Flood retention lakes (RLs) are widely employed in rural-urban catchments with low impacts on the natural environment. While hydrologic models have been commonly applied to evaluate RLs’ performances, insights are lacking over the consequences of a wide climatic variability, particularly those corroborated by 2D hydrodynamic models. Thus, this study aims to conduct a systematic catchment-scale evaluation of RL effectiveness; blueprinted RLs with various geographic configurations are also considered in addition to the individual RL. A 2D hydrodynamic model is verified to be capable of simulating flood events in the rural-urban catchment interacted with RLs. Under a wide range (1- to 15- hour duration and 1- to 100-year return period) of rainstorm scenarios, the RL has a satisfactory performance within an L-shaped band in the frequency-duration diagram. The L-shaped band coincides with moderate return periods (10- to 25- year) and large durations (> 6 hours) or small durations (< 4 hours) and large return periods (> 25-year), whereas different criteria yield different optimal combinations. With the increase of event size, 4 typical stages of RL-river interactions are characterized, which projects the catchment-scale performances. Blueprinted RLs with distributed and parallel connections mitigate larger areas of inundation in sub-watersheds than aggregated and series ones. Upstream controls are less effective than downstream controls under moderate events while the relation is reversed for extreme events. Critical changing factors concerning RL-river interactions and spatiotemporal rainstorm variabilities prompt comprehensive considerations of both local-scale RL configurations and catchment-scale climate characterizations.