Abstract: The frequent abnormal river regimes in the braided reach of the Lower Yellow River since 1985 have significantly complicated rational management of the Yellow River. This study aims to characterize the spatiotemporal distribution and evolution laws of abnormal river regimes in the braided reach of the Lower Yellow River from 1985 to 2023. Deep learning methods were applied to historical remote sensing images to identify the changes to the three typical abnormal river sections in the upper, middle, and lower reaches of the braided segment. A novel approach, Deviation degree from the proposed river alignments (D_D), was introduced to quantitatively assess the degree of deviation of the mainstream from the proposed river alignments, following which the evolution of abnormal river regimes was examined at three morphological scales: local river sections, individual river bends, and cross-sections. The results show that operationalization of the Xiaolangdi Reservoir has shifted the locations of abnormal river regimes upstream, predominantly in the upper section of the braided reach. Abnormal regimes are more likely to develop under low-flow and high-sediment conditions. River sections with a D_D ＜ 0.2 will not develop abnormal regimes. Riverbed material influences the evolution of abnormal river regimes, with the cut-off of abnormal river bends initiating scouring at the deepest river point, which subsequently progressively expands laterally. The outcomes of this study can guide improved scientific management of the braided reach of the Lower Yellow River.