Christopher Zuccaro, B.A.S.c University of Toronto
Robert J. Muir, M.A.Sc., P.Eng., City of Markham
Zahra Parhizgari, M.Sc., PMP, P.Eng., City of Markham
Jack Zi, P.Eng., City of Markham
INTRODUCTION
The concept of “likelihood” in applied hydrology plays numerous roles in the design and performance of stormwater, wastewater, and natural drainage systems. Likelihood is typically represented in terms of a return period, an abstract representation of how often an event of at least a particular size is expected to occur on average over an extended time. This might sound straightforward, but within this basic concept are a variety of difficulties of both perception and reality that reward some study and reflection, especially in the context of recent extreme rainfall events that highlight system capacity limitations.
This paper explores how suggested changes in extreme rainfall return periods have occasionally been mischaracterized compared to measured long-term data trends in Environment and Climate Change Canada’s (ECCC’s) Engineering Climate Datasets. Trends across Canada, southern Ontario, and the Greater Toronto Area (GTA) are reviewed, including clarifications following reviews by Canada’s Minister of the Environment and CBC Ombudsman offices. Mischaracterizations may stem from faulty intuitive perceptions or cognitive biases based on limited available data, such as statements on the rarity of multiple extreme events that fail to recognize the proliferation of rainfall monitoring stations in the GTA. This paper statistically explores the rarity of such events, considering the areal extent of storms and the independence/dependence of adjacent gauges. It also shares examples of often-overlooked historical great rainfalls and floods, based on newspaper archives dating back to 1878 and ECCC records to 1840. These historical records, along with 2024 event data, are used to assess trends in Toronto’s 100-year rainfall.
While the data does not suggest increases in extreme rainfall intensities in Toronto or southern Ontario, other factors such as antecedent moisture conditions (defined by multi-day rainfall) affect storm runoff and extraneous wastewater flow responses. Thus, multi-day trends in Toronto and the western Lake Ontario basin, based on long-term ECCC records that are typically over 150 years in length, are assessed.
The spatial and temporal variability of 2024 extreme rainfall is
analyzed using event total volumes and selected short-duration intensity data
for several events. Data for
Lastly, while storm events are often characterized by rain gauge total event volumes and their return periods, these statistics do not align with the return period of system peak flows. For wastewater systems in particular, peak flows are more closely correlated with short-durations intensities. It is proposed that short-duration rainfall metrics be considered a valuable supplement to total event characterizations.
INTUITIVE PERCEPTIONS OF PROBABILITY VS. DATA TRENDS
Following extreme rainfall events, media commentary often demonstrates common biases in estimating an event’s rarity or severity, usually without rigorous statistical analysis. Psychologist and Nobel laureate Daniel Kahneman identified various biases in his book Thinking Fast and Slow (2013), and local examples in the context of flooding and extreme rainfall have been described by others (Muir, 2018). Politicians across Canada have routinely declared trends in severe storms, for example:
i)
Prime
Minister Trudeau, after the 2017 Gatineau/Ottawa River flood, stated:
"The
frequency of extreme weather events is increasing, and that's related to
climate change. We're going to have to understand that bracing for a 100-year
storm is maybe going to happen every 10 years. Or every few years."[1]
ii)
Toronto
Mayor Tory stated to the CBC, following a federal grant for flood relief works:
"Toronto is
experiencing more severe storms, with more rain falling over a short amount of
time.”[2]
Requests for data on storm frequency and severity have supported these claims. In 2019, Environment and Climate Change Canada Minister Catherine McKenna responded on behalf of the PMO, writing that "the observational record has not yet shown evidence of consistent changes in short-duration precipitation extremes across the country"[3]. The General Manager of Toronto Water responded that the City would adjust its messaging to not describe actual storms:
“We apologize for the delay in getting back to you
with regards to your question about the statement on storms being 'more severe'
and occurring over a 'short amount of time' and the City's messaging. We will
utilize language that describes what can happen to City infrastructure during a
rainfall rather than describing the actual rainstorm.”
(Personal communication, L. Di Gironimo to R. Muir, July 29, 2019)
To improve media messaging, the authors of
this paper have highlighted these biases in media coverage through the Financial
Post[4]
and in a ‘Backgrounder’ document for the CBC Ombudsmen, to assist CBC and
Radio-Canada editors scrutinize reports on extreme weather and flooding. This
CBC Backgrounder was intended to improve reporting accuracy, given several
recent violations of the CBC’s Journalistic Standards and Practices in 2019 and
2020 regarding extreme rainfall trends[5] [6]
[7].
To contrast statements by political figures and the media, ECCC’s long-term datasets have been used to characterize national and local trends in extreme rainfall. These trends are reported in the National Research Council of Canada (NRC) Guidelines on Undertaking a Comprehensive Analysis of Benefits, Costs and Uncertainties of Storm Drainage Infrastructure and Flood Control Infrastructure in a Changing Climate (2021). The findings align with Minister McKenna’s statement above, namely, that there is no overall increase in extremes at 226 climate stations across Canada when the last 10 years of data area added, in southern Ontario’s local IDF statistics when comparing 1990 IDF statistics with v3.1 Engineering Climate Datasets, or in local Toronto-area extreme rainfall IDF analyses.
Media and layperson statements suggesting changes in extreme rainfall often arise from an ‘availability bias’, a cognitive short-cut defined by Kahneman whereby judgments rely on easily recalled information rather than comprehensive data. This tendency overlooks many historical flood events that reveal no clear upward trend. For example, Toronto’s significant rainfall and flooding events stretch back over 150 years, with newspaper archives documenting numerous “great floods”. FIGURE 1 compiles sample newspaper clippings from extreme events in 1878, 1897, and 1905. The “Great Flood”[8] of 1878 resulted in the loss of 4 lives, or 1/20,000 of the population of about 80,000[9], equivalent to a staggering 233 lives lost in 2024 based on today’s population.
FIGURE 1 – HISTORICAL FLOOD REPORTS PER NEWSPAPER ARCHIVES (LATE 1800’s AND EARLY 1900’s)
Note: [Left] “The
Great Rainstorm”, The Globe and Mail, Sept. 14, 1878; [Centre] “Torrents of
Water”, The Globe and Mail, July 28, 1897, [Right] “Worst Storm of Year Sweeps
Over Toronto”, The Toronto Star, August 15, 1905.
Notwithstanding the actual data trends, following the July 16th, 2024 extreme rainfall in the GTA, claims of increasing storm frequency and severity were widely reported. For example, a post-storm City of Toronto council motion stated: “As a result of climate change, Toronto is experiencing more frequent and severe storms, resulting in flooding events that impact our road and transit network, our homes and businesses, and our infrastructure.”[10] A Freedom Of Information (FOI) request to the City of Toronto for documentation showing more frequent and/or more severe storms in Toronto indicated that “despite a thorough search, they [Toronto Water Division] were unable to locate any records responsive to your request[11]. The following sections analyze the probability of multiple extreme events and review southern Ontario trends, including the impact of 2024 extreme rainfall on Toronto-area IDF statistics. This analysis encourages use of the more reflective “System 2” for analytical thinking, rather than the fast and intuitive “System 1” when assessing claims of increasing severity and frequency.
PROLIFERATION OF RAINFALL GAUGES AND THE PROBABILITY OF MULTIPLE EXTREME EVENTS
Over
the past 40 years, municipalities in Ontario have invested heavily in
wastewater system monitoring and capacity to accommodate high extraneous flows
during wet weather. To support this
work, continuous, permanent monitoring of rainfall rates and wastewater flow
rates has proliferated across southern Ontario.
For example, the City of Toronto operated 17 rain gauges in 1982[12],
equivalent to one every 36 sq.km, similar to a 6 km grid. By 2013[13],
the City operated 35 gauges. In late 2024[14]
Toronto operated
Across the GTA, other municipalities and regions have also increased the
number of permanent rain gauges as well:
● York Region increased their number of gauges
from 10 before 2008, to 18 in 2008[15], and 44 in 2024. The
combined total number of Regional, lower-tier municipality and Conservation
Authority gauges in 2024 is 72.
Lower-tier municipality, the City of Markham, increased from limite
● Peel Region increased from 8 before 2013, 28 in 2013, and 30 in 2024[17]. Combined with others including the City of Mississauga, and Conservation Authorities, Peel Region has 64 rain gauges in 2024.
Combined, Toronto, Peel Region and York Region operate a total of 186
gauges in 2024, compared to half that number in earlier years pre-2008. A total
of 34 Halton Region gauges were in operation during the August 2014 Burlington
severe storm, according to Conservation Halton open data sources[18].
This brings the total number of gauges
TABLE 1 summarizes the number of gauges across Toronto, Peel and York reporting over 86.3 mm of rainfall, equivalent to the York Region 24-hour 100-year rainfall totals, during three 2024 events.
TABLE 1 – RAINFALL GAUGES WITH 100-YEAR
RAINFALL VOLUME DURING JUNE, JULY AND AUGUST 2024 SEVERE STORM EVENTS
|
Region/Municipality |
2024 Rainfall Event |
||
|
June 19 |
July 16 |
August 17 |
|
|
Toronto (50
gauges)[19] |
0 |
3 |
12 |
|
York (61 gauges) |
2 |
0 |
10 |
|
Peel (42 gauges) |
0 |
8 |
3 |
While 100-year rainfall totals were observed during each event, the
extent of extreme totals varied considerably across the events with only 2 of
153 gauges reporting over 86.3 mm on June 19th in York Region,
compared to 25 of 153 gauges on August 17th. It is suggested that the size of dependent
100-year gauge clusters is between 2 and 25 gauges. Accordingly, there may be a broad range of 7
to 93 independent samples among the
The insurance industry has suggested (argued?) that the frequency of extreme rainfall events has increased due to climate change effects[20]. However, many of these claims are based on theoretical as opposed to observed data (Muir, 2018). A 2019 City of Toronto Council Member Motion referenced the Insurance Bureau of Canada in stating that the GTA has had six 100-year storms from 2005 to 2018 and that there were “a direct result of climate” and sought to pursue legal option to recoup flood-related costs from GHG emitters.
The risk of multiple extreme events over a
given period can be estimated to assess whether six 100-year storms would be
rare (or not) over 12 years. Assumptions
must first be made about how statistically independent or dependent rain gauges
records are. This affects the effective
number of samples observed. Lower risks
would be estimated when most rain gauges are assumed to be independent, such
that in each year of observation each of the gauge clusters provides a
sample.
Using the online Stat Trek tool,[21]
probabilities of multiple 100-year events were calculated for various scenarios
over 12 to 21 years. Scenario 1 and 2 are based on six 100-year storms over 12
years (2005 to 2018) in the GTA, as quoted by the IBC[22]. A review indicates the reported storms
include both small areal extent (e.g., Mississauga Aug. 4, 2009 and Toronto
Aug. 7, 2018) and large areal extent storms (e.g., GTA Aug. 19 2005, July 8,
2013, and Aug. 4, 2014
TABLE 2 – PROBABILITY OF MULTIPLE STORMS
ACROSS THE GTA
|
Gauge
Cluster Size |
Number
of Independent Clusters (= 220 / cluster size) |
Scenario
1 a |
Scenario
2 b |
Scenario
3 c |
|
|
2 (Jun. 2024) |
110 |
99.1% d |
N/A |
N/A |
|
|
11 (Jul. 2024) |
20 |
3.5% (6 events) 9.5 % (5 GTA events) e |
22.1% f |
53.9% g |
|
|
25 (Aug. 2024) |
8.8 |
N/A |
10.1% h |
N/A |
|
Notes:
b Extent = Large; # Years = 12; # Events = 4 (2005, 2013, 2014, 2024)
c Extent = Small & Large; # Years = 21; # Events = 12 (2004x2. 2005x2, 2009, 2013, 2014, 2013, 2017x2, 2018, 2024)
The areal extent of the storm (small vs. large) critically influences overall probability. Multiple small-area storms, each recorded by only a couple of gauges, are much more likely than multiple small plus large-areal extent storms. And simply adding up small and large-area storms can distort the true probability of multiple extreme storms affecting the region at once. Observing six 100-year events over twelve years, including small extent storm, has a notably high probability of over 99% considering small gauge clusters in Scenario 1. Six moderate-size storms with cluster sizes of 11 gauges would have only a 3.5% chance of occurring – while this is very rare, the GTA did not experience six such events over the 2005-2018 period as one event occurred in Orangeville outside the GTA. The probability of five GTA events would be higher at almost 9.5%.
Considering large areal extent storms, the risk of 4 events such as those in 2005, 2013, 2014 and 2024 would be in the range of 10.1% to 22.1% over 12 years (see Scenario 2). When the areal extent and number of gauges increases, the probability decreases such as in Scenario 3. The number of gauges has been factored up to represent the larger area of southern Ontario in which twelve 100-year events have been reported over 21 years (2004-2024). Assuming an average gauge cluster size of 11 to represent both the small and large areal extent storms, the probability of 12 100-year events over 21 years is quite high at over 50%.
The probability should account for both event volume and its spatial coverage. As noted by Adams and Howard (1986) in their paper Design Storm Pathology:
Obviously, a natural hydrological event containing many characteristics cannot be fully described by statistics of only one, or at most a few, of the characteristics of the natural event.
The
challenge of characterizing the complexities of individual storms would support
the Toronto Water position to “utilize language that describes what can happen
to City infrastructure during a rainfall rather than describing the actual
rainstorm”.
To this end, this paper concludes by contrasting infrastructure responses compared to rainfall inputs. Yet, practitioners must make assumptions regarding the areal extent of design storms in hydrologic analyses and so later sections evaluate storm size and areal reduction curves considering 2024 events, local hydrology studies, and industry references that may implicitly account for the probability of storm areal extent.
SOUTHERN ONTARIO IDF TRENDS
In the 2021 NRC Flood Cost Benefit Guideline, IDF statistics from 1990 were compared with those from ECCC’s v3.1 dataset (up to 2017), where no overall increase in extreme rainfall intensities for 21 southern Ontario stations were found. TABLE 3 updates that analysis to 2021 (v3.3), where results show minor or insignificant increases or decreases across various durations and return periods.
TABLE 3 – AVERAGE
CHANGE IN SOUTHERN ONTARIO IDF - ECCC IDF TABLES PRE V.1 DATA (TO 1990) VS.
V3.30 DATA (TO 2021)
|
Duration |
Return
Period |
||||||
|
2-Year |
5-Year |
10-Year |
25-Year |
50-Year |
100-Year |
All |
|
|
5 min |
-2.1% |
-1.6% |
-1.6% |
-1.6% |
-1.4% |
-1.4% |
-1.6% |
|
10 min |
-0.1% |
0.0% |
0.0% |
0.2% |
0.2% |
0.2% |
0.1% |
|
15 min |
-0.2% |
0.1% |
0.3% |
0.5% |
0.6% |
0.7% |
0.3% |
|
30 min |
-0.1% |
0.3% |
0.5% |
0.6% |
0.7% |
0.8% |
0.5% |
|
1 hr |
0.0% |
0.4% |
0.5% |
0.6% |
0.8% |
0.8% |
0.5% |
|
2 hrs |
-1.3% |
-0.9% |
-0.8% |
-0.7% |
-0.5% |
-0.5% |
-0.8% |
|
6 hrs |
-1.5% |
-1.4% |
-1.5% |
-1.5% |
-1.5% |
-1.5% |
-1.5% |
|
12 hrs |
-1.1% |
-0.4% |
-0.2% |
0.0% |
0.2% |
0.3% |
-0.2% |
|
24 hrs |
-0.4% |
-0.3% |
-0.3% |
-0.3% |
-0.2% |
-0.2% |
-0.3% |
|
Avg. |
-0.8% |
-0.4% |
-0.3% |
-0.2% |
-0.1% |
-0.1% |
-0.3% |
Stations:
Sarnia Airport, Chatham WPCP, Delhi CS, Port Colborne, Ridgetown RCS, St Catharine’s
Airport, St. Thomas WPCP, Windsor Airport, Brantford MOE/Airport, Fergus Shand
Dam, Guelph Turfgrass CS, London CS, Mount Forest (Aut), Stratford WWTP,
Waterloo Wellington Airport, Bowmanville Mostert, Hamilton Airport, Hamilton
RBG CS, Oshawa WPCP, Toronto City, Toronto International Airport (Pearson).
Overall, these minor changes are neither statistically nor practically significant to design. The analysis considers 1026 station-years of data with an average record period of 49 years.
TORONTO-AREA IDF TRENDS
Toronto-area IDF trends can be further explored considering the effect
of the July 16 and August 17, 2024 storms.
ECCC’s v3.3 datasets (to 2021) for the long-term stations Toronto City
(i.e., “Bloor Street” ID 6158355) and Toronto Intl. Airport (i.e., Pearson
Airport, ID 6158731) indicate 100-year 24-hour rainfall volumes of 97.3 mm and
117.3 mm, respectively. Adding subsequent years maxima and the July 2024
rainfall of 83.6 mm at Toronto City[23] and
August 2024 rainfall of
TABLE 4 – TORONTO CITY AND PEARSON AIRPORT
100-YEAR 24-HOUR RAINFALL STATISTICS
|
Record Period |
Pearson
A. |
Ratio
to 1990 |
Toronto
City |
Ratio
to 1990 |
|
1990 (Pre v1) |
121.5 mm |
100.0 % |
95.9 mm |
100.0 % |
|
ECCC v3.3 (to 2017/2021) |
117.3 mm |
96.5 % |
97.3 mm |
101.5 % |
|
ECCC v3.3 extended to 2024
storms |
|
100.6 % |
98.4 mm |
102.6 % |
|
ECCC 1-30 Day IDF (1840-1910) |
N/A |
N/A |
102.9 mm |
107.3 % |
These findings suggest that previous, long-term Toronto rainfall intensities were higher than current values, but that the changes are insignificant relative to their underlying uncertainty. For example, the confidence intervals reported by ECCC for these 100-year statistics are typically large (Toronto City 95% confidence limits is +- 0.6 mm/hr, equivalent to +- 14.4 mm per day). This 28.8 mm confidence band is much wider than the changes in mean statistic values over time and suggests that designers should be as focused on core uncertainty in rainfall statistics as any changes in statistics (e.g., non-stationarity due to climate effects over time).
Toronto short-duration rainfall statistics above include 5-minute to 24-hour values and consider data as early as 1940 for Toronto City and 1950 for Pearson A. stations. ECCC’s 1–30-day IDF dataset extends significantly further back to 1840 and have been used in ECCC analyses (Pollock, 1974). This Toronto data has been analyzed to determine a long-term 100-year 24-hour rainfall statistic, using the same Gumbel extreme value distribution used for the short-duration datasets. FIGURE 2 illustrates this long-term series.
The statistic for the early Toronto City record of 1840 to 1910 is 102.9 mm, which is 4.6% higher than the v3.3 short duration dataset with 2022-2024 added (i.e., 1940-2024). This can be explained by high rainfall events in 1843, 1878, 1897 and 1905.
The changes in southern Ontario and Toronto-area statistics do not
support the many strong statements by political figures and the media that have
claimed increases in storm intensity.
Those statements are likely based on System 1 thinking and limited
analysis, and can demonstrate an availability bias with a lack of awareness of
many earlier extreme events noted previously. Long-term trends in southern
Ontario daily maximum rainfall have mixed results based on ECCC data[26].
FIGURE 2 - TORONTO CITY MAXIMUM 1-DAILY RAINFALL - COMBINED ECCC 1-30 DAY, v3.3 IDF DATASETS & 2022-2024 DAILY MAXIMA
TRENDS
IN ANTECEDENT MOISTURE CONDITIONS
Runoff and extraneous flow in wastewater systems are complex and depends not only on peak intensities but also on soil and system moisture before a storm. The Soil Conservation Service (SCS) Curve Number (CN) Method accounts for this by using the total rainfall of the preceding five days (the Antecedent Moisture Condition, or AMC). According to this approach:
·
AMC I (dry): Less than 35 mm in
the previous five days.
·
AMC II (normal): Between 35 and
53 mm.
· AMC III (wet): Over 53 mm.
In Ontario, ‘return-period’ convective storms usually assume AMC II conditions, while historical storms like Hurricane Hazel use AMC III because of high preceding rainfall.
The ECCC 1-30 Day IDF annual series may also be used to assess trends in multi-day rainfall in southern Ontario. FIGURE 3 illustrates a 5-day maximum rainfall trend for Toronto from 1840 to 2002. It is noted that recent years’ data up to 2016 have been omitted due to incomplete/partial ECCC data[27].
FIGURE 3 – TORONTO CITY 5-DAY MAXIMUM
RAINFALL (ECCC 1-30 DAY IDF DATASET 1840-2002)
The data shows a slight decrease in 5-day rainfall over 162-years, with some of the highest totals in the 1800’s (1841, 1843, 1878, 1894 and 1897).
Outside of Toronto, records for Woodstock, Welland, and Belleville indicate slight increases in multi-day totals. For example, Welland’s 10-day rainfall has increased by 0.16 mm per year since 1872. York Region flow monitoring shows that the 10-day antecedent rainfall can affect Rainfall Derived Inflow and Infiltration (RDII) in the wastewater system. Over a 20-year planning horizon, a Welland-type increase of 3.2 mm in 10-day rainfall (i.e., from 100.3 mm today to 103.5 mm in 2045) could raise RDII volumes by about 3%[28]. In regions with past and projected increases this factor could be considered in wastewater system planning, including as part of sensitivity analysis on system response.
In practice, the key take-away for hydrologic modelers and infrastructure designers for surface water systems is that absolute AMC values, rather than minor trends, matter most. FIGURE 3 illustrates that the majority of years experienced AMC III (wet) conditions, not AMC II (normal) conditions applied in return-period storm hydrologic modelling. Given the more prevalent ‘wet’ AMC conditions, higher CN values and more conservative runoff potential could be considered in return-period event simulations, such as for the 100-year storm.
SPATIAL
VARIABILITY OF EXTREME RAINFALL
The July 16th, 2024 storm in Markham demonstrated notable spatial variability. TABLE 5 (adapted from AECOM’s York Region summary) illustrates how rainfall intensity return periods varied across 18 Markham gauges for various durations.
TABLE 5 – MARKHAM RAIN GAUGE RETURN PERIOD
VARIABILITY ON JULY 16, 2024
|
Site
ID |
Rainfall Intensity by Duration |
|
||||||||
|
5
min |
10
min |
15
min |
30
min |
1
hr |
2
hrs |
6
hrs |
12
hrs |
24
hrs |
Total Rain |
|
|
R-MUN-MA-05 |
30 |
30 |
25 |
19.5 |
17.5 |
13.5 |
4.96 |
2.5 |
1.25 |
30 mm |
|
R-MUN-MA-07 |
36 |
32.4 |
28.8 |
23.6 |
18.2 |
14 |
5.03 |
2.53 |
1.27 |
30.4 mm |
|
R-MUN-MA-08 |
69 |
55.5 |
47 |
46.5 |
33.5 |
23.1 |
9.08 |
4.56 |
2.28 |
54.75 mm |
|
R-MUN-MA-09 |
40.8 |
28.8 |
24.8 |
20.8 |
15 |
10.1 |
3.5 |
1.75 |
0.88 |
21 mm |
|
R-MUN-MA-11 |
96 |
88.8 |
71.2 |
54.8 |
35.4 |
22.1 |
8.3 |
4.18 |
2.09 |
50.2 mm |
|
R-MUN-MA-12 |
60 |
60 |
53.6 |
37.6 |
30.2 |
17.3 |
6.3 |
3.23 |
1.62 |
38.8 mm |
|
R-MUN-MA-13 |
98.4 |
72 |
63.2 |
54.8 |
39.8 |
23 |
8.2 |
4.1 |
2.05 |
49.2 mm |
|
R-MUN-MA-14 |
111 |
102 |
88 |
64.5 |
42.8 |
25.12 |
8.83 |
4.42 |
2.21 |
53 mm |
|
R-MUN-MA-15 |
63 |
55.5 |
49 |
45 |
32.8 |
21.75 |
7.88 |
3.96 |
1.98 |
47.5 mm |
|
R-MUN-MA-16 |
66 |
63 |
56 |
40 |
29 |
17.75 |
6.21 |
3.1 |
1.55 |
37.25 mm |
|
R-MUN-MA-18 |
26.4 |
24 |
19.2 |
15.6 |
12.8 |
8.6 |
3.03 |
1.53 |
0.77 |
18.4 mm |
|
90 |
85.5 |
82 |
73 |
53.8 |
33.38 |
12.8 |
6.44 |
3.22 |
77.25 mm |
|
|
R-MUN-MA-22 |
21.6 |
14.4 |
14.4 |
14 |
11.6 |
7.9 |
3.17 |
1.63 |
0.82 |
19.6 mm |
|
R-TR-MA-06 |
105.6 |
90 |
80.8 |
70.8 |
44.6 |
25.4 |
8.97 |
4.5 |
2.26 |
54.2 mm |
|
R-YR-MA-01 |
69.6 |
57.6 |
48 |
46.4 |
33.2 |
23.4 |
8.93 |
4.53 |
2.27 |
54.4 mm |
|
R-YR-MA-03 |
64.8 |
56.4 |
52.8 |
44.4 |
33.4 |
22.1 |
8.6 |
4.37 |
2.18 |
52.4 mm |
|
R-YR-MA-19 |
86.4 |
79.2 |
68 |
54.4 |
36.8 |
23.6 |
8.57 |
4.33 |
2.17 |
52 mm |
|
R-YR-MA-20 |
100.8 |
81.6 |
70.4 |
59.6 |
40.6 |
23.4 |
8.37 |
4.3 |
2.15 |
51.6 mm |
|
Return Period |
Buttonville Airport IDF Intensities |
|
||||||||
|
2 Year Event |
106.2 |
74.5 |
60.2 |
36.7 |
21.2 |
11.8 |
5.5 |
3.2 |
1.8 |
|
|
5 Year Event |
136.4 |
95.6 |
77.8 |
51.2 |
31 |
16.6 |
7.2 |
4.1 |
2.3 |
|
|
10 Year Event |
156.4 |
109.7 |
89.5 |
60.8 |
37.5 |
19.8 |
8.4 |
4.8 |
2.6 |
|
|
25 Year Event |
181.7 |
127.4 |
104.3 |
72.9 |
45.6 |
23.7 |
9.8 |
5.6 |
3 |
|
|
50 Year Event |
200.4 |
140.5 |
115.2 |
81.9 |
51.7 |
26.7 |
10.9 |
6.2 |
3.3 |
|
|
100 Year Event |
219 |
153.6 |
126.1 |
90.8 |
57.7 |
29.6 |
12 |
6.7 |
3.6 |
|
|
Return Period |
Number of Exceedances |
% of all stats |
||||||||
|
2 Year Event |
0 |
1 |
5 |
6 |
2 |
2 |
2 |
2 |
10 |
19% |
|
5 Year Event |
0 |
0 |
2 |
4 |
6 |
2 |
4 |
9 |
0 |
17% |
|
10 Year Event |
0 |
0 |
0 |
2 |
4 |
8 |
6 |
0 |
0 |
12% |
|
25 Year Event |
0 |
0 |
0 |
1 |
0 |
2 |
0 |
0 |
1 |
2.5% |
|
50 Year Event |
0 |
0 |
0 |
0 |
1 |
0 |
0 |
1 |
0 |
1.2% |
|
100 Year Event |
0 |
0 |
0 |
0 |
0 |
1 |
1 |
0 |
0 |
1.2% |
|
|
|
|
|
|
|
|
|
|
|
|
|
Key: |
|
more
prevalent return period for rainfall duration |
||||||||
FIGURE 5 maps the spatial variability of total rainfall volumes and categorizes gauge intensities based on their highest return period.
FIGURE 5 – MARKHAM RAIN GAUGE MAXIMUM INTENSITY RETURN PERIOD AND TOTAL EVENT VOLUME JULY 17, 2024
Although this event was described as “100-year”, only 1.2% of all gauge-duration statistics exceeded the 100-year threshold. Specifically, only the 2 and 6-hour intensities at R-MUN-MA-21 exceeded the Buttonville Airport’s 100-year IDF statistics. Most prevalent return periods ranged from 2 to 10 years. Shorter-duration intensities (less than 1 hour), which govern local storm sewer performance, generally matched 2 to 25-year return period intensities. The response of the wastewater collection system to this and other events is described in a subsequent section, comparing peak flow return periods to average and maximum rainfall return periods.
TEMPORAL
VARIABILITY OF EXTREME RAINFALL FOR DIFFERENT STORM TYPES
While the previous section illustrates spatial variability over an entire storm, this section focuses on temporal changes in intensity. FIGURE 6 compares two events in Markham: a convective storm on August 17th, 2024 and remnants of Hurricane Beryl on July 10th, 2024. The chart on the right shows highly variable accumulations across gauges for the convective event (coefficient of variation, or COV, of 0.41), while the chart on the left shows a more consistent accumulation for the hurricane remnant (COV of 0.11). The temporal lags between the highest intensities, suggest movement of highest intensity cells across the city.
FIGURE 6 – TEMPORAL VARIABILITY OF CONVECTIVE AND HURRICANE STORM TYPES IN MARKHAM IN 2024
Looking more broadly across York Region, the August convective storm generally displayed the highest COV, averaging 0.51 across 69 gauges, whereas the July hurricane remnant was more uniform (average COV of 0.18). A “spring” event on April 2, 2024 had intermediate variability (average COV of 0.30).
When IDF statistics are derived from Annual Maximum Series (AMS), the type of storm is not considered. That is, rainfall data resulting from convective storms (e.g., August 2005 and July 2013 Toronto events) is combined with rainfall data resulting from hurricane-type events (e.g., October 1954). As these events have different origins and patterns, separate series and statistics could be considered. Others have noted the importance of considering seasonal extreme rainfall statistics that may be explained by the variable storm types and seasonal factors. For instance, Dickinson (1976) noted “The implications of this seasonal variability are significant for the estimation of flood peaks and their frequency of occurrence.” Future research could assess the joint probability of factors like seasonal intensity variations and AMC conditions, particularly if rising AMC (wet) intervals align with higher storm intensities.
AREAL
REDUCTION FOR 2024 CONVECTIVE STORMS
To
characterize the areal extent of 2024 convective storms, rainfall event volumes
were compiled from York Region, Peel Region and Toronto gauges. Peel Region data includes data for gauges
operated by the Region and the City of Mississauga. York Region data includes
Region-, municipality- and Conservation Authority-operated gauge data.
FIGURE 7A – JUNE 19, 2024 GTA RAINFALL
DISTRIBUTION
FIGURE 7B – JULY 16, 2024 GTA RAINFALL DISTRIBUTION
FIGURE 7C – AUGUST 17, 2024 GTA RAINFALL DISTRIBUTION
Areal reduction curves were then derived for circular areas around each peak gauge: 96.8 mm at Markham R-MUN-MA-16 (June 19th), 104.2 mm at Peel Region RG 07 (July 16th), and 128 mm at Toronto RG-046[29] (August 17th). FIGURE 8 shows how average rainfall drops relative to the peak gauge volume as the area expands.
FIGURE 8 illustrates a wide variation in areal reduction around the peak gauge, consistent with the 100-year gauge counts noted earlier. The June event (only two gauges over 100-year totals) had a steep drop-off in mean areal rain, whereas the more widespread August event maintained higher volumes over a larger area. Mean rainfall for selected TRCA and CVC watersheds is also shown, illustrating how watershed size can affect average totals. Despite that trend the mean Mimico Creek July rainfall was high relative to the maximum rainfall in that watershed, despite its moderate watershed size.
FIGURE 8 – AREAL REDUCTION FOR THREE 2024
GTA EXTREME RAINFALL EVENTS COMPARED TO DESIGN GUIDANCE
Design areal reduction factors have been added FIGURE 8 including the 1-to-24-hour ranges reported by the WMO and by NWS/NOAA in the US mid-west (Huff and Angel, 1992). The WMO values are one source of factors for return period storms recommend by OMNR (2002). While OMNR prescribes factors for historical storms (Hurricane Hazel and Timmins Storm) based on observed characteristics, on thunderstorms in small areas in notes data limitations:
“When analyzing thunderstorm type rainfall in urban areas, storm distribution becomes an important factor. Variation in rainfall intensity over small areas, such as 5 km2, can be significant. Unfortunately, due to lack of rainfall data, no guidelines are available on storm distribution within small urban areas.”
The WMO factors have been applied in the GTA such as the Don River Hydrology Update (AECOM, 2018) applying the 1-hour reduction factor to the 12-hour design storm. The August event factor observed around the Toronto RG-046 gauge, and July event Mimico Creek factor would appear to exceed the Don Watershed study design values. The Mimico Creek July factor aligns with the WMO 24-hour factor, suggesting that high conservative factors can be realized in real storms. As the Huff and Angel factors are based on over 600 observed storms and can exceed the long duration WMO factors cited by OMNR, practitioners should consider the possibility of large area extreme storms. The 2024 storm observations confirm that high rainfall volumes may be sustained over very small study areas (e.g., local wastewater collection sewersheds or small storm drainage areas). For example, the Carolyn Creek July storm factor is 0.95 over an equivalent circular area of 13 sq.km[30]. Larger wastewater collection areas, e.g., that may be 250 sq.km in size[31], may have limited reduction factors over long durations based on the upper bound of the WMO and Huff-Angel 24-hour curves.
OBSERVED
FLOW RATE RETURN PERIOD COMPARED TO AVERAGE AND PEAK RAINFALL RETURN PERIOD
Peak flows in Markham’s wastewater system were compared to design storm outputs from the City’s calibrated InfoWorks model. Four monitoring sites[32] are near rain gauge MA-016 (June 19, 2024 event), and two sites[33] are near gauges MA-01 and MA-07 (August 17, 2024 event). These older Unionville and West Thornhill communities exhibited both high total volumes and periods of high intensity.
The return period of wastewater flow peaks at the four sites on June 19, 2024 was typically in the 5 to 10-year range, with one site recording less than a 2-year peak flow. Even though the 2 and 24-hour intensities at MA-016 exceeded 100-year return periods, shorter durations (2 and 25-year return periods) actually governed the peak flows. So, while the storm event exhibited some 100-year characteristics, the critical characteristics that govern wastewater system response and performance risk were not nearly as unlikely.
Similarly, the return period of flows for the August 17, 2024 event were in the 2 to 10-year range. While the MA-01 2 hour and MA-07 6-hour intensities both exceeded 100-year values, the shorter duration intensities (2 and 25-year return periods) were less severe and aligned with observed peak flows.
This discrepancy shows that long-duration metrics (i.e., total event volumes) do not always correlate with wastewater system flow peaks. In many cases, short-duration intensities are a more accurate indicator of potential infrastructure stress. Yet, storm mapping and news reports often highlight total volumes instead.
CONCLUSIONS
Claims of changes in extreme storm severity and frequency are often inconsistent with observed and reported trends documented in ECCC’s Engineering Climate Datasets. For instance, v3.3 IDF files for southern Ontario do not show any overall increase in extreme rainfall intensities compared to pre-1990 values. Even when considering data up to the 2024 extreme events, the Pearson Airport’s 24-hour, 100-year rainfall volume is only 0.6% higher than in 1990, and Toronto City’s 2.6% increase still remains 0.1% below the long-term value, including an additional 100 years of data from 1840 to 1940.
The probability of recording multiple extreme storms over multiple years in the GTA has increased with the growing number of municipal rainfall gauges. With 220 gauges, observing six 100-year events in 12 years becomes statistically probably (over 99%) if only one gauge in the network is affected by a small-sized storm cell. However, observing four moderate-sized storms over that same period has a lower probability (around 22.1%) considering 11 affected gauges in each independent cluster. This is consistent with ECCC’s reporting following the extreme May 16, 1974 storm, that a 4.2 inch or greater rainfall could be observed every two years in southern Ontario. The areal extent of storms should therefore be considered an additional factor to assess the probability of events affecting large systems, which may be combined with the isolated peak volumes often used to characterize events and could be reflected in locally-derived areal-reduction factors for convective storms in urban areas. An assessment of 2024 extreme rainfall patterns suggests that areal reduction factors applied in practice can sometimes underestimate both these observed and literature reference values.
In several Markham catchments, wastewater system peak flows showed lower design return periods than event’s volume return periods (i.e., long-duration intensities). These peaks aligned more closely with short-duration intensities, rather than 24-hour volumes. This suggests that the severity of storms should be more critically addressed for different systems considering more than the event volume (e.g., governing short duration intensities).
Extreme storms in the GTA in 2024 included both convective events, featuring high spatial and temporal variability, and remnants of tropical storms (hurricanes) with less variability. These events have independent causes yet are combined in single annual series and IDF frequency analyses implying identical distributions, which is unlikely. Modified IDF analysis excluding hurricane-type events from annual series should be evaluated. This could assess how high daily hurricane-derived rainfall volumes affecting short duration design intensities when daily IDF volumes are used derive short duration intensities in traditional design hyetographs.
As desirable as simplifications are, design storms inherently simplify complex hydrologic processes, especially when dealing with varying antecedent moisture conditions or mixed storm populations. The lack of dedicated “urban hydrometric networks” historically justified relying on design storms that can be somewhat arbitrary (Marsalek and Watt). However, some jurisdictions, such as the City of Ottawa, have begun using frequency analyses of wastewater system RDII flows, reducing the need for design storms assumptions. This approach deserves continued consideration to overcome what could be an intractable challenge of defining the 100-year design storm.
PRESENTATION
BIBLIOGRAPHY
Adams, B.J., Howard, C.D.D., (1986) Design Storm Pathology, Canadian Water Resources Journal, https://doi.org/10.4296/cwrj1103049
AECOM, (2018), Don River Hydrology Update, Toronto and Region Conservation Authority.
City of Ottawa (2008), Sanitary Sewer Extraneous Flow Analysis https://docs.google.com/document/d/0B9bXiDM6h5VianROT1EtV2c5UFU/edit?usp=sharing&ouid=115169455109461543967&resourcekey=0-sIt--IZF8kevusIkk76iMg&rtpof=true&sd=true
Dickinson, T. (1976), Season variability of rainfall extremes, Atmosphere, Volume 14, Number 4, https://doi.org/10.1080/00046973.1976.9648424
Huff, F.A., Angel, J.R. (1992) Rainfall Frequency Atlas of the Midwest, Bulletin 71, Midwestern Climate Centre (CAA, NWS, NOAA), Illinous State Water Survey. https://drive.google.com/file/d/1-jY16dLUu1Pi_9SDWd_kp8Sof3HMehSs/view?usp=sharing
Marsalek, J, W.E. Watt (1983), Environment Canada, National Water Research Institute, Design Storms for Urban Drainage Design. https://drive.google.com/file/d/1N4xc7qJAJv0ZhY24KDVOcSrLV0NbMn72/view?usp=sharing
Muir, R., (2018) Evidence Based Policy Gaps in Water Resources: Thinking Fast and Slow on Floods and Flow, Journal of Water Management Modeling, https://www.chijournal.org/C449
Ontario Ministry of Natural Resources, (2002), Technical Guide River & Stream Systems: Flooding Hazard Limit
Pollock, D.M, (1976), Environment Canada, Atmospheric Environment, The Rainstorm of May 16, 1974, In Southern Ontario. https://drive.google.com/file/d/1MRXf36wCUtA6ynzMl-3UMot4tNzVDM5T/view?usp=sharing
National Research Council of Canada (NRC) (2021) Guidelines on Undertaking a Comprehensive Analysis of Benefits, Costs and Uncertainties of Storm Drainage Infrastructure and Flood Control Infrastructure in a Changing Climate.
[8] Keating Channel Flood Inquiry
Report (Lorant, 1981)
https://www.torontopubliclibrary.ca/detail.jsp?Entt=RDM209381&R=209381
[11] City of Toronto Access Request Number FOI-2024-03265
[12] Toronto Open Data / Historic
Rain Gauge Locations and Precipitation
https://open.toronto.ca/dataset/historic-rain-gauge-locations-and-precipitation/
[13] Toronto Open Data / Rain Gauge
Locations and Water Collected (July 8, 2013 storm)
https://open.toronto.ca/dataset/rain-gauge-locations-and-water-collected/
[14] Toronto Open Data / Rain Gauge
Locations and Precipitation
https://open.toronto.ca/dataset/rain-gauge-locations-and-precipitation/
[15] York Regional and Municipal
Inflow & Infiltration Assessment Reduction Assessment, Region of York, Rain
Gauges Installation Report (AECOM, 2008)
[16] R. Muir, City of Markham,
personal communication December, 2024
[17] M. Faye, Region of Peel,
personal communication October, 2024
[18] Conservation Halton Open Data
/ Rainfall Rain Gauge 20140804
https://conservationhalton-camaps.opendata.arcgis.com/datasets/d17cc3ff48a940549a8f19c399e0939f_0/explore?location=43.449139%2C-79.868000%2C10.69
[19] Note that Toronto rainfall
data reflects reported 3-hour total. Additional gauges may observe over 86.3 mm
for longer durations.
[20] Insurance Bureau of Canada,
Telling the Weather Story (2012),
https://www.iclr.org/wp-content/uploads/PDFS/telling-the-weather-story.pdf
[22] August 4, 2009 Mississauga even had extreme rainfall for only one gauge, Station 6, per Cooksville Creek - August 2009 Flooding (EWRG, 2010) https://www.mississauga.ca/file/COM/Cooksville_Creek_Flooding_v03_Draft.pdf. August 7, 2018 Toronto extreme rainfall was recorded for only one Toronto gauge per City of Toronto Open Data https://open.toronto.ca/dataset/rain-gauge-locations-and-precipitation/
[24] August 17, 2024 per
rainfallhttps://www.theweathernetwork.com/en/news/weather/severe/when-it-rains-it-pours-toronto-records-its-wettest-summer-on-record-ontario.
Toronto City 2018-2021 rainfall added to series plus 2022 and 2023 trendline
estimates.
[26] Welland, Belleville and daily
maximum rainfall has increased by 2.7 mm, 2.2 mm and 6.9 mm per century based
on data back to 1873, 1866 and 1870, respectively. Peterborough daily maximum
has decreased by 5.9 mm per century based on data back to 1866.
[27] Incomplete/partial year data
in the 1-30 Day dataset was replaced with short-duration IDF data for the
previous 100-year statistical analysis.
[28] York Region RDII antecedent
rainfall chart indicates approximate linear increase in RDII volume with
antecedent rainfall totals grouped for total below 12.5 mm, 12.5-38.1mm and
above 38.1 mm with a strong correlation between RDII volume storm volume for
these groups of antecedent rainfall (i.e., reported R-squared values of 0.904,
0.8866 and 0.9622 respectively).
[29] Note – highest rainfall was
135 mm at Toronto RG-052, however RG-046 is situated more centrally within a
higher rainfall area.
[30] Actual watershed area is 5.3
sq.km. Maximum rainfall in watershed was 102 mm while mean watershed rainfall
was 97 mm.
[31] City of Toronto Ashbridge’s
Bay Wastewater Treatment Plant approximate sewershed size is 25,000 ha (2021
Annual Report, www.toronto.ca)
[32] Monitor ID’s MA006a, MA006b_10,
MA006c_2 and MA083_10
[33] Monitor ID’s MA044_10 and MA046_10
