Abstract: The changing environment has brought about time-varying high-order moments of a series of river stage. Therefore, those design lowest navigable water levels considering only variations in the series mean are inconsistent with reality. To solve this problem, we develop a variance variability-based guarantee-rate frequency analysis method based on the idea of decomposition and integration and a second-order moment variation analysis, and demonstrate its application in a case study for the Gaodao hydrological station. Results show that when variance variation is considered, design lowest navigable water levels at guarantee rates of 95% and 98% under the current condition are 2.15 m and 1.60 m higher than those of traditional designs respectively (at frequency of 80%). And the differences in the design values among three cases ? considering variance variation, considering mean variation only, considering neither of them ? are prominent, confirming the necessity of analyzing both mean variation and variance variation.

Key words: lowest navigable water level, non-stationarity, variance variability, guarantee-rate water level frequency method, decomposition and integration