from PIA remain valid only in relative rather than absolute terms due to inevitable processing errors. Importantly, when multiple LSMs are available, decision-making should preferentially use the statistically significant mode map to ensure robust, consensus-based hazard assessment. This standardized framework enables reproducible, transparent comparisons, and can help guide both future LSM development and operational risk management.