Alabdan Rana, Sharmila C, Alruwais Nuha, Alshahrani Haya Mesfer, Anbukkarasi S, Sujatha M, Vivek S
Department of Information Systems, College of Computer and Information Science, Majmaah University, 11952, Al-Majmaah, Saudi Arabia.
Department of Computer Science and Engineering, SRM Institute of Science and Technology, Tiruchirappalli, India.
Sci Rep. 2025 Jul 10;15(1):24796. doi: 10.1038/s41598-025-08912-4.
Floods are among the most destructive natural disasters, causing significant harm to human lives, infrastructure, and the environment. This study addresses the critical challenge of early flash flood prediction in dynamically vulnerable regions. The primary aim of this research is to develop a robust methodology for assessing flood vulnerability in Chennai, Tamil Nadu, India, using advanced machine learning techniques. To achieve this, we employed two ensemble models-artificial neural network (ANN) and random forest (RF)-within a GIS framework, analyzing data from 280 historical flood sites and twelve flood-related parameters. Information gain ratio and multicollinearity diagnostic tests were applied to identify and quantify the influence of key factors contributing to flood occurrences. The predictive performance of the models was compared using statistical criteria, including the "Friedman" test. The findings revealed that both ANN and RF models effectively simulated flood susceptibility, with ANN categorizing areas as 18% very low, 16% low, 13% moderate, 22% high, and 31% very high in vulnerability, while RF classified 7% as very low, 11% low, 34% moderate, 31% high, and 18% very high. The study highlights actionable strategies, including strengthening drainage systems, adopting regulated construction practices in sensitive zones, implementing early warning systems, and enhancing public awareness. The novelty of this research lies in integrating machine learning models with GIS for flood prediction and comparing ensemble model efficacy in a coastal metropolitan setting. The results provide critical insights for policymakers, local authorities, and disaster management agencies to formulate sustainable mitigation strategies. By ensuring effective collaboration among government bodies, NGOs, and local communities, this study contributes to building resilience against floods in Chennai through infrastructure improvement, proactive planning, and community engagement.
洪水是最具破坏性的自然灾害之一，对人类生命、基础设施和环境造成重大损害。本研究应对动态脆弱地区早期山洪暴发预测这一关键挑战。本研究的主要目的是利用先进的机器学习技术，开发一种用于评估印度泰米尔纳德邦金奈市洪水脆弱性的稳健方法。为此，我们在地理信息系统（GIS）框架内采用了两种集成模型——人工神经网络（ANN）和随机森林（RF），分析了来自280个历史洪水地点的数据以及12个与洪水相关的参数。应用信息增益比和多重共线性诊断测试来识别和量化导致洪水发生的关键因素的影响。使用包括“弗里德曼”检验在内的统计标准对模型的预测性能进行了比较。研究结果表明，ANN和RF模型都有效地模拟了洪水易发性，ANN将区域的脆弱性分类为18%极低、16%低、13%中等、22%高和31%极高，而RF将7%分类为极低、11%低、34%中等、31%高和18%极高。该研究突出了可采取的策略，包括加强排水系统、在敏感区域采用规范的建设做法、实施预警系统以及提高公众意识。本研究的新颖之处在于将机器学习模型与GIS集成用于洪水预测，并在沿海大都市环境中比较集成模型的有效性。研究结果为政策制定者、地方当局和灾害管理机构制定可持续的减灾策略提供了关键见解。通过确保政府机构、非政府组织和当地社区之间的有效合作，本研究通过改善基础设施、积极规划和社区参与，为金奈市增强抗洪能力做出了贡献。
