Living with Floods Using State-of-the-Art and Geospatial Techniques: Flood Mitigation Alternatives, Management Measures, and Policy Recommendations
Living with Floods Using State-of-the-Art and Geospatial Techniques: Flood Mitigation Alternatives, Management Measures, and Policy Recommendations
复制标题
使用最先进的地理空间技术应对洪水：防洪替代方案、管理措施和政策建议
作者：
Rabin Chakrabortty;S. Pal;Dipankar Ruidas;Paramita Roy;A. Saha;Indrajit Chowdhuri
洪水是一种独特的自然灾害，近几十年来在全球范围内发生的频率越来越高，往往是一种意想不到的、不可避免的自然灾害，但通过采取有效的措施，其损失和破坏是可以管理和控制的。近年来，洪水风险易感性测绘已成为将这一全球威胁的最坏影响降至最低的首要考虑因素；但几个致洪因素与风险水平的动态性之间的非线性关系使其变得复杂，并面临可靠评估的巨大挑战。因此，我们考虑了支持向量机、射频和神经网络--这是地理信息系统平台中的三种不同的最大似然算法--来描绘印度西孟加拉邦亚热带康萨巴蒂河流域的洪水风险区域；由于整个季风季都有强烈的降雨，该流域经历了频繁的洪水事件。在我们的研究中，所有采用的最大似然算法都能更有效地解决洪水风险评估中的所有非线性问题；采用多重共线性分析和皮尔逊相关系数技术来识别所有采用的15个洪水致因因素中的共线性问题。本研究通过六种显著可靠的统计方法(“AUC-ROC、特异度、敏感度、PPV、NPV、F-Score”)和一种图形化(泰勒图)技术对预测结果进行评估，结果表明，人工神经网络是最可靠的建模方法，其次是RF模型和支持向量机模型。训练数据集和验证数据集的ANN模型的AUC值分别为0.901和0.891。结果表明，约有7.54%和10.41%的区域处于高和极高洪水危险区域。因此，这项研究可以帮助决策者在区域和国家层面上制定适当的战略，以减轻特定区域的洪水灾害。这类信息可能有助于各当局在决策的各个领域落实这一成果。除此之外，未来的研究人员也可以通过在洪水敏感性评估中考虑这种方法来进行他们的研究。
Flood, a distinctive natural calamity, has occurred more frequently in the last few decades all over the world, which is often an unexpected and inevitable natural hazard, but the losses and damages can be managed and controlled by adopting effective measures. In recent times, flood hazard susceptibility mapping has become a prime concern in minimizing the worst impact of this global threat; but the nonlinear relationship between several flood causative factors and the dynamicity of risk levels makes it complicated and confronted with substantial challenges to reliable assessment. Therefore, we have considered SVM, RF, and ANN—three distinctive ML algorithms in the GIS platform—to delineate the flood hazard risk zones of the subtropical Kangsabati river basin, West Bengal, India; which experienced frequent flood events because of intense rainfall throughout the monsoon season. In our study, all adopted ML algorithms are more efficient in solving all the non-linear problems in flood hazard risk assessment; multi-collinearity analysis and Pearson’s correlation coefficient techniques have been used to identify the collinearity issues among all fifteen adopted flood causative factors. In this research, the predicted results are evaluated through six prominent and reliable statistical (“AUC-ROC, specificity, sensitivity, PPV, NPV, F-score”) and one graphical (Taylor diagram) technique and shows that ANN is the most reliable modeling approach followed by RF and SVM models. The values of AUC in the ANN model for the training and validation datasets are 0.901 and 0.891, respectively. The derived result states that about 7.54% and 10.41% of areas accordingly lie under the high and extremely high flood danger risk zones. Thus, this study can help the decision-makers in constructing the proper strategy at the regional and national levels to mitigate the flood hazard in a particular region. This type of information may be helpful to the various authorities to implement this outcome in various spheres of decision making. Apart from this, future researchers are also able to conduct their research byconsidering this methodology in flood susceptibility assessment.