Abstract: Global climate change and intensified human activities have disrupted the stationarity of hydrological time series, and consideration of sample non-stationarity may generate more uncertainty in frequency analyses. Taking the Yangtze River Estuary as the representative area, this study conducts a non-stationary hydrological frequency analysis （NHFA） on the tide level series of the estuary. To consider properly the possible occurrences of change points in a time series, a non-stationary bootstrap uncertainty estimation method based on conditional probability distributions is proposed in this study. The results indicate that the design tide levels at representative stations in the Yangtze River Estuary increase under different return periods due to the non-stationarity of the tide level series. With the capability of accurately identifying change points, the proposed method can quantitatively assess the parameter uncertainty of the non-stationary hydrological frequency analysis results. It was found that under non-stationary conditions, the uncertainty intervals of design values at lower return periods are relatively larger, and the range of affected return periods decreases as the magnitude of change increases.