Comparison between Different Distributed Methods for Flood Susceptibility Mapping

Abstract

Flooding is one of the main natural hazards in Northern Europe and several areas of the Northern Boreal Hemisphere, where during intense rainfall events, several river basins are affected by a fast water level rise that may cause severe damage to human lives and properties. For these reasons, the development of flood models to identify susceptible areas is essential for decision-makers. Geographic Information Systems (GIS) are currently accurate and valuable support tools for defining flood susceptibility maps at different spatial scales. In this study, the prediction accuracy of different GIS-based procedures in the identification of flooding susceptibility is tested and compared. These procedures include the frequency ratio, a combination of the frequency ratio and logistic regression, a combination of the frequency ratio and Shannon’s entropy index, and the statistical index. Ten conditioning parameters of flooding susceptibility are considered: elevation, slope, curvature, land use, Topographic Wetness Index, Stream Power Index, hydrogeology, stream distance, flow direction and average annual rainfall. The comparison analysis is carried out by applying these methods to the study area of Devon County in Southwest England. A total of 225 flood events are used to define the models. For model validation, 1000 randomly selected training and testing sub-datasets have been used in the definition of the receiver operating characteristic (ROC) curves. The results show that the procedure based on the statistical index provides the highest accuracy and reliability in flood susceptibility predictions.