Can We Use Daily Rainfall Models To Predict Short Duration Trends? Not Always - Observed Daily and Short Duration Trends Can Diverge

One can assess trends in rainfall intensities over various durations and return periods using Environment Canada's Engineering Climate Datasets.  National trends based on updating 226 station IDF curves were shown in an earlier post.

What are the trends in regions of Canada that have experienced significant flooding in the past?  And do the trends projected by models for long durations (1 day precipitation) match observed data trends?  No - some 24-hour trends are decreasing despite models estimating they will go up (or have gone up because of increasing temperatures).

Also, what is happening with observed short duration intensities, the ones responsible for flooding in urban areas, compared to the observed 1-day trends?

The data show short duration and long duration trends diverge. Therefore relying on models of 1 day precipitation to estimate what is happening with short duration, sudden, extreme rainfall should be done with caution.

A couple charts help illustrate these observed data trends and show what is wrong with relying directly on models to project local extreme rainfall.

This is the trend in observed rainfall for southern Ontario climate stations, using median changes in IDF statistics:

Southern Ontario Extreme Rainfall Trends

Long duration intensities are decreasing and short duration intensities are decreasing even more.  The extreme intensities (red dots = 100 year, orange dots = 50 year) decrease more than the small frequent storm intensities (green dots = 2 year).  Observed data diverges from Environment Canada models that suggest intensities are going up due to a warmer climate (see recent CBC article).

These are the trends for Alberta observed rainfall when new data are added and are reflected in the most current v3.10 datasets:

Alberta Extreme Rainfall Trends

In Alberta, long duration intensities decrease significantly (100 year is down by 4% on average).  Meanwhile the short duration intensities increase.  The long duration decrease is contrary to Environment Canada's simulation models that estimate 1 day rainfall at a sub-continental scale.

In northern Ontario, trends are different than in southern Ontario as shown below:

Northern Ontario Extreme Rainfall Trends

In northern Ontario the long duration intensities have increased but short duration intensities have decreased on average.  So we see short and long duration rainfall trends are diverging when we consider new data.

Climate modellers may suggest that simulated 1 day precipitation can guide what happens during short durations too.  Observed data suggest otherwise.  Trends actually diverge.

In brief, for this sample of regions shown above, we see these trends:

Location                 Short Duration Trend         Long Duration Trend

Southern Ontario      Larger Decrease                        Decrease
Northern Ontario            Decrease                              Increase
Alberta                            Increase                               Decrease

Remember "All models are wrong, some are useful".  Climate models do not accurately project changes in extreme rainfall in Canada based on observed data.  Furthermore, simulated 1 day precipitation trends from models cannot be used to assume short duration trends related to flooding in urban areas - short and long duration rainfall trends are observed to change in opposite directions in sample regions across Canada.

When using 1 day rainfall trends to estimate short duration trends, given the actual observed data trends above, it may be appropriate to conduct sensitivity analysis on potential shorter duration trends, especially if those shorter durations influence system behaviour (e.g., 'flashy' urban drainage systems).  Those short duration trends trends may be in an opposite direction or magnitude than the 1 day trends. For example, in Northern Ontario the 1 day 100-year intensities have increased 2% as a result of the most recent IDF data updates, however the intensities for durations of 2 hours or less have mostly decreased.

The following chart compares the 30 minute, 1 hour and 2 hour 100-year intensity trends with the 24 hour 100-year trends at 226 climate stations across Canada.

The correlation of short duration trends with 24 hour trends is weak with R-squared value of 0.12 for 2 hour trends, 0.06 for 1 hour trends and 0.006 for 30 minute trends.  This suggests that short duration trends are not correlated with 24 hour trends.   

***

Given recent flooding in British Columbia, it is worthwhile looking at trends in design rainfall intensities in BC. This chart shows that extreme rainfall intensities (red dots with 100-year return period, orange dots with 50-year return period) have not increased for most durations - the 12 hour duration intensities are up slightly on average (less than 0.5% increase), while other duration intensities have decreased by more than 2 % on average (5-minute 100-year intensities).

For the longest duration of 24 hours, intensities have decreased on average - typical 2-year intensities are unchanged on average while the moderate and extreme intensities (5-years, 10-year, 25-year, 50-year and 100-year) have decreased on average.

The following tables shows trends in observed annual maximum rainfall over various durations at BC climate stations with long-term records. These are called the Annual Maximum Series (AMS). Trends in derived design rainfall intensities above (e.g., 2-year to 100-year rainfall rates) follow these trends in AMS.

Extreme Rainfall Trends in Canada - Engineering Climate Datasets for Long Term Climate Stations Show Increases and Decreases

Environment and Climate Change Canada's Engineering Climate Datasets includes trends in observed annual maximum rainfall over durations of 5 minutes to 24 hours.  Version 3.00 of the data was released in early 2019 (see Intensity-Duration-Frequency (IDF) Files https://climate.weather.gc.ca/prods_servs/engineering_e.html and Google Drive link to trend charts grouped by province and territory https://drive.google.com/drive/folders/1VzJdW7DUIA3mpqz8mA8jqG6LdCqy_UOF).  The following table shows trend direction and significance for stations across Canada.  It represents 3993 station-years of data, with an average of 47 years of data at 85 stations.

Some observations:

- out of 85 stations with trends over 9 durations, 7.9% of trends are statistically significant increases

- 1.8% of trends are statistically significant decreases

- the total of significant increases and decreases (7.9+1.8=9.7%) is mostly explained by chance (5% could be explained by random chance, due to the natural variability of the data)

- there are more increases than decreases with the exception of Ontario where southern Ontario has more decreases than increases, while northern Ontario has more decreases

- southern Ontario has 50% more significant decreases than increases

- Alberta is almost even with increases and decreases, and has no statistically significant increases, and just one significant decrease

- statistically significant increases are more prevalent for long durations over 1 hour (10%), than for short durations of 1 hour or less (6.4%) .. so significant increases for short durations are slightly above the % explained by randomness in the natural variability, in contrast, long durations have more significant increases than would be expected by chance

- statistically significant decreases are more prevalent for short durations of 1 hour or less (2.1%), than for long durations of over 1 hour (1.5%)

A review of these trends based on earlier v2.30 datasets, specifically stations with 20 years of record between 1965 and 2005, was presented by Shephard et. al in Atmosphere-Ocean in 2014:

"Summary statistics in Table 6 show that for all durations fewer than 5.6% and 3.4% of the total number of stations have significant increasing and decreasing trends in the AMS amounts, respectively. The highest percentage of stations with significant trends from any duration is 7.8%
(5.6% + 2.2%) for the 24-hour duration, which is close to the nominal 5% significance level. Based on this IDF single station analysis, and the more general single station climate results from the 1965–2005 period presented in Section 4a, we conclude that the annual maximum short duration rainfall values across Canada typically do not show a significant trend. Thus, for most of the single station IDF stations across Canada there is no evidence indicating that the stationarity assumption used in the traditional national EC IDF calculations has been violated. These results are not unexpected given the typical high variability and relatively short time series of the extreme short-duration rainfall observations."

Therefore Environment and Climate Change Canada find 'no evidence' that data used in IDF calculations is changing (values are stationary), and significant trends are generally no more than the natural variability would suggest.

***

The version 3.1 datasets have just been released.  An assessment of trends at all stations is included in a new post: https://www.cityfloodmap.com/2020/05/annual-maximum-rainfall-trends-in.html - it also shows how trends have changed from older data sets (right chart) to the most recent sets (left chart) - no appreciable change.

Canadian Annual Maximum Rainfall Trends and Statistical Significance

The version 2.30 dataset was updated with data up to 2013 in 2014.  Version 3.00 was updated in early 2019 with some stations updated to 2017 but some with last update as far back as 2007.  The version 3.10 fills in many recent gaps and adds more stations - there were 565 in v2.30, 596 in v3.00 and now 651 stations in v3.00.  The v3.0 trends across Canada are shown below.

Annual Maximum Rainfall Trends in Canada - Environment Canada Engineering Climate Datasets v3.00 (released 2020) - trends per Environment Canada file idf_v3-10_2019_02_27_trends.txt

Southern Ontario Observed Rainfall Intensities Decreasing - Annual Maximum Values Lower In Environment and Climate Change Canada's Engineering Climate Datasets (Version 3.0)

Long term southern Ontario observed maximum rainfall trends,
according to Environment and Climate Change Canada's Version 3.0
Engineering Climate Datasets - decreasing trends in rain intensity and
more significant decreases than increases. 

Good news! Rainfall intensities have been decreasing in Ontario, Canada's most-populated province according to newly-released data. Less intense rain means lower urban flooding risk, contrary to many media reports that have confused future predictions of more extreme weather as a climate change effect with actual observed changes in the past. 

Maximum annual rainfall amounts over short durations at Ontario climate stations are used to derive engineering design intensities used in design of infrastructure such as sewers, culverts, channels, and ponds - the things that help convey rainfall runoff safety away from otherwise vulnerable people and property.

Environment and Climate Change Canada (ECCC) has recently updated its Engineering Climate Datasets that include a statistical analysis of observed trends in maximum values observed each year. The newest data are identified as Version 3.0 and are available as part of the Intensity-Duration-Frequency (IDF) Files on the ECCC website:
http://climate.weather.gc.ca/prods_servs/engineering_e.html

The previous Version 2.3 datasets showed decreasing annual maximum values at 21 southern Ontario climate stations with at least 30 years of observations - see previous post.

The updated Version 3.0 datasets continue this decreasing trend, showing that at the same 21 climate stations with an average observation period of 47 years:

1. There are 42% more decreasing trends than increasing ones across all durations and stations (55.6% decreasing trends vs. 39.2% increasing ones).
2. There are 75% more statistically significant decreases than increases (7 significant decreases vs. 4 significant increases).

This table shows the station name, ID, trends for each duration of 5 minutes to 24 hours, as well as the length of record and the most recent year in the Version 3.0 dataset.

Southern Ontario Observed Maximum Rainfall Trends - Environment and Climate Change Canada
Engineering Climate Datasets - Version 3.0
Trend Direction and Significance for 21 Climate Stations with Long Period Records (Greater than 30 Years)

Other observations:

1. There are no statistically significant increases for durations less than 6 hours - that means the short duration convective storms burst that can lead to urban flooding related to most infrastructure systems do not show any appreciable increases.
2. Overall downward trends are contrary to insurance industry statements, particularly the disproved "Telling the Weather Story" claim that there has been a one standard deviation increase in the probability of extreme rainfall according to Environment Canada data (the "Story" was only a theory/concept incorrectly cited and widely misreported as real data).
3. Overall downward trends are contrary to many media reports citing a new normal of wild weather. Fortunately, some media, lead by the the Financial Post's Terence Corcoran are engaged in a critical review of urban flood drivers including extreme rainfall and the means to mitigate flood damages:
• Article clarifying the "Weather Story" and discussing effective means of mitigation.
• Article highlighting media inaccuracies on extreme rainfall reporting.
• Article questioning City of Toronto statements on more extreme rainfall as a cause of flooding that are not supported by data or statistical analysis of frequency or trends.
4. CBC staff and the CBC Radio Canada Ombudsman have helped focus on facts Environment and Climate Change Canada data and corrected many stories on increasing storm frequency or intensity as noted here:
• CBC Ombudsman finds violations of journalistic standards of practice regarding reporting on 100 Year storm trends,
• CBC Windsor corrects story on the number of storms affecting insurance premiums,
• CBC News corrects story on Environment Canada report acknowledging no change in extreme precipitation linked to urban flooding:

5. Analysis by the School of Engineering at the University of Guelph, published in the International Journal of Environmental Research in 2015, looked at monthly trends and suggested that "The decrease in August extremes seems to have a significant impact on the annual extremes in the southwest and southeast regions": https://drive.google.com/file/d/1AngUYFFlm-RqQlmSC0gZqxy8nV61BW8J/view

Urban flooding is certainly an important issue to be addressed. And there are many factors that affect today's flood risks as explored in a previous post. While the insurance industry has suggested a link between increasing flood damages to increasing rain extremes due to climate change, given the wealth of evidence pointing to other quantifiable factors like increasing hydrologic and hydraulic stresses - and no change in rainfall extremes! - means that there is not even a correlation much less a causation relationship between flood damage and rain extreme trends (i.e., damages are up but rain intensities are down). This was pointed out in my Financial Post OpEd. 

Effective flood mitigation strategies must recognize the intrinsic capacity limitations in the vast amount of legacy infrastructure built over 30 years ago, and focus on reducing risks by addressing any level of service gaps through adaptation. Cost-effective and timely methods can include increasing the conveyance capacity of grey infrastructure, as opposed to mitigating rain/weather stresses that have not appeared to change, based on official, national engineering datasets. While such infrastructure investments should consider potential future climate effects, and we have many examples of analyzing stormwater and wastewater systems for such effects, past trends do not point to an increase to date in rainfall extremes. As a result, derived intensity duration frequency values for the stations reviewed above, based on values in the Version 3.0 datasets, shows an overall decrease in design intensities for small frequent and large rare storms across southern Ontario - those results were presented in a previous post, as shown below:

Ontario Intensity Duration Frequency (IDF) Trends - 2 Year to 100 Year for all Durations
Environment and Climate Change Canada Engineering Climate Datasets - Version 3.0

Recognizing trends in observed rainfall maximum values and the derived design intensities will support data-driven, evidence-based policies and programs for achieving flood resilience through strategic infrastructure investments.

***

The following table explores annual maximum values at Ontario climate stations with over 50 years of record:

Ontario Observed Maximum Rainfall Trends - Environment and Climate Change Canada
Engineering Climate Datasets - Version 3.0
Trend Direction and Significance for 11 Climate Stations with Long Period Records (Greater than 50 Years)

The table expands into higher latitude eastern Ontario communities including Kingston and Ottawa as well as to northern Ontario. The eastern Ontario climate stations show an overall consistent trend in decreasing observed rainfall maxima over the shortest durations. Another eastern Ontario station, the Ottawa Airport also shows decreasing trends over short durations, including several statistically significant decreases (i.e., lower observed rainfall intensities) for durations of 10 minutes, 15 minutes and 1 hour.

Previous analysis of the Version 2.3 datasets showed the differences in southern and northern Ontario trends. Increases in intensities in the north, beyond Ontario's largest urban centres, could reflect a shift toward more rainfall events instead of snowfall as a result of warming temperatures. 

Environment Canada Report Confirms No Overall Change in Extreme Rainfall - Generally Random Ups and Downs - Stated Certainty of Future Shifts Contradicts American Society of Civil Engineer's "Significant Uncertainty"

A new Environment and Climate Change Canada (ECCC) report Canada’s Changing Climate Report https://changingclimate.ca/CCCR2019/ reviews past, observed rainfall extremes https://changingclimate.ca/CCCR2019/chapter/4-0/ and confirms there are no observed changes in extreme rainfall across the country:

"For Canada as a whole, there is a lack of observational evidence of changes in daily and short-duration extreme precipitation."

ECCC predicts increases showing a theoretical probability density function shift (Figure 4.21) where the blue line probability density function represents today's/yesterday's eventt magnitudes and frequencies without climate effects, and red represents with effects (shift right means higher magnitude for any frequency):

Engineering Climate Datasets in some regions show trends in the magnitude of rain intensity magnitudes (reality) going the other way however:
https://www.cityfloodmap.com/2019/03/idf-updates-for-southern-ontario-show.html .

This image shows the difference between the theory and the local data reality - the green line is the REALITY showing for any given frequency (2, 10, 50, 100 Year events) the magnitude is going down in southern Ontario:

ECCC suggests there is insufficient data to observe the changes in extremes expected: "Estimating changes in short-duration extreme precipitation at a point location is complex because of the lack of observations in many places and the discontinuous nature of precipitation at small scales." - while that MAY be accurate for extreme events that are rare and elusive, why do 2 Year rain intensities, derived from many, many yearly observations at all long term rain gauges, show the clearest decline, across all durations from 5 minutes to 24 hours?

Surely, we have DO enough point locations and observations to see the change in these small storms. But if these small frequent storm intensities are no higher with today's temperature shifts, why do we expect the extremes to be higher either? Data we do have shows in southern Ontario these 100 year intensities are 0.2% LOWER on average. So extremes are shifting shifting along with the means.... shifting lower.

A theoretical probability density function shift has been promoted in the past by ICLR and IBC in the 2012 Telling the Weather Story report:

This has been shown to be 'made-up' and not related to real data (ECCC IDF tables and charts mistakenly cited as the source of the 40 year to 6 year frequency shift) - this chart shows the theoretical 1 standard deviation shift widely circulated by IBC and real data shifts:

See the difference between theory and data? It is pretty clear.

Given the lack of past trends, and uncertainty in future noted in the ECCC report ("It is likely that extreme precipitation will increase in Canada in the future, although the magnitude of the increase is much more uncertain"), we must follow the American Society of Civil Engineer's recommended "Observational Method" approach see 2015 report Adapting Infrastructure and Civil Engineering Practice to a Changing Climate at http://theicnet.org/wp-content/uploads/2015/07/2015-07-ASCE-Practice-to-Climate-Change-2015.pdf, and also see https://ascelibrary.org/doi/book/10.1061/9780784415191?utm_campaign=PUB-20181023-COPRI%20Alert&utm_medium=email&utm_source=Eloqua# for the new 2018 manual on engineering practice Climate-Resilient Infrastructure, Adaptive Design and Risk Management.

The ASCE 2018 manual promotes incorporating any no-regret, now cost measures in design today considering most probable future conditions, and allowing design flexibility to adapt in the future if and when performance is shown to be inadequate or affected by future changes - this is a practical approach intended to avoid costly over-design, and over-investment in potentially unnecessary and cost-ineffective infrastructure today.

While the ASCE 2015 report notes the high degree of uncertainty "However, even though the scientific community agrees that climate is changing, there is significant uncertainty about the location, timing and magnitude of the changes over the lifetime of infrastructure."

In contrast, the ECCC report appears to asset a high degree of confidence in future changes saying "For Canada as a whole, there is a lack of observational evidence of changes in daily and short-duration extreme precipitation. This is not unexpected, as extreme precipitation response to anthropogenic climate change during the historical period would have been small relative to its natural variability, and as such, difficult to detect. However, in the future, daily extreme precipitation is projected to increase (high confidence). - how can ECCC assert high confidence when there are no observed trends? How can ECCC contradict ASCE's statement on high "signifcant uncertainty'?

ECCC reports that summer precipitation is expected to decrease: "Summer precipitation is projected to decrease over southern Canada under a high emission scenario toward the end of the 21st century, but only small changes are projected under a low emission scenario." - how can that be if the summer temperatures are going up? Does this not violate the Clausius-Clapeyron theory cited in the ECCC report states that "increased atmospheric water vapour in this part of the world should translate into more precipitation, according to our understanding of physical processes" - so that is a theory - what about the real data? What does it show? the Clausius-Clapeyron relationship does not stand up to scrutiny as shown in a previous post.

Given highest rainfall extreme are in the summer (see the work of Dr. Trevor Dickinson on seasonal extremes), a summer decrease in precipitation could potentially mean lower flood risks. The data for southern Ontario already show a decrease in the annual maximum series (reflecting lower means and typical 2 Year design intensities in derived IDF curves) and the extreme 100 Year design intensities are decreasing slightly as well.

Overall, many in the media have over-hyped concerns about changing rainfall severity. Data and ECCC's report shows there has been no change, beyond random fluctuation. Looking ahead the American Society of Civil Engineers indicates that future changes have "significant uncertainty"- this contracts the ECCC's statement on "high confidence" on future extremes.

Toronto Island Flooding 2017 - Were Lake Ontario Levels Extreme? No, Barely Above Historical Maximum Levels

Blockbuster Release: Toronto Island Flood 2017..
really just a sequel to 1947, 1973, 1952 lake levels ...
and to storms back in the 1800's too (note - original
record years corrected Feb. 2019)

See 2019 update at bottom of the post

It is easy to 'think fast' and make observations about hydrologic and hydraulic events without looking at objective data. For example the document Blueprints for Action Minimizing Homeowner Flood Risk in the GTHA July 2017 notes the Toronto Island flooding:

"For many Torontonians in general, 2017 could be called “the year of the flood.” Higher amounts of rainfall during the spring caused Lake Ontario to swell by 55 centimetres, submerging not only parts of the Toronto Islands, but also affecting regional shorelines..."

And concludes:

"The cause of these kinds of flooding isn’t a secret. Climate change is having major local impacts across the Greater Toronto and Hamilton Area (GHTA) including major flooding. Storms and extreme weather that typically happened every 20, 40, 80, to 100 years are now happening with increasing frequency."

What does data show us, however? Are Lake Ontario levels impacted in a major way by climate change? Not really. Below is a review of May to August Lake Ontario levels showing previous levels back to 1918:

Correction: August 2017 did not set a record - that is held by 1947 levels. April record remains 1973.

May 2017 Lake Ontario levels were a few centimetres above historical maximum values in 1973 - June levels in 2017 (not shown) were similarly only a few centimetres above previous maximum monthly average values.

We broke a previous 1952 record by a couple inches in June according to Environment and Climate Change Canada's LevelNews;

"“Lake Ontario’s level at the start of June … set a new record for the highest beginning-of-June level on Lake Ontario in the period of record (1918 to present) breaking the previous record set in June 1952 by 5 cm.”

By August 2017 Lake Ontario levels were no greater than previous maximum values in 1947. So 2017 Lake Ontario levels are not a blockbuster climate change impact but rather a slow prodding sequel to previous fluctuations in levels.

So what is different in 2017 compared to previous years with high water levels? Is the risk of flooding on Toronto Island greater today than before? Yes, and this has to do with land use planning and risk management as opposed to climate change risks. This is a short chronology of development on the Toronto Islands:

• 1858 – Storm separates Toronto Islands from the mainland. Quinn's Hotel and Parkinson's Hotel are destroyed.
• 1950’s – 630 Cottages / Homes on Toronto Islands
• 1970’s – 250 Cottages / Homes , by 1978 Metro Toronto had writs of possession for these homes and planned removal
• 1981 - Province of Ontario passed a law legalizing the Islanders to stay until  2005
• 1993 - Ontario Government passed Toronto Islands Residential Community Stewardship Act, Islanders to purchase 99-year land leases from a Land Trust

Was granting 99-year leases a good land use planning idea considering storm and flooding risks? Or did it unnecessarily add flood risk to a known high risk area, that had high lake levels in the past and significant storm damages even back in the mid 1800's? Obviously history shows us that there is not a significant climate change impact affecting flood risk but rather deliberate land use planning increasing the investment in a vulnerable area, as opposed to removing the risk (as was the plan through the 1970s).

And on the conclusion that storms and extreme weather are happening with greater frequency? Not at all. As Environment and Climate Change Canada has corrected the insurance industry on several occasions including recently in Canadian Underwriters:

"Associate Editor’s Note: In the 2012 report Telling the Weather Story, commissioned to the Institute for Catastrophic Loss Reduction by the Insurance Bureau of Canada, Professor Gordon McBean writes: “Weather events that used to happen once every 40 years are now happening once every six years in some regions in the country.” A footnote cites “Environment Canada: Intensity-Duration-Frequency Tables and Graphs.” However, a spokesperson for Environment and Climate Change Canada told Canadian Underwriter that ECCC’s studies “have not shown evidence to support” this statement."

The Toronto Island flooding is not unlike the Gatineau flooding in 2017. Buildings have been historically been built in areas that are at a high risk of flooding, and given enough time the risk will manifest themselves. Over time, old records will be broken - that is fundamental of statistical observations of random events, and not necessarily climate change impacts pushing old records higher. And the Toronto Island risks are not unlike high profile locations like High River, Alberta in 2013 or low profile locations like Cedar Beach Road, Clarington, Ontario in 2017. All these locations have intrinsic flood vulnerabilities based on their location and land use planning decisions. It is easy to call this a climate change impact but it it not accurate as the risks are long-standing ... yes back to the 1800's there were storm risks too.

Here is a summary of historical water levels (per August 2017 Bulletin and June 2017 LevelNews) and land use planning decisions at Toronto Islands (per Wikipedia):

Historical Lake Ontario water elevations and 2017 records.

Responsible journalism should not make sensational headlines out of every flood event that occurs and recognize that there is not a 'new normal' in weather extremes but rather the 'old normal' showing up again in a sequel. For example:

1) GO Train Flooding - Not New:

2) Toronto Road and Basement Flooding - Not New:

3) Toronto Island flooding and high Lake Ontario levels - not new either.

2017 levels nudged above previous records for May 1973, June 1952 and July 1947. It is not reasonable to expect that after 70 years very old records may be surpassed, like the July 1947 record being surpassed in 2017? Or that newer records would be surpassed too, like May 1973 after 44 years? After all its only a 100 year period of data and the Cornwall and Long Sault dams have been in place for just over half that time.

***

In 2019, levels have approached previous maximum levels and in June are expected to exceed previous monthly average levels by several centimetres. This chart shows May 2019 levels that are not above the 2017 maximum average monthly level - 2017 was 7 centimetres above the previous maximum in 1973.

The projected June 2019 average monthly level is expected to be 5 centimetres above the 2017 maximum. The 2017 maximum was 5 centimetres above the previous 1952 maximum.

So the records have been "nudged" higher in 2017 (May record) and in 2019 (projected June record).

How can one adjust to such water level fluctuations? Retreating from the known hazard is one approach. PARA stands for "Protect, Accommodate, Retreat or Avoid" and is discussed in the Canadian context in the paper Protect, accommodate, retreat or avoid (PARA): Canadian community options for flood disaster risk reduction and flood resilience by University of Waterloo researchers. In the paper it is noted that "In Toronto, post-Hurricane Hazel, the City used retreat as an effective strategy to reclaim and rezone flood-prone land." and that in the five years after the storm at least 530 properties were expropriated and over 9000 hectares of land rezoned to disallow housing.

In some cases iconic buildings like the Leuty Avenue Lifeguard Station in Toronto's eastern Beaches neighbourhood have been moved to accommodate changing water levels. The station has recently been raised to protect it from recent high lake levels and has been relocated 4 times since its construction in 1920: