Do We have Enough Climate Stations in Canada To Track Trends in Extreme Rainfall?

Some have suggested that we have lost so many climate stations due to cut backs in the 1990's that we can't accurately detect trends in extreme rainfall.  But many are confusing manual climate stations with the stations that collect rainfall intensity data, often automatically.  The number of stations measuring extreme rainfall has been increasing since 1990.

Declining number of stations was noted in the ECO's report 2018 GREENHOUSE GAS PROGRESS REPORT CLIMATE ACTION IN ONTARIO: WHAT'S NEXT? - (see Appendix D)

CBC New has also referred to this concept in responding to a complaint to the CBC Ombudsman regarding accuracy in reporting on extreme weather trends.  What has been cited as evidence of that decline is the chart in Appendix D in the ECO report above. CBC's Director of Journalistic Standards Paul Hambleton wrote:

"The report suggests several possible reasons for this inconsistency, including issues with data collection: There simply are not enough rain gauges. Rainfall data is collected using rain gauge buckets that can record both amount and intensity of rainfall. After a series of federal budget cuts in the 1990s, there are fewer rain gauge stations across the country than there were 60 years ago."

Fewer rain gauge stations? Or fewer "manual" rain gauge stations?  Yes there is a difference.

What does that chart show?  It summarizes declining manual stations in Canada and is a excerpt from the paper in Atmosphere-ocean An Overview of Surface-Based Precipitation Observations at Environment and Climate Change Canada (Mekis et al., 2018) -

The chart of manual station count in Canada is Figure 2a in the paper on the left below.
Number of Manual Climate Stations in Canada

This chart has been referred to in discussions on extreme rainfall trends.  For example, in the ECO report this chart has been related to intensity-duration-frequency of isolated localized storms as in the excerpt at right:

Readers of this blog will have seen extensive analysis of the trends in extreme rainfall across Canada, including annual maximum series and intensity-duration-frequency (IDF) trends.  The data used is that of Environment and Climate Change Canada, distributed in the Engineering Climate Datasets.

What do Engineering Climate Datasets show us in terms of number of stations that collect and analyze extreme rainfall and IDF trends - they have been increasing!  And the number of station-years of data has been increasing - that means more long-term data to support more reliable statistical analysis.  Good news. The following table summarizes the trends:

Rainfall Intensity Data in Canada
Number of Climate Stations in Canada With Rainfall Intensity Analysis

The newer datasets include more stations, a 22% increase in station count since 1990. And the number of station-years has increased by 48% since 1990 - that's almost 50% more data to analyze and derive IDF design curves since I graduated and started working in this field.

How have the number of stations with extreme rainfall analysis, increasing since 1990, compared to the number of manual stations decreasing since 1990? See chart below:

Climate stations in Canada - trends, count, type
Number of Climate Stations in Canada - Manual and Intensity-Duration-Frequency Stations.  Manual stations decreasing while IDF stations and number of station-years of data increasing. (note: v2.00 (557 stations) and v3.00 (596 stations) not shown on chart)

The Mekis et al. figure is shown in blue and the IDF station trends in orange. Obviously the decline in manual stations does not relate at all to the trends in IDF stations.  As noted in other blog posts, municipal IDF stations have also proliferated over past decades, complementing the IDF stations charted above.

So when CBC's Paul Hambleton writes: "After a series of federal budget cuts in the 1990s, there are fewer rain gauge stations across the country than there were 60 years ago" he missed an important detail - yes manual stations that are expensive to operate have declined, as we expect.  It makes sense that we have fewer manual climate stations since 1990. 
Technology changes.  A good summary of the changes in equipment is described by Mekis et al. - image above are from the website that describes the history of rain gauges and their evolution.

But what about automated weather stations? And what about the number of stations used to collect extreme rainfall information and rainfall intensities? Has the number of stations that define extreme rainfall decreased since 1990? No.

IDF stations have increased from 532 to 651 stations since 1990, many with longer periods of record - we have more extreme weather data to rely on today!  The CBC and others should clearly be more careful when interpreting data on climate station and extreme rainfall  monitoring.  

Yes, we're getting more extreme rainfall, and it's due to climate change, study confirms .. well not so fast

CBC News has a new report "Yes, we're getting more extreme rainfall, and it's due to climate change, study confirms"

The byline is "Federal scientists predict more frequent and severe rainfall in future", referring to this research paper Human influence has intensified extreme precipitation in North America by Megan C. Kirchmeier-Young and Xuebin Zhang

The research paper refers to "heavy rainfall", i.e., Kirchmeier-Young the lead author and research scientist at Environment and Climate Change Canada stated "We're finding that in North America, we have seen an increase in the frequency and severity of heavy rainfall events."

Kirchmeier-Young also refers to "extreme rainfall" and makes a connection to urban flooding in the CBC article:

"And as we continue to see warming, we will continue to see increases in the frequency and severity of extreme rainfall," Kirchmeier-Young said. "And heavy rainfall is one of the major factors in flash flooding, particularly in urban areas."

The CBC relates extreme weather to rising insured flood damage trends in Canada since the early 1980's.

Let's review:

1) What 'heavy rainfall' events were reviewed in Kirchmeier-Young's research paper?

2) Is 'heavy rainfall' for a climate researcher the same as 'extreme rainfall' for an engineer?

3) Do 'heavy rainfall' and precipitation trends follow 'extreme rainfall' trends used in engineering design?

4) Do 'heavy rainfall' events studied in the research paper cause damaging flood events, and flash flooding 'particularly in urban areas?

5) What are the trends in 'extreme rainfall' in Environment and Climate Change Canada's Engineering Climate Datasets, the data used by engineers to analyze and design infrastructure to manage flash flooding risks in urban areas?

6) What does Kirchmeier-Young's research paper reveal about previous extreme rainfall and flooding events in Canada - has climate change increased runoff that could aggravate flood damages?

1) What 'heavy rainfall' events were reviewed in Kirchmeier-Young's research paper?

The research paper abstract indicates "Here, we address the question of whether observed changes in annual maximum 1- and 5-d precipitation can be attributed to human influence on the climate."

What does "1- and 5-d precipitation" mean?  This is the amount of rainfall over one to five days, so 24 to 120 hours. While precipitation can include snowfall too, the focus is on rain.

Note, the research paper actually refers to 'heavy precipitation' and not 'heavy rainfall'.

The authors have confirmed that short-duration rainfall was not reviewed, only annual maximum daily rain.

2) Is 'heavy rainfall' for a climate researcher the same as 'extreme rainfall' for an engineer?


The research paper states:

"We focus on the annual maxima of 1-d (Rx1day) and 5-d (Rx5day) rainfall. Rx1day is important for flash floods as well as infrastructure design. Rx5day is relevant to large-scale river flooding."

A training session on the use of rainfall intensity design curves from a climate scientist (link: indicates that shorter times influence flooding (underline and all-caps emphasis are in the original material, not added here):

 "An urban centre could experience flooding from heavy rains falling over a SHORT period of time, such as A 5 TO 30 MINUTE PERIOD."

"• A rural highway with deep ditches on its shoulders would not likely be impacted by an intense rainfall lasting only 5 to 15 minutes, although the paved road itself would see ponding of water.
• A heavy rainfall event lasting 1 to 6 hours might be more significant for filling the ditches and overflowing the roadway."

So short durations are important for flooding.

From the insurance industry perspective, an Institute for Catastrophic Loss Reduction paper in Journal of Flood Risk Management notes the importance of short-duration rainfall (link: states

"Subdivisions built before the 1970s are less likely to be serviced by major systems (Watt et al., 2003), and are thus more vulnerable to overland flooding from extreme short-duration rainfall events."

A civil engineer will tell you that rarely that the a 1-day rainfall is not 'important for flash floods'.  Why? Because urban flooding is caused by short-duration rainfall.  Designers of storage facilities such as stormwater manage ponds may consider design rainfall events up to 24 hours.

In the Canadian Water Resources Journal, authors of Flood processes in Canada: Regional and special aspects (link: representing six universities across Canada, INRS-ete, and Environment Canada review "key processes that generate floods in Canada":

"Similarly, floods can be generated across most of the country by rainstorms with large depths and/
or intensities (Figure 1). Thus, convective and frontal systems can generate large short-duration rainfall intensities (Alila 2000) which can occur in all regions (Table 1). Nevertheless, the significance of such storms to flood generation varies across the country, with the greatest
depths and intensities for short-duration events in southern parts of Canada and the smallest in the Arctic. These short-duration events are often responsible for flood generation in relatively small drainage basins, given the greater chance of high-intensity rainfall occurring over
the entire basin (Watt et al. 1989)."

"Short-duration events are often responsible for flood generation".

"Small drainage basins" is equivalent to urban drainage systems. In the municipality where I was Manager, Stormwater our storm sewer drainage systems averaged just over 50 hectares in size.  Urban drainage systems that are 'flashy', responding quickly to rainfall running off hard surfaces, are characterized in engineering design by a 'time of concentration' that is the response time of the drainage area, and which is used to determine the extreme rainfall durations relevant to infrastructure design.  It is never 24 hours or one day.  Typical times of concentration are measured in minutes and up to hours.

The Ontario Ministry of Transportation describes the design rain storms that may be used to analyze rural and urban areas, including the duration of the storm (link:

Storms of duration up to 24 hours are applicable to rural land uses.  Storms of up to 4 hours (including flashy Chicago hyetograph temporal distributions) are applicable to urban areas.  The SWMM Knowledge Base, a discussion forum for the standard U.S.EPA Stormwater Management Model and other modelling platforms, provides insight into what storm durations practicing civil engineering / urban system modelling professionals use.  In the discussion thread "Design storm duration" (link: a duration of 24 hours is deemed by one practitioner to be 'ultra conservative' ("ultra conservative choice of a 24-hour storm but it hardly can be justified when no detention storage is involved"), another states that in small urban systems the 5-minute rainfall governs peak flows ("a small (25 acre) urban, very impervious, drainage area was that the peaks were almost the same no matter the duration, and that they were driven by the peak 5-minute rainfall"), and Ben Urbonas, Ben Urbonas,
President of Urban Watersheds Research Institute and Owner, Urban Watersheds, LLC (LinkedIn:  notes the use of durations of 2-6 hours ("All of our design storms are front loaded intensity types and range from 2-hour to 6-hour durations depending on watershed area.").

Marsalek and Watt's paper Design storms for urban drainage design in the Canadian Journal of Civil Engineering shows design storm durations of often 1 hour duration, sometimes up to 6 hours (US Soil Conservation Service (SCS) for rural areas, as highlighted in their Figure 1.

Marsalek and Watt tablulate design storms with duration and categorize the use of the storms for different hydrological studies, including urban/sewer design and other applications, such as the study of large rural basins. Table 2 from their urban drainage review shows durations of up to 1- 12 hours for Canada's Atmospheric Environment Service's (AES) storm, and 1, 3 and 4 hour storms for sewer sizing in other jurisdictions (see highlights below).

Practitioners in Ontario, Canada will know that longer duration storms are considered for large regional wastewater systems that have a slow response to long-high volume storm events.  These govern large trunk sewer system performance, but not local sewer system performance that is dominated by short duration rainfall.  Even small wastewater system trunks may be governed by short duration rainfall intensities where there are direct inflows, which is common for many flood prone systems.  Analysis of trunk system response in the Kitchener-Waterloo Region showed wastewater trunk peaks flows for most-highly correlated to the 5-minutes rainfall intensities in on Master Plan study (i.e., more than longer durations).

Rivard's paper in the Journal of Water Management Modelling entitled Design Storm Events for Urban Drainage Based on Historical Rainfall Data: a Conceptual Framework for a Logical Approach (link: summarize early work on characterizing storms in Canada and in the highlighted excerpt notes that 1- 12 hour durations represented convective (thunderstorm) and synoptic scale events. See the highlight to the right.

Rivard also summarized what storm durations are of interest for urban design graphically as follows:

So in small to medium basin, up to a 3 hour duration is critical, and for a very large urban basin, up to 6 hours.  Twenty four hour durations and longer are critical to large rural basins.

The statement "Rx1day is important for flash floods as well as infrastructure design." is therefore inconsistent with professional engineering practice in Canada.

Environment and Climate Change Canada publishes Engineering Climate Datasets including Intensity-Duration-Frequency statistics describing rainfall, both common, moderate and extreme, used from infrastructure design.  The durations analyzed are from 5-minutes to 24-hours.

So again, no, 'heavy rainfall' in a climate research paper is not the same as 'extreme rainfall' an engineer uses for infrastructure analysis and design. Rainfall over 1-5 days periods is not the same as extreme rainfall over minutes to hours used to design conveyance systems in urban areas - those 'flashy' systems with short 'time of concentration' characteristics.

The statement in the research paper "Rx1day is important for flash floods as well as infrastructure design." is questionable.  One-day rainfall is way at the fringe of influence on flash flooding.

3) Do 'heavy rainfall' and precipitation trends follow 'extreme rainfall' trends used in engineering design?

Kirchmeier-Young's research found that 1-day duration simulated precipitation from various models has increased over past decades, and this trend follows observations from HadEX2 (a global gridded dataset).

We can compare the HadEX2 trends across North America, and subregions shown in the research paper, with extreme rainfall trends based on Canadian climate station observations.  Let's start with the 1-day, 24-hour annual maximum rainfall trends across Canada.

The chart below shows how annual maximum rainfall has changed according to Environment and Climate Change Canada's version 3.10 Engineering Climate Datasets for all storm durations from 5-minutes to 24-hours.

For 24-hour durations, 4.9% of all stations have a significant increase, 91.2% have no significant change, 2.3% have significant decreases and 1.5% of stations had no data.

Comparing to earlier datasets:

                                                     Version 2.30           Version 3.00            Version 3.10

No significant 24-hour trend            91.5%                    91.1%                         91.2%

Significant 24-hour increase              5.3%                      5.4%                           4.9%           

So the percentage of data that has no significant trend is relatively steady, and represents over 90% of the data. The percentage of data that has a significant increase in 24-hour rainfall is decreasing relative to the earlier datasets.

Canadian Engineering Climate Dataset trend data does not show increases consistent with the research paper.

4) Do 'heavy rainfall' events studied in the research paper cause damaging flood events, and flash flooding 'particularly in urban areas?

No.  Flash flooding is due to short duration, high-intensity rainfall.

The severe thunderstorms that are responsible for urban flooding and that occur over minutes to hours are different than the storms that occur over hours to days as indicated in the RSI IDF training presentation noted above:

For this reason, those interested in turban flooding drivers should look at short duration rainfall extremes - see below.

5) What are the trends in 'extreme rainfall' in Environment and Climate Change Canada's Engineering Climate Datasets, the data used by engineers to analyze and design infrastructure to manage flash flooding risks in urban areas?

Short duration rainfall is responsible for urban flash flooding.  Environment and Climate Change Canada's Engineering Climate Datasets indicate the following on annual maximum rainfall trends across Canada:

The short durations from minutes to a couple hours have low percentages of significant increase, just like the 24-hour data noted above.  The amount of significant increases expected due to chance is 2.5% increasing and 2.5% decreasing.

In a review of an earlier dataset by Environment Canada's Shephard et. al in 2014 (link: these amounts of changes were deemed not significant:

"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."

And more recently in Canada’s Changing Climate Report, such changes in short duration extreme precipitation were explained by chance (link:

"There do not appear to be detectable trends in short-duration extreme precipitation in Canada for the country as a whole based on available station data. More stations have experienced an increase than a decrease in the highest amount of one-day rainfall each year, but the direction of trends is rather random over space. Some stations show significant trends, but the number of sites that had significant trends is not more than what one would expect from chance (Shephard et al., 2014; Mekis et al., 2015; Vincent et al., 2018)."

The short duration intensities used for infrastructure design, derived based on annual maximum series, have not increased in many regions based on compiled studies (see previous post:  A review of design intensities in southern Ontario shows overall increases in short duration values (see previous post:

So no change in how infrastructure is designed based on short-duration design intensities (that is, not including checks or 'stress tests' for future changes).

6) What does Kirchmeier-Young's research paper reveal about previous extreme rainfall and flooding events in Canada - has climate change increased runoff that could aggravate flood damages?

Nothing.  The storms that lead to widespread urban flooding are not addressed in the research paper.  The processes driving 1-5 day rainfall are different than those driving short-duration rainfall.  There are no significant increases in the short-duration rainfall that causes flooding based on Engineering Climate Datasets as shown above.

Why then have damages increased over decades? Possible reasons are:

a) growth in net written premiums: more insured properties = more losses

b) urbanization: more pavement means more runoff and impacts

The landmark case Scarborough Golf Country Club Ltd v City of Scarborough et al. (Ontario Court of Appeal, 1988, decision indicates that Toronto-area urbanization markedly increased runoff stresses that caused runoff, erosion and flooding:

“Expert evidence confirmed the effect of the city's rapid urbanization and water control plans on the creek.”

“It is important to note that the case is not presented primarily as a complaint against flooding but
rather that the markedly increased flows and increased velocity of flow have caused and continue to
cause damage to the creek bed and the adjacent tableland.” and

“There can be no doubt that the storm sewer facilities and urbanization of the lands to the north of the Club are the cause of the effects just described and that the difference in flow and velocity of flow is very substantial.”

Cities are growing and there is more runoff as shown here in some regions:

Urbanization and Flood Risks

c) inconsistent data: the data source for losses cited by CBC changed from 2008 onward

Changing data methods can lead to different results (see previous post on this:


The research paper makes a reference to an attribution study for the 2013 Alberta flood.  It states:

"Additionally, event attribution studies have identified an increased probability of some individual extreme precipitation events in this region due to anthropogenic influence (4, 5)"

Reference 4 is:

B. Teufel et al., Investigation of the 2013 Alberta flood from weather and climate
perspectives. Clim. Dynam. 48, 2881–2899 (2017). (link:

So we have one Canadian storm assessed. Findings are:

"Event attribution analysis suggests that greenhouse gas increases may have increased 1-day and 3-day return levels of May–June precipitation with respect to pre-industrial climate conditions. However, no anthropogenic influence can be detected for 1-day and 3-day surface runoff, as increases in extreme precipitation in the present-day climate are offset by decreased snow cover and lower frozen water content in soils during the May–June transition months, compared to pre-industrial climate."

So greenhouse gases may have increased precipitation, but that is offset by less snow, resulting in no change in runoff, compared to pre-industrial climate.

So with no change in runoff, can there be a change in flood damages attributed to the precipitation change?  The net effect is no increase in risk.


To wrap it up, CBC has relied on a research paper that looks at rainfall events (1-5 day precipitation) that are not related to urban flash flooding and are not related to the events that lead to significant damages (convective thunderstorms with peak intensities over minutes to hours).  The research does not review short-duration rainfall that is relevant to infrastructure design governed by short 'times of concentration' - i.e., they are 'flashy'.  The research does not appear to be consistent with trends in 24-hour annual maximum rainfall observed at Canadian climate stations and as published in Environment and Climate Change Canada's Engineering Climate Datasets - data show few statistically significant increases and the percentage of significant increases is decreasing slightly for 24-hour rainfall across Canada.  The short-duration rainfall intensities responsible for urban flooding show no consistent changes, and any significant changes are explained by chance, according to Environment Canada.

Many factors go into increasing flood damages.  Changes in rainfall does not appear to be one of those factors.  Media should take the time to dive deeper into the technical details they reference to improve the accuracy of reporting, so that the public is better informed about complex issues.
Urban flooding is a complex issue, and an important challenge to address that requires significant funding and attention.  A better understanding of the causes of flooding, and any changes in design rainfall, is required to mitigate flooding in the most objective, cost-effective manner.  CBC has relied more on model predictions than on actual data in the past, even confusing the two (see previous post:  In this recent report it has not met its JSP principle for accuracy by confusing longer-term precipitation and short-duration extreme rainfall.


BONUS - Trends in short-duration rainfall, based on annual maximum observations, from Environment Canada's version 3.10 Engineering Climate Datasets are summarized below (link:  These tables consider stations with a long period of record and recently updated data for regions across Canada.