Flood scaling study is an important means to solve the problem of flood prediction in ungauged and poorly gauged basins. With the impact of climate change and human activities, the mechanism and process of flood are constantly changing. However, in many areas, there are only simple scaling results that can be used to guide daily work. Taking the Daqinghe River basin as an example, a fixed flood scaling exponent determined in 1974 (before the change point 1979) is still used all over the basin, which is apparently no longer appropriate. Therefore, in this paper, we aim to explore: (1) the scale relationship between the peak flows and the basin area under changing environment; (2) the validation of the scale invariance theory; (3) the physical relationship between the event-based scaling theory and the annual flood quantile-based scaling theory in the mesoscale non-nested and partly nested basins; and (4) the modification of the existing uniform flood scaling exponent in the study area. To achieve these objectives, eight simultaneous observed flood events in seven non-nested and partly nested mesoscale sub-basins of the Daqinghe River basin were selected to analyze the flood scaling theory. The results showed that there was a scaling relationship between the flood peaks and watershed area for the flood events, and the scale invariance theory was also supported herein. To analyze the effect of the environmental conditions on the flood scaling in the Daqinghe River basin, the flood events were reconstructed after the change point (the year of 1979). It has been found that the flood scaling exponents of the reconstructed flood events are larger than those of the observed events after change point. The flood scaling exponent changed with flood events, which varied from 0.65 to 1.26 when considering the basin area as the independent variable, and it decreased with a minimum of 0.36 when taking the rainfall characteristics into consideration. It has also been found that the mean of the event-based scaling exponents is larger than the annual flood quantile-based scaling exponents.