Flood forecasting in data-scarce regions poses significant challenges due to irregular rainfall patterns and limited hydrological monitoring networks, particularly in semi-arid regions in Africa, South America, and Asia. However, despite significant efforts and advancements, there remains a substantial gap in the accurate prediction of flood events necessary for effective risk management and mitigation, evidenced by the recurrence of devastating floods in middle to low-income countries in recent years. Here, we address this problem by testing advanced modeling techniques in a local African case, using a combination of statistical methods for extreme event prediction, hydrodynamic modeling, and remote sensing data, to recommend the most adapted and accurate approach under a variety of settings. Our case study is an emerging urban area in Northern Morocco, situated in a triangular plain interposed between adverse geomorphological and precipitation settings, and unregulated expansion flow, creating an exceptionally overwhelming context for disastrous floods. In the absence of previous studies, we integrate frequency distribution analysis to predict extreme rainfall events and flood flow modeling to simulate floodplain inundation. Data sources included high-resolution remote sensing, local hydrological measurements, fine topographical data, and interviews with stakeholders. We found the Pearson Type 3 distribution to be the most suitable for modeling extreme precipitation in coastal areas, whereas the Generalized Extreme Value (GEV) distribution better fits inland areas. For flood flow assessment, the Gradex method proved to be the most accurate, while other empirical methods outlined critical limitations. Findings reveal that advanced hydrodynamic models significantly enhance flood hazard assessments, even in regions with limited data, showing outstanding correlations with previous flood records and stakeholder feedback. The outcomes carry critical implications for highlighting the importance of selecting appropriate models based on geographical and climatic conditions to inform more resilient urban planning and disaster management practices. We anticipate that these insights will support local decision-makers and urban planners in developing strategies that enhance community resilience and reduce the adverse impacts of flooding. Our work contributes to the broader field of flood risk management, providing a foundation for future developments and practical applications in similar regions worldwide.

中文翻译：

---

通过先进的建模方法提高数据稀缺地区的洪水预报准确性


由于降雨模式不规则和水文监测网络有限，数据稀缺地区的洪水预报带来了重大挑战，尤其是在非洲、南美洲和亚洲的半干旱地区。然而，尽管做出了重大努力和进步，但在准确预测有效风险管理和缓解风险所需的洪水事件方面仍然存在巨大差距，近年来中低收入国家再次发生的毁灭性洪水就证明了这一点。在这里，我们通过在非洲当地案例中测试先进的建模技术来解决这个问题，结合使用极端事件预测的统计方法、水动力建模和遥感数据，以推荐在各种设置下最合适和最准确的方法。我们的案例研究是摩洛哥北部的一个新兴城市地区，位于三角形平原上，夹杂着不利的地貌和降水环境，以及不受控制的扩张流，为灾难性的洪水创造了异常压倒性的背景。在没有先前研究的情况下，我们整合了频率分布分析来预测极端降雨事件，并整合了洪水流建模来模拟洪泛区洪水泛滥。数据来源包括高分辨率遥感、当地水文测量、精细地形数据以及与利益相关者的访谈。我们发现 Pearson Type 3 分布最适合模拟沿海地区的极端降水，而广义极值 （GEV） 分布更适合内陆地区。对于洪水流量评估，Gradex 方法被证明是最准确的，而其他实证方法则概述了关键的局限性。 研究结果表明，即使在数据有限的地区，先进的水动力学模型也能显著增强洪水灾害评估，与以前的洪水记录和利益相关者反馈显示出出色的相关性。这些结果具有重要意义，可以强调根据地理和气候条件选择适当模型的重要性，以便为更具弹性的城市规划和灾害管理实践提供信息。我们预计这些见解将支持当地决策者和城市规划者制定增强社区复原力和减少洪水不利影响的策略。我们的工作为更广泛的洪水风险管理领域做出了贡献，为全球类似地区的未来发展和实际应用奠定了基础。