Scholarlink

<p>This study develops a comprehensive flood susceptibility assessment framework that integrates machine learning techniques with climate change scenarios in the Taquari River Basin, Southern Brazil. Using 23 environmental variables, including topographic, hydrological, and meteorological factors, the performance of five machine learning algorithms was evaluated. The Random Forest model demonstrated superior performance, achieving the highest accuracy (0.997) and ROC AUC (0.999) scores. Feature importance analysis revealed that the five-day maximum rainfall (14.25%), elevation (11.59%), and one-day maximum rainfall (10.39%) were the most influential predictors of flood occurrence. The trained Random Forest model was then applied to analyze flood probability under four Shared Socioeconomic Pathways (SSPs) scenarios using MIROC6 climate projections. Results showed consistent spatial patterns across all scenarios, with the southwestern coastal region exhibiting persistently high flood probabilities (80–100%), while interior and eastern regions maintained minimal risk. Notably, the progression from the most optimistic (SSP1-2.6) to the most severe scenario (SSP5-8.5) showed only marginal increases in flood probability, suggesting that regional flood vulnerability is predominantly influenced by geographical and topographical characteristics rather than climate variations. This study provides valuable insights for flood risk management and climate adaptation strategies in the Taquari River Basin. The developed framework offers a robust methodology for flood susceptibility assessment that can be adapted for other river basins, contributing to improved flood risk management and climate change adaptation planning.</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

Machine learning and climate scenario integration reveals controls on flood susceptibility in the Taquari-Antas Basin, Brazil

  • 期刊论文

  • 发表日期：

    2025年9月8日

查看全文
  • Enner Alcântara,
  • Cheila Baião,
  • Yasmim Guimarães,
  • José Mantovani

摘要

This study develops a comprehensive flood susceptibility assessment framework that integrates machine learning techniques with climate change scenarios in the Taquari River Basin, Southern Brazil. Using 23 environmental variables, including topographic, hydrological, and meteorological factors, the performance of five machine learning algorithms was evaluated. The Random Forest model demonstrated superior performance, achieving the highest accuracy (0.997) and ROC AUC (0.999) scores. Feature importance analysis revealed that the five-day maximum rainfall (14.25%), elevation (11.59%), and one-day maximum rainfall (10.39%) were the most influential predictors of flood occurrence. The trained Random Forest model was then applied to analyze flood probability under four Shared Socioeconomic Pathways (SSPs) scenarios using MIROC6 climate projections. Results showed consistent spatial patterns across all scenarios, with the southwestern coastal region exhibiting persistently high flood probabilities (80–100%), while interior and eastern regions maintained minimal risk. Notably, the progression from the most optimistic (SSP1-2.6) to the most severe scenario (SSP5-8.5) showed only marginal increases in flood probability, suggesting that regional flood vulnerability is predominantly influenced by geographical and topographical characteristics rather than climate variations. This study provides valuable insights for flood risk management and climate adaptation strategies in the Taquari River Basin. The developed framework offers a robust methodology for flood susceptibility assessment that can be adapted for other river basins, contributing to improved flood risk management and climate change adaptation planning.