Comparison of analytic network process and artificial neural network models for flash flood susceptibility assessment

Abstract

Floods are natural risks with devastating consequences for the environment, people and the economy. Gabes Catchment is regularly devastated by floods owing to urban sprawl, population growth, unregulated municipal systems and indiscriminate land use. However, mitigation of flood impacts can be achieved via flood implementation forecasting systems. In this study, Analytical Network Process (ANP) and Artificial Neural Network (ANN) models were integrated into a Geographic Information System (GIS) to identify and classify flood-prone areas. A geographic information database was derived from the existing geological map, digital elevation model (DEM), precipitation and land-use data. The evaluation of different factors can affect the flood analysis. The ranked and normalized indicators were then weighted and classified with an ANP model to establish the training database. The normalized layers, combined with the training site maps, were then fed to a multilayer perceptron neural network (MLP) to yield a flood risk map. Using a field survey, historical flood data, and satellite imagery, 226 flood locations were identified and classified into 70% training data sets and 30% validation data. Results obtained from ANP and ANN showed that 10% and 14% of all areas were classified as high and very high flood susceptibility, respectively. The performance of both models was assessed using the operational characteristic of the ROC model. The Area Under the Curve ‘AUC’ of ANP and ANN models were 0.861 and 0.876, respectively. The obtained results show the similarity and comparability of the used methods. These results corroborate the perception of susceptibility in the population of the city of Gabes. The study outcomes are of great value to policy makers and state authorities in order to achieve greater awareness and adopt strategies for the preparation and management of the environment in the future for the city of Gabes.