Recently, the deleterious effect of climate change has sprung up, leading to unprecedented environmental catastrophes. Climate change and intensive human activity have resulted in broad changes to streamflow regimes necessary to meet world socio-economic requirements and ecological health (Poff, 2018). Therefore, cities may become more susceptible to floods if climate change alters the precipitation patterns (Kim et al., 2017). Moreover, various detrimental effects of flash floods can be witnessed, such as disruption in people’s live, damage in urban traffic, failure of houses, and spread of pollution (Norallahi and Kaboli, 2021). Therefore, to avoid the detrimental repercussions of flooding, it is essential to identify the flooded regions (Wahba et al. 2022). Consequentially, watershed hydrology can be altered by expanding imperviousness and tubing or channeling natural drainage ways in urban areas (Li et al., 2013; Hester and Bauman, 2013; Liu et al., 2015).
Additionally, the study area has no stations to record the rainfall rate; therefore, overcoming this scarcity in recorded rainfall is essential. Meanwhile, the geomorphological index can be beneficial to figure out how the watershed responds to an abundance of rainfall. This will be advantageous since we investigate the hazards of floods in unrecorded areas (Martini and Loat 2007; Muzik 1996; Lastra et al. 2008).
Numerous models were utilized to generate flood prone mapping either physical or statistical models, regarding the flash floods modeling. Physical models like HEC-RAS (Wiles and Levine 2002; Rangari et al. 2019), Storm Water Management Model SWMM (Laouacheria et al. 2019), and Hydrological Simulation Program-FORTRAN (HSPF) (Fonseca et al. 2018; Bicknell et al. 1993), while statistical models are like logistic regression, generalized linear models, entropy, frequency ratio, weighting factors, weights of evidence, flexible discriminant analysis, statistical index, and multivariate statistical methods (Youssef et al. 2016; Giovannettone et al. 2018).
In machine learning techniques, there are several approaches of machine learning employed to investigate the flash flood hazard, such as classification and regression trees, the approach of support vector machines, least-squares SVM, random forest, Naïve Bayes, method of adaptive neuro-fuzzy inference system, decision trees, genetic algorithm rule-set production (GARP), and quick unbiased efficient statistical tree (Darabi et al. 2019, 2020; Hong et al. 2018; Chen et al. 2017; Yan et al. 2018; Lee et al. 2017; Rahmati et al. 2019; Hosseini et al. 2020; Eini et al. 2020).
Meanwhile, as far as hazard mapping is concerned, machine learning algorithms have shown auspicious results in locations where adequate hydraulic and hydrological data is unavailable (Mahmood et al., 2019; Mahmood and Rahman 2019).
As a result of the negative consequences of flash floods in Egypt, in October 2016, a rainstorm ravaged Egypt's northern area, killing and injuring 73 and 30 citizens, respectively (Elnazer, Salman, and Asmoay, 2017; FloodList, 2016; IFRC, 2017).
Furthermore, during years of 2018 and 2019, multiple devastating floods damaged the city of New Cairo, causing displacement of a considerable number of residents, possessions devastation, loss of significant monetary, and the lives of 7 individuals. (FloodList, 2018; FloodList, 2019; FloodList, 2020). In Alexandria and Beheria, respectively, at least six and twenty-five people were killed in 2015. Moreover, New Cairo city witnesses acceleration in rainfall rate as according to the Egyptian Meteorological Authority, the recent torrential rains in Egypt were caused by climate change, which is impacting the whole planet (Egypt Today, 2018). In addition, the studied area has been founded recently on major streamlines, which leads to a dramatic increase in runoff.
The current study aims to predict vulnerable urban areas to flash floods using machine learning techniques. The novelty of this research is to estimate the maximum runoff depth and to utilize it as a conditioning factor to produce the flood susceptibility map.