Decoding flood resilience in riverine Bangladesh: A concordance–discordance framework integrating AHP–TOPSIS and PROMETHEE II

Flood resilience assessment in riverine systems is characterized by methodological uncertainty arising from divergent multi-criteria decision analysis (MCDA) frameworks. This study presents the first systematic concordance–discordance analysis comparing Analytical Hierarchy Process–Technique for Order Preference by Similarity to an Ideal Solution (AHP–TOPSIS; distance-based) and Analytical Hierarchy Process–Preference Ranking Organization Method for Enrichment Evaluation II (AHP–PROMETHEE II; outranking-based) approaches for flood resilience mapping across 1000 spatial locations in northern riverine Bangladesh. A total of 56 multidimensional indicators were evaluated using identical AHP–derived weights (CR = 0.153). PROMETHEE II employed Type V preference functions with indifference and preference thresholds (q = 0.20σ, p = 0.60σ). A two-level validation system was used, which comprised concordance analysis and extensive robustness testing. There was high concordance in overall resilience rankings (p < 0.001) and 62.0% of locations had the same resilience classification. However, large disparities in rank (mean = 112.3 positions; maximum = 464) and spatial randomness of disagreement (Global Moran I = -0.0099, p = 0.5541) demonstrated location-specific disagreements that could be used to prioritize intervention. Marked methodological contrasts were found in the robustness analysis. PROMETHEE II was found to have rank-reversal rates that were 80% lower and had a very high level of weight stability, whereas TOPSIS had 100% weight sensitivity, with all indicators. The indicator-level concordance of 96.4% was used to establish that 96.4% of the indicators were very highly correlated (Δρ < 0.1) due to cumulative aggregation effects. TOPSIS offers computational simplicity suited to exploratory analysis where indicator weights are well-established and rank stability is a lesser concern, whereas PROMETHEE II's demonstrated robustness to weight uncertainty (rank-reversal and weight-sensitivity testing) makes it preferable when stakeholder-elicited weights carry inherent uncertainty and ranking stability is operationally critical, such as in high-stakes flood resilience planning.