flash flooding, stormwater, and decision...
TRANSCRIPT
Flash Flooding, Stormwater, and Decision Making
http://graham.umich.edu/climate
Photo: Richard Deming Photography
Flash Flooding, Stormwater, and Decision Making for Cities in the Great Lakes
Megan Krajewski Climate Center Research Assistant, Earth and Environmental Science Student
University of Michigan
Daniel Brown Climatologist
University of Michigan Climate Center
Elizabeth Gibbons Director
University of Michigan Climate Center
This project was sponsored by the Climate Center and University of Michigan’s Undergraduate Research Opportunity Program.
PROJECT BRIEF: FLASH FLOODING, STORMWATER, AND DECISION MAKING
2
http://graham.umich.edu/climate Last updated: 9/25/2015
Recommended Citation:
Krajewski, M., D. Brown, E. Gibbons, 2015. Flash Flooding, Stormwater, and Decision Making for Cities in the Great Lakes. Available from the University of Michigan Climate Center.
For further questions, please contact Daniel Brown, [email protected]
PROJECT BRIEF: FLASH FLOODING, STORMWATER, AND DECISION MAKING
3
http://graham.umich.edu/climate Last updated: 9/25/2015
Contents
Introduction .......................................................................................................................................................................................................................................... 3
Definitions:............................................................................................................................................................................................................................................. 5
Method ..................................................................................................................................................................................................................................................... 5
Challenges .............................................................................................................................................................................................................................................. 5
Results...................................................................................................................................................................................................................................................... 6
Population Density .................................................................................................................................................... Error! Bookmark not defined.
Number of Flash Floods per Year/Average Precipitation of a Flash Flood by City ...................................................................................... 7
Average Precipitation ............................................................................................................................................................................................................... 7
Difference between Combined and Separate Sewer Event Totals by Precipitation Threshold .............................................................. 9
Sensitivity ...................................................................................................................................................................................................................................... 9
Statistical Tests for Population Density ........................................................................................................................................................................ 10
Statistical Tests for Sensitivity .......................................................................................................................................................................................... 12
Damages ...................................................................................................................................................................................................................................... 15
Discussion ........................................................................................................................................................................................................................................... 18
Acknowledgments ........................................................................................................................................................................................................................... 18
References ........................................................................................................................................................................................................................................... 18
COMBINED SEWER SEPARATE SEWER
1. Albany, NY 16. Akron, OH
2. Aurora, IL 17. Ann Arbor, MI
3. Buffalo, NY 18. Bloomington, IN
4. Chicago, IL 19. Duluth, MN
5. Cleveland, OH 20. Eau Claire, WI
6. Detroit, MI 21. Erie, PA
7. Fort Wayne, IN 22. Grand Rapids, MI
8. Harrisburg, PA 23. Green Bay, WI
9. Lafayette, IN 24. Madison, WI
10. Milwaukee, WI 25. Marquette, MI
11. Philadelphia, PA 26. Minneapolis, MN
12. Pittsburgh, PA 27. Oswego, NY
13. Saginaw, MI 28. Springfield, IL
14. South Bend, IN 29. St. Cloud, MN
15. Toledo, OH 30. Traverse City, MI
Flash flood data from 15 cities with combined sewer systems and 15 cities with separate sewer systems was
analyzed for this report.
4
http://graham.umich.edu/climate Last updated: 9/25/2015
Introduction
Heavy precipitation events have been increasing in frequency and intensity over time. The amount of
precipitation falling in the most intense 1% of precipitation events increased by 37% in the Midwest and 71% in
the Northeast from 1958 through 2012 (Walsh et al., 2014). The amount of precipitation falling during week-
long, once a year precipitation events has also increased by 25% to 100% across the Great Plains and Upper
Midwest (Kunkel et al., 1999). While projections of future heavy precipitation events greatly vary, most models
project that daily extreme precipitation events will continue to become more frequent and more intense for
many areas of the Great Lakes region (Kunkel et al. 2013). Areas that are currently vulnerable to heavy
precipitation events will likely become more vulnerable in the future.
This work evaluates trends in flash flooding and historical precipitation totals from 1996-2011 in the eight states
that border the Great Lakes. The National Oceanic and Atmospheric Administration’s (NOAA) Storm Events
Database (https://www.ncdc.noaa.gov/stormevents/) includes detailed records of flash floods organized by their
county or location of occurrence since 1996. GLISA staff maintains quality-controlled NOAA NCEI Global
Historical Climate Network-Daily observational data (GHCN-Daily) from the Great Lakes region in order to
inform climate adaptation efforts at the local, state, and regional level. This data is currently available and in
accessible formats to users of varying backgrounds. (http://glisa.umich.edu/resources/great-lakes-climate-
stations) While GLISA staff and affiliates use this quality-controlled subset of GHCN-Daily data widely
throughout the region to provide quantitative, locally-relevant references of historical climate for stakeholders
near the observation sites, it remained unclear if daily precipitation totals included in this data could be used as
a proxy for quantifying the relative vulnerability and sensitivity to damage from precipitation-related storm
events for nearby communities across the region.
Decision-makers in the region are interested in where vulnerabilities to stormwater overflows and unplanned
discharges are greatest. Cities with combined sewers are often assumed to have a greater risk of overflows, as
they direct stormwater into the same conveyance system that carries untreated wastewater. During heavy
precipitation, combined sewer systems operating near capacity are forced to discharge untreated wastewater at
pre-determined points or risk an uncontrolled overflow elsewhere. Separate sewer systems convey stormwater
through a separate conveyance system, reducing or eliminating the risk of wastewater being discharged and the
associated public health risks. Combined sewers are most common in the Northeast and Great Lakes regions of
the US in cities.
While the public health and stormwater management benefits of a separate system are well known, it remained
unclear if GHCN-Daily data could be used in conjunction with the NOAA Storm Events database to quantify an
increased capacity to cope with heavy precipitation in cities with separate sewer systems versus those with
combined systems.
The 30 cities analyzed in this study were chosen with an equal number of separate and combined sewer cities.
The cities also cover a wide area throughout the Great Lakes states and information about each of the city’s
sewers was clearly available online. The chosen cities had either mostly separate or mostly combined sewer
systems.
In this study, we attempted to 1) test the effectiveness of GHCN-Daily data in quantifying regional sensitivity to
the frequency of flash flooding following heavy precipitation, and 2) test the effectiveness of GHCN-Daily data
in quantifying potential increases in the capacity of separate sewer systems. Understanding how GHCN-Daily
data can be used to identify past and future vulnerabilities will help decision makers plan for changing weather
patterns in the region.
5
http://graham.umich.edu/climate Last updated: 9/25/2015
Definitions:
Flash flood (according to NOAA’s Storm Events Database):
A rapid rise in water level in places that are normally dry or have much lower water levels
Must pose a threat to life or property
Must be within a 6 hour period of the causative event (e.g., intense rainfall, dam failure, ice jam-related)
Cannot exist for 2-3 consecutive days
Combined Sewer: Water removal systems that collect precipitation runoff, domestic sewage, and industrial
wastewater in the same conveyance system. Most of the time, combined sewer systems transport wastewater to
a sewage treatment plant where it is treated and then discharged to a water body. During periods of heavy
precipitation, however, the wastewater volume in a combined sewer system can exceed the capacity of the
sewer system or treatment plant. In these cases, combined sewer systems are designed to overflow and
discharge excess wastewater directly to water bodies. (EPA, http://water.epa.gov/polwaste/npdes/cso/)
Separate Sewer: Water removal systems that collect stormwater, domestic sewage, and industrial wastewater
in separate conveyance systems. During periods of heavy precipitation, sanitary sewage and industrial
wastewater flows remain unaffected, and only precipitation runoff is discharged into local water bodies.
Methods
The code was written in Python to extract data for 30 cities from NOAA’s Storm Events Database and GLISA’s
daily precipitation record for 1996-2011. For each of the chosen cities, the flash flood date and damage that
was recorded for the city’s county by NOAA was paired with GLISA’s daily precipitation data. The damages
were adjusted for inflation according to the US Bureau of Labor Statistics data. Great Lakes cities were chosen
based on the quality of the matching precipitation data and the availability of information for each city’s sewer
type. This data was further used for examination in Excel.
Challenges
NOAA’s storm events data was given by county, but GHCN-Daily precipitation data is point-based.
Daily precipitation data is rarely collected exactly where storms are most intense and does not capture
fluctuations in precipitation rate throughout a given day.
Complete data from the Storm Events Database was not available for 2012 or 2014, and 2013 was
omitted for data continuity.
NOAA’s damage records greatly vary between cities; some cities have hardly any damages recorded
while others have damage for nearly every storm. This could be a product of different data recorders
and methods.
Flash flooding can be caused by snow melt or infrastructure failure, which are classified as a 0 inch
precipitation threshold.
6
http://graham.umich.edu/climate Last updated: 9/25/2015
Results
Population Density There is a strong correlation between population density and how often a city has a flash flood. Additionally,
many of the cities with highest population densities have combined sewers. The cities that flood the most are
generally cities with population densities greater than 6000 people per square mile that would have a difficult
and expensive time replacing their old, combined sewer system with a new, separate one.
In general, more densely-populated cities have more impervious land cover than less dense cities. Natural land
cover (forests, grasslands, pervious soil or vegetation) has the ability to absorb rainfall with much less runoff
than impervious land cover such as buildings or parking lots. In a setting with high impervious land cover,
runoff is transported to streams much more quickly than it would be in a more permeable setting, which can
result in more frequent and significant flooding (Perlman, 2015; Flinker, 2010). Sewers in bigger cities must be
able to handle a significant amount water during a precipitation event to avoid sewer overflows.
One obstacle faced during this study was comparing and contrasting the cities with separate and combined
sewers because of other variables such as impervious surface, age of infrastructure, topography, and soil type,
could also impact precipitation impact in the city. The majority of cities in the Great Lakes region with high
population densities also rely on combined sewer systems. Because of this commonality it is difficult to find
cases to compare similarly dense cities with separated versus combined systems.
y = 0.0984x + 0.0098R² = 0.4798
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
0
2000
4000
6000
8000
10000
12000
14000
16000
18000
Du
luth
St.
Clo
ud
Tra
vers
e C
ity
Ma
rqu
ette
Sp
ring
field
Ea
u C
laire
Gre
en
Ba
y
Ft.
Wa
yne
Osw
eg
o
La
yfa
yette
So
uth
Be
nd
Sa
gin
aw
Ma
dis
on
Akr
on
Blo
om
ing
ton
Tole
do
An
n A
rbo
r
Gra
nd
Ra
pid
s
Au
rora
Alb
an
y
Erie
Ha
rris
bu
rg
Milw
au
kee
Min
ne
ap
olis
Cle
vela
nd
De
tro
it
Ph
ilad
elp
hia
Bu
ffalo
Pitt
sbu
rgh
Ch
ica
go
Av
era
ge
Nu
mn
ber
of F
las
h F
loo
ds
Pe
r Y
ea
r
Po
pu
lati
on
De
ns
ity
(pe
op
le p
er s
qu
re m
ile)
City
Population Density and Average Number of Flash Floods per Year
People per square mile (separate sewer cities)Average flash floods per year
Linear (Average flash floods per year)
People per square mile (combined sewer cities)
7
http://graham.umich.edu/climate Last updated: 9/25/2015
Number of Flash Floods per Year/Average Precipitation of a Flash Flood by City In the 30 cities, the average number of flash floods per year in combined sewer cities is 2.02 floods versus 1.05
floods for separate sewer cities.
Each city can take a different average amount of precipitation before it floods. There are several possible
factors that may affect this number including amount of impervious surface, age of infrastructure, topography in
and around the city, sewer type, soil type, and soil saturation before the flood occurs (Holton, 2003).
In order to compare the frequency of flash flooding in all 30 cities, the average number of flash floods a city has
on a yearly basis was divided by the average precipitation the city gets before it floods.
Cities that rank the lowest on this graph must either have less than 1 flood per year, on average, or be able to
handle a high amount of precipitation before a flood occurs. In this graph, 8 out of 10 of the lowest ranking
cities have separate sewers. On the other hand, the highest ranking cities flood relatively often at lower
precipitation thresholds.
Average Precipitation Many separate sewer cities have a low number of flash floods per year divided by their average precipitation.
To try to explain this trend, the average precipitation of all combined and all separate sewer cities was
investigated. Of the cities sampled, there was no significant difference in the average amount of rain a
0
1
2
3
4
5
6
7
Tra
vers
e C
ity
Gre
en
Ba
y
Sa
gin
aw
Ea
u C
laire
So
uth
Be
nd
Osw
eg
o
An
n A
rbo
r
Du
luth
Gra
nd
Ra
pid
s
Ma
rqu
ette
St.
Clo
ud
Blo
om
ing
ton
De
tro
it
Tole
do
Ft.
Wa
yne
Au
rora
La
yfa
yette
Milw
au
kee
Akr
on
Ma
dis
on
Min
ne
ap
olis
Ha
rris
bu
rg
Sp
ring
field
Ph
ilad
elp
hia
Cle
vela
nd
Alb
an
y
Erie
Bu
ffalo
Ch
ica
go
Pitt
sbu
rgh
Nu
mb
er
of F
las
h F
loo
d E
ve
nts
pe
r Y
ea
r
City
Average Number of Flash Floods Per Year
Combined Sewer
Separate Sewer
0123456789
Gre
en B
ay
Tra
vers
e C
ity
Sou
th B
end
Sag
inaw
Gra
nd R
apid
s
Eau
Cla
ire
Dul
uth
Ann
Arb
or
Osw
ego
St.
Clo
ud
Det
roit
Tole
do
Mad
ison
Milw
auke
e
Lafa
yett
e
For
t Way
ne
Aur
ora
Phi
lade
lphi
a
Min
neap
olis
Har
risbu
rg
Akr
on
Spr
ingf
ield
Alb
any
Blo
omin
gton
Cle
vela
nd
Erie
Chi
cago
Mar
quet
te
Buf
falo
Pitt
sbur
gh
Nu
mb
er o
f Fla
sh F
loo
d E
ven
ts
per
Yea
r / A
vera
ge
Pre
cip
itat
ion
of a
Fla
sh F
loo
d
City
Number of Flash Floods per Year/Average Precipitation of a Flash Flood by City
Combined Sewer
Separate Sewer
8
http://graham.umich.edu/climate Last updated: 9/25/2015
combined or separate sewer city receives before a flash flood occurs. However, it is important to note that
combined sewer cities as a whole have more flash flood events than separate sewer cities.
After analysis of precipitation totals for 1, 2, 3, and 7 days, the most statistically significant data comes from the
1 day total. Precipitation totals for 2 and 3 days are somewhat significant, but 7 day totals do not correlate well
with how often a city has a flash flood.
Because flash floods must occur within hours of a storm, they are typically caused by more than 1 inch of
precipitation on the same day as the flood. Duration and intensity of rainfall are important factors to flood
causation and vary from city to city (Montz, 2002). Additionally, the topography in and around the city plays a
role in how much rain can cause a flood. According to Kelsch et al. (2001), “High intensity rainfall is more
important than the total accumulation on small, fast-response basins. Basin characteristics are easily as
important as the rainfall characteristics for determining the nature of the runoff.”
0
0.5
1
1.5
2
2.5
3
3.5M
arqu
ette
Buf
falo
Blo
omin
gton
Pitt
sbur
gh
Osw
ego
Aur
ora
Cle
vela
nd
Erie
For
t Way
ne
St.
Clo
ud
Spr
ingf
ield
Chi
cago
Tole
do
Akr
on
Min
neap
olis
Lafa
yett
e
Har
risbu
rg
Tra
vers
e C
ity
Det
roit
Ann
Arb
or
Dul
uth
Alb
any
Sag
inaw
Sou
th B
end
Phi
lade
lphi
a
Gre
en B
ay
Milw
auke
e
Mad
ison
Gra
nd R
apid
s
Eau
Cla
irePre
cip
ita
tio
n (i
nc
he
s)
City
Average Precipitation per Flash Flood Event
Combined Sewer
Separate Sewer
9
http://graham.umich.edu/climate Last updated: 9/25/2015
Difference between Combined and Separate Sewer Event Totals by Precipitation
Threshold Nearly across the board, combined sewer cities have more flash flood events than separate sewer cities by
threshold. Combined sewers appear to be less effective in handling any size precipitation event than combined
sewers.
Percent difference of the combined minus the separate flash flood events divided by the total number of events
per threshold shows a substantially larger percentage of flash flood events occurring in combined sewer cities
than separate sewer cities in 11 of13 thresholds. The average of all the percent difference thresholds show
combined cities have 26.98 percent more flash flood events than separate cities.
Sensitivity Flash flood sensitivity is how likely a city will flood if it reaches a certain precipitation threshold. In this study,
the sensitivity was found by dividing the number of flash flood events in a precipitation threshold by the total
number of days a city reaches that threshold. Then all of the combined sewer cities and separate sewer cities
were averaged. The most significant results come from precipitation totals accumulated on the same day as the
-10
0
10
20
30
40
50
60
70
80
0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.25 2.5 2.75 3 3+
Dif
fere
nce
Bet
wee
n T
ota
l
Co
mb
ined
Sew
er E
ven
ts -
To
tal
Se
pa
rate
Se
we
r E
ve
nts
Precipitation Threshold (Inches)
Difference Between Combined and Separate Sewer Event Totals by Precipitation Threshold
-20
-10
0
10
20
30
40
50
60
70
0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.25 2.5 2.75 3 3+
Pe
rce
nt D
iffe
ren
ce
Precipitation Threshold (Inches)
Percent Difference Between Combined and Separate Events by Threshold
10
http://graham.umich.edu/climate Last updated: 9/25/2015
flood. Somewhat significant results were seen when looking at sensitivities when precipitation totals from the
day of the flood and the day before. The 3 and 7 day precipitation totals did not provide significant sensitivity
trends.
The most notable result found when looking at sensitivities was a significantly higher chance of flooding in
combined sewer cities above the 2.25 inch precipitation threshold. This may indicate that sewer type does not
have an impact on flash flooding until very large amounts of precipitation are accumulated in one day.
Additionally, maximum sensitivities were higher for combined sewers when precipitation gets exceptionally
high.
Statistical Tests for Population Density As a statistical check of the effect of population density on flash flooding, each city’s population density was
divided by the total number of flash flood events for that city, the number of events caused by greater than 1.25
inches of precipitation in one day, and the number of events caused by greater than 1.75 inches of precipitation
in one day.
Clay-based soil types typical of the Midwest, such as alfisols and mollisols, can absorb about 1.25 inches of
precipitation in one day. More than 1.25 inches begins cause soil saturation and runoff (Takle, 2011). Heavily
urban areas can absorb less precipitation, but also actively manage water. Less urban and rural areas with less
or no water management system would begin to see flash floods occur near this 1.25 inch threshold.
At 1.75 inches of precipitation in 24 hours, green infrastructure begins to get overwhelmed (Cruce, 2011).
Additionally, The Milwaukee Metropolitan Sewerage District estimated in 2011 that 1.75 inches of
precipitation in 24 hours was the lower threshold at which combined sewer overflows into Lake Michigan
began to occur (Cruce, 2011).
An issue with comparing the separate and combined sewer cities is that the top 6 densest cities in this study
have combined sewers. These cities are Chicago, Pittsburgh, Buffalo, Philadelphia, Detroit, and Cleveland. For
comparisons sake, trends were found for all separate cities, all combined cites, and combined cities within the
range of population densities as separate cities (Adjusted Combined on graph). The trend of all 15 combined
cities dramatically changed when data for the 6 highest population densities were discarded for floods caused by
1.25 inches or greater precipitation days. In order to get an accurate idea of how population density and number
of flash flood events are related, more cities need to be added to the study.
0
0.5
1
1.5
2
2.5
3
0 0.5 1 1.5 2 2.5 3 3.5 4
Ev
en
ts P
er
Th
res
ho
ld/
Da
ys
in T
hre
sh
old
Precipitation Threshold (Inches)
1 Day Sensitivity
Separate Sewer
Combined Sewer
Combined Cities Average
Combined Cities Maximum
Separate Cities Average
Separate Cities Maximum
11
http://graham.umich.edu/climate Last updated: 9/25/2015
When looking the trendline for all storm events for separate, combined, and the adjusted combined sewer cities,
there is no significant statistical difference for flash flood frequency.
As population density increases, combined sewer cities were less likely to flood than separate sewer cities for
flash floods caused by more than 1.25 inches of precipitation in one day. After discarding the 6 highest
population density cities, the adjusted combined trend was not significantly different than the separate sewer
trend.
y = 0.0002x + 0.3122R² = 0.5165
y = 0.0003x + 0.2381R² = 0.3176
y = 0.0002x + 0.3426R² = 0.3019
0
1
2
3
4
5
6
7
0 5000 10000 15000 20000
Av
era
ge
Nu
mb
er
of F
las
h F
loo
d
Ev
en
ts P
er
Ye
ar
People Per Square Mile
Flash Flood Events by Population Density
Combined Sewer
Separate Sewer
Linear (Combined Sewer)
Linear (Separate Sewer)
Linear (Adjusted Combined)
Combined
Separate
y = 5E-05x + 0.4761R² = 0.2314
y = 0.0002x + 0.069R² = 0.3474
y = 0.0002x + 0.1316R² = 0.2427
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
0 5000 10000 15000 20000Ave
rag
e N
um
ber
of
Fla
sh F
loo
d E
ven
ts P
er
Yea
r G
reat
er t
han
1.2
5 In
ches
of
Pre
cip
itat
ion
People Per Square Mile
Flash Flood Events >1.25 Inches by Population Density
Combined Sewer
Separate Sewer
Linear (Combined Sewer)
Linear (Separate Sewer)
Linear (Adjusted Combined)
Combined
Separate
12
http://graham.umich.edu/climate Last updated: 9/25/2015
Combined sewer cities are not as likely as separate sewer cities to flood as population increased for flash flood
events caused by more than 1.75 inches of precipitation in one day. After discarding the 6 highest population
density cities, the adjusted combined trend was more likely to flood as population density increased.
These graphs again suggest that sewer type does not have a huge impact on flash flooding frequency until a
certain threshold of rain is reached.
Statistical Tests for Sensitivity The 1.25 and 1.75 inch precipitation threshold sensitivities are important as the thresholds for beginning to
cause runoff and overwhelming green infrastructure. After removing one outlier from each type of city for
having a very high average number of precipitation days greater than 1.25 or 1.75 inches per year, the linear
trends for separate and combined cities were found.
y = 3E-05x + 0.3821R² = 0.0916
y = 1E-04x + 0.1101R² = 0.219
y = 0.0002x - 0.1324R² = 0.3907
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
0 5000 10000 15000 20000
Ave
rag
e N
um
ber
of
Fla
sh F
loo
d E
ven
ts P
er
Yea
r G
reat
er t
han
1.7
5 In
ches
of
Pre
cip
itat
ion
People Per Square Mile
Flash Flood Events >1.75 Inches by Population Density
Combined Sewer
Separate Sewer
Linear (Combined Sewer)
Linear (Separate Sewer)
Linear (Adjusted Combined)
Separate
Combined
13
http://graham.umich.edu/climate Last updated: 9/25/2015
Bloomington and Philadelphia were outliers that were discarded in this graph because they had a high average
number of precipitation days greater than 1.25 inches per year. Without these cities, the combined sewer
trendline changes slope from 0.2985 to 0.1482 and the separate sewer trendline changes slope from 0.026 to
0.285. The separate sewer trend with all of the cities has a lower sensitivity to flooding than combined.
The combined sewer trend is more sensitive than the separate sewer trend. The separate sewers are less
sensitive when a city has fewer average days of precipitation greater than 1.25 inches (around 3 days per year).
y = 0.2049x - 0.0487R² = 0.2985
y = 0.0256x + 0.4324R² = 0.0104
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
0 2 4 6 8 10 12
Av
era
ge
Nu
mb
er
of F
las
h F
loo
d E
ve
nts
G
rea
ter
tha
n 1
.25
Inc
he
s P
er
Ye
ar
Average Number of Precipitation DaysGreater than 1.25 Inches Per Year
Average Number of Flash Flood Events Per Number of Precipitation Days >1.25 Inches by Year
Combined Sewer
Separate Sewer
Linear (Combined Sewer)
Linear (Separate Sewer)
Combined
Separate
y = 0.1854x + 0.0269R² = 0.1482
y = 0.2851x - 0.5473R² = 0.3193
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
0 2 4 6 8
Av
era
ge
Nu
mb
er
of F
las
h F
loo
d E
ve
nts
G
rea
ter
tha
n 1
.25
Inc
he
s P
er
Ye
ar
Average Number of Precipitation DaysGreater than 1.25 Inches Per Year
Average Number of Flash Flood Events Per Number of Precipitation Days >1.25 Inches by Year (Highest Number of
Precipitation Days Discarded for Separate and Combined Sewers)
Combined Sewer
Separate Sewer
Linear (Combined Sewer)
Linear (Separate Sewer)
Combined
Separate
14
http://graham.umich.edu/climate Last updated: 9/25/2015
Separate sewers and combined sewers have about the same sensitivities when a city has a larger number of
precipitation days greater than 1.25 inches per year (around 6 days per year).
y = 0.3196x + 0.0502R² = 0.26
y = 0.0726x + 0.281R² = 0.0453
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
0 1 2 3 4 5
Av
era
ge
Nu
mb
er
of F
las
h F
loo
d E
ve
nts
G
rea
ter
tha
n 1
.75
Inc
he
s P
er
Ye
ar
Average Number of Precipitation DaysGreater than 1.75 Inches Per Year
Average Number of Flash Flood Events Per Number of Precipitation Days >1.75 Inches by Year
Combined Sewer
Separate Sewer
Linear (Combined Sewer)
Linear (Separate Sewer)
Combined
Separate
y = 0.4137x - 0.0825R² = 0.233
y = 0.4275x - 0.2133R² = 0.5386
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
0 0.5 1 1.5 2 2.5 3Av
era
ge
Nu
mb
er
of F
las
h F
loo
d E
ve
nts
G
rea
ter
tha
n 1
.75
Inc
he
s P
er
Ye
ar
Average Number of Precipitation DaysGreater than 1.75 Inches Per Year
Average Number of Flash Flood Events Per Number of Precipitation Days >1.75 Inches by Year (Highest Number of
Precipitation Days Discarded for Separate and Combined Sewers)
Combined Sewer
Separate Sewer
Linear (Combined Sewer)
Linear (Separate Sewer)
Combined
Separate
15
http://graham.umich.edu/climate Last updated: 9/25/2015
The data from Bloomington and Philadelphia were again removed in this graph because they were high outliers
in average number of precipitation days greater than 1.75 inches per year. Combined sewers are slightly more
sensitive than separate sewers greater than the 1.75 inch precipitation threshold.
Since the trends had a significant change in slope when one point from both categories was removed, more
cities are needed to verify the 1.25 and 1.75 sensitivity trends are accurate.
Damages Damages from all flash flood events were analyzed and adjusted the amounts for inflation. After a number of
different analyses, no correlation was found between separate and combined sewers and damage.
When comparing the average damage per storm and sewer type, no correlation was found. Most cities had an
average amount of less than $1 million per storm. A few storms had outrageous damages that made the average
damage amount unrealistic for cities like Green Bay and Detroit, and therefore not a good comparison between
all cities.
Storms that caused more than $1 million in damages represent exceptionally damaging storms, and this level of
damage is not typical of a flash flood. Discarding these outlier storms, the trend for damages and sewer type
still did not yield an obvious correlation.
02468
10121416
Tra
vers
e C
ity
Ma
rqu
ette
Osw
eg
o
Bu
ffalo
Du
luth
Ha
rris
bu
rg
La
faye
tte
Erie
Au
rora
Milw
au
kee
Blo
om
ing
ton
Cle
vela
nd
Ch
ica
go
Alb
an
y
Pitt
sbu
rgh
Fo
rt W
ayn
e
Sp
ring
field
Ma
dis
on
Ea
u C
laire
Gra
nd
Ra
pid
s
St.
Clo
ud
So
uth
Be
nd
Ph
ilad
elp
hia
Sa
gin
aw
Akr
on
Min
ne
ap
olis
An
n A
rbo
r
Tole
do
Gre
en
Ba
y
De
tro
it
US
D (
Mil
lio
ns
)
City
Average Damage per Storm
Combined Sewer
Separate Sewer
16
http://graham.umich.edu/climate Last updated: 9/25/2015
Precipitation amounts larger than 2 inches are more likely to overwhelm a sewer system, whereas smaller
amounts are more likely to be managed or cause a smaller flood and less damage. After examining flash floods
caused by less than 2 inches of precipitation in one day, no correlation between sewer type and damage amount
was found.
One factor to keep in mind is this graph does not take soil saturation into account. A city’s soil could be near
saturation from a previous day of rain and a small amount of precipitation could cause a flash flood. However,
in our analysis this is a rare occurrence.
To see if certain precipitation thresholds were handled better by either sewer type, average damages by
precipitation threshold were analyzed. Initially, a few outlier cities were found that caused unrealistic average
damage amounts that are not typical of a flash flood. Any storm that causes over $1 million in damages is a
storm that likely was too big for any sewer system to manage. About 90 percent of the average damages per
threshold were less than $1 million.
0
20000
40000
60000
80000
100000
120000
140000
160000
Ma
rqu
ette
Ph
ilad
elp
hia
Tra
vers
e C
ity
So
uth
Be
nd
Gre
en
Ba
y
Fo
rt W
ayn
e
Min
ne
ap
olis
An
n A
rbo
r
Co
lum
bu
s
Sa
gin
aw
Blo
om
ing
ton
Ha
rris
bu
rg
La
faye
tte
Au
rora
Pitt
sbu
rgh
Osw
eg
o
Du
luth
Sp
ring
field
St.
Clo
ud
Alb
an
y
Milw
au
kee
Erie
De
tro
it
Ch
ica
go
Ea
u C
laire
Gra
nd
Ra
pid
s
Bu
ffalo
Cle
vela
nd
Ma
dis
on
Akr
on
Tole
do
Da
ma
ge
(U
SD
)
City
Average Damage for Floods with Less than $1 Million in Damage
Combined Sewer
Separate Sewer
0
500000
1000000
1500000
2000000
2500000
3000000
3500000
4000000
Du
luth
Ma
rqu
ette
Tra
vers
e C
ity
La
faye
tte
Gra
nd
Ra
pid
s
Gre
en
Ba
y
Sa
gin
aw
So
uth
Be
nd
An
n A
rbo
r
Min
ne
ap
olis
De
tro
it
Ha
rris
bu
rg
Osw
eg
o
Co
lum
bu
s
Alb
an
y
Blo
om
ing
ton
Ea
u C
laire
Ch
ica
go
Erie
Fo
rt W
ayn
e
Tole
do
Bu
ffalo
Akr
on
Sp
ring
field
Pitt
sbu
rgh
St.
Clo
ud
Ph
ilad
elp
hia
Au
rora
Cle
vela
nd
Ma
dis
on
Milw
au
kee
Da
ma
ge
(U
SD
)
City
Average Damage for Floods with Less than 2 Inch Precipitation
Combined Sewer
Separate Sewer
17
http://graham.umich.edu/climate Last updated: 9/25/2015
Looking only at the average damages per threshold that were less than $1 million, there appears to be no
precipitation threshold where substantial damages are reported. There is no apparent trend in damages
compared to precipitation. Additionally, there is no correlation between sewer type and damage amount.
0
5
10
15
20
25
30
35
40
45
50
0 0.5 1 1.5 2 2.5 3 3.5
US
D (
mill
ion
s)
Precipitation Threshold (Inches)
Average Damages per ThresholdDISASTERS
Separate Sewer
Combined Sewer
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.5 1 1.5 2 2.5 3 3.5
US
D (
mill
ion
s)
Precipitation Threshold (Inches)
Average Damages per ThresholdSTORMWATER MANAGEMENT
Separate Sewer
Combined Sewer
18
http://graham.umich.edu/climate Last updated: 9/25/2015
Discussion
There are a number of factors that can help predict when a city will have a flash flood and how severe that flash
flood will be. Factors that appear to strongly impact flooding include population density and sensitivity to
flooding at high precipitation thresholds. As population density increases, frequency of flooding increases,
likely due to more impervious surface in large cities. Also, as cities receive over 2 inches of precipitation in one
day, there is a reasonable chance a flash flood will occur, regardless of sewer type. Using the GHCN-Daily data
to analyze regional sensitivity to frequency of flash flooding following heavy precipitation proved an effective
source for data analysis.
On the other hand, using the GHCN-Daily data to quantify potential increases in the capacity of separate sewer
systems is uncertain. There is no significant difference in the amount of precipitation an average combined or
separate sewer city receives, yet combined sewers have a higher percent difference of flash floods by threshold
for nearly all precipitation thresholds. Additionally, after more than 2.25 inches of precipitation in one day,
combined sewers were more sensitive than separate sewers. However, these results are impacted by population
density, which shows that sewer type probably does not play the most important role in flooding frequency.
The combined sewer cities in this report have an average population density of 7,131 people per square mile,
versus 3,047 people per square mile in the average separate sewer city. Many combined sewer cities have a
high population density and flood more often due to higher impermeable surfaces that cause more runoff and
are more likely to overwhelm a sewer system. Cities with separate sewers may not have as many events on
average because they typically have lower population density and cover a smaller area. Lower population
density cities have less impervious surface and allow more rainwater to enter the soil instead of a sewer system.
There were also some variables found that did not show correlation with flash floods. Damages had no
connection between sewer type or precipitation threshold. Also, sensitivities generated with multi day
precipitation did not yield a strong correlation to sewer type. One day sensitivities from floods caused by less
than 2.25 inches of precipitation also did not differ much between combined and separate average sensitivities.
In order to be able to help cities determine their individual thresholds for reasonable stormwater management,
there are additional variables that need to be taken into account. These include amount of impervious surface,
age and capacity of infrastructure, topography in and around the city, soil type, and so on. More research needs
to be done to determine how these additional factors impact each city and what each city can do to be the most
prepared for heavy precipitation in the future.
In general, more cities need to be added to this research in order to get accurate statistics about all of the factors
that impact flash flooding occurrences.
Acknowledgments
The authors would like to thank the University of Michigan’s Undergraduate Research Opportunity Program,
for the opportunity to begin this project. The authors would also like to thank the University of Michigan
Climate Center and the Great Lakes Integrated Sciences and Assessments for continuing to supervise and
support this project.
19
http://graham.umich.edu/climate Last updated: 9/25/2015
References Cruce, T., & Yurkovich, E. (2011). Adapting to climate change: A planning guide for state coastal managers–a Great
Lakes supplement. Silver Spring, Maryland: NOAA Office of Ocean and Coastal Resource Management. Retrieved
from http://coastalmanagement.noaa.gov/climate/docs/adaptationgreatlakes.pdf
Flinker, P. (2010). The need to reduce impervious cover to prevent flooding and protect water quality. Ed. Millar, S.
Providence, RI: State of Rhode Island Department of Environmental Management. Retrieved from
http://www.dem.ri.gov/programs/ bpoladm/ suswshed/pdfs/ imperv.pdf
Holton, J. R., Curry, J. A., & Pyle, J. A., (2003). Flooding. In Encyclopedia of Atmospheric Sciences. (Vol:1-6).
Elsevier. Retrieved from http://app.knovel.com/web/toc.v/cid:kpEASV0002/viewerType:toc/root_slug:encyclopedia-
atmospheric/url_slug: encyclopedia-atmospheric
Kelsch, M., Caporali, E., & Lanza, L. G. (2001). Hydrometeorology of flash floods. In Coping with Flash Floods, 19-35.
Retrieved from http://www.sciencedirect.com.proxy.lib. umich.edu/science/article/pii/S1464286702000116
Kunkel et al. (1999). Long-Term Trends in Extreme Precipitation Events over the Conterminous United States and
Canada. American Meteorological Society. 2515-2527. Retrieved from
http://journals.ametsoc.org/doi/pdf/10.1175/1520-442(1999)012%3C2515%3ALTTIEP%3 E2.0.CO%3B2
Melillo, Jerry M., Terese (T.C.) Richmond, and Gary W. Yohe, Eds., 2014: Climate Change Impacts in the United
States: The Third National Climate Assessment. U.S. Global Change Research Program, 841 pp.
doi:10.7930/J0Z31WJ2.
Montz, B. E., & Gruntfest, E. (2002, March). Flash flood mitigation: recommendations for research and
applications. Global Environmental Change Part B: Environmental Hazards, 4(1), 15-22. Retrieved from
http://www.sciencedirect.com.proxy.lib.umich.edu/science/ article/pii/S1464286702000116
Perlman, H. (2015, May 5). Impervious surfaces and urban flooding. US Geological Survey. Retrieved from
http://water.usgs.gov/edu/impervious.html
Storm events database (2015). National Climatic Data Center, National Oceanic and Atmospheric Administration.
Retrieved from https://www.ncdc.noaa.gov/stormevents/
Takle, E. S. (2011, August). Climate change in Iowa - part II. AgMRC Renewable Energy & Climate Change Newsletter.
Retrieved from http://www.agmrc.org/renewable_energy/climate_change_and_agriculture/climate-changes-in-iowa-
part-ii/
Walsh, J., Coauthors (2014). Ch. 2: Our Changing Climate. Climate Change Impacts in the United States: The Third
National Climate Assessment, U.S. Global Change Research Program, 19-67. Retrieved from
http://s3.amazonaws.com/nca2014/low/NCA3_Full_Report_02_Our_Changing_Climate_LowRes.pdf?download=1