Regional flood frequency analysis (RFFA) is widely used to estimate flood quantiles at ungauged sites. RFFA involves the pooling of gauged stations to form homogeneous groups. A homogeneous group can be formed based on geographic boundaries or in catchment attributes space. In this study, we have pooled stations to form groups of catchments based on the ‘nearby basin concept’, which assumes that a group is to be formed based on basin proximity. Flood and catchment data from southeast Australia are used in this study, and several catchment groups are formed based on this concept. In Group 1, gauged catchments from Basins 201, 203, 204, 206, 207, 208, 209 and 210 are pooled together as they are adjacent to each other although they are not on the same rive network. Group 2 considers gauged catchments from nearby basins 211, 212, 215, 218, 219, 220, 221 and 222. In Group 3, catchments from basins 416, 418 and 419 are pooled together. In Group 4, catchments from basins 405, 406, 407 and 421 are pooled together. In Group 5, catchments from all these basins are pooled together to form a supergroup. For each of these groups, a quantile regression technique (QRT) is applied to develop prediction equations. Based on leave-one-out validation, it is found that the basin neighbourhood concept improves the model prediction in RFFA.