Author(s): Henk Van Den Boogaard; Herman Gerritsen
Linked Author(s):
Keywords: Flood risk; Accelerated sea level rise; Trend break assessment; Maximum Likelihood; Uncertaint ies; Resampling techniques; Skew probability distr ibutions
Abstract: The present paper descr ibes a technique for trend assessment in water level t ime ser ies, which can detect and quant ify significant changes in trend, such as due to accelerated sea level rise. Assuming that at some point, a trend break occurs in a particular measurement series, the technique provides jo int est imates of the most likely po int in time when the change occurred, the assoc iated level and the linear trends before and after the trend break, including appropriate quantitative measures for the uncertaint ies in these estimates. In the actual determinat ion of the uncertaint ies resampling techniques are used, which allow to fully determine the probability density distr ibutions of the estimates avo iding assumptions on the uncertainty such as absence of skewness. The technique is illustrated on an applicat ion for Saint Petersburg, Russia. Long period historic water level series since 1703 show that the local flood level (+160 cm) has a return period 0. 98 year, with the data unt il 1980 show ing a largely linear trend. Based on only the data since 1980, however, this return period is twice as large. This suggests a substant ia l change in trend or trend break, possibly due to accelerated sea level rise. Application o f the method and evaluat ion of the est imates of the break point and the associated two linear trends plus the 95% confidence intervals show that a trend break in 1969 gives the best jo int estimates. The ident ified probability distribut ions reveal that skewness indeed plays a role. Moreover, a further analysis showed that the trend break is significant in statist ical sense.
Year: 2007