Both short- and long-term dependence of the peak flow series of 90 Canadian rivers were analyzed two decades ago. It showed that although short-term dependence was practically absent for most of the flow series, significant long-term dependence was present for a large number of rivers tested. With 20 or more years of additional data available today, the authors analyzed 57 rivers (only 57/90 were suitable for analysis) for: a) short-term dependence using several parametric and non-parametric tests; b) long-term dependence using a resampling-based Hurst’s K; and c) trend using Mann-Kendall’s test. Results showed that as expected, short-term dependence is practically absent in all rivers before or after the additional data. However, the percentage of rivers showing long-term dependence remains high. The trend test showed that most of the rivers showed no trends before or after the additional records were added. However, several rivers showed a downward trend, and a few showed an upward trend before and after the additional records were added. This study showed that sample statistics and the associated statistical significance tests can change unpredictably over time. Hence engineering decisions made in the past need to be re-visited and cannot assumed to remain unchanged especially when dealing with natural phenomenon such as annual peak flows.
Lye and Lin (1994): Long-term dependence in annual peak flows of Canadian rivers, Journal of Hydrology, 160, pp 89-103.