Windsor and Tecumseh Flood Reporting Exposes Gaps in Media Meteorology Math and Memory

Reporting on the Windsor and Tecumseh flooding in September 2016 highlights the gaps in extreme weather reporting in Canada .. and amnesia regarding past events... and the unwillingness to explore other factors affecting flooding, like hydrologic changes.

The Facts: CBCNews reports 190 mm fell in Tecumseh 80.8 mm in Windsor

The Fishy: CBCNews reports "I've never seen anything that intense in the 35 years I've been in this region," (Tecumseh Mayor Gary ) McNamara said. "This is unprecedented."

But the event was not unprecedented in the region. Water resource management and civil engineering professionals know that the region has recently experienced a much greater storm. Even CBCNews reported on the anniversary of the Harrow Storm 27 years ago : they reported in July 2014 on the July19-20, 1989 storm:

"A massive, stationary storm dumped 450 mm (17 inches) of rain on Harrow and Colchester South in the matter of hours in southwestern Ontario.|

The Facts: CBCNews reports 63.4 mm of rainfall fell at Windsor Airport

The Fishy: "That beat the record for that date in history, which was formerly 36.8 mm set back in 1973." ... this is a fishy statement because drainage systems are not designed to specific calendar day rainfall characteristics. In fact annual series of maximum daily, hourly, and 5 minute intensities are used to characterize weather. This is a typical headline-grabbing technique such as when the July 8, 2013 storm in Mississauga was reported to be triple the maximum record ... ummm... for a calendar day that usually does not receive hardly any rain. This inaccurate reporting sets up the 'anchoring bias' in the readers' perception of the rarity of the event.

The Weather Network's Chris St. Clair says on the October 1, 2016 morning broadcast that the Windsor storm was "unprecedented" ... its not clear if he meant unprecedented for September 29th's, which is irrelevant, or unprecedented for the region on any calendar day which is plain wrong. In later report he calls the rain event 'unbelievable'. The Weather Network started broadcasting in September 1988, so perhaps it should be aware of the Harrow Storm in 1989.

The Weather Network's Kim MacDonald reports that Tecumseh received twice its average monthy rainfall in 15 hours on the October 1, 2016 broadcast - this reporting is another example of an 'anchoring bias' - comparing a rainfall statistic to another statistic that is irrelevant to drainage design - no drainage systems are designed for monthly totals, instead they are designed based on annual extreme time series and derived extreme value statistics to project rare conditions that may have not even occurred yet. See our post on heuristic biases in rainfall reporting.

Kim MacDonald notes that 1700 flooded homes were reported in this 2016 storm. In the 1989 Harrow Storm CBCNews reported 2000 flooded homes. Unprecedented?

Big picture questions for Essex Region could include whether the 100 year regional storm design standard is big enough for the region. Other conservation authorities regulate to larger storms - Upper Thames uses the 250 year storm - most other use Hurricane Hazel, which is perhaps a 500 year storm. Other regions use large historical storms like the Timmins Storm. Perhaps it is time for the "Windsor Storm" based on the Septmeber 2016 storm.

And are local flood hazard limits keeping up with hydrologic changes to the region over decades? We explore this for many Ontario cities here - but here is a view of the growth from the 1960's to late 1990's.

Consider this: when your design standard is so low (Essex Region has the minimum 100-year flood standard in Ontario according to the Provincial Policy Statement on natural hazards), the importance of expanded development and intensification in existing development areas is an even more important factor when considering increased runoff and flood risk. Why? Because pervious land uses can absorb some fraction of 'small' 100 year storm but not much of the large storms used in other regions. Those using Hurricane Hazel design storms which saturate the ground with anecedent moisture conditions (US Soil Conservation Service AMC III conditions to be exact) do not have as significant an increase in runoff following development. Those using 100 year storm, like Essex Region, use the drier AMC II conditions in hydrologic models, meaning that the soil-vegetation surface can absorb relatively more ... until it is paved over.

Does Climate Change Cause More Severe Storms? NOAA Says Yes In Louisiana But Not in Colorado. Canada Mixed?

The 2016 extreme rainfall in Louisiana has been deemed by NOAA to be more likely and severe due to change climate, while the 2013 extreme rainfall in Colorado has been deemed less likely. Why the different behaviour? Like in real estate, its all about location, location, location. Trends are not uniform across the entire continent - Louisiana is influenced by the Gulf of Mexico, and Colorado isn't.

The 2016 Louisiana rainfall event dropped more than six inches of rain in some regions August 11, 2016 and Baton Rouge received over 11 inches of rain August 12. Such an event has been analyzed to be more likely, with a shorter 'return period', meaning a higher probability of occuring. Specifically the 2016 Louisiana storm was assessed to have a 30 year return period today (i.e., 1/30 = 3.3% chance of occurring), while only a 50 year return period in the past (i.e., a lower 1/50 = 2% chance of occurring). Details are here.

Louisiana rain frequency shift - 3 day rainfall.
Details are on the NOAA website, including the graphic above.

But a NOAA-led study published in the Bulletin of the American Meteorological Society indicated that "The probability for an extreme five-day September rainfall event over northeast Colorado, as was observed in early September 2013, has likely decreased due to climate change." That storm dropped 17 inches of rain.

Details of the Colorado rainfall analysis are here, indicating a 12% decrease over the last century:

"Despite a warmer and moister climate, the frequency of September heavy five-day rain events does
not increase in the simulations but substantially declines in northeast Colorado (Fig. 5.2c). Using the model’s 95th percentile of five-day rainfall totals, we find a 12% decline in occurrence during recent decades compared to the late 19th century."

And the following:

"... a slight decline in intensity of the maximum five-day precipitation over the central Great Plains
during summer is also projected (Sillman et al. 2013), emphasizing that global and annual perspectives of climate change may not always pertain to events at a specific place and time."

Research by Peterson shows regional trends in flood magnitude over decades that support the decreasing Colorado rainfall trends:

How to Interpret: "Trend magnitude (triangle size) and direction (green = increasing trend, brown = decreasing trend) of annual flood magnitude from the 1920s through 2008. Flooding in local areas can be affected by multiple factors, including land-use change, dams, and diversions of water for use. Most significant are increasing trends for floods in Midwest and Northeast, and a decreasing trend in the Southwest."

These USA extreme events and trends highlight the difference in regional impacts of a changing climate. In Canada, precipitation analysis by Vincent and Mekis showed less extreme daily rain, although more days with rain:

"The analysis of the precipitation indices for 1950–2003 reveals more days with precipitation, a decrease in daily intensity and a decrease in the maximum number of consecutive dry days."

Research has shown regional trends in annual maximum observed rainfall amounts as well. Environment Canada scientists indicate the following in Atmosphere-Ocean indicating increases and decreases:

"Both the southwest and the east (Newfoundland) coastal regions generally show significant increasing regional trends for 1- and 2-hour extreme rainfall durations. For the shortest durations of 5–15 minutes, the general overall regional trends in the extreme amounts are more variable, with increasing and decreasing trends occurring with similar frequency; however, there is no evidence of statistically significant decreasing regional trends in extreme rainfall amounts. The decreasing regional trends for the 5- to 15-minute duration amounts tend to be located in the St. Lawrence region of southern Quebec and in the Atlantic provinces."

It is important to distinguish between 5-day rainfall trends, noted in the Louisiana and Colorado analyses, 5 minute to 24 hour rainfall trends in the Environment Canada analysis. In the Canadian data the trends can be mixed across a single weather station depending on the duration. This table shows that variability in long term Canadian climate stations:

The data shows, for example, decreasing very short duration and longer duration trends at Toronto Pearson Airport (5 minute, 6, 12, 24 hour), but increasing mid duration trends (10 minute to 2 hour). Ottawa has decreasing short duration trends - suggesting less severe urban flooding potential - and increasing longest duration trends.

Overall, Environment Canada indicates "a general lack of a detectable trend signal" in short duration rainfall (meaning 5 minute to 24 hour) that is important for engineering infrastructure design. In Southern Ontario, there are more statistically significant decreases in annual maximum rainfall than increases:

In Southern Ontario, the increases in extreme rainfall are generally over longest durations (6, 12, and 24 hours) that is not critical for urban flash flooding - in contrast, the shortest duration trends show decreasing extreme rainfall.

Previous posts suggest that the increase in flood damages in Ontario cities can be related to multiple factors such as land use change (also noted by Peterson regarding USA Floods). For example:

Urbanization and runoff explain Ontario flooding

Ontario, Canada city land use change and flood risk

Urbanization, Runoff, Overland Flow and Flooding - How Sprawl of Ontario Cities Drives Flood Risk and Insurance Losses in Urban Areas

Readers of this blog have seen these basic process described several times: (1) rain transforms into runoff when it hits the ground, (2) runoff accumulates and flows in rivers or municipal drainage infrastructure, (3) the capacity of the flow systems determines whether flow "backs up", "surcharges", "spills", or generally flows uncontrollably to where we don't want it to go, causing flooding.

Using Environment Canada's data and research, we have shown that rainfall intensities have not increased in southern Ontario here. In fact there are more statistically significant rain intensity decreases than increased south of 44 degrees. So the rainfall influence on runoff is not increasing. But runoff has been increasing after decades of urbanization under the today's stable or decreasing rainfall intensities.

The following maps show urban expansion in Mississauga, Oakville and Burlington Ontario from 1966 to about 2000 (data varies from 1999 to 2002). The overland flow system path based on Ontario conditioned digital elevation model is superimposed on the land use map so that the impact of urbanization and runoff into the drainage system can be considered.

 The effect of urbanization in Mississauga on runoff would be most acute in the smaller watersheds (e.g., not the Credit), where the upstream urban area has increased significantly since 1966.

Likewise in Oakville - Bronte Creek, a large watershed more slightly influenced by the city's sprawl, has not been affected to the same degree as the smaller Fourteen Mile Creek to the west, where a high relative change in land use over that smaller watershed has occurred throughout the city.

Same in Burlington - many small creek watersheds originating off the escarpment have dramatically increased urbanization over three decades. Burlington is characterized by creeks that have been realigned, straightened and encroached upon. These can be expected to be more sensitive to increased runoff rates due to expanded urbanization.


Parts of Hamilton have been urbanized up to the watershed divide (black line) by the late 1990's / early 2000's. How does this affect runoff into the old 'core' built to pre-1960's standards?

Hamilton, wider perspective. Some wetlands remaining upstream of Dundas? :

Richmond Hill (Lake Wilcox near upper middle of map). Some urbanization around the lake flows to the Humber where flow impacts would be muted, while other areas to the south flow flow to headwater tributaries of the Don and Rouge:

Land Use Change Drives Urban Flood Risk .. Yet Hydrologists Become "Useless Appendage" To More Virile Sciences

"The unsatisfactory state of hydrology is, in the final analysis, the result of the dichotomy between the theoretical recognition of hydrology as a science in its own right and the practical impossibility of studying it as a primary discipline but only as an appendage of hydraulic engineering, geography, geology, etc." Vit KlemeŇ° in Dilettantism in hydrology: Transition or destiny?, Water Resources Research, Vol. 22, 1986.

That explains it! Nobody studies hydrology in its own right, and so this limp appendage of other scientific pursuits has been ignored. Pity - because hydrologic science and simple quantifiable land use changes can explain increased runoff and increased urban flood damages in Ontario municipalities. Why is it ignored? Because hydrology is just the foreplay to other things like dynamic hydraulic simulations and then infrastructure construction. Everybody is always stampeding to the Priestman slot, frothing over pressure head - hydraulic hussies! A veritable orgy of climate simulators, and media stimulators, rages on while hydrologists are left alone :(

So we need some "tantric hydrology" people. Slow and savoured and reflected upon. Maybe if Sting was an engineering professor instead of an english teacher he could have promoted it, and enlightened the masses to the joy of runoff coefficients. Alas, he wasn't and it is so up to to spell it out - how increased urbanization has affected hydrologic responses in Ontario cities and magnified runoff potential, flood damages, and insurance losses.

Get ready.

Just open your peepers and gaze upon these Ontario urban area land use maps, and see how the limits of urban areas have increased over thirty years or so from 1966 to the late 1990's. It can be dramatic. For some catchments, the increase in urban land coverage can be an order of magnitude.

Sure, some of these areas developed between 1966 and 2000 have stormwater management controls up to a point. But it is commonly accepted that the on-site local stormwater controls do not completely address cumulative volumetric runoff impacts at larger scales. And the controls are finite - not controlling the rarest storms that drainage infrastructure could see in its design life. So expanded urban areas can drive downstream flood risk. Yes, every catchment is different and maybe no property is at risk downstream of some urbanized area (i.e., dedicated drainage easements and wide regulated flood plains have maintained a wide major flow path) - but in may cities the overland flow path has not been mapped managed or preserved, so increased runoff and flood risk can result.
Tantric Hydrology
Brought to you by the Flood Guru.
OK, could a movie about hydrology be any worse than The Love Guru? I guess.....

Secret is Out ! Urbanization and Runoff Explain Increased Urban Flood Risk in Southern Ontario

Ashley Madison ad: "Middle-aged
hydrologist seeking good listener,
someone, anyone, interested in science".
"Only the small secrets need to be protected. The big ones are kept secret by public incredulity."(attributed to Marshall McLuhan)

Could it be that we have ignored the obvious "secret"? That is, the effect of increased urbanization on urban flood risk in Ontario. It seems so.

Mapping of urbanization patterns in that Greater Toronto, Golden Horseshoe, and the Big K-W areas shows obvious, dramatic increases in watershed urbanization over a thirty year period from 1966 to the late 1990's. In some Toronto watersheds, urbanization increased by 986%, and that does not even consider intensification within previously developed areas. This analysis begs the question: "Why is there such a disproportionate focus on global climate change impacts to local extreme weather (now disproved) to explain flooding, compared to a focus on the obvious land use changes on hydrology, and increased runoff under historical climate conditions?"

The following map shows the increase in urbanization from 1966 according GIS mapping converted and shared by Ducks Unlimited (source is the Canada Land Inventory Land Use 1:50,000 scale mapping for Southern Ontario), and to 1999-2002 per the SOLRIS Version 1.2 land cover GIS mapping, as compiled in the Ontario Land Cover Compilation Version 2.0 and available through Land Information Ontario.
Greater Toronto Area Urban Area Growth in TRCA watersheds and Flood Risk Influence on Urban Flooding
Greater Toronto Area Urbanization and Flood Risk

Urban Growth in TRCA watersheds and Flood Risk Influence on Urban FloodingClearly, the amount of urbanization has increased dramatically in southern Ontario watersheds, including many that have experienced flooding in the past decades. In about 30 years, urbanization has increased by 986% over and above the 1966 amounts in the Rouge River Watershed, and by 696% in the Duffins Creek Watershed. In Toronto area watersheds, unlike Rouge and Duffins, the increase is also dramatic - not as much for the percentage increase, but for the relative coverage. The Highland Creek Watershed increased by an added 124% over the 1966 coverage, resulting in over 88% urbanization coverage. Similarly, Mimico Creek Watershed urban areas expanded by 135% more, also resulting in over 88% urbanization by 1999-2002. The following table shows the TRCA watershed breakdown.

How about Mississauga where there was flooding in 2013? Well, Etobicoke Creek Watershed added 341% urban coverage resulting in 64% coverage by 1999-2002. Yes it should seem greater, but we have not counted the valley areas or other large open or vegetated areas within the urban areas (e.g., hydro corridors, etc.). We have also subtracted these 'greenspaces' from the 1966 data to avoid over-reporting the 1966 coverage, as the 1:50,000 source data was not as detailed as the the SOLRIS data that classified land used down to a 15 m cell resolution. And the Cooksville Creek area in Mississauga? In 1966 urban development extended generally up to Dundas Street, while 30 years later it extended to Highway 407 as shown in the image below.

Mississauga Urbanization 1966 to 1999-2002.

Do the math. Nearly an order of magnitude increase in urban coverage in one large watershed. Doubling of urban coverage in all but one other TRCA jurisdiction watershed. And even more intensification within the old urban areas. Runoff potential has increased dramatically as a result. This explains increased flood risk in urban areas.

While stormwater management controls have been put into place for some later developments to mitigate hydrologic impacts, these are finite, perhaps up to 100 year storm standards - extreme weather events can be above or below that threshold, meaning flood damages can occur for the large events. Furthermore, many infrastructure components put in place decades ago are not sized to handle 100 year storms, and have been designed to overtop for high frequency, lower return period events (e.g., underpasses, local roadway crossings).

So why have folks jumped on the now-discredited conclusion that climate change has been driving local flood damages and losses, but ignored basic hydrology and effects of urbanization? Perhaps we can learn from Vit Klemes, perhaps our greatest hydrologist. He had reiterated his piece to the Canadian Water Resources Association (Implications of possible climate change for water management and development. Water News (CWRA), 11, 1, S2-S3) in his address to the International Interdisciplinary Conference on Predictions for Hydrology, Ecology, and Water Resources Management: Using Data and Models to Benefit Society,15–18 September 2008, Prague, Czech Republic. In this address, entitled "Political Pressures in Water Resources Management. Do They Influence Predictions?",  he states:

"[the theorists] find it easier to play trivial scenario-generating computer games while the [managers] find these games much easier to finance... And so by happy collusion of interests, an impression is created that 'something is being done for the future' while the real problems are quietly allowed to grow through neglect of the present"

Neglect obvious present hydrology. Instead, generate theoretical climate change storm impacts .. with computer "games", i.e., models, or even basic high school math probability density function manipulation. And happily fund every climate change prediction or adaptation study.

NB - the newer urban areas have higher densities and greater runoff potential than the older ones:

York Region Urban Development Intensification 1952-2002 - Impervious Surface Area Coverage Doubles Over 50 Years

And intensification within established urban areas can increase runoff potential as well. This shows changes since the early 1970's across the 7 lots beside my house in east Toronto:

Redevelopment and increased urbanization in east Toronto - no residential, single lot stormwater management controls to mitigate runoff impacts to existing drainage system.
And this shows intensification in York Region in an area developed in the early 1950's but undergoing redevelopment / intensification, expanding impervious area coverage, to this day: