A framework for determining lowest navigable water levels with nonstationary characteristics

The Lowest navigable water level (LNWL) is an important indicator used for navigation design to balance the relationship between navigation safety and economic benefits of a waterway. However, it is a challenge of accurately estimating LNWLs due to the nonstationary characteristics of observed water level data series. In this study, a comprehensive framework was developed for handling this issue. In this framework, inter-annual variabilities in both the mean and variance of water level series were described by decomposing original series and were eliminated by composing new series. Intra-annual variability was determined by detecting indicators describing intra-annual water level distributions. Considerations of inter- and intra-annual variabilities were combined by designing annual water level processes for the past and current environments. Shipping risks during both annual and multi-annual periods were considered in the framework as well. The framework was demonstrated in estimating LNWLs at the Gaodao and Shijiao stations in the North River basin, southern China. The recommended LNWLs at the Gaodao station were 22.32 m for 95% guaranteed rate and 21.84 m for 98% guaranteed rate; LNWLs at the Shijiao station were 0.27 m for 95% guaranteed rate and 0.15 m for 98% guaranteed rate. The impact of variance variability on estimations of LNWLs was also evaluated. Results indicated that the recommended LNWLs would have errors of 0.11 ~ 0.48 m at the Gaodao station and 0.03 ~ 0.04 m at the Shijiao station if the variance variability was not considered. The proposed framework was then compared with Nonstationary synthetic duration curve (NSDC) method, and results illustrated that the duration curves plotted by NSDC method were unreasonable, leading to inaccurate design values. Overall, the developed framework is more reasonable and suitable for designing LNWLs of waterways where the variabilities of the water levels at different time scales are different or where the historical water level data contain various variations .