Flood Risk Perception of Households in Adama City of Ethiopia

Abstract Flood risk perception is a critical determinant of household preparedness and response behavior in flood-prone communities. Understanding the factors that shape risk perception is essential for designing effective risk communication strategies and targeted interventions. Limited empirical evidence exists from rapidly urbanizing African cities where flood hazards are escalating. This study aimed to identify behavioral predictors of household flood risk perception in Adama City, Ethiopia, and to assess the predictive performance of a binary logistic regression model incorporating four behavioral indicators: immediate evacuation intention, unplugging appliances, checking river levels, and home protection measures. A cross-sectional survey was conducted with 384 households in flood-prone areas of Adama City. Data were analyzed using binary logistic regression with simultaneous entry of all predictors. Model fit was evaluated through the Omnibus test, Hosmer-Lemeshow goodness-of-fit test, pseudo-R 2 measures, and classification accuracy. The model was highly significant (χ² = 483.250, df = 4, p < 0.001) and explained 95.6% of the variance in flood risk perception (Nagelkerke R 2 = 0.956). All four predictors were statistically significant: Immediate Evacuation (OR = 126.4, p < 0.001), Unplug Appliances (OR = 24.2, p = 0.001), Check River levels (OR = 15.2, p = 0.001), Home protection measures (OR = 6.6, p = 0.032). The model correctly classified 98.7% of cases, with excellent sensitivity (98.9%). The Hosmer-Lemeshow test confirmed good model fit (x 2 = 1.299, df = 2, p = 0.522). Flood risk perception is strongly associated with observable household behaviors. Behaviors requiring immediate cognitive engagement (evacuation intention) demonstrate the strongest links to risk perception, followed by appliance protection and river monitoring, with home protection showing a weaker but significant association. The excellent predictive accuracy suggests that behavioral indicators can serve as practical surrogate markers for flood risk perception, enabling targeted risk communication and preparedness interventions without extensive surveys.