Khajehali Marzieh, Safavi Hamid R, Nikoo Mohammad Reza, Najafi Mohammad Reza, Alizadeh-Sh Reza
Department of Civil Engineering, Isfahan University of Technology, Isfahan, Iran.
Department of Civil and Architectural Engineering, Sultan Qaboos University, Muscat, Oman.
Sci Rep. 2025 Jan 2;15(1):146. doi: 10.1038/s41598-024-84543-5.
Floods are among the most severe natural hazards, causing substantial damage and affecting millions of lives. These events are inherently multi-dimensional, requiring analysis across multiple factors. Traditional research often uses a bivariate framework relying on historical data, but climate change is expected to influence flood frequency analysis and flood system design in the future. This study assesses the projected changes in flood characteristics based on eight downscaled and bias-corrected General Circulation Models (GCMs) that participated in the Coupled Model Intercomparison Project Phase 6. The analysis considers two emission scenarios, including SSP2-4.5 and SSP5-8.5 for far-future (2070-2100), mid-term future (2040-2070), and historical (1982-2012) periods. Downscaled GCM outputs are utilized as predictors of the machine learning model to simulate daily streamflow. Then, a trivariate copula-based framework assesses flood events in terms of duration, volume, and flood peak in the Kan River basin, Iran. These analyses are carried out using the hierarchical Archimedean copula in three structures, and their accuracy in estimating the flood frequencies is ultimately compared. The results show that a heterogeneous asymmetric copula offers more flexibility to capture varying degrees of asymmetry across different parts of the distribution, leading to more accurate modeling results compared to homogeneous asymmetric and symmetric copulas. Also it has been found that climate change can influence the trivariate joint return periods, particularly in the far future. In other words, flood frequency may increase by approximately 50% in some cases in the far future compared to the mid-term future and historical period. This demonstrates that flood characteristics are expected to show nonstationary behavior in the future as a result of climate change. The results provide insightful information for managing and accessing flood risk in a dynamic environment.
洪水是最严重的自然灾害之一，会造成重大破坏并影响数百万人的生命。这些事件本质上是多维度的，需要对多个因素进行分析。传统研究通常使用基于历史数据的双变量框架，但预计气候变化将在未来影响洪水频率分析和洪水系统设计。本研究基于参与耦合模式比较计划第六阶段的八个降尺度和偏差校正的全球气候模型（GCM）评估洪水特征的预测变化。分析考虑了两种排放情景，包括远未来（2070 - 2100年）、中期未来（2040 - 2070年）和历史时期（1982 - 2012年）的SSP2 - 4.5和SSP5 - 8.5。降尺度的GCM输出被用作机器学习模型的预测因子来模拟日流量。然后，基于三变量Copula的框架从持续时间、流量和洪峰方面评估伊朗卡恩河流域的洪水事件。这些分析使用了三种结构的分层阿基米德Copula进行，最终比较它们在估计洪水频率方面的准确性。结果表明，异质不对称Copula在捕捉分布不同部分的不同程度不对称性方面具有更大的灵活性，与同质不对称和对称Copula相比，能产生更准确的建模结果。还发现气候变化会影响三变量联合重现期，特别是在远未来。换句话说，在某些情况下，与中期未来和历史时期相比，远未来洪水频率可能会增加约50%。这表明由于气候变化，未来洪水特征预计将呈现非平稳行为。研究结果为在动态环境中管理和评估洪水风险提供了有见地的信息。
