The independent storm event is identified from precipitation time series based on a criterion of a specific minimum dry period. The selection of the criterion is arbitrary due to different research purposes. The minimum dry period is set to 6 h in this study, which is corresponding to the average flow concentration time of CRB. An independent storm event could be identified if it separated from the preceding and succeeding precipitation by the minimum dry period or longer. The storm duration, inter-storm period, average storm intensity, and the within-storm pattern could be calculated once a single storm event is identified. Then, the average storm duration (inter-storm period) could be determined by total duration (inter-storm period) divided by the number of storm events (inter-storm periods), and the average storm intensity could be calculated by total precipitation divided by the total duration of the selected storm events, and the within-storm pattern could be characterized using cumulative normalized mass curves (Equation 4 in Figure 2). Storm properties during a certain period, for example, the MFS and TFS, are statistically analyzed from the observed station data and CMIP6 precipitation projections (Figure 2). The differences between each storm property are analyzed for storm events of different magnitudes. To further assess the potential changes in the future, we divided the whole CMIP6 simulation period into three parts: hist (1981–2010), future1 (2021–2050), and future2 (2061–2090). The percentage changes in storm properties were computed by comparing the results from future runs with those from historical runs. The significance of the changes is tested using a permutation resampling method (Yu et al. 2015). Sets of years (30 years) are randomly sampled without replacement from both the historical and future periods. Monthly storm properties are calculated for each set of years. The sampling is repeated for 1,000 times and 1,000 sets of storm properties are obtained. Then it is considered significant at the 0.1 level if the magnitude of the storm properties exceeds the 95th or 5th percentile of that determined from the random permutations. We also evaluate the uncertainty of storm properties during the two rainy seasons under four climate change scenarios (Table 1) through the combination of shared socioeconomic pathways (SSPs) and representative concentration pathways (RCPs).