Stationarity of AMAX series of short-duration rainfall for long-term Canadian stations: detection of jumps and trends

Short-duration rainfall data are key inputs for design of stormwater infrastructure. To address concerns raised regarding the validity of relations based on these data, 13 Canadian climate stations were selected: one station for each province and territory, having a "long" record length (27 to 63 years), minimal missing years and potential significance for urban infrastructure design. For each station, a state-of-practice time series analysis was completed for the nine intensity duration frequency (IDF) durations (5 min - 24 h) to identify non-stationarities in the form of trends or jumps. Approximately, 30% of the time series showed statistically significant trends or jumps. Data sets not satisfying the stationarity criterion should be critically reviewed before computed quantiles are applied for design. Arbitrary split sample analysis was investigated as an alternative method for identifying non-stationarities and was found to be an unreliable indicator of significant changes.