Scholarlink

<p>Assessing flood risk in estuarine and coastal zones is challenging when river discharge, tidal peaks, and weather-driven forces coincide. Traditional models–based on fixed dependency patterns or isolated extreme events–tend to overlook the rapidly evolving, non-linear interactions among these drivers over short time scales, leading to systematic underestimation of compound flooding severity. The Compound Dynamic Risk Index overcomes these limitations by integrating daily-resolution probabilistic dependence modeling (via copula-derived joint exceedance probabilities of river discharge and storm surge) with a curvature-based diagnostic that captures second-order temporal dynamics. Unlike conventional indicators activated only upon threshold exceedance, the curvature component enables early detection of shifts in hazard trajectory, providing an anticipatory signal of risk escalation before peak intensity is reached. CDRI values are classified into four categories: <i>Low</i> (CDRI&#xa0;<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(&lt;0.30\)</EquationSource> </InlineEquation>), <i>Medium</i> (0.30–0.60), <i>High</i> (0.60–0.90), and <i>Severe</i> (CDRI&#xa0;<InlineEquation ID="IEq4"> <EquationSource Format="TEX">\(\ge 0.90\)</EquationSource> </InlineEquation>). Consequently, the CDRI functions both as a descriptive index and an anticipatory indicator of emerging compound floods. Application to two hydroclimatically distinct regions–the Fraser River in British Columbia and the Potomac River in the eastern United States–demonstrated that curvature-driven transitions generally occurred ahead of subsequent increases in CDRI, indicating their role as an anticipatory precursor of compound-risk intensification. For predictive evaluation, daily CDRI time series were supplied to four supervised learning algorithms: Random Forest, XGBoost, LSTM, and Deep Echo State Network. Among these, the DeepESN achieved the highest classification accuracy: 88.95% in the Fraser River and 81.87% in the Potomac system. By focusing on curvature-driven escalation, daily-scale resolution, and predictive reliability, the index addresses critical gaps in early warning systems and supports more adaptive flood management under growing hydroclimatic uncertainty.</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

Dynamic assessment of compound flooding through a risk index approach

  • 期刊论文

  • 发表日期：

    2026年2月12日

查看全文
  • Farhad Yazdandoost,
  • Neda Izanloo

摘要

Assessing flood risk in estuarine and coastal zones is challenging when river discharge, tidal peaks, and weather-driven forces coincide. Traditional models–based on fixed dependency patterns or isolated extreme events–tend to overlook the rapidly evolving, non-linear interactions among these drivers over short time scales, leading to systematic underestimation of compound flooding severity. The Compound Dynamic Risk Index overcomes these limitations by integrating daily-resolution probabilistic dependence modeling (via copula-derived joint exceedance probabilities of river discharge and storm surge) with a curvature-based diagnostic that captures second-order temporal dynamics. Unlike conventional indicators activated only upon threshold exceedance, the curvature component enables early detection of shifts in hazard trajectory, providing an anticipatory signal of risk escalation before peak intensity is reached. CDRI values are classified into four categories: Low (CDRI  \(<0.30\) ), Medium (0.30–0.60), High (0.60–0.90), and Severe (CDRI  \(\ge 0.90\) ). Consequently, the CDRI functions both as a descriptive index and an anticipatory indicator of emerging compound floods. Application to two hydroclimatically distinct regions–the Fraser River in British Columbia and the Potomac River in the eastern United States–demonstrated that curvature-driven transitions generally occurred ahead of subsequent increases in CDRI, indicating their role as an anticipatory precursor of compound-risk intensification. For predictive evaluation, daily CDRI time series were supplied to four supervised learning algorithms: Random Forest, XGBoost, LSTM, and Deep Echo State Network. Among these, the DeepESN achieved the highest classification accuracy: 88.95% in the Fraser River and 81.87% in the Potomac system. By focusing on curvature-driven escalation, daily-scale resolution, and predictive reliability, the index addresses critical gaps in early warning systems and supports more adaptive flood management under growing hydroclimatic uncertainty.