Abstract: Sediment flow behavior depends mainly on the runoff erosivity in the Loess Plateau, which is one of the most seriously eroded areas in the world. It is necessary to effectively identify the flow-sediment relationship in the erosive runoff events, thereby understanding the water and sediment movement. In this study, two energy indicators were proposed to determine the erosive runoff events, including the stream power (ω, W/m) and stream energy factor (SE, W). The sediment-runoff yield data of 37 slope runoff events was selected through statistical frequent. The energy parameters were also established to investigate the flow-sediment relationship at inter- and intra-event time scales, according to the identification of erosive-runoff events. The results showed that the general standards of parameters in the erosive runoff events at the slope scale were: the stream power or energy was greater than10.5 W/m or 0.059 W, respectively; the area-specific sediment yield was above 0.6 kg/m2 (equivalent to 600 t/km2), the sediment load above the reference standard was accounted for 95% of the total, and the erosive runoff events was about 37% of the total. 16 erosive and 21 non-erosive runoff events were then selected to determine the flow-sediment relationship at the inter- and intra-event time scales. A power function was found between the area-specific sediment yield, sediment concentration, and energy parameters, whether the erosive- or non-erosive-runoff events. The area-specific sediment yield was primarily controlled by the stream energy factor, where the mean sediment concentration was well characterized by the stream power. There was a better fitting in the flow-sediment relationship for the erosive runoff events, compared with the non-erosive. A power function was also found between instantaneous sediment concentration and instantaneous stream power in the erosive runoff events at the intra-event time scale. The sediment concentration tended to be stable, when the instantaneous stream power exceeded a certain critical (among 50-80 W/m) in the erosive runoff events. However, there was no change in the non-erosive runoff events. A positive linear was found between the sediment discharge and instantaneous stream power, as well as between the increments in the area-specific sediment yield and the increments in the stream energy factor. Specifically, the area-specific sediment yield was reduced by up to 50%, if per 10 W/m reduction in the stream power, whereas, there was the reduction of up to 71%, if per 0.1 W decrease in the stream energy factor. The sediment delivery capacity by the erosive runoff events was 2.5 times that of the non-erosive, whereas, the increment in the area-specific sediment yield by per unit increment in the stream power for the non-erosive runoff events was 4.8 times that for the erosive. There were more powerful runoff energy, and the capacity of sediment transport in the erosive runoff events, compared with the non-erosive. Correspondingly, the total flow-sediment relationship can be represented by the erosive runoff events, rather than all the events. As such, the feasible discrimination standard was constructed to identify the erosive runoff events. Overall, the flow-sediment relationship can be expected to serve as the typical runoff events on the loess slope from an erosive energy perspective. Therefore, the runoff regulation on the slopes can also be implemented to reduce the stream energy factor, where the connectivity of erosive energy flow was removed for the sediment control. The findings can provide a strong reference for the decision making on the regulation of erosive runoff energy, as well as the design and the implementation of anti-erosion measures on the hill slopes in the Loess Plateau.