Estimating extreme water levels with long-term data by GEV distribution at Wusong station near Shanghai city in Yangtze Estuary

In this study, a 91-year data set at the Wusong Station near Shanghai City in Yangtze Estuary has been used to estimate the 100-year Annual Maximum Water Level (AMWL). The performances of four common distribution models have been evaluated. The GEV model provides the best estimates of an AMWL. It results in the minimum difference (0.04 m) compared to the observed 92-year AMWL, with the high correlation coefficient (0.99) and minimum root-mean-square-error (0.045 m) value. Predictions from other distribution models cause non-negligible deviations, underestimating the 92-year AMWL by 0.57, 0.38, and 0.15 m for Weibell, Lognormal, and Gumbel distribution models, respectively. In order to examine the effects of a shorter data set, a 59-year data set was investigated. Model predictions using 59-year data set underestimates the observed 60-year AMWLs. By comparing to the 100-year AMWL estimated by the GEV distribution, using the 91-year data set, results using the shorter 59-year data set lead to underestimates of the 100-year AMWL by 0.78 m for Weibull, 0.58 m for Lognormal, 0.38 m for Gumbel, and 0.39 m for GEV distributions. Therefore, one should be cautious when estimating the 100-year AMWL if the data set covers a period much shorter than 100 years. Selecting an appropriate distribution model can improve prediction accuracy. (C) 2010 Elsevier Ltd. All rights reserved.