Sydney, Nova Scotia Record Rainfall Not Record for the Region - Trends Decreasing - No Climate Change Impact

Expanded post with regional trend data.

CBC News published a story describing record rainfall due to the remnants of Hurricane Matthew. While the 225 mm of rainfall recorded over the short Sydney rain gauge  monitoring period is a record, the record in the region in Halifax, Nova Scotia was 239.5 mm back in September 1942. Environment Canada's Engineering Climate Datasets provide historical records - an excerpt of Halifax data is shown below.

The full data is available at the following site and can be downloaded by region:

Trends in annual maximum extreme data in Halifax are up and down according to data included in the Environment Canada datasets - trends are up for short durations of less than 1 hour and down for longer durations as shown below:
It is likely that if the 1942 event was not recorded then longer duration events would have increasing trends too in Halifax. This helps illustrate the random nature of recording records. Records should be reviewed with caution unless the observation period is very, very long.

Sydney records only began in 1961 missing the 1942 event, suggesting the 2016 event is a record for that area ... well it is, but only because the observation period is relatively short to assess extreme events ..... if a tree falls in the forest and noone is there to hear it we think it really didn't occur.


In a recent report by CTV News Atlantic about flood damages in Sydney, Nova Scotia, an incorrect statement on extreme rainfall trends is made.

"(Mayor of the Cape Breton Regional Municipality Cecil) Clarke said the weather is a reminder that climate change is contributing to more intense storms on the East Coast."

This contradicts Environment Canada scientists' analysis of extreme rainfall trends in the region, as published in Atmosphere-Ocean in 2014. In that review of the Engineering Climate Datasets scientists say for short duration rainfall intensities that govern urban flash flooding, trends are decreasing:

"The decreasing regional trends for the 5- to 15-minute duration amounts tend to be located in the St. Lawrence region of southern Quebec and in the Atlantic provinces." and then

"The decreasing regional trends for the 5- to 15-minute durations are mainly located in southern Quebec, most of the Atlantic provinces, and in southern Manitoba and Alberta."

Environment Canada reports the "maximum non-significant decreasing trend value in the AMS (Annual Maximum Series) amounts of −4.4% per decade (-0.44% per year) for the 5-minute duration."

Drilling down into the local area, Sydney has no statistically significant trends in annual maximum rainfall intensity, based on Sydney CS Station Number 8205702 data trends.
(see file idf_v2-3_2014_12_21_820_NS_8205702_SYDNEY_CS_t.pdf in the dataset)

It is important to note that increasing observations of more extreme rainfall events over time does not necessarily indicate a change in the underlying rainfall behaviour, e.g., due to climate change or even other factors, but rather 'regression to the mean'. Short rainfall records of several decades fail to capture extreme events as a simple matter of statistical probability - longer records, and more observations capture more extreme events. For natural phemonema that have skewed probability density functions describing the rare nature of extreme events, it can takes many decades to accurately characterize the behaviour. Generally, short records with high coefficients of variation in the random variable (rainfall intensity) will underestimate the true extreme values the most. In the field of catastrophic loss estimation, and as published by Fleming in the journal Variance, a peer-reviewed journal published by the Casualty Actuarial Society:

"Also,the most likely sample average value for any small sample from a skewed population will be below the mean of the skewed population being sampled. Experienced actuaries are aware of these issues. However, we have to be on guard and not fall back on easy assumptions that are appropriate for results from symmetrical distributions."

Environment Canada Engineering Climate Datasets show that it took twenty years for a 100 year statistic to be exceeded in Sydney - observations started in 1961, and 100 year 30-minute to 24-hour intensities were observed in 1981. That raised the bar on the baseline conditions, making it harder to exceed the 100 year statistic. Then in 2012 the 100 year statistics were exceeded for only 1 and 2 hour durations, meaning it is becomes harder to exceed over time as a longer record better reflects the underlying characteristics and extreme value probabilities. The 2016 Hurricane Matthew statistics should exceed the 100 year statistics again - raising the statistics to better reflect the underlying distribution of extremes. But with only 53 years of record since 1961 (1992 and 2010 are incomplete), the chances that a 200 year or 500 year extreme storm has been observed are rare. This is a reminder that regional storms can be considered in floodplain management and urban drainage design. If Sydney drainage systems (floodplains, infrastructure) were designed considering the 1942 Halifax storm as extreme regional conditions, perhaps flood damages would not be so severe.


Over time there is a regression to the mean. The incidence of hurricanes has been in a lull as pointed out by Pielke Jr. in the graph below:

is hurricane frequency increasing in US - no

Hurricane Harvey a category 4 storm is expected every 9 years or so. So it is due given the gap in days since a hurricane.