From little things big things flow: the drivers of flood risk adaptation

• Meaningful engagement with flood-prone households prompts adaptation. • Prior adaptations are the strongest statistical predictor of subsequent adaptation. • Easy, low-effort adaptations trigger behavioral spillovers into larger adaptations. • Embedding prior adaptations and spillovers in models better captures human agency. • Prior adaptation and spillovers boost model realism beyond rational-actor assumptions. Climate change and the urbanization of floodplains are intensifying flood risk internationally, driving demand for adaptation strategies and socio-hydrological models to more accurately capture human agency. Existing household-scale adaptation research often overlooks how adaptation evolves over time, prioritizes perceptions and intentions over observed actions, and assumes awareness straightforwardly translates into action. This study makes an original contribution by answering: what are the predictors of flood risk adaptation as implemented by flood-prone households following participation in a meaningful engagement , which prioritizes dialogue and critical self-reflection over one-way awareness raising. We analyze longitudinal, mixed-methods data that was collected during initial and follow-up engagements with 641 households across two flood-prone regions of Melbourne, Australia. Regression analyses examining flood exposure, awareness, risk perception, lived experience, and prior adaptations indicate that previous adaptive actions are the strongest predictor of subsequent adaptation. In contrast, other commonly cited determinants show no statistically significant effects. Qualitative analyses reveal that households often perceive small, actionable adaptations as catalysts for subsequent, and sometimes larger-scale adaptations. These findings suggest that meaningful engagement, rather than awareness, is a critical driver of behavioral ‘spillovers,’ whereby households’ implementation of one adaptation may set in motion a sequence of further adaptations that over time strengthen household flood resilience. By embedding prior adaptations and behavioral spillover dynamics into socio-hydrological models, modeled outputs can move beyond rational-actor assumptions to more accurately capture diverse forms of human agency. This quantification and incorporation of human agency enhances the realism of socio-hydrological models by reflecting how households actually adapt following engagement. These findings highlight the need to broaden engagement strategies toward supporting practical, incremental actions rather than relying primarily on awareness-raising.