Search

Flash flood is one of the most dangerous natural phenomena because of its high magnitudes and sudden occurrence, resulting in huge damages for people and properties. Our work aims to propose a state-of-the-art model for susceptibility mapping of the flash flood using the decision tree random subspace ensemble optimized by hybrid firefly–particle swarm optimization (HFPS), namely the HFPS-RSTree model. In this work, we used data from a flood inventory map consisting of 1866 polygons derived from Sentinel-1 C-band synthetic aperture radar (SAR) data and a field survey conducted in the northwest mountainous area of the Van Ban district, Lao Cai Province in Vietnam. A total of eleven flooding conditioning factors (soil type, geology, rainfall, river density, elevation, slope, aspect, topographic wetness index (TWI), normalized difference vegetation index (NDVI), plant curvature, and profile curvature) were used as explanatory variables. These indicators were compiled from a geological and mineral resources map, soil type map, and topographic map, ALOS PALSAR DEM 30 m, and Landsat-8 imagery. The HFPS-RSTree model was trained and verified using the inventory map and the eleven conditioning variables and then compared with four machine learning algorithms, i.e., the support vector machine (SVM), the random forests (RF), the C4.5 decision trees (C4.5 DT), and the logistic model trees (LMT) models. We employed a range of statistical standard metrics to assess the predictive performance of the proposed model. The results show that the HFPS-RSTree model had the best predictive performance and achieved better results than those of other benchmarks with the ability to predict flash flood, reaching an overall accuracy of over 90%. It can be concluded that the proposed approach provides new insights into flash flood prediction in mountainous regions. [ABSTRACT FROM AUTHOR]

• A new QPSO-CDTreeEns is proposed for flash flood modeling. • Ten flash flood indicators were considered. • QPSO-CDTreeEns has a high performance and better than benchmarks SVM, CART, and LR. • LULC, slope, curvature, and TWI are the most important indicators. • QPSO-CDTreeEns is a new tool for flash flood study. Flash flood is considered as one of the most destructive natural hazards worldwide, especially in tropical countries, where tropical cyclones with torrential rains are recurrent problems yearly. Therefore, an accurate prediction of susceptible areas to flash floods is crucial for developing measures to prevent, avoid, and minimize damages associated with flash floods. The aim of this research is to propose a new state-of-the-art model based on hybridizing Quantum Particle Swarm Optimization (QPSO) and the Credal Decision Tree (CDT) ensemble, namely the QPSO-CDTreeEns model, for spatial prediction of the flash flood. The concept of the proposed model is to build a forest tree of the CDT established through the Random Subspace ensemble. Therein, QPSO is integrated to optimize the three parameters, the subspace size, number of trees, and the maximum depth of trees. A district suffered from a high frequency of flash floods in the north-western mountainous area of Vietnam was selected as a case study. In this regard, a geospatial database that includes a total of 1698 flash flood and inundation polygons derived from Sentinel-1 C-band SAR images and ten input indicators were used to construct and to verify the proposed model. The result shows that the QPSO-CDT-Ens model performed well (Overall accuracy = 90.4, Kappa coefficient = 0.807) and outperformed the five machine learning algorithms in flash flood susceptibility mapping. Among the ten factors, the land-use/land-cover (LULC), the slope, the curvature, and the TWI are the most important indicators. We conclude that the proposed model is a promising tool for flash flood susceptibility mapping in the tropics and may assist decision-makers in sustainable land-use planning in the national disaster mitigation strategies. [ABSTRACT FROM AUTHOR]

Flash flood is one of the most dangerous hydrologic and natural phenomena and is considered as the top ranking of such events among various natural disasters due to their fast onset characteristics and the proportion of individual fatalities. Mapping the probability of flash flood events remains challenges because of its complexity and rapid onset of precipitation. Thus, this study aims to propose a state-of-the-art data mining approach based on a hybrid equilibrium optimized SysFor, namely, the HE-SysFor model, for spatial prediction of flash floods. A tropical storm region located in the Northwest areas of Vietnam is selected as a case study. For this purpose, 1866 flash-flooded locations and ten indicators were used. The results show that the proposed HE-SysFor model yielded the highest predictive performance (total accuracy = 93.8%, Kappa index = 0.875, F1-score = 0.939, and AUC = 0.975) and produced the better performance than those of the C4.5 decision tree (C4.5), the radial basis function-based support vector machine (SVM-RBF), the logistic regression (LReg), and deep learning neural network (DeepLNN) models in both the training and the testing phases. Among the ten indicators, elevation, slope, and land cover are the most important. It is concluded that the proposed model provides an alternative tool and may help for effectively monitoring flash floods in tropical areas and robust policies for decision making in mitigating the flash flood impacts. • HE-SysFor is proposed for flash flood modeling. • Ten flash flood indicators were considered. • Elevation, slope, land cover, and rainfall are the most important indicators. • HE-SysFor has a high performance, better than benchmarks, SVM-RBF, LReg, C4.5, and DeepLNN. • HE-SysFor and wrapper technique are new tools for flash flood study. [ABSTRACT FROM AUTHOR]