Determining the sensitivity of flood risk assessment to different multiple criteria analysis approaches

Abstract

Floods, as a prevalent and destructive natural hazard, have globally inflicted substantial human, economic, and environmental losses. With floods affecting approximately 1.6 billion people and accounting for 44% of all disaster events from 2000 to 2021, there is a clear risk associated with residing in areas known for susceptibility to seasonal or flash flooding, as it can result in the loss of lives or property damage. Due to that, the urgency to address flood related risks is paramount. One approach is through the identification of flood risk areas through multi-criteria decision analysis (MCDA) such as the Analytical Hierarchy Process (AHP) and Triangular Fuzzy Analytical Hierarchy Process (FAHP). To date, numerous flood risk studies have utilized the Analytical Hierarchy Process (AHP) and Triangular Fuzzy Analytical Hierarchy Process (FAHP) but often overlooked the sensitivity of these methods to variations in weighting factors, leading to increased subjectivity in their findings. Consequently, this study focuses on evaluating the sensitivity of both AHP and FAHP to changes in criteria values, aiming to enhance the objectivity and reliability of flood risk assessment. A GIS-based MCDA framework was employed for flood risk estimation, excluding considerations of social vulnerability and exposure, for the eThekwini Metropolitan Municipality in South Africa. The criteria, derived from a literature review, included soil type, elevation, rainfall, slope, drainage density, distance to rivers, and land use/land cover. AHP and FAHP were utilized to assign weights to these criteria, with sensitivity analysis conducted through twenty-eight simulations, involving a 20% increase and decrease in the weights of each parameter. Results, evaluated using ROC-AUC analysis, demonstrated high accuracy for both AHP and FAHP, with AUC values of 94.4% each, placing them within the "Good (Very Good)" range. FAHP demonstrated greater stability compared to AHP in response to sensitivity testing involving changes in weightings. Based on these findings, the study recommends prioritizing FAHP over AHP for flood hazard mapping studies.