There is an evident increase in the importance that remote sensing sensors play in the monitoring and evaluation of natural hazards susceptibility and risk. The present study aims to assess the flash-flood potential values, in a small catchment from Romania, using information provided remote sensing sensors and Geographic Informational Systems (GIS) databases which were involved as input data into a number of four ensemble models. In a first phase, with the help of high-resolution satellite images from the Google Earth application, 481 points affected by torrential processes were acquired, another 481 points being randomly positioned in areas without torrential processes. Seventy percent of the dataset was kept as training data, while the other 30% was assigned to validating sample. Further, in order to train the machine learning models, information regarding the 10 flash-flood predictors was extracted in the training sample locations. Finally, the following four ensembles were used to calculate the Flash-Flood Potential Index across the Bâsca Chiojdului river basin: Deep Learning Neural Network–Frequency Ratio (DLNN-FR), Deep Learning Neural Network–Weights of Evidence (DLNN-WOE), Alternating Decision Trees–Frequency Ratio (ADT-FR) and Alternating Decision Trees–Weights of Evidence (ADT-WOE). The model’s performances were assessed using several statistical metrics. Thus, in terms of Sensitivity, the highest value of 0.985 was achieved by the DLNN-FR model, meanwhile the lowest one (0.866) was assigned to ADT-FR ensemble. Moreover, the specificity analysis shows that the highest value (0.991) was attributed to DLNN-WOE algorithm, while the lowest value (0.892) was achieved by ADT-FR. During the training procedure, the models achieved overall accuracies between 0.878 (ADT-FR) and 0.985 (DLNN-WOE). K-index shows again that the most performant model was DLNN-WOE (0.97). The Flash-Flood Potential Index (FFPI) values revealed that the surfaces with high and very high flash-flood susceptibility cover between 46.57% (DLNN-FR) and 59.38% (ADT-FR) of the study zone. The use of the Receiver Operating Characteristic (ROC) curve for results validation highlights the fact that FFPI_DLNN-WOE is characterized by the most precise results with an Area Under Curve of 0.96.

High-accuracy flood susceptibility maps play a crucial role in flood vulnerability assessment and risk mitigation. This study assesses the potential application of three new ensemble models, which are integrations of the adaptive neuro-fuzzy inference system (ANFIS), analytic hierarchy process (AHP), certainty factor (CF) and weight of evidence (WoE). The experimental area is the Trotuș River basin in Romania. The database for the present research consisted of 12 flood-related factors and 172 flood locations. The quality of the models was evaluated using root mean square error (RMSE) values and the ROC curve (AUC). The results showed that the ANFIS-CF model and the ANFIS-WOE model have a high prediction capacity (accuracy > 91.6%). Therefore, we concluded that ANFIS-CF and ANFIS-WoE are two new tools that should be considered for future studies related to flood susceptibility modelling.

The present study aims to enrich the specialized literature by proposing and calculating a new flash-flood propagation susceptibility index (FFPSI). Thus, firstly the Flash-Flood Potential Index (FFPI) using the ensembles of the next models was calculated: Weights of Evidence (WOE), Analytical Hierarchy Process (AHP), Logistic Regression (LR), Classification and Regression Trees (CART), and Radial Basis Function Neural Network-Weights of Evidence (RBFN-WOE). A number of 255 flash-flood locations, split into training (70%) and validating (30%) samples, along with 10 predictors were used as input in the five models. The Receiver Operating Characteristics (ROC) Curve and several statistical metrics were used to evaluate the Flash-Flood Potential Index results. LR-WOE and AHP-WOE were the most performant models. Nevertheless, all the applied models performed very well (AUC > 0.85). Further, the FFPSI was determined by integrating the FFPI results into a Flow Accumulation procedure. Over 55% of the valleys identified are characterized by high and very high values of FFPSI. Graphical abstract

Floods have become more and more severe and frequent with global climate change. The present study focuses on the Black Sea’s immediate riparian area over which the Danube Delta extends. Due to the accelerated increase in the severity of floods, the vulnerability of the deltaic areas is augmenting. Therefore, it is very important to adopt measures to mitigate the negative effects of these phenomena. The basis of the measures to limit the negative effects is the activity of identifying areas prone to flooding. Thus, this research paper presents a methodology for estimating flood susceptibility using the Analytical Hierarchy Process (AHP) and Fuzzy-Analytical Hierarchy Process (FAHP) models. To determine the susceptibility to these natural risk phenomena, the following eight flood predictors were taken into account: slope, elevation, altitude above channel, land use, hydrological soil group, lithology distance from the river, and distance from water bodies. Furthermore, the weights that each flood predictor has in terms of determining flood susceptibility were determined through the previously mentioned models. The results revealed that the slope is the most important predictor, followed by elevation, distance from the river, and land use. These weights were used in the GIS environment to evaluate the susceptibility to floods from a spatial point of view. The areas with a high/very high value for these phenomena occupy over 70% of the surface of the Danube Delta.