ROAD DEICING SALT IMPACTS ON URBAN WATER QUALITY A THESIS SUBMITTED TO THE FACULTY OF THE GRADUATE SCHOOL OF THE UNIVERSITY OF MINNESOTA BY Eric Vladimir Novotny IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF Doctor Of Philosophy September, 2009
205
Embed
ROAD DEICING SALT IMPACTS ON URBAN WATER QUALITY · 2016-05-18 · Chloride retention in urban areas where road salt is applied should cause much concern. Mitigation measures, best
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
5.4 Specific conductance for uncorrected and corrected data . . . . . . . . . 112
5.5 Recorded temperature and specific conductance . . . . . . . . . . . . . . 113
5.6 Specific conductance data and weather data for buoy system. . . . . . . 114
5.7 Modeled and observed water temperature and specific conductance . . . 117
5.8 Temperatures and specific conductance at the surface, at mid-depth and
at the bottom of Tanners Lake . . . . . . . . . . . . . . . . . . . . . . . 118
5.9 Root mean square error between modeled and observed data . . . . . . 119
5.10 Data used as initial conditions for simulations . . . . . . . . . . . . . . . 120
5.11 Time vs. depth plot of model simulation results . . . . . . . . . . . . . 121
5.12 Profiles of density gradients caused by salinity and temperature . . . . . 123
5.13 Daily hypolimnetic eddy diffusion coefficients near the lake bed . . . . . 124
5.14 Results from the case study dissolved oxygen simulation . . . . . . . . . 125
xii
Chapter 1
Overview
1
2
1.1 Problem Definition
In the snow-belt regions of the U.S. and other northern countries deicing agents are
applied to remove snow and ice from roadways in winter in order to increase driving
safety. The primary agent used for this purpose is rock salt consisting mainly of sodium
chloride (NaCl). Other agents in the road salt mixture, such as ferrocyanide, which is
used as an anti-clumping agent, and impurities consisting of trace elements (phosphorus,
sulphur, nitrogen, copper and zinc), can represent up to 5% of the salt weight [1].
In the U.S. annual rock salt use for road de-icing increased from 163,000 tons in
1940 to over 23 million tons in 2005 according to the United Stated Geological Survey
(USGS) mineral yearbooks [2, 3]. In the state of Minnesota annual rock salt purchases
increased during the same time period from 60,000 tons to over 900,000 tons [2](Figure
1.1. Other deicing agents are available (e.g. calcium or magnesium chloride (CaCl2)
or potassium acetate), but because of a large difference in cost NaCl is applied most
frequently [4].
Road salt applications keep roads free of ice for safe winter travel in northern climate
zones, however this practice comes at a cost to the infrastructure and the environment,
especially in urban areas with high road densities. The rock salt applied to the roads
dissolves in the melting snow and ice separating into sodium and chloride ions. The
salt- containing water runs off into streams, lakes or storm sewers or infiltrates into the
soil eventually reaching the groundwater. It affects the chemistry and biota in the soil
and water [5]. The overall scope of my PhD research project was to analyze the fate and
transport of road deicing salt (NaCl) in the environment and the water quality issues
associated with road salt applications in a major metropolitan area (Minneapolis/St.
Paul Twin Cities Metropolitan Area, Minnesota, United States).
1.2 Background Information
1.2.1 Biota
Chloride and sodium levels can influence terrestrial and aquatic biota. Chloride levels
of 1000 mg/l can have lethal or sub-lethal affects on aquatic plants and invertebrates
[6]. Continuous levels as low as 250 mg/L have been shown to be harmful to aquatic life
3
0.0
0.2
0.4
0.6
0.8
1.0
1.2
0
5
10
15
20
25
1930 1940 1950 1960 1970 1980 1990 2000 2010 M
inn
es
ota
Ro
ck
Sa
lt U
se
(M
illi
on
to
ns
)
Un
ite
d S
tate
s R
oa
d S
alt
Us
e (
Mil
lio
n t
on
s)
United States Road Salt Use
Minnesota Rock Salt Use
Figure 1.1: Total amount of sodium chloride (NaCl) used as deicing salt throughoutthe United States and the annual amount of rock salt (NaC) purchased by the state ofMinnesota [2].
4
and to render water non-potable for human consumption [6]. In the state of Minnesota
chloride standards of 860 mg/L for acute exposure events and 230 mg/L for chronic
exposure have been established by the Minnesota Pollution Control Agency (MPCA)
for surface waters designated as important for aquatic life and recreation (Minnesota R.
Ch. 7050 and 7052). The groundwater standard for chloride has been set at 250 mg/L
by the USEPA and is a secondary standard relating to the taste of the water.
Increases in sodium and chloride concentrations have been show to decrease the bio-
diversity in wetland areas and waterways [7, 8]. Wood frog species richness in wetlands
in northwestern and southwestern Ontario have been negatively impacted by increased
stress, increased mortality, and altered development resulting from acute and chronic
exposure to road salts [9]. Fish diversity and richness also decreased as sodium and
chloride concentrations increased along with an increase in impervious surfaces in river
watersheds in the Twin Cities area [10]. Macroinvertebrates on the other hand are not
affected by levels of chloride found in wetlands [11, 12]. Chloride ranks third among
chemical ion species for the regulation of diatom species, and is therefore used by paleo-
limnologists to reconstruct chloride levels in lakes from sediment cores [13, 14].
Microorganisms and bacteria can also be influenced by the salinity of the water.
Microorganisms are classified into four groups based on how they react to salt concen-
trations [15]. The first two groups are classifications for nonhalophiles. Nonhalophiles
can be either salt sensitive or salt tolerant. Salt sensitive bacteria can only grow in
media containing less that 2 percent salt (20,000 mg/L). On the other hand salt tol-
erant bacteria grow best in media containing less that 2 percent salt, but will grow in
media containing more that 2 percent. Halophiles can also be classified in two groups:
facultative and obligate. Facultative halophiles will grow in media containing less that
2 percent salt, but will grow best in media containing more that 2 percent. Obligate
on the other hand can only grow in media containing more that 2 percent growth.
These four groups are also known as nonhalophiles, halotolerant bacteria, halophiles
and extreme halophiles respectively [16].
The presence of salts influences the waters activity (amount of water available for
biological processes). Water activity is defined as the ratio between the vapor pressure
of a solution to the vapor pressure of pure water resulting in a value of 1 for pure
water and 0.98 for seawater (3.5% salinity) [16]. Water flows from regions of low solute
5
concentrations to regions of high solute concentration through osmosis [16].
In most bacteria the cytoplasm of a cell has a higher concentration of solutes than
the surrounding environment resulting in a flow of water into the cell[16]. If the salinity
outside the cell causes the activity of the water to be lower than that of the cytoplasm
inside the cell, the flow will reverse will happen causing dehydration of the bacteria.
The loss of water from the cytoplasm results in the fastest and most lethal consequence
high salinity can have on a bacteria [17]. Most freshwater and marine organisms are
stenohaline, i.e. they can not handle wide fluctuations in salinity, but some organisms
have the ability to withstand increases in salinity from their normal environment with
a reduction in growth rates (halotolerant)[17].
In order to adjust to the decrease in water availability in the media around the cell,
most halotolerant and halophilic bacteria must maintain a cytoplasm with a much lower
salt concentration then their surrounding environment. The bacteria accomplish this
by creating low molecular weight organic compounds that help provide osmotic balance
between the medium and the cell cytoplasm [18]. The low molecular weight solutes
accumulate in the cytoplasm creating a high intracellular concentration allowing for the
flux of water to continue to flow into the cell. These solutes are preferred due to their
protective nature against inactivation, inhibition, and denaturalization of enzymes and
macromolecular structures [17].
The solutes that respond to external osmotic pressures are called osmotica. Osmotica
share three main qualities, they are polar, highly soluble molecules and show only
limited interactions with proteins. Polyols (glyceral, arabitol, mannitol, erythritol)
In the Twin Cities metropolitan area (TCMA) of Minneapolis/St.Paul, Minnesota, USA,
an estimated 317,000 metric tonnes (t) of road salt were used annually for road de-icing
between 2000 and 2005. To determine the annual retention of road salt, a chloride
budget was conducted for a 4150 km2 watershed encompassing the populated areas of
the TCMA. In addition to inflows and outflows in the major rivers of the TCMA, mul-
tiple sources of chloride were examined, but only road salt and wastewater treatment
plant (WWTP) effluents were large enough to be included in the analysis. Accord-
ing to the chloride budget 235,000 t of chloride entered the TCMA annually with the
Mississippi and Minnesota Rivers, and 355,000 t exited through the Mississippi River.
Of the 120,000 t of chloride added annually to the rivers inside the TCMA watershed
boundaries, 87,000 t came from WWTPs and 33,000 t came from road salt. Of the
142,000 t of chloride applied annually in the TCMA watershed as road salt (241,000 t
NaCl), only 23% (33,000 t) were exported through the Mississippi River and 109,000 t
or 77% were retained in the TCMA watershed. Chloride budgets for 10 sub-watersheds
within the TCMA analyzed in a similar way, gave an average chloride retention rate of
72%. The retention is occurring in the soils, surface waters (numerous lakes, wetlands
and ponds) and in the groundwater. Chloride concentrations in many of these urban
water bodies are now considerably higher than the pre-settlement background levels of
less than 3 mg/L with concentrations as high as 2000 mg/L in shallow groundwater
wells. The continued accumulation of chloride in the groundwater and surface waters is
a cause for concern.
21
2.2 Introduction
It has been reported that 21 million metric tonnes (t) of road salt were used in the United
States in 2005 to improve driving safety in the winter [2]. In the seven county Twin
Cities metropolitan area (TCMA) of Minneapolis/St.Paul, Minnesota, an estimated
317,000 t of road salt were used annually for road de-icing between 2000 and 2005 [60].
Road salt (mostly NaCl) is highly soluble in water resulting in the sodium and chloride
ions dissociating from one another when snowmelt occurs. Chloride and sodium ions are
both transported from roads to receiving waters along three pathways: 1) a rapid runoff
pathway from impervious surfaces, 2) a shallow subsurface pathway through the soil, and
3) a deeper and slower pathway through underground aquifers [4]. All three pathways
can result in the retention of sodium and chloride in the surface water and groundwater
of a watershed [61]. In addition to the accumulation that can occur in the surficial and
deep groundwater aquifers, small amounts of chloride and more so sodium could also
be retained through interactions with soils and organic matter [40, 62, 50, 63]. The
major factors influencing retention in soils and groundwater include soil permeability,
vegetation cover, topography, and roadside drainage techniques [64].
The accumulation of sodium and chloride ions in the environment degrades the wa-
ter quality in a watershed [64, 6, 61]. Increased chloride concentrations decrease the
biodiversity of waterways and roadside vegetation [6]. If chloride reaches the groundwa-
ter it can contaminate drinking water supplies [65]. Not only has chloride been shown
to affect organism, it can also increase the transport and bioavailability of heavy metals
such as cadmium, lead, chromium, copper and even mercury in the environment, which
are also harmful to biota [4]. Other secondary consequences of road salt applications
in lakes include the ability to inhibit or delay natural mixing events limiting the oxy-
genation of benthic waters and sediments and facilitating the release of heavy metals,
mercury and phosphorus stored in the sediments [64].
Elevated and/or increasing chloride concentrations attributable to road salt applica-
tions are present in groundwater and surface waters in urban environments in northern
climate regions[66, 67, 68, 69, 22, 23, 70, 38, 1, 71, 5]. Mass balance studies on in-
dividual streams with watershed areas less than 450 km2 indicate that between 27%
and 65% of the road salt applied was retained within the individual urban watersheds
22
[72, 73, 74, 75, 76].
Unlike previous studies, which analyzed small watersheds located in urban environ-
ments, this study examines an entire metropolitan area encompassing a surface area of
4,150 km2 including rural, suburban, and urban communities. By examining an entire
metropolitan area, a better understanding of the total retention and the holistic effect
of road salt applications on groundwater and surface waters can be obtained. This
study also used data collected over an 8-year period. Potential environmental damage
will affect the entire metropolitan area, and policies on road salt applications cannot
be developed by extrapolation from small sub-watersheds. Overall this study draws at-
tention to a developing problem that will affect around 3 Million people in a 4,150-km2
area and provides insight for other major metropolitan areas on the affects of road salt
applications.
The purpose of the study was to examine the spatial and temporal chloride transport
dynamics and to develop a chloride budget for an entire metropolitan area. This analysis
was used to estimate how much of the road salt applied annually is exported from the
watershed by the Mississippi River and how much is retained in the soils, groundwater
and surface waters.
2.3 Methods
2.3.1 Metro area chloride balance
The Minneapolis/St. Paul Twin Cities metropolitan area (TCMA) is an urbanized area
with a population of 2.8 million [77], many watercourses and 950 lakes. Located in the
north central U.S., the TCMA experiences cold climate with an average annual snowfall
of 1420 mm between November and April [78]. The hydrologic drainage system of the
TCMA includes many small streams, lakes and wetlands along with an extensive storm
sewer systems and hundreds of detention and infiltration basins. Under the TCMA
is a system of several aquifers, some of which are used for urban water supply. Two
major rivers, the Minnesota and the Mississippi, flow through the TCMA. The combined
watersheds of these two rivers encompass 4,150 km2 of the seven-county metropolitan
area, and provide the natural boundaries for a control volume to be used in a chloride
mass balance (Figure 2.1).
23
Figure 1: Watershed boundaries (bold gray lines) of the Twin Cities metropolitan area and major
rivers (Mississippi, Minnesota and St. Croix). Numbers label each of the data collection or
sampling points listed in Table 1. Data from unlabeled chloride sampling points were not used in
the budget analysis, but to plot Figure 2.
Figure 2.1: Watershed boundaries (bold gray lines) of the Twin Cities metropolitanarea and major rivers (Mississippi, Minnesota and St. Croix). Numbers identify each ofthe data collection or sampling points listed in Table 2.1. Data from unlabeled chloridesampling points were used in Figure 2.2, but not used in the budget analysis
24
Table 2.1: Names of sampling points and data collection organizations for flow ratesand chloride concentrations used in the budget analysis. Locations are shown in Figure2.1.Number Name Organization
1 05330000 Minnesota River at Jordan U.S. Geological Survey2 05288500 Mississippi River at Anoka U.S. Geological Survey3 05331000 Mississippi River at St. Paul U.S. Geological Survey4 Upper Mississippi River Mile 871.6 Metropolitan Council5 Minnesota River Mile 39.6 Metropolitan Council6 Upper Mississippi River Mile 815.6 Metropolitan Council7 Blue Lake WWTP Metropolitan Council8 Seneca WWTP Metropolitan Council9 Metro WWTP Metropolitan Council10 Eagle Point WWTP Metropolitan Council
The chloride ion, known to be harmful to biota [64, 61] and more conservative in
water than sodium [40, 62, 50, 63], was chosen to develop an annual chloride mass
balance (Eq. 2.1) for the TCMA control volume.
I +Mp +Mnp −O = S(t/yr), (2.1)
Where I is the total annual inflow (t/yr) through the Mississippi River at Anoka and
the Minnesota River at Jordan, Mp and Mnp represent the total annual mass of chloride
added inside the watershed from point sources and non-point sources respectively, O is
the total annual mass of chloride exported through the Mississippi River at Hastings
and S is the annual retention of chloride (t/yr) in the TCMA. Chloride is highly soluble
in water, and has been treated as conservative once it is in solution.
Once the water reaches the Minnesota or Mississippi Rivers storage potential is
limited. It is expected that the mass of chloride entering the TCMA at the inflow
stations will exit at the outflow stations. Likewise, all of the chlorides discharged directly
to the rivers through point sources are expected to reach the outflow station. However,
the hydrological transport from nonpoint sources of chloride is unknown. Non point
sources of chloride can infiltrate into soils, can accumulate in wetlands or lakes, can
travel through storm sewers or can reach the groundwater. Only some of the salt
applied as a non-point source in the TCMA is expected to reach the Mississippi River.
Therefore, chloride can only be retained in the watershed if it comes from a non-point
25
source. This assumption allows for the rearrangement of Eq. 2.1 to (Eq. 2.2).
O − I −Mp = mnp(t/yr) (2.2)
Where mnp = Mnp S. We first calulated mnp and then S knowing Mnp. The value of
mnp represents the amount of chloride from non-point sources that entered the river
system and was flushed out of the watershed system at the Mississippi River outflow
station.
2.3.2 Inflows and outflows
Sodium and chloride concentrations as well as river flow rates were measured by state
and federal agencies at gauges and sampling points on the Mississippi and the Min-
nesota Rivers (Figure 2.1). The Metropolitan Council Environmental Services (MCES)
collected grab samples for chemical analyses at six locations along the Mississippi River
and two locations on the Minnesota River two to five times per month between January
2000 and December 2007. For the same time period, daily average, and monthly average
flow rates were obtained from the United States Geological Survey (USGS).
The watershed area upstream from the grab sampling station at Anoka is 48,900 km2,
while the watershed area upstream from the USGS stream gauging station is 49,500 km2,
a difference of only 1.2%. Therefore, the flow rates and the grab sample concentrations
were used together without adjustment. The USGS St. Paul gauging station is located
38 km upstream from the outflow grab sample location at Hastings. The watershed
areas for these two locations are 95,300 km2 and 95,900 km2 respectively, or a difference
of 0.6%. The only major inflow between these two points is the Metro wastewater
treatment plant (WWTP). The river outflow rate from the TCMA was therefore taken
to be the flow rate at St. Paul plus the outflow from the Metro WWTP. Using the daily
flow data and grab sample concentrations from 2000 to 2007, flow weighted monthly
average chloride concentrations were calculated. Each of these concentrations (mg/L)
were multiplied by the associated mean monthly flow rate (m3/s) to estimate the mean
monthly chloride mass transport rate (t/yr).
26
2.3.3 Chloride sources
Sodium and chloride sources include natural weathering of minerals, natural deposition
from rainfall, processing of agricultural products, industrial production of chemicals
and food, processing in the metal-, paper-, petroleum-, textile-, and dying-industries,
household uses, water softening, and road salt applications of NaCl [79]. Point sources
of chloride in the watershed include WWTP effluents and industrial discharges to the
Mississippi or Minnesota Rivers. Most industrial sources of chloride are connected to
the WWTPs and are included in the point-source discharges from the WWTPs. Non-
point sources of chloride are natural sources (atmospheric deposition, weathering), rural
household septic systems, fertilizer applications, and snowmelt runoff containing road
salt.
Point sources: Wastewater treatment plants (WWTPs). Effluents from
WWTPs are a significant source of chloride from domestic (foods and water softening)
and industrial NaCl uses. Chloride concentrations in effluent grab samples were mea-
sured by the Metropolitan Council (MCES) from June 2007 to June 2008 in two-week
intervals at the four major wastewater treatment plants within the TCMA watershed.
Locations of the WWTPs are shown in Figure 1. The annual average chloride con-
centrations and the annual average flow rates determined from daily WWTP effluent
flow data were used to determine the mean annual rates (t/yr) of chloride input to
the rivers from the four WWTPs. The majority of industries and 2.4 million of the
2.8 million people living in the TCMA contribute wastewater through sanitary sew-
ers to the four-wastewater treatment plants discharging within the boundaries of the
TCMA watershed. The majority of the other 400,000 people are either located outside
the boundaries of the watershed or contribute to one of the three other wastewater
treatment plants that do not discharge within the control volume. Combined sewers in
the TCMA have been reduced to a minimum. Therefore, all household and industrial
sources of chloride within the control volume boundaries in the TCMA are included in
the effluent values from the four major wastewater treatment plants.
Non-point sources: Natural sources. Natural sources of chloride and sodium,
in general, include mineral salt deposits, weathering of geological formations, and wet
deposition from ocean evaporation[3]. Mineral salt deposits and geological sources of
chloride are negligible in the TCMA. The annual average concentration of chloride in
27
rainwater was measured at a site in the northern part of the TCMA from 2000 to 2007
to be 0.07 mg/L [80]. This value combined with an annual average rainfall of 747 mm
was used to estimate natural source loads of chloride.
Septic systems of rural households. In 1997 around 70,000 households in the 7
county TCMA were using a septic system [81]. Of those 70,000 household around 60%
were located outside of the watershed boundaries. This population estimation was used
to determined chloride loads from private septic systems.
Agricultural sources of chloride. Farming in the outskirts of the TCMA is lim-
ited to 19% (77,000 ha) of the total watershed area. This percentage was calculated
using land use data from 2005 provided by the Metropolitan Council. For the produc-
tion of corn 36 kg/ha of potassium chloride are required according to the Minnesota
Department of Agriculture [82] A value of 49.9 kg/ha per year was used in a study of
fertilizer and manure contributions to streams in Sweden [5].
Road salt. The total mass of road salt (NaCl) applied annually (2000 to 2005 aver-
age) to roads and parking lots within the seven-county TCMA by government agencies
and commercial/private users was estimated to be 317,000 t/yr [60]. 241,000 or 76%
came from public applications (i.e. city, county and state agencies), the other 24% or
76,000 t was estimated to come from commercial/private applications. The commercial
percentage provided by the American Salt Institute for the years 2005 and 2006 was
based on market share information for road salt purchases. Commercial/private appli-
cations were determined to be the amount of bulk salt purchased by non-government
agencies plus the amount of package deicing salt purchased by both homeowners and
commercial applicators.
The TCMA watershed boundaries do not coincide with the political TCMA bound-
aries (Figure 2.1). For that reason the amount of road salt applied within the political
seven county TCMA had to be adjusted using road kilometers. City, county and state
road data were obtained from the GIS database of the Metropolitan Council. Road
lengths were divided by government entity into total kilometers of city roads for each
city, of county roads for each county and state road. Fractions (percent) of road lengths
inside the watershed vs. the total length inside a municipality, county, or state jurisdic-
tion were multiplied by the total mass of road salt applied by each individual agency
to estimate the total mass of salt applied in the watershed. The total mass of road salt
28
applied by private and commercial uses on parking lots, sidewalks etc. was added to the
public road salt applications by using a value equal to 24% of the total salt applications.
2.3.4 Chloride balances in sub-watersheds
The chloride budget study for the entire TCMA was supplemented by a chloride budget
study for 10 sub-watersheds of small streams located entirely within the TCMA. This
study was done to determine if salt retention rates obtained at a geographic scale of less
than on tenth the TCMA were comparable to those obtained for the entire TCMA.
The analysis of the sub-watersheds was based on Eq. 2.1. The inflow (I) for the
sub-watersheds was based on the estimated chloride concentration in the streams that
would be expected if road salts were not applied in the watershed. This value is different
from the overall TCMA study and was defined as the background concentration in the
stream. The background concentration was estimated from a linear relationship between
the annual average chloride concentrations in the streams vs. the mass of chloride from
road salt applied per ha per year within the watersheds. The background concentration
was found by extrapolating the linear relationship to a chloride application rate of zero.
The background concentration was multiplied by the annual average discharge (flow)
from the watershed to obtain the inflow loading (I).
The only internal chloride source (M) in the watershed was from road salt appli-
cations. No WWTP effluents or other chloride sources besides road salts were being
applied in the sub-watersheds. The mass of road salt applied to the roads was deter-
mined using the methods described for the TCMA watershed analysis explained above.
The percentage of the watershed covered by impervious surfaces and the total watershed
areas were found using GIS data from the University of Minnesota Remote Sensing and
To calculate the annual export rate of chloride mass (O) exiting each sub-watershed,
daily average flow data and grab sample data, collected by the Metropolitan Council
Environmental Services (MCES) two to five times per month between 1/1/2000 and
12/31/2007, at the outflow from each sub-watershed were used.
29
2.4 Results and Discussion
2.4.1 Inflows and outflows of chloride in the major rivers
The hydrologic transport of sodium and chloride in the TCMA was inferred from the
concentrations in its major rivers. Average annual concentrations measured in these
rivers show significant changes with distance through the TCMA (Figure 2.2). Increased
Na+ and Cl− concentrations in the Mississippi River were most pronounced downstream
from the confluence with the Minnesota River and the Metro WWTP where mean annual
chloride concentrations increased from 16 mg/L to 33 mg/L. The Minnesota River
arrived in the TCMA with higher concentrations than the Mississippi River. Discharges
into the Minnesota River include effluents from the Blue Lake and the Seneca WWTPs
as well as surface runoff from populated areas, resulting in a mean annual chloride
concentration increase from 30 mg/L to 42 mg/L. The St. Croix River is outside the
TCMA and carries much lower chloride concentrations ( 5 mg/L) because the watershed
is largely undeveloped. Inflow from the St. Croix River caused a decrease in chloride
concentrations in the Mississippi River downstream from the TCMA. In all three rivers
sodium and chloride follow similar concentration distributions with distance.
!"
#!"
$!"
%!"
&!"
'!"
()!")!!")$!")&!")*!"))!"
!"#$%#&'(&)"#*+,-./0*
122%'*3)44)44)22)*5)6%'*3)7%*3('8%'4*
3)44)44)22)*5)6%'*
+,--./012"3,4.5"
+.150"6678" 91:";50,<"3,4.5"
!"
#!"
$!"
%!"
&!"
'!"
!"$!"&!"*!")!"
!"#$%#&'(&)"#**+,-./0*
9)4&(#$%*:'",*3)44)44)22)*5)6%'*+;,0*
;=>05,?." 90?,@A"
3)##%4"&(*5)6%'*
B>@."C2D."6678"9.-.E2"6678"
Figure 2: Median chloride concentrations (2000-2008) of sodium and chloride in grab samples (2000-2007)
from the two major rivers of the Twin Cities Metropolitan Area. Arrows denote locations of WWTP
discharges or river junctions. Figure 2.2: Median concentrations of sodium and chloride in grab samples (2000-2007)from the two major rivers of the Twin Cities Metropolitan Area. Arrows denote loca-tions of WWTP discharges or river junctions.
Where the Mississippi enters the TCMA at Anoka mean monthly flow weighted
chloride concentrations were between 15 and 20 mg/L (Figure 2.3). Individual grab
30
samples showed only small variations for a particular month with the exception of Jan-
uary, February and November. In the Minnesota River inflow to the TCMA at Jordan,
mean monthly chloride concentrations ranged from 20 to 40mg/L (Figure 2.3). Flow
weighted mean monthly concentrations were highest between December and February
and lowest from March to August. Variability between individual grab samples for a
given month were highest in February.
!"
#!!"
$!!"
%!!"
&!!"
'!!!"
!"
#!"
$!"
%!"
&!"
'!!"
!" #" $" %" &" '!" '#"
!"#$%&'()%*+
,-./%
01"#&23)%*+4-5/%
6#7(1%
()*+,-*"./01*02+,23/0"
4+,5"6,789*:"
;/02<=9">9/?"+,2*"
62..2..2882%92:)&%;7<"#$%
!"
#!!"
$!!"
%!!"
&!!"
'!!!"
!"
#!"
$!"
%!"
&!"
'!!"
!" #" $" %" &" '!" '#"
!"#$%&'()%*+
,-./%
01"#&23)%*+4-5/%
6#7(1%
6277).#('%92:)&%;7<"#$%
!"
#!!"
$!!"
%!!"
&!!"
'!!!"
!"
#!"
$!"
%!"
&!"
'!!"
!" #" $" %" &" '!" '#"
!"#$%&'()%*+
,-./%
01"#&23)%*+4-5/%
6#7(1%
62..2..2882%92:)&%%
=>(<"#$%
Figure 3: Flow-weighted average monthly chloride concentrations and flow rates at the inflow and
outflow of major rivers in the TCMA watershed (2000 - 2007). Figure 2.3: Flow-weighted average monthly chloride concentrations and flow rates atthe inflow and outflow of the major rivers in the TCMA watershed (2000 - 2007).
At the Mississippi River outflow station in Hastings mean monthly concentrations
ranged between 20 and 50 mg/L with a strong seasonal variation (Figure 2.3). The
high concentrations occurred between January and March, and the lows between April
to July. Individual grab sample concentrations varied the most in March, and the least
in June.
Estimates were made for the total mass of chloride passing through the inflow and
31
outflow observation points in Figure 2.3 for every month. The annual mass (rates)
of chloride entering the TCMA by the Minnesota and Mississippi River inflows were
determined to be 119,000 and 116,000 t/yr, respectively. The mass (rate) of chloride
flowing out of the TCMA watershed with the Mississippi River was found to be 355,000
t/yr. Roughly 50% more chloride was found to be exiting the watershed than what was
entering with the Mississippi and Minnesota Rivers combined.
2.4.2 Metro area chloride sources
Point sources: The only point source of chloride in the TCMA is the wastewater
discharged from four wastewater treatment plants 2.4. Effluent chloride concentrations
increased during the winter months at the Metro WWTP. Domestic and industrial
waste loads to the wastewater treatment system were expected to remain fairly con-
stant throughout the year, as was shown for the other three WWTPs. Therefore, the
increase at Metro during the winter was attributed to the addition of road salt to the
system through car washes, a small number of combined sewers and possible seepage
into the sanitary sewer system. To avoid double counting road salt inputs, the average
Cl− concentration between June and November was used for the entire year in the an-
nual chloride budget. Although the other three WWTPs did not display a significant
concentration increase in winter, the same procedure was used for consistency. The
(June 2007-Nov 2007) average chloride concentrations in the WWTP effluents and the
annual average effluent flow rates are shown in Table 2.2. The associated mass (rate)
of chloride entering the river system from the four WWTPs, including domestic and in-
dustrial wastewater, but excluding chloride from road salt applications, was estimated
to be 87,000 t/yr.
Non-point sources: Non-point chloride loads from natural sources, septic systems,
agricultural sources and road salt were evaluated. Using recorded rainfall amounts
and measured concentrations of chloride in precipitation in the TCMA, the chloride
contribution from natural sources was estimated to be around 220 t/yr.
The majority of people within the TCMA watershed boundaries are connected to one
of 7 WWTPs by sanitary sewers and a majority of the households using a private septic
system are located outside the watershed boundaries. If as many as 60,000 people,
i.e. the number of people connected to the Eagle Point WWTP, were using septic
32
!"#$
%##$
%"#$
&##$
&"#$
'##$
'"#$
'(%(#)$ )(&(#)$ !#(%(#)$ !(!(#*$ '(%(#*$ )(%(#*$
!"#$%&'()*+,-./)
+,-.$/01.$ 203,.$45678$ 9.8:5$ ;.7.<0$
Figure 4: Chloride concentrations from grab samples from the effluents
of the four WWTPs. Dates are MM/DD/YY. Figure 2.4: Grab sample chloride concentrations from the effluents of the four WWTPs.Dates are MM/DD/YY.
Table 2.2: Estimates of average chloride concentrations, flow rates and total mass of Cl-in effluents from major WWTPs in the TCMA each year. Average chloride concentra-tions are flow weighted and for the period June 2006 to June 2007 excluding the wintermonths. Average flow rates are from 2000 to 2007.Name of WWTP Chloride (mg/L) Flow (m3/s) Mass (t/yr)Metro 227 8.63 62000Blue Lake 387 1.18 14000Seneca 280 1.01 9000Eagle Point 348 0.18 2000Total 11.00 87000
33
systems within watershed boundaries the 2000 t of chloride discharged annually from
this WWTP would be a reasonable estimate of chloride releases from septic tanks (Table
2.2). For agricultural loads, using a value of 49.9 kg/ha of chloride from fertilizer and
manure results in only 3,800 t of chloride added to the watershed. If all of the designated
agricultural land were instead only used to grow corn this value would only be 2,800
t/yr of chloride.
The largest non-point source of chloride in the TCMA was road salt. It was deter-
mined that 241,000 t of the 317,000 t of road salt applied annually in the seven county
metropolitan area was applied within the watershed boundaries shown in Figure 2.1.
This translates to a non-point source input of 142,000 t/yr of chloride from road salt
applications. It was determined that natural deposition, agricultural inputs and septic
sewer systems would only contribute an additional 1-3%, depending on the calcula-
tion method, to the total chloride load (from point and non-point sources). Therefore
the loads from these sources were neglected, leaving road salt applications as the only
significant non-point source of chloride within the watersheds boundaries.
2.4.3 Metro area chloride balance calculation
The individual chloride balance components (Mississippi River inflow, Minnesota River
inflow, Mississippi River outflow, road salt application, WWTP effluents) presented in
the previous sections were combined in a chloride mass balance using equation (2.1) on
a monthly timescale.
Mean monthly mass fluxes (t/month) of chloride entering or exiting the TCMA wa-
tershed from the two major rivers and the four WWTPs are illustrated in Figure 2.5A.
Dashed lines in Figure 2.5A give cumulative contributions by month from WWTP efflu-
ents, the Minnesota River and the Mississippi River, up to the total Inflow + WWTP
curve. The Outflow curve is from the Mississippi River outflow station in Hastings. Fig-
ure 2.5A is a graphical representation of Equation 2.2, showing the difference between
the mass of chloride exiting the watershed (O) and the sum of the amount entering the
watershed (I) and the amount added by point sources (Mp). Figure 2.5B represents the
amount of chloride added to the river system from non-point sources (mnp in Eq 2.2.).
Since road salt is the only significant non-point source of chloride inside the watershed
Figure 2.5B also represents the monthly mass of chloride from road salt applications
34
exported by the Mississippi River from the TCMA watershed. The monthly non-point
source chloride contribution to the Mississippi River outflow was highest (Figure 2.5B)
during the winter months when road salt was being applied to the roads. The high De-
cember to April values can be interpreted as the direct impact of road salt applications
and snowmelt water runoff through systems of storm sewers and small streams to the
big rivers. Delays occur because winter is a low flow season, and only when snowmelt
sets in does the routing processes accelerate. The small contributions in late summer
can be attributed to the flushing of chloride from lakes and wetlands [71] or to delayed
transport to the rivers through interflow or groundwater flow [4].
!"
#!!!!"
$!!!!"
%!!!!"
&!!!!"
'!!!!"
(!!!!"
)*+"
,-."
/*0"
120"
/*3"
)4+-"
)453"
146"
7-28"
9:8"
;<="
>-="
/*??"<@":A5<0BC-"D8E30F"
948@5<G"
H+@5<G"I"JJKL"
JJKL"
/B++-?<8*"
MB=-0"
/B??B??B22B"
MB=-0"
!"
%!!!"
(!!!"
N!!!"
#$!!!"
#'!!!"
)*+"
,-."
/*0"
120"
/*3"
)4+-"
)453"
146"
7-28"
9:8"
;<="
>-="
/*??"<@"OA5<0BC-"D8E30F"
!"#"
Figure 5: (A) Monthly rates (mass) of chloride fluxes (t/yr) from point sources entering or exiting the Twin cities metropolitan
area watershed. Shaded areas between dashed lines give monthly contributions from four wastewater treatment plant (WWTP)
effluents, the Minnesota River and the Mississippi. Values are additive up to the total "Inflow + WWTP" curve. The
"Outflow" curve is from the Mississippi River outflow station in Hastings, downstream from the Twin Cities metropolitan
area. (B) Monthly differences between the “Outflow” and the “Inflow + WWTP” (solid lines) from panel A. Panel B gives the
algebraic sum of monthly inflow, outflow and point source loads of chloride to the river. By virtue of the mass balance in Eq.
(1) the plotted values also give the monthly amounts of non-point source chloride exported by the river as well as the
uncertainties in the mass balance.
Figure 2.5: (A) Monthly chloride fluxes (t/yr) from point sources entering or exitingthe Twin Cities metropolitan area watershed. Shaded areas between dashed lines givemonthly contributions from four WWTP effluents, the Minnesota River and the Mis-sissippi. Values are additive up to the total ”Inflow + WWTP” curve. The ”Outflow”curve is from the Mississippi River outflow station in Hastings, downstream from theTwin Cities metropolitan area. (B) Monthly differences between the Outflow and theInflow + WWTP (solid lines) from panel A. By virtue of the mass balance in Eq.(2.2)the plotted values give the monthly amounts of non-point source chloride exported bythe river as well as the uncertainties in the mass balance.
In March the difference between the mass of chloride exported and the mass im-
ported had a maximum. Road salt applied in the TCMA watershed accounted for
34% of the chloride passing the Mississippi River outflow station in March, WWTPs
contributed 19%, and the inflows from the Mississippi and Minnesota Rivers into the
TCMA watershed provided 47% (Figure 2.5A). The total amount of chloride exported
from the system during this month was 13,000 t (Figure 2.5B). By adding the values
35
for all months in Figure 2.5B, the annual mass of chloride from road salt exiting the
control volume (the TCMA watershed) was found to be 33,000 t/yr. It had previously
been estimated that 142,000 t of chloride/yr were applied as road salt to the TCMA
watershed area. If 33,000 of the 142,000 t were carried away by the Mississippi then
109,000 t or 77% of the chloride applied annually had to stay behind in the TCMA
watershed system (S in Eqs 2.1 and 2.2).
2.4.4 Sub-watershed chloride balance calculation
The average annual chloride concentrations in the 10 small streams within the larger
TCMA watershed ranged from 37 mg/L in Carver Creek to 185 mg/L in Shingle Creek;
annual average flow rates were from 0.093 m3/s in Riley Creek to 1.685 m3/s in Min-
nehaha Creek (Table 2.3). The total watershed area, the percentage covered by imper-
vious surfaces and the mass of road salt applied annually were also determined (Table
2.3). Chloride application rates in the 10 sub-watershed ranged from 0.08 to 0.82 t/ha
per year.
36T
able
2.3:
Smal
lstr
eam
wat
ersh
edin
form
atio
nan
dro
adsa
ltap
plic
atio
nra
tes
wit
hin
each
wat
ersh
ed.
Ave
rage
conc
en-
trat
ions
and
aver
age
flow
rate
sar
efr
om20
00to
2007
.M
ass
(rat
e)of
chlo
ride
expo
rted
from
each
stre
amw
ater
shed
int/
yran
das
ape
rcen
tage
ofth
ech
lori
deap
plie
das
road
salt
inth
ew
ater
shed
.Im
per
vio
us
An
nu
alA
nnu
alT
otal
Cl-
Cl-
app
lied
Mas
sC
l-C
l-A
rea
Su
rfac
eA
vera
geF
low
Ave
rage
[Cl-
]ap
plied
per
area
exp
orte
dre
tain
edC
reek
Nam
e(h
a)(%
)(m
3/s)
(mg/
L)
(t/y
r)(t
/yr
ha)
(t/y
r)(%
)B
asse
tt11
,100
340.
9713
88,
100
0.73
3,60
056
Bat
tle
3,00
032
0.22
147
2,40
00.
890
063
Blu
ff2,
300
110.
1165
600
0.26
200
67C
arve
r21
,600
40.
9837
1,80
00.
0860
067
Cre
dit
Riv
er13
,300
90.
544
1,80
00.
1440
078
Fis
h1,
300
270.
0910
060
00.
4620
067
Min
neha
ha46
,100
151.
6968
17,7
000.
382,
600
85N
ine
Mile
9,90
029
0.71
744,
700
0.47
1,20
074
Rile
y3,
400
180.
1152
600
0.18
100
83Sh
ingl
e10
,800
350.
4918
57,
000
0.65
2,60
063
Tot
al45
,300
12,5
0072
37
A background concentration of 18.6 mg/L (+/- 34.9 mg/L at the 95% confidence
interval) was determined using a linear regression analysis between annual average Cl-
concentrations in the stream (mg/L) and total Cl- applied (t/yr) per watershed area
(Figure 2.6; R2 = 0.79). The background concentration of 18.6 mg/L matched with
concentrations in the Mississippi River before it entered the TCMA. Retention of chlo-
ride in each individual sub-watershed was calculated to be from 55% to 83% (Table 2.3).
The total combined mass of chloride applied in the 10 sub- watersheds was 45,300 t/yr,
which represents 32% of the amount applied in the entire TCMA watershed. The total
amount of chloride exported from the 10 sub-watersheds was estimated at 12,500 t/yr.
This means that only 28% of the salt applied was exported from the 10 sub-watersheds
each year, and therefore 72% was retained. Due to the large confidence interval around
the background concentrations an analysis was conducted by setting the background
concentration to 0 mg/L. This represents a scenario where all of the chloride exiting the
watershed in a given year is from road salt applications during that year. This analysis
resulted in the lowest possible retention rate. If the background concentration was set
to 0 mg/L the amount of road salt exiting the watershed would be raised to 15,900 t/yr
reducing the retention rate to 65%. The retention estimates of 72% and 65% are lower
than the 77% value obtained for the entire TCMA watershed, but comparable.
C83>"48"J4<L759C2>"<92<32?54?192"12"?F3"8C>I4?358F3>"JC>73?"424A!818'"Figure 2.6: Average annual chloride concentrations vs amount of chloride applied perwatershed area for the 10 sub-watershed streams. Intercept was used as backgroundconcentration in the sub-watershed chloride budget analysis
38
2.4.5 Sodium retention
The retention of chloride from road salt applications in a watershed encompassing the
Twin Cities metropolitan area was determined to be around 77%. While an analysis was
not conducted on the other ion in rock salt, sodium, due to its slightly less conservative
behavior a value equal to or higher than 77% would be expected. Sodium interacts
more readily with soils through ion exchange allowing for the possibility that higher
amounts could be stored in the soil column [40, 50]. The assessment of chloride (road
salt) retention was made with the best information available, however assumptions,
which had to be made, do influence the results obtained. The sensitivity of the findings
to the assumptions was therefore investigated.
2.4.6 Sensitivity of the results
The assessment of chloride (road salt) retention was made with the best information
available, however assumptions, which had to be made, do influence the results ob-
tained. The sensitivity of the findings to three assumptions and procedures used was
therefore investigated: the sampling frequency of the grab samples at the inflow and
outflow stations, the method of calculations for chloride exported from the watershed,
and the estimations method for the non-point sources of chloride.
Sampling Frequency
An analysis was conducted to determine if continuous monitoring or more frequent
sampling was needed to capture chloride concentrations in snowmelt events of short
duration at the inflow and outflow stations on the Mississippi and Minnesota Rivers
[75]. To test the sensitivity to sampling frequency, records of daily average specific
conductance in the Mississippi River near Hastings (outflow station) were used from a
continuous monitoring station maintained by the Metropolitan Council (Figure 2.7A).
Although a direct relationship between chloride and specific conductance could not be
obtained, due to the dampening of the chloride signal in relation to other ions from the
high flow rates, snowmelt events could be clearly detected by fluctuations in specific
conductance during the winter. A Comparison was conducted between the continuous
time series of daily specific conductance values and the specific conductance and chloride
39
values recorded on the grab sample dates. This analysis provided evidence that a suit-
able representation of the chloride dynamics was obtained with the biweekly sampling
frequency used (Figure 2.7A). Furthermore, the cumulative distributions of specific con-
ductance values obtained from the continuous daily time series and the values on the
days when chloride grab samples were taken were virtually identical (Figure 2.7B). It
was concluded that the sampling frequency used was adequate to estimate the annual
load of Cl- (salt) exiting or entering the TCMA watershed over the study period.
6378$69.76&'1./$&2'37/0'43%'8.02'?@&7'.'6@/3%"8&'#%.A'2.:5/&'?.2'9.B&7>'I@&'8.2@&8'/"7&'$2&2'.//'8."/0'1./$&2'2@3?7'"7'+>'Figure 2.7: (A) Daily averages of specific conductance from 15-minute continuous mon-itoring (solid line). Days when chloride grab samples were taken (top). Chloride con-centrations from grab samples (bottom). Data are from the Mississippi River outflowstation in Hastings. (B) Cumulative normalized distribution functions of daily specificconductance values. Specific conductance values only for days when a chloride grabsample was taken (solid line) and all daily values shown from panel A (dashed line).
Chloride export
Other pathways of chloride export such as airborne transport of salt particles or chlori-
nation of natural organic matter were not analyzed. Salt particles can become airborne
behind vehicles and during strong winds, however they are typically deposited within
100 m of the road [83]. Cl- has also been found to interact with natural organic matter
forming chlorinated organic matter when transported through soils [84]. It is therefore
possible that some of the Cl− from road salt applied in the TCMA was exported in the
Mississippi River while attached to organic matter and is thus not accounted for. Our
40
budget analysis did not include airborne or organic matter transport mechanisms. If
incorrect, this assumption would cause an overestimation of chloride retention in the
watershed. If as much as 10% of the road salt applied in the TCMA were exported
by alternative and not included transport mechanisms, the total export would rise to
47,800 t/yr from 33,000 t/yr, and the road salt retention estimate for the TCMA would
be lowered to 64%.
Non-point chloride sources in the TCMA
Only road salt was considered as a significant non-point source of chloride in the water-
shed. Others sources, such as natural deposition, septic tank seepage from residences
not connected to a WWTP or fertilizer applications, were assumed negligible. If an
additional 3% (4,500 t) of chloride were added to the system by outside processes, the
estimated Cl- retention in the watershed would increase slightly.
The 24% commercial road salt application rate was taken from market share infor-
mation provided by the American Salt Institute for the entire United States. While
this value is based on national statistics it is the best estimate for a large metropolitan
area for a number of reasons. The diversity in area including suburban, urban and rural
land uses and the size of the watershed studied allow for a comparison with the national
scale. National trends in municipal salt purchases match Minnesota trends in rock salt
purchases [60]. Finally, obtaining information from commercial road salt appliers and
estimating application rates or areas where road salt was applied commercially is very
difficult. The results from this analysis would likely include more error than accurate
national sales data. If the 24% commercial road salt application rate were lowered to
10%, a value reported in a Canadian study[6], the chloride contribution from road salt
applications would be reduced to 122,000 t/yr, the water-borne export rate from the
TCMA would be raised to 27%, and the retention rate in the watershed would only
drop to 73%.
If all effects reducing the retention rate were combined (commercial application
rate reduced from 24 to 10% of total application rate; increased export of 10% of the
applied amount) the application rate would be lowered to 122,000 t/yr, and the export
rate would be raised to 33,000 + 12,200 = 45,200 t/yr. The total export rate would
then be 37% and the retention rate 63%. In other words, even with extremely favorable
41
assumptions for road salt flushing from the TCMA the estimated retention rate remains
high.
2.4.7 Comparison to other metro areas
Other studies have shown significant retention of chlorides in urban watersheds, but
the estimated retention percentages vary significantly. Most studies were conducted in
small, urbanized watersheds ranging in size from 104 to 435 km2. Retention percentages
were found to be 55% in a stream watershed in the greater Toronto, Canada area [73],
50 - 65% in Helsinki, Finland [75], 28 - 45% in Chicago, Illinois, USA [76], 59% in
Rochester, New York, USA [72] and 35% in Boston, Massachusetts, USA [74].
In Toronto, Ontario, Canada the accumulation of salt in the watershed due to deicing
practices has seriously compromised the shallow aquifers [65]. In Waterloo, Ontario
chloride concentrations in the aquifers had not reached equilibrium after 57 years of
road salt applications. It was estimated that on the order of another 100 years will be
required to reach equilibrium concentrations under current conditions [85]. In shallow
aquifers near Chicago, Illinois, chloride concentrations have increased since 1960. 24%
of the wells studied in the 1990s had concentrations above 100 mg/L, when median
values before 1960 were less than 10 mg/L, and 15% of the wells had rate increases
greater than 4 mg/L per year [70]. The accumulation of salt in shallow groundwater
also affects base flow concentrations in streams [1]. Baseline salinity in urban streams
and streams near roadways have been increasing in the northeastern part of the United
States [22, 23], in the Greater Toronto Area, Canada [68] and in Sweden[38, 5].
2.4.8 Chloride retention in the TCMA watershed
Evidence of significant chloride retention within the TCMA watershed is provided in
streams lakes and aquifers. Four streams (Minnehaha Creek, Battle Creek, Shingle
Creek and Nine Mile Creek) in the TCMA are on the MPCAs 2008 list of impaired
waters for chloride [20]. Pulses of very high chloride concentrations occur in these and
other small streams of the TCMA during the winter months (Figure 2.8).
Volume-weighted average chloride concentrations in 38 lakes of the TCMA have
increased from 1984 to 2005 by an average of 1.5 mg/L per year (range of 0.1 to 15 mg/L
42
!"
#!!"
$!!"
%!!"
&!!"
'!!!"
'#!!"
'$!!"
#!!#"#!!("#!!$"#!!)"#!!%"#!!*"#!!&"#!!+"
!"#$%&'()*+,-./)
0&1()2&#()!%((3)
!"
#!!"
$!!"
%!!"
&!!"
'!!!"
'#!!"
'$!!"
#!!#"#!!("#!!$"#!!)"#!!%"#!!*"#!!&"#!!+"
!"#$%&'()*+,-./)
4566#()!%((3)
Figure 8: Chloride concentrations in two streams of the Twin Cities metropolitan area from grab
samples obtained by the Metropolitan Council Environmental Services (MCES). Both streams are
on the 2008 list of chloride impaired waters (Minnesota Pollution Control Agency 2008).
Figure 2.8: Chloride concentrations in two streams of the Twin Cities metropolitanarea from grab samples obtained by the Metropolitan Council Environmental Services(MCES). Both streams are on the 2008 list of chloride impaired waters [20].
per year), following a pattern similar to the mass of road salt purchased by the state
of Minnesota over that same time period [71]. Median concentrations, in the surface
waters of the 38 lakes from 2001-2005, of 87 mg/L (range 31 to 505 mg/L) [71] were
much higher than the estimated presettlement concentrations of 3 mg/L [14]. In urban
lakes of the TCMA chloride concentrations tend to fluctuate seasonally, with maximum
in February or March and minimum in October or November. In the long term, a mean
annual equilibrium concentration is reached when all of the chloride added to a lake
during the winter season is flushed out during the summer and fall season. There was
an indication that smaller lakes with high summer flushing rates have already reached
equilibrium, while larger lakes and lakes with low flushing rates can be expected to have
rising mean annual chloride concentrations for years to come. If salt applications were
stopped completely, the recovery of many urban lakes would take from 10 to 30 years
[86].
Chloride concentrations in surficial sand and gravel aquifers throughout the state of
Minnesota vary substantially with land-use. Median chloride values were of 46 mg/L
were found in urban areas, 17 mg/L in agricultural areas, and 1.2 mg/L in forested area
[69]. 3% of the water samples taken from wells in the TCMA were found to exceed
the USEPA secondary chloride standard of 250 mg/L [69]. In a cross section of the
surficial aquifer in a northwestern subburb of Minneapolis, directly down gradient from
a high traffic roadway, chloride concentrations ranged from 200 mg/L at the watertable
3m below the soil surface, to 590 mg/L at a depth of 13.5m below the soil surface
43
1
10
100
1000
10000
0 20 40 60 80 100 120
Chlo
ride (
mg/L
)
Well Depth (m)
2004 2005
Figure 9: Chloride concentration in wells located throughout the
Twin Cities Metropolitan area in 2004 and 2005. Information was
collected by the Minnesota Pollution Control Agency. Figure 2.9: Chloride concentration in wells located throughout the Twin CitiesMetropolitan area in 2004 and 2005. Information was reported by the Minnesota Pol-lution Control Agency.
[67]. Concentrations of 380 470 mg/L were also measured downgradient from a major
Interstate Highway (I-94) in summer and late fall pointing towards a longterm storage
of road salt in the surficial aquifer [67]. Data collected by the Minnesota Pollution
Control Agency (MPCA) throughout the seven-county TCMA showed that the highest
chloride concentrations in groundwater, up to 2000mg/L, were found in shallow wells
(Figure 2.9).
Elevated chloride concentrations in aquifers may be delayed due to storage in the
subsurface [85, 70]. Shallow wells respond first resulting in some to have already reached
concentrations above state standards. The storage potential of the TCMA aquifer sys-
tem is very large. If road salt applications were completely stopped today, chloride
concentrations in deep wells may continue to increases for many years until subsurface
saline transport has reached equilibrium [85, 70]. Residence times in the aquifers can
be high ranging from tens to hundreds of years in the upper and surficial aquifers to
thousands of years in the deep aquifers. Residence times in lakes are smaller, on the
order of 3 to 14 years, but still provide a means of chloride storage [86]. The TCMA has
hundreds of lakes and wetlands, and hundreds of man-made detention and infiltration
44
basins. Policy has been to delay runoff from rainfall and snowmelt, and to increase infil-
tration by routing storm sewers into these systems. Although very useful in stormwater
management, these practices could be adding to the accumulation of road salt in the wa-
tershed by promoting retention of surface runoff and/or infiltration of the contaminated
water into the groundwater.
2.5 Summary and Conclusions
Road salt is used to increase driving safety in the Twin Cities metropolitan area (TCMA)
of Minnesota. A chloride budget for the TCMA watershed (Figure 2.1) with data from
2000 to 2007 revealed the final destinations of the road salt after it was dissolved in the
snowmelt water. The TCMA watershed analyzed covered an area of 4150 km2 with a
population of 2.8 Million. According to the annual chloride budget 235,000 t of chloride
entered the TCMA annually in the Mississippi and Minnesota Rivers, and 355,000 t
exited through the Mississippi River resulting in 120,000 t being added to the Mississippi
and Minnesota Rivers as they traveled through the TCMA. Of the 120,000 t of chloride
added annually 87,000 t came from the four WWTPs (point sources) and 33,000 t came
from road salt (non-point source). Of the 142,000 t of chloride applied annually in the
TCMA as road salt, only 23% (33,000 t) were exported through the Mississippi River
and 109,000 t or 77% were retained in the TCMA watershed. Chloride budgets for 10
sub-watersheds within the TCMA analyzed in a similar way, gave a retention rate of
72% for road salt.
Evidence of chloride retention is widespread in the TCMA. Four streams in the
TCMA are on the MPCAs 2008 list of Cl- impaired waters, 38 lakes in the TCMA
had a rising mean annual Cl- concentration over a 22-year period [71], and elevated
Cl- concentrations in groundwater have been measured [67, 69] Chloride retention in
urban areas where road salt (NaCl) is applied should cause much concern. Mitigation
measures, best management practices (BMPs) for road salt application and alternatives
to NaCl need to be examined.
Aknowledgements
45
We acknowledge and thank the following individuals and institutions: the Minnesota
Local Road Research Board (LRRB) in cooperation with the Minnesota Department of
Transportation (Mn/DOT) and the University of Minnesota for providing the funding
for this research; the Technical Advisory Panel, lead by Wayne Sandberg of Washington
County, for input and suggestions to our research; Karen Jensen of the Metropolitan
Council Environmental Services, the Metropolitan Council (MCES) and the United
States Geological Survey (USGS) for providing data used in this study; numerous indi-
viduals in cities, counties and Mn/DOT for providing data on road salt applications.
Chapter 3
Increase of urban lake salinity by
road deicing salt
Eric V. Novotny, Dan Murphy and Heinz G. Stefan
St. Anthony Falls Laboratory,
Department of Civil Engineering, University of Minnesota
Over 317,000 tonnes of road salt (NaCl) are applied annually for road de-icing in the
Twin Cities Metropolitan Area (TCMA) of Minnesota. Although road salt is applied to
increase driving safety, this practice influences environmental water quality. Thirteen
lakes in the TCMA were studied over 46 months to determine if and how they respond
to the seasonal applications of road salt. Sodium and chloride concentrations in these
lakes were 10 and 25 times higher, respectively, than in other non-urban lakes in the
region. Seasonal salinity/chloride cycles in the lakes were correlated with road salt
applications: high concentrations in the winter and spring, especially near the bottom
of the lakes, were followed by lower concentrations in the summer and fall due to flushing
of the lakes by rainfall runoff. The seasonal salt storage/flushing rates for individual
lakes were derived from volume weighted average chloride concentration time series.
The rate ranged from 9 to 55% of a lakes minimum salt content. In some of the lakes
studied salt concentrations were high enough to stop spring turnover preventing oxygen
from reaching the benthic sediments. Concentrations above the sediments were also
high enough to induce convective mixing of the saline water into the sediment pore
water. A regional analysis of historical water quality records of 38 lakes in the TCMA
showed increases in lake salinity from 1984 to 2005 that were highly correlated with
the amount of rock salt purchased by the State of Minnesota. Chloride concentrations
in individual lakes were positively correlated with the percent of impervious surfaces
in the watershed and inversely with lake volume. Taken together, the results show a
continuing degradation of the water quality of urban lakes due to application of NaCl
in their watersheds.
48
3.2 Introduction
In the snow-belt regions of the United States de-icing agents are used to increase driving
safety on public roads in the winter. The primary agent used for this purpose is rock
salt consisting mainly of sodium chloride (NaCl). Its cost is considered moderate, and
storage, handling and dispersing on surfaces are relatively easy [4]. Other agents in
the road salt mixture, such as ferrocyanide, which is used as an anti-clumping agent,
and impurities consisting of trace elements (phosphorus, sulphur, nitrogen, copper and
zinc), can represent up to 5% of the salt weight [1].
In the U.S., annual rock salt use for road de-icing increased from 163,000 tons in
1940 to over 23 million tons in 2005 according to the United States Geological Survey
(USGS) mineral yearbooks [2, 3]. In the state of Minnesota annual rock salt purchases
increased during the same time period from 60,000 tons to over 900,000 tons [2]. Other
deicing agents are available (e.g. calcium or magnesium chloride (CaCl2) or potassium
acetate), but because of a large difference in cost, NaCl is applied most frequently [4].
Road salt applications can keep roads free of ice for safe winter travel; however
this practice comes at a cost to the infrastructure and the environment, especially in
urban areas with high road densities. Much of the rock salt applied to the roads is
dissolved in the melting snow and ice. The salt containing water runs off into streams,
lakes or storm sewers or infiltrates into the soil eventually reaching the groundwater
affecting the chemistry and biota in the soil and water [5]. Organisms in streams and
shallow, small lakes and ponds are particularly vulnerable to road salt application and
chloride pollution [6]. Chloride standards of 860 mg/L for acute events and 230 mg/L
for chronic pollution have been established for surface waters designated as important
for aquatic life and recreation by the Minnesota Pollution Control Agency (MPCA)
Minnesota Rules Chapter 7050 and 7052.
Small streams flowing through urban areas are known to exceed the chronic and
acute chloride levels periodically [23, 19]. If a stream receives snowmelt runoff directly
from roadways treated with salt (NaCl), concentrations of sodium and chloride will
spike during the winter months and spring thaw, and decline quickly once the salt
application has stopped [19]. In the state of Minnesota, data are available showing high
concentrations of chloride traveling through streams and storm sewers during the winter
49
months (Figure 3.1). Four streams (Shingle Creek, Nine Mile Creek, Beavens Creek and
Battle Creek) in the Minneapolis/St. Paul Twin Cities Metropolitan Area (TCMA) have
been designated as impaired waters and placed on the Clean Water Act section 303d
Total Maximum Daily Load (TMDL) list [20] because of salt pollution. Other regions
have shown similar elevated chloride concentrations[22, 4] including the northeastern
United States where chloride concentrations in an urban stream were recorded as high
as 5000 mg/L [23]. In Sweden chloride concentrations in runoff from roadways has also
reached 3500 mg/L in the winter compared to values averaging 15.6 mg/L during the
summer [24].
0200400600800
100012001400
2001 2002 2003 2004 2005 2006 2007 2008 2009
Chl
orid
e (m
g/L)
Figure 3.1: Chloride concentrations in Battle Creek (3.5 km from its outlet into Mis-sissippi River). Battle Creek drains portions of East St. Paul (Metropolitan Councildata).
Streams and storm sewers capturing snowmelt water from roadways are likely to
cause seasonal salinity variations in lakes into which they discharge. Road salt applica-
tions have been shown to increase the salinity in lakes near major roadways of urban
watersheds [14, 5, 59, 51, 56]. Road salt applications in rural areas have been found
to affect lakes several hundred meters away [6]. In urban environments where runoff is
collected in storm sewers, the impact could be felt at much larger distances. The ap-
plication of road salt has been show to cause chemical stratifications in small lakes and
ponds strong enough to prolong or prevent lake mixing [87, 88]. In this study multiple
50
lakes throughout a metropolitan area will be examined to determine regional trends and
how different lakes react to road salt applications.
The Minneapolis/St. Paul Twin Cities Metropolitan area (TCMA) in the state of
Minnesota provides a perfect setting to study how lakes are influenced by road salt
applications. The seven county TCMA is an urbanized area with a population of 2.7
million that contains many watercourses and 950 lakes. Located in the north central
U.S., the TCMA experiences cold climate in the winter with a total annual snowfall of
1.4 meters [78], falling between the months of November to March. Due to these climate
conditions an estimated amount of 350,000 short tons (317,000 tonnes) of road salt/year
is used in the TCMA for winter road maintenance [60]. With an expanding population
and road system, more and more lakes in the TCMA are susceptible to pollution from
storm water and snowmelt runoff. This paper focuses on how road salt applications are
changing lake water quality; it has five objectives: 1) develop a relationship between
chloride and specific conductance in lake waters, and to determine if sodium and chloride
are the dominate ions causing observed changes and fluctuations in specific conductiv-
ity; 2) determine if observed concentrations of sodium and chloride in lakes receiving
runoff from major roadside environments are elevated above background concentrations,
and if these elevations can be contributed to road salt applications; 3) understand how
individual lakes are influenced by road salt applications in terms of chemical stratifica-
tion, seasonal salinity cycles and convective transport of NaCl into lake sediments; 4)
examine regional trends in lake chloride concentrations and 5) determine if relationships
exist between these concentrations and watershed/lake parameters.
3.3 Methods
3.3.1 Data collection
Two sets of lake water quality data from the TCMA were analyzed to meet the ob-
jectives of the study. The first set was collected by the authors and the second was a
historical data set. In addition, sediment cores were extracted from two lakes.
Lake data collection (Data Set 1) The first lake data set includes 13 lakes which
51
were selected to meet four criteria: 1) receive runoff from a major highway or road-
way through storm sewers, streams or overland flow; 2) have a maximum depth large
enough so that stratification, i.e. a seasonal thermocline and/or chemocline could form;
3) have previously been monitored by a public agency, e.g. by the Metropolitan Coun-
cil, Minnesota Pollution Control Agency (MPCA) or area watershed district, so that
long term unbiased information is available; 4) have data on bathymetry and water-
shed available. Locations of the selected lakes and their watersheds in the TCMA are
shown in Figure 3.2. Lakes are listed in Table 3.1 along with characteristics of each lake
and its watershed. Bathymetric data were obtained from the Minnesota Department of
Natural Resources and watershed delineations were gathered using GIS data obtained
from the Metropolitan Council. Impervious surface areas in the lakes watersheds in
2002 were obtained using GIS data from the University of Minnesota Remote Sensing
and Geospatial Analysis Laboratory (http://land.umn.edu/index.html).
Bryant
Cedar Isles
Brownie
Parkers
Bass
Medicine SweeneyGervais
McCarron
Johanna
Tanners
LegendMajor Roads
Lakes
Lakesheds´
Ryan
Minneapolis
St Paul
10Kilometers
Figure 3.2: Locations and watersheds of the 13 lakes sampled (Data Set 1) from 2004-2007 in the Minneapolis-St. Paul Twin Cities Metropolitan Area (TCMA).
52T
able
3.1:
Lak
ean
dw
ater
shed
info
rmat
ion
for
the
13la
kes
sam
pled
from
2004
to20
07(D
ata
Set
1).
Sam
plin
gpe
riod
repr
esen
tsth
eye
ars
inw
hich
each
lake
was
sam
pled
.04
/05
refe
renc
esla
kes
sam
pled
betw
een
2/15
/200
4to
4/13
/200
5,06
/07
repr
esen
tsla
kes
sam
pled
betw
een
1/14
/200
6to
11/1
5/20
07an
d04
/07
repr
esen
tsla
kes
sam
pled
duri
ngbo
thti
me
peri
ods.
Max
Su
rfac
eS
urf
ace
Wat
ersh
edW
ater
shed
/P
erce
nt
Sam
plin
gD
epth
Are
aV
olu
me
Are
a/D
epth
Su
rfac
eS
urf
ace
Are
aIm
per
vio
us
Per
iod
(m)
(Ha)
(m3)
(ha/
m)
(ha)
(%)
(yea
r/ye
ar)
Bas
s9.
470
.467
3,00
07.
511
3116
215-
Apr
Bro
wni
e14
.35
200,
000
0.3
136
2733
7-A
prB
ryan
t13
.765
.23,
245,
000
4.8
901
1424
7-Ju
nC
edar
15.5
68.4
4,43
3,00
04.
453
78
287-
Apr
Ger
vais
12.5
94.7
4,82
3,00
07.
611
4412
307-
Jun
Joha
nna
13.1
86.2
4,27
4,00
06.
611
8814
395-
Apr
Isle
s9.
444
.11,
120,
000
4.7
252
629
5-A
prM
cCar
ron
17.4
27.6
2,15
1,00
01.
654
920
247-
Apr
Med
icin
e14
.935
8.6
18,5
89,0
0024
4380
1229
5-A
prP
arke
rs11
.336
.91,
414,
000
3.3
340
927
7-Ju
nR
yan
117.
629
5,00
00.
777
1034
7-A
prSw
eene
y7.
626
.795
2,00
03.
515
1257
377-
Jun
Tan
ners
1428
.31,
848,
000
221
48
337-
Jun
53
These 13 lakes were sampled every 4-6 weeks during two sampling periods. All
of the lakes studied are natural lakes with inflows coming from streams and/or storm
sewers. Two of the lakes have special conditions: Brownie Lake and Sweeney Lake.
Brownie Lake has been known to be meromictic since 1925. It is a very small and
wind sheltered lake. The construction of a channel connecting Brownie Lake to Cedar
Lake and road construction drastically reduced the lakes surface area, resulting in a
low surface area to depth ratio[89, 90]. Road salt, while not the original cause, has
contributed to the current meromictic conditions. Sweeney Lake, on the other hand,
has an aeration system, which makes it an artificially mixed lake. In April 2007 this
system was shut off for the purpose of conducting a phosphorus TMDL study.
From 2/15/2004 to 4/13/2005 eight lakes were sampled: Bass, Cedar, Lake of the
Isles, Johanna, McCarron, Medicine, Ryan and Brownie. From 1/15/2006 to 11/15/2007
four of the previous lakes were sampled (Ryan, Cedar, Brownie and McCarron) and
5 new lakes were added (Tanners, Parkers, Bryant, Gervais, and Sweeney). Parkers
Lake was added on 5/7/2006 and Sweeney Lake, Tanners Lake and Lake Gervais were
added on 8/8/2006. Lakes that did not show strong salinity stratification after the
first sampling period were dropped, and other lakes with a high likelihood of salinity
stratification were added. Lake selection was thus biased towards lakes that would
receive high salinity runoff. Cedar Lake was kept as a reference lake that showed little
stratification.
In total, 173 specific conductivity/temperature profiles were measured in the 13 lakes
at 4- to 6- week intervals over a 46-month period. These measurements were made in
the water column every 0.5 meters at approximately the deepest location in each lake
using an YSI Model 63 probe [91]. Water samples were collected during the winter
(22 February 2007) when sodium and chloride concentrations are highest and in the
fall (15 November 2007) when maximum flushing of the lakes due to summer rainfall
has occurred. Water samples were taken 1 m below the waters surface and 1 m above
the bottom at approximately the deepest location in the 9 lakes sampled from 2006 to
2007. Samples were analyzed for major ion concentrations in the laboratory of Geology
and Geophysics at the University of Minnesota-Twin Cities. Anions were analyzed on a
Dionex ICS-2000 ion chromatography system consisting of an AS19 analytical column,
ASRS Ultra II suppressor, AS40 autosampler, and integrated dual piston pump and
54
conductivity detector. Cation samples were acidified with HCl to a pH of 2 and ana-
lyzed on a Dionex ICS-2000 ion chromatography system consisting of a CS16 analytical
column and guard column, CSRS Ultra suppressor, AS40 autosampler, and integrated
dual piston pump and conductivity detector.
Historical lake data(Data Set 2) The second set of lakes examined for this study
includes lakes sampled by watershed districts, consulting companies, and government
agencies in the TCMA, typically during the ice free months of May November with
limited data available during the winter months. This data set is available on the
MPCA Environmental Data Access website (http://www.pca.state.mn.us/data /eda/)
used to store environmental data from water bodies around the state. Historical chloride
concentrations in 38 TCMA lakes were selected based on the length and continuity of
the data set. Information on individual lakes was gathered from the same sources as
data set 1.
Lake sediment cores In order to examine if the salt water concentrating at the
bottom of the lakes in winter and spring is seeping into the lake sediments, and possibly
into the groundwater, sediment cores were extracted from two lakes. Tanners lake,
which displayed high concentrations of chloride at the bottom during the winter months
and Lake McCarron, which displayed much lower chloride concentrations. These two
lakes were chosen to make a comparison between the two conditions. One 1.2 m long
sediment core was extracted from each lake at the deepest part of the lake during the
winter (2/28/2007). The cores were separated into 4-centimeter sections and the pore
water was extracted using a centrifuge and filtered through a standard 0.045 µm filter.
The 4-centimeter sections allowed for the extraction of enough pore water to be used in
ion analysis. Five samples evenly distributed through the sediment core were analyzed
to give a representation of how ion concentrations are changing with depth.
3.3.2 Data analysis
A relationship was developed between specific conductance and all the major ions in
the water samples to determine which ions were dominant. Chloride concentration
and specific conductance data from different lakes, water depths and times throughout
the year were available (data set 2, in addition to our own water samples taken on
2/22/2007 and 11/15/2007 in 9 lakes) to create a robust relationship between specific
55
conductance and chloride concentration. The relationship was used to convert specific
conductance profiles measured in the lakes to chloride concentration profiles. Chloride
is a conservative ion with minimal natural sources in the TCMA and therefore useful
as an indicator of road salt.
Data set 1 was used to analyze objectives 2 and 3. The chloride vs. depth profile time
series were used to determine stratification patterns as well as seasonal cycles in the 13
monitored lakes. Volume-weighted average concentrations, using the bathymetric data,
were determined for each measured chloride profile in each lake and each measurement
date. For a more comprehensive view of the salinity cycles in the lakes, the volume-
weighted average concentrations for each survey date were normalized. The average
concentrations for each lake from May 2004 to April 2005 or from Sept 2006 to August
2007 were used as the reference for normalization of the 2004/2005 and 2006/2007 data
sets, respectively. This normalization was calculated by dividing the time series by
the reference value. The normalized data sets for all lakes were then combined, i.e.
averaged, for each sampling date to get a seasonal cycle for all the lakes combined.
Volume-weighted average concentrations were also used to determine the seasonal
storage and flushing of salt in the lakes. This normalized flushing rate is equal to
the maximum minus the minimum volume-weighted average chloride concentrations of
a lake for a particular year. This difference in concentrations multiplied by the lake
volume would give an idea of how much salt (tonnes/year) is being flushed through
a particular lake in the course of a year. This amount of salt was also expressed as
a fraction (percentage) by dividing the difference of maximum and minimum average
concentrations by the minimum concentration.
Profiles in specific conductance, temperature and dissolved oxygen were analyzed
for Tanners Lake. This lake was chosen because it changed from monomictic behavior
in 2006 to dimictic behavior in 2007. We sampled Tanners Lake starting on 8/8/2006
taking profiles of temperature and specific conductance. Profiles of dissolved oxygen
for the entire time period and temperature and specific conductance from 5/11/2006
to 7/12/2006 were available from the Minnesota Pollution Control Agency (MPCA)
Environmental Data Access; the measurements were made by the Ramsey-Washington
Watershed District.
Data set 2 was used for objectives 4 and 5. Annual average specific conductivity
56
values for the top 3 meters and annual maximum values were determined for each lake
and date of survey to obtain an estimate of how the salinity of the 38 lakes has been
changing over time. The time series for the top 3 meters between 1984 and 2005 was
used for a trend analysis. Each lake was normalized individually using the average
annual concentrations from 2001-2005 as a reference. Once a time series was developed
for each individual lake a combined time series was calculated by averaging each of these
series together resulting in a single time series of normalized specific conductance for
all 38 lakes. Correlations were also calculated between lake watershed and bathymetric
parameters with chloride concentrations.
3.4 Results
3.4.1 Ionic composition and relationship to specific conductance
The ionic composition in the lake water samples collected in 2007 (Table 3.2) shows
a difference between the winter (February) after some snowmelt water has entered the
lake and the fall (November) after the summer flushing of the lake by rainfall events.
Correlation coefficients between the specific conductance values and the individual ionic
concentrations for chloride, sodium, sulfate, potassium, calcium and magnesium were
0.99, 0.97, -0.09, 0.93, 0.12 and 0.28, respectively.
Table 3.2: Median ionic concentrations of water samples from 9 lakes in data set one.Water samples were taken in February and in November 2007. All concentrations arein mg/L. SC represents specific conductance (µs/cm). Top represents samples taken 1meter below the water surface and bottom represents samples taken 1 meter from thebottom.
Using historical data from numerous other lakes in the TCMA in addition to the
above water sample data, a linear relationship between chloride and specific conductance
(Eq.3.1 with R2 = 0.94) was created (Figure 3.3).
[Cl−] = 0.25 ∗ SC − 37.25, (3.1)
where [Cl-] is the chloride concentration in mg/L and SC is the specific conductance
in µS/cm. This equation was used to convert specific conductance measurements in the
lakes to chloride concentrations.
y = 0.25x - 37.25
0
200
400
600
800
1000
1200
0 1000 2000 3000 4000 5000
Ch
lori
de (
mg
/L)
Specific Conductance (µS/cm)
Parkers McCarron
Cedar Carver
Diamond Gervais
Spring Pond Tanners
Loring pond Powderhorn
Como Battle Creek
Snail Valentine
2/22/2007 11/15/2007
Figure 3.3: Relationship between chloride and specific conductance. Individual lakevalues were obtained from the Minnesota Pollution Control Agency (MPCA) Environ-mental Data Access website and were sampled by government agencies or consultingcompanies in the TCMA. Data points marked 11/15/2007 and 2/22/2007 are watersamples from the 9 lakes monitored between January 2006 and September 2007.
58
3.4.2 Seasonal salinity cycles and salinity stratification
Lake surface waters had often lower specific conductance than lake bottom waters.
To document this stratification we plotted chloride concentrations measured at 0.5 m
depth below the lake surface and at 0.5 m above the lake bottom of each lake. Figure
3.4 documents the measurements taken from the 9 lakes sampled between 2006 and
2007. Volume-weighted average chloride concentrations were also plotted. For the lakes
sampled between 2004 and 2005 plots similar to Figure 3.4 are presented and discussed
in detail in [92]. Salinity stratification can be seen in all 13 lakes studied. The strongest
salinity stratification was found in Brownie Lake, Parkers Lake, Tanners Lake and Ryan
Lake; the least in Bass Lake and Cedar Lake.
0
200
400
Chl
orid
e C
once
ntra
tion
(mg/
L)
0
200
400
1/06 7/06 1/07 7/070
200
400
1/06 7/06 1/07 7/07 1/06 7/06 1/07 7/070 400 800
BottomTopAverage
Sweeney Tanners Ryan
GervaisMcCarronParkers
Cedar Bryant Brownie
Figure 3.4: Chloride concentrations 0.5 m below the surface and 0.5 m above the bottomof a lake, and volume-weighted average chloride concentrations in each lake sampledfrom January 2006 to September 2007. The dark lines on the x-axis represent monthswhen snowfall can be expected to accumulate on the ground (November March).
59
The combined normalized volume-weighted average chloride concentration time se-
ries gives a comprehensive view of seasonal salt accumulation and flushing from a lake
(Figure 3.5). How the numbers were obtained is described in the methods section. Al-
though the data are for two different time periods and lake sets, the results converge
nicely: the highest normalized concentrations occur in the winter - when road salt is
being applied - and the lowest concentrations occur in the summer and fall when fresh
rainwater runoff entered the lakes and flushed some of the salt away, except for the
meromictic Brownie lake where the seasonal salinity dynamics occur only above the
chemocline.
0.7
0.8
0.9
1
1.1
1.2
Jan-04 Jan-05 Jan-06 Jan-07 Jan-08Nor
mal
ized
Vol
ume-
Wei
ghte
d [C
l-]
MeanMedian
Figure 3.5: Seasonal salinity (Cl-) cycles illustrated by normalized (volume-weighted)average chloride concentrations, averaged for all lakes in each time period (Data set 1).Bars represent the standard deviation for the set of lakes.
Seasonal salinity cycles, expressed in terms of the amount of salt passing through a
lake in a year relative to its minimum salt content in that year, were present in all lakes
studied, but were more pronounced in some of the lakes than others. The strength of
the seasonal salinity cycle was calculated and analyzed (Table 3.3). Brownie Lake, Ryan
Lake, Lake of the Isles and Sweeney Lake had the strongest seasonal salinity cycles of all
the lakes studied. The highest annual salt turnover (accumulation and flushing) rates in
three years of observation were 54% and 55% of the minimum salt content in Sweeney
Lake and Lake of the Isles. The lowest annual salt turnover rates were obtained for
Bryant Lake and Parkers Lake; they were 9% and 11%.
60T
able
3.3:
Salin
ity
(Cl-
)cy
cles
inT
CM
Ala
kes.
Per
cent
chan
ge=
((M
ax-M
in)/
Min
)*10
0%Ja
n20
04-
Nov
2004
Jan
2006
-N
ov20
06Ja
n20
07-
Nov
2007
Min
Max
Per
cent
Min
Max
Per
cent
Min
Max
Per
cent
(mg/
L)
(mg/
Lch
ange
(mg/
L)
(mg/
Lch
ange
(mg/
L)
(mg/
Lch
ange
Bas
s76
9424
––
––
––
Isle
s88
135
54–
––
––
–Jo
hann
a11
212
714
––
––
––
Med
icin
e10
112
827
––
––
––
Bro
wni
e27
038
643
279
381
3625
633
332
Ced
ar88
106
2096
109
1410
111
817
McC
arro
n10
212
321
113
132
1712
113
915
Rya
n88
128
4585
103
2192
123
34B
ryan
t–
––
8997
910
011
010
Ger
vais
––
––
––
113
146
29P
arke
rs–
––
––
–14
716
311
Swee
ney
––
––
––
172
266
55T
anne
rs–
––
––
–13
116
727
61
3.4.3 Effects of salinity on seasonal mixing and dissolved oxygen
In addition to the salinity dynamics in the surface and bottom waters of each lake
it is of interest to consider the complete measured temperature, chloride and dissolved
oxygen profiles, and to study chemoclines, thermoclines and dissolved oxygen picnoclines
in relation to each other (Figure 3.6). Tanners Lake, with a chemocline capable of
preventing a full lake turnover in the spring but not in the fall, can be used to illustrate
the seasonal stratification cycles in more detail. Tanners Lake does not appear to be
meromictic, but monomictic at times. During the early spring and summer of 2006,
chemical stratification was present in Tanners Lake. Erosion of the chemocline through
the fall, due to cooling of the lakes surface, allowed for some of the chloride to be flushed
from the lake. During the winter months the chemocline reformed in the deepest portion
of the lake. The highest concentrations of chloride and the thickest layer of saline water
are seen in April. This is significant because the thermocline had already begun to
form. If the spring overturn of the lake were triggered by density differences due to
temperature only, the lake would have fully mixed by April. Similar patterns are seen
in Parkers Lake (Novotny et al. 2007).
Chloride concentration profiles in the other lakes studied (McCarron, Cedar, Bryant,
Ryan, Gervais) indicate that these lakes received salt, but not enough, to prevent full
mixing either in the spring or fall [93]. Inflows of salt water to the bottom of these
lakes during the winter can be inferred from the observed changes in chemical strati-
fication. This chemical stratification is not as strong as in Tanners Lake and Parkers
Lake. Sweeney Lake on the other hand acts more like a well-mixed body of water with
concentrations in the whole lake increasing in the winter and decreasing in the summer
and fall [93]. This behavior is caused by an artificial lake aeration/mixing system.
3.4.4 Salinity in lake sediment cores
To see if the saline water layer at the lake bottom is connected to the pore water in the
lake sediments, sediment cores were extracted from Lake McCarron and Tanners Lake.
In Tanners Lake the saline layer above the bottom of the lake is up to 7 m thick and
has chloride concentrations reaching 400 mg/L compared to maximum concentrations
of 240 mg/L chloride in Lake McCarron. The cores were sectioned and the pore water
62
!
Chloride (mg/l)
0 250
-12
-8
-4
0
De
pth
(m
)
5/1
1/2
00
6
0 250
6/1
4/2
00
6
0 250
7/1
2/2
00
6
0 250
8/8
/20
06
0 250
9/2
5/2
00
6
0 250
11
/8/2
00
6
0 2501
/24
/20
07
0 250
2/2
1/2
00
7
0 250
4/1
/20
07
0 250
5/1
7/2
00
7
0 250
7/1
3/2
00
7
0 250
9/1
7/2
00
7
0 250
11
/16
/20
07
Figure 3.6: Chloride concentration, dissolved oxygen concentration and water tempera-ture vs. depth in Tanners Lakes from May 2006 to Nov 2007. The Ramsey WashingtonWatershed District gathered all dissolved oxygen profiles along with specific conductiv-ity and temperature profiles between May and July 2006. The authors monitored allother data points for specific conductance and temperature between August 2006 andNovember 2007.
63
was extracted and analyzed for major ions. The profiles (Figure 3.7) show sodium
and chloride concentrations decreasing more or less exponentially with depth in the
sediments of both Lake McCarron and Tanners Lake. The chloride profiles in the pore
water start at the sediment surface with concentrations equal to those of the saline
water layer at the bottom of the lake. The concentrations of the other ions in the pore
water appear to stay about constant with depth into the sediment, but sodium and
chloride do not. The concentrations of sodium and chloride start at around 250 and
410 mg/L, respectively, in Tanners Lake and around 113 and 186 mg/l, respectively, in
Lake McCarron. Over the length of the core (1.2 m) these values are reduced to 36 and
78 mg/L in the Tanners Lake core and to 14 and 34 mg/L in the Lake McCarron core,
respectively.
-120
-70
-20
30
80
0 100 200 300 400Concentration (mg/L)
Dis
tanc
e ab
ove
and
belo
w
sedi
men
t wat
er in
terfa
ce (c
m) Tanners Lake
0 100 200 300 400Concentration (mg/L)
ChlorideSodiumCalciumMagnesiumSulfatePotasium
Lake McCarron
Figure 3.7: Ionic composition of pore water in sediment cores extracted from the deepestpoints of Lake McCarron and Tanners Lake. Values of chloride concentrations rightabove the sediments are also shown. Sediment cores were extracted on 2/28/2007.
3.4.5 Salinity trends in TCMA lakes
To determine if long-term changes in lake salinity are occurring, the average annual
specific conductance values of the 38 lakes from 1984 to 2005 (data set 2) were plotted
against time along with the amount of rock salt purchases by the state of Minnesota
from 1930 to 2005 (Figure 3.8). The trends for both the specific conductance (salinity)
of TCMA lakes and for the rock salt purchases by the state are strikingly similar. Both
64
time series show an increase from 1984 to 2005 and a correlation coefficient of 0.72.
Road salt applications vary somewhat from year to year with the number of snowfall
events [60]. To remove this variability, the correlation analysis was repeated with 5-year
running averages of the two time series, resulting in a correlation coefficient of 0.93.
y = 0.018x - 34.794
0.3
0.5
0.7
0.9
1.1
1.3
1930 1940 1950 1960 1970 1980 1990 2000 2010
Ave
rage
Nor
mal
ized
Spe
cific
Con
duct
ance
0
0.2
0.4
0.6
0.8
1
1.2
Min
neso
ta R
ock
Sal
t Use
(Mill
ion
tons
)
TC Metro area lakesMinnesota rock salt use
Figure 3.8: Time series of average normalized specific conductance in 38 Twin CitiesMetro Area lakes (Data Set 2) and total rock salt purchases by the State of Minnesota.
3.4.6 Relationships between lake salinity and watershed characteris-
tics
By collecting lake bathymetry, watershed information and chloride concentrations for
each of the 38 lakes in data set 2 (Table 3.4) correlations can be made with chloride
concentrations (Table 5). Lake surface area and lake depth, and even watershed area,
taken as single independent variables, have a very low correlation with chloride concen-
tration parameters in the lakes. The highest correlation was between lake surface area
65
and the chloride trend with a correlation coefficient of -0.44. Chloride concentrations
correlate better with the percentage of impervious surface areas in the watershed (Table
3.5). The ratio of impervious surface area in the watershed to a proxy for lake volume
(expressed as the product of lake surface area * lake depth) has the strongest correlation
with lake chloride concentrations (Table 3.5).
Table 3.4: Historical average, trend and maxima of chloride concentrations, and bathy-metric and watershed data for 38 TCMA lakes (Data Set 2). Chloride concentrations forthe top 3 meters are annual average value for the period 2001-2005. Trend is based onthe time series of annual average concentrations from 1984 - 2005 for the top 3 metersof each lake and normalized with the average value from 2001-2005 to get a percentchange per year. Annual Max chloride concentration is the average of the maximumconcentration in the lake for each year between 2001 and 2005, measured at any depth.
Years of [Cl-] top 3 Lake Max Watershed Percentdata meters Trend [Cl-] max area Depth area impervious
Table 3.5: Correlation coefficients of chloride concentrations with lake and watershedparameters (Data Set 2). PI (percent impervious), SA (lake surface area), D (lake maxdepth).
Lake watershed PI/ PI *area Depth area Percent (SA*D) D/SA(ha) (m) (ha) Imperv. (1/m3) (1/m)
[Cl-] top 3 m -0.25 -0.29 -0.18 0.67 0.94 0.57[Cl-] ave. annual max -0.28 -0.26 -0.24 0.66 0.79 0.78Trend -0.44 -0.40 -0.18 0.54 0.43 0.26
3.5 Discussion
3.5.1 Comparison of ionic composition with other freshwaters
Are sodium and chloride concentrations measured in the urban lakes really different
from those measured in lakes and rivers outside the metropolitan area? If differences
in measured sodium and chloride concentration can be documented, are there similar
differences in other major ions? Answers to these two questions can be provided by
comparing the results in Table 3.2 for 9 urban lakes to similar measurements in other
water bodies. The comparison with six different natural waters is made in Table 3.6.
The ionic strengths measured in the 9 urban lakes after a summer and partial fall
season, are listed in the last column of Table 3.6 for comparison. It is apparent that
the urban lakes have much higher sodium and chloride concentrations than any of the
other waters, including the Mississippi and Minnesota Rivers which are by no means
pristine even before they enter the TCMA. Ionic strengths of other ions, such as calcium,
magnesium and potassium, sulfate and nitrate in the 9 urban lakes are comparable to
or lower than those in other surface water bodies. It can be concluded that the sodium
and chloride concentrations measured in the urban lakes are not found in other surface
water bodies.
67T
able
3.6:
Ioni
cco
mpo
siti
on(m
g/L
)of
sele
ctiv
esu
rfac
ew
ater
sin
Nor
thA
mer
ica
and
the
nine
lake
sst
udie
din
2006
/200
7.a
From
[55]
.b
From
Lon
gT
erm
Eco
logi
calR
esea
rch
(LT
ER
)si
te(N
orth
Tem
pera
teL
akes
(Tro
ut)
Reg
ion)
.c
Ave
rage
s(2
000-
2007
)fr
omM
etro
polit
anC
ounc
ilda
taba
se.
Loc
atio
nsar
ebe
fore
the
rive
rsen
ter
the
TC
MA
.dA
ver-
ages
from
lake
sin
Min
neso
ta[9
4].
eM
edia
nco
ncen
trat
ions
from
area
lake
sst
udie
ddu
ring
2006
/200
7.Sa
mpl
esta
ken
11/1
5/10
07. aC
onti
nen
tal
bD
ilu
tedM
inn
esot
aaN
orth
cM
issi
ssip
pi
cM
inn
esot
aeL
akes
Rai
nW
isco
nsi
nL
akes
Am
eric
anR
iver
atR
iver
atS
tud
ied
Lak
eR
iver
sA
nok
aJo
rdan
Ca2
+0.
2-
413
2921
5010
342
Mg2
+0.
05-
0.5
316
517
4714
Na+
0.2
-1
26
912
3259
K+
0.1
-0.
5–
31
35
3N
H4+
0.1
-0.
5–
––
––
0.1
SO42
-1
-3
–14
2017
162
13C
l-0.
2-
20.
34
817
3410
9N
O3-
0.4
-1.
3–
–1
0.8
60.
6
68
3.5.2 Indicators of the salinity sources
What is the source of the elevated sodium and chloride concentrations in the urban
lakes? The TCMA has minimal natural sources of chloride. Consequently, under natural
conditions Cl- concentrations would be expected to be low. Indeed, the median Cl-
concentration in TCMA lakes in 1800 and 1750 was estimated to be 3 mg/L by using
diatom assemblages in sediment cores [14]. Under current land use conditions lakes
in the North Central Hardwood Forests ecoregion of Minnesota, which includes the
TCMA as well as less developed areas surrounding the TCMA, have 4-10 mg/L of
chloride based on the inter-quartile 25th 75th percentile [95]. The increased sodium and
chloride concentrations in urban lakes of the TCMA are also apparent in a comparison
with lakes located throughout the state of Minnesota [94]. This set includes 91 lakes
overall, excluding lakes in western Minnesota where high specific conductance is due to
rich sulfur bearing minerals (gypsum and pyrite) and lakes in northeastern Minnesota
(Canadian shield area) with very low specific conductivity values. Chloride and sodium
concentrations in the TCMA lakes are between 10 and 25 times higher than in these
other lakes (Table 3.6). By comparison, calcium concentrations in the TCMA lakes are
only 1.4 times higher than in the other lakes throughout Minnesota. Overall, the nine
urban lakes in the TCMA appear to be chloride and sodium dominated waters differing
from other water bodies in the region, which are either calcium or sulfate dominated
[94, 96].
Because the TCMA has minimal natural sources of chloride, an anthropogenic salt
source has to be suspected. Since both Cl- and Na+ ions in urban lake waters are
elevated, it can be suspected that the common source is sodium chloride (NaCl). If Cl- is
derived solely from NaCl, a stoichiometric requirement is that the molar concentrations
of Cl- and Na+ should be in a 1:1 relationship. In the lake water samples gathered on
2/22/2007 and 11/15/2007 the molar ratio of Cl- to Na+ was 1.13:1. It is speculated
that the molar relationship is not exactly 1:1 because Na+ ions are known to adsorb
onto particles whereas Cl- is not (Lofgren 2001; Norrstrom and Bergstedt 2001; Oberts
2003). The molar ratio is, however close enough to 1: 1 to confirm that NaCl is the
likely source of elevated Na+ and Cl- concentrations in the TCMA lakes.
In addition to road salt use, domestic salt use for water softening and industrial/commercial
uses have to be considered. A major use of NaCl in the TCMA is for water softening but
69
a majority of the saline water from water softeners in the TCMA is discharged through
sanitary sewers into WWTPs and eventually into the Mississippi River bypassing any
lakes.
The presence of a seasonal lake salinity cycle points towards road salt as the source
of lake salinity. Snowmelt runoff containing dissolved road salt in winter and spring,
and rainfall runoff without road salt content in summer and fall would be expected to
cause a seasonal salinity cycle in lakes and streams. This pattern has been observed in
all 13 lakes studied (Figures 3.4 and 3.5 and Table 3.3). This same pattern was seen in
a lake in Sweden and modeled using estimated road salt applications, precipitation and
evaporation in the lakeshed [97].
3.5.3 Salinity, temperature and dissolved oxygen stratification
In all of the lakes studied the formation of a chemocline occurred in the winter coin-
cident with observed peak salt concentrations in streams of the TCMA (Figure 3.1).
The location and strength of the chemocline varied between individual lakes and from
year to year. The variability is associated with many parameters including the number
of snowfall events and snowfall amounts in a winter season, influencing the amount of
road salt applied, in addition to climate parameters such as air temperatures and solar
irradiance, influencing the timing and amount of runoff. Wind speed and direction as
well as surface cooling, which cause convective circulation to break the chemical strati-
fication in fall or spring when thermal stratification is weak, can also add to temporal
inconsistencies. Lake stratification simulations operating at high temporal resolution
can be used to relate salt inputs, lake mixing, temperature stratification and weather
patterns.
Among the nine individual lakes in Figure 3.4 five (Parkers, Tanners, Ryan, Brownie,
and McCarron) appear to have a stronger chemical stratification then the other four.
These five lakes also have the smallest surface area to depth ratios. It is interesting
to note, that a very similar parameter, the lake geometry ratio defined as the ratio of
maximum lake depth to the fourth root of the lake surface area was introduced [98] as
an indicator of the strength of temperature stratification of lakes. This parameter has
been very useful to distinguish between dimictic and polymictic temperature stratified
lakes [99, 100].
70
It is clear from the data that the chemical stratification in some of the lakes can be
strong enough to prevent mixing during either the fall or spring turnover periods or at
least delay the lake from fully mixing. In two lakes, Tanners Lake and Parkers Lake,
monomictic behavior was observed. In Tanners Lake mixing was prevented in the spring
of 2006 resulting in anoxia of the benthic waters (Figure 3.6). When the lake mixed in
the fall oxygen was circulated again over the full lake depth. In the following spring
of 2007 the turnover period was delayed due to the existence of chemical stratification,
but occurred later in the season.
The prevention or delay of the full lake turnover (vertical mixing), when the water
reaches maximum density of freshwater at 4oC, can adversely affect the lakes water
quality, especially near the lakebed. Oxygenation of the benthic waters is an impor-
tant process in the health of a lake. It is well known that phosphorus is released from
anoxic sediments into the water above at much higher rates than when the sediments
are well aerated. With longer anoxic periods more phosphorus could be released from
the sediments stimulating algal blooms in the surface mixed layer when the lake fi-
nally overturns. Certain fish typically migrate to deeper waters during the hot summer
months. If mixing is prevented in the fall the anoxic zone could be increased reducing
the inhabitable space for fish in the lake. Prolonged anoxic periods are shown in Figure
3.6 for the year 2006 when mixing was prevented in the fall. The next full mixing of
Tanners Lake occurred before the 5/17/2007 sampling period resulting in oxygenation
of the benthic waters.
3.5.4 Seasonal flushing of salt from the TCMA lakes
The strength or amplitude of the seasonal salinity cycle in the urban lakes (Table 3.3)
appear to be loosely related to the size of a lake relative to its depth, watershed area
and impervious area in the watershed (Tables 3.3 and 3.4). Sweeney, Brownie, Isles,
and Ryan Lakes have the highest seasonal flushing rate (percent change). The strongest
salinity cycle is seen in Sweeney Lake, which has a large watershed area to lake area ratio
and drains Interstate 394 and Highway 100, both heavily traveled roadways. Sweeny
Lake is artificially mixed throughout the year and has a small depth to surface area ratio
and a larger watershed area to lake surface area ratio, all of which appear to increase
the flushing (salt removal) from the lake. Seasonal salinity cycles had already been
71
found in grab samples collected from the surface of lakes near highways in the TCMA
from 1982-1987 [101]. That study was, however, limited to grab samples from the lakes
surface and did not include volumetric average concentrations or concentrations near
the bottom of the lake.
Seasonal fluctuations in chloride concentrations can also influence the aquatic life in
a lake. The chronic standard of 230 mg/L chloride required for the protection of aquatic
life was exceeded at some point in time in 5 of the 13 lakes studied (Tanners, Parkers,
Brownie, Ryan, Sweeney). These elevated concentrations were typically found during
the winter and spring months and occurred in the deepest portion of the lakes. Only
one lake (Sweeney) displayed chloride concentrations above the standard throughout the
entire water column because it was artificially mixed by an aeration system to control
eutrophication. The presence of this high salinity water for prolonged periods of time
could be changing the aquatic community in each of these lakes. Increases in sodium
and chloride concentrations have been shown to decrease the biodiversity in wetland
areas and waterways [7, 8]. Shifts in diatom communities to more salt tolerant species
have also been observed with increasing chloride concentrations in lakes [14].
3.5.5 Convective mixing of saline lake water with pore water in the
sediments
The presence of a high salinity layer at the bottom of the lake causes the convective pen-
etration of saltwater into the pore system of the lake sediments. It is the same process,
induced by density instabilities, that mixes the surface waters of a lake during cooling,
except that in the surface waters the density differences are caused by temperature and
not salinity differences, and no pores are present. Leakage of saline water into the less
dense pore water of lake sediments can be inferred from Figure 3.7. Similar chloride
profiles were found in a study of the benthic sediments of a stormwater detention pond
receiving runoff from a major roadway. In this detention pond chloride concentrations
were 3000 mg/L at the sediment-water interface and 1500 mg/L at a depth of 0.4m into
the sediment core [44]. The convective mixing in the sediment pore system of a lake or
pond receiving road salt runoff is driven by density instability. If the density of the saline
water above the sediments is higher than the density of the water in the pore system,
finger-like intrusions of the denser saline water into the fresher (lighter) pore water will
72
form; in turn the fresher (lighter) water will rise in finger-like formations through the
pore system. This process is slow, but persistent, and referred to as convective mixing.
As a result pore water with a low salt concentration and density moves upward and is
replaced by water with a higher salinity moving downwards into the sediment pores.
Convective transport of solutes into the sediment bed occurs also in estuaries when
the density of the overlaying water is greater then the sediment pore water [102]. Convec-
tive transport has been observed in laboratory experiments and in numerical simulations
of saline lakes. Instabilities created by density differences cause increased transport of
salt into the groundwater [103, 104, 105, 106]. The process was also found in tests of
sulfate reduction in sediment cores from Devil Lake, South Dakota [107]. Due to the
density differences between the overlying water and the sediment pore water the effec-
tive diffusion (dispersion) coefficient was found to be much larger then the molecular
diffusion coefficient.
These observations suggest that a loss of chloride into the sediments occurs during
the winter and early spring when saline water accumulates at the bottom of a lake. The
loss would be expected to continue until an impermeable sediment layer or stability
between the layers of water is reached. The penetration of saline water into the lake
sediments can cause the release of metals from the solids. Increased transport of heavy
metals (Zn, Cd, Cu and Pb) coincident with road salt applications have been observed
in road side soils in Germany, Sweden and the United States due to chloride complexes,
ion exchange, and even increased colloidal dispersion [39, 37, 38, 36, 40]. High NaCl
concentrations decrease the partitioning coefficients between the dissolved and particu-
late phases of these metals resulting in higher mobility and bioavailability [35, 34, 36].
By increasing the concentrations of metals in the dissolved phase they are free to move
with the water either further into the sediments due to convective mixing or to diffuse
into the overlying lake waters.
3.5.6 Salinity trends in TCMA lakes
Salt concentrations in the urban lakes are increasing. Short-term trends of chloride
concentrations in individual lakes can be seen in Table 3.3. Two of the four lakes
(Cedar Lake and Lake McCarron) analyzed for the entire field study period (2004-2007)
show increases in the annual minimum and maximum concentrations from year to year.
73
This increase represents an accumulation of salt in both lakes over time. Bryant Lake
displays the same pattern, but has only been studied for two years. These findings
give only a hint of increasing lake salinities. Long-term and regional trends are seen
in Figure 3.8. The slope of the normalized specific conductance in Figure 3.8 is 0.018
or 1.8% per year, i.e. specific conductivity in lakes of the TCMA increases annually
and on average by 1.8% (The reference value is the average chloride concentration for
the period 2001 to 2005). Trend values for individual lakes range from 0.1% to 3.0%
(Table 3.4). The average trend seems reasonable because hind-casting projects the year
of minimum (near zero) chloride concentration to be about 1950. This is the time when
road salt application began to increase dramatically throughout the state (Figure 3.8).
By extending the trend into the future a doubling of specific conductivity, and
therefore chloride concentrations in the 38 lakes, would occur in about 50 years, i.e.
around 2060. Chloride concentrations would increase from a current median value of
87 mg/L to 174 mg/L. In 100 years the average chloride concentration in many TCMS
lakes would exceed the current chronic chloride standard year round and throughout the
water column. Current efforts to reduce road salt use by public agencies, commercial
and private users, may be balanced by continued urban growth and possible increased
snowfall due to climate change (Seeley 2003) so that the trends seen in Figure 3.8 could
continue for some time into the future.
3.5.7 Relationships between lake salinity, lake bathymetry and water-
shed characteristics
Although trends and increased chloride concentrations are occurring throughout the
region, the magnitude of these parameters appear to be dependent on both lake and
watershed characteristics of the individual lakes. Table 3.4 gives parameters for the 38
individual lakes in database 2. Chloride concentrations in the surface layer throughout
the TCMA ranged from 31 mg/l in White Bear Lake, a large lake located in the northern
suburbs to 505 mg/l in Spring Lake which is a small lake receiving large amounts of
runoff from I-394 in downtown Minneapolis. The highest chloride concentrations all
occurred during the winter months and near the bottom of the lakes. These values
ranged from 43 mg/L in White Bear Lake to 1,018 mg/L in Spring Lake. Each of the
38 individual lakes shows an increasing trend in average annual chloride concentrations
74
(Table 3.4). Although small lakes, and lakes near major roadways seem to have the
strongest upward trends, the increase in chloride and sodium concentrations in urban
lakes due to road salt applications appears to be regional and not specific to a small
number of lakes located close to highways.
The strongest correlations between chloride parameters and lake or watershed pa-
rameters were found by examining what percentage of the watershed is impervious.
This is to be expected since road salt is applied to impervious surfaces such as streets
and parking lots and as this percentage is increased not only is more salt being applied
but a more direct route is available for the runoff to reach the lakes. When including
the proxy for lake volume the correlation was even higher. As the volume of the lake
increases there is more water to dilute the snowmelt runoff causing lower concentrations.
3.6 Conclusions
The water quality of urban lakes in the Twin Cities Metropolitan Area (TCMA) of Min-
nesota has been investigated to identify and quantify impacts of road salt applications in
their watersheds. Lakes in the TCMA especially near major roads have elevated chloride
and sodium concentrations compared to other non-urban Minnesota lakes. The almost
1:1 molar relationship between chloride and sodium in the lake waters points towards
sodium chloride (NaCl) as the source. Since natural sources of NaCl in the geology of
the TCMA and Minnesota in general are very limited or rare, the source of the NaCl
must be anthropogenic. In the TCMA sodium chloride (NaCl) is used predominantly
for water softening and road deicing. The seasonal cycles of lake chloride concentrations
(high in winter/early spring and low in fall) point to road salt applications as the cause
of lake salinity in the urban lakes.
Specific conductance in the TCMA lakes is strongly correlated with Cl- and Na+
ions. Specific conductance profiles showed chemical stratification in almost all urban
lakes investigated. Over the course of a year, specific conductance, Cl- and Na+ con-
centrations are cyclic, both in the surface and the bottom waters of the urban lakes.
Volume-weighted average chloride concentrations give the amount of seasonal salt stor-
age/flushing. In the 13 urban lakes investigated, the annual storage/flushing rate ranged
from 9 to 55% of the minimum salt content in the lake. Smaller lakes with larger
75
watershed areas and a higher percentage of impervious surfaces had higher seasonal
storage/flushing rates, as to be expected.
There are physical, chemical and biological consequences of increased lake salinity. In
two of the lakes (Tanners and Parkers Lake) chemical stratification near the lake bottom
was strong enough to prevent the lake overturn in spring of 2006. This behavior prevents
dissolved oxygen from reaching the lake sediments. Complete lake mixing resumed in fall
2006 and spring 2007 In several individual lakes chloride concentrations were observed
to exceed the chronic water quality standard for aquatic life. High salt concentrations
were also found in the sediments of Tanners Lake which can cause the release of heavy
metals from the solids into the pore water from where they are transported further into
the sediments or into the overlaying water.
Historical chloride concentration data from 38 lakes in the TCMA show an annual
average increase of 1.8% (range from 0.1 to 3.0%) throughout the TCMA. The increase
is strongly correlated with the amount of road salt purchased annually by the state of
Minnesota since the 1950s. Chloride concentrations in individual lakes are positively
correlated with the percent of impervious surface area in the watershed, and inversely
correlated to lake volume. Overall, the results show a progressive degradation of the
water quality of urban lakes due to application of NaCl in their watersheds. Road salt
is used to increase driving safety in winter, but current road salt application practices
do impair lake water quality in urban lakes.
Aknowledgements
We acknowledge and thank the Minnesota Local Road Research Board (LRRB) and
the Minnesota Department of Transportation (Mn/DOT) for providing the funding for
this research. We also thank the Technical Advisory Panel, lead by Wayne Sandberg of
Washington County, for input and suggestions to our research. We thank Amy Myrbo
and Kristina Brady of the Limnological Research Center (LacCore Facility) at the Uni-
versity of Minnesota, Department of Geology and Geophysics, for providing equipment
and expertise for the extraction and sectioning of the lake sediment cores. Karen Jensen
of the Metropolitan Council (MCES) provided valuable information on lake watershed
delineations and water quality information.
Chapter 4
A 0-D modeling approach to
study long-term chloride
concentration in lakes receiving
runoff containing road salt
Eric V. Novotny and Heinz G. Stefan
St. Anthony Falls Laboratory,
Department of Civil Engineering, University of Minnesota
Minneapolis, Minnesota 55414
76
77
4.1 Abstract
Chloride concentrations in lakes located in urban environments using road deicing salts
are increasing to levels that are changing natural lake mixing behavior and influencing
aquatic life. A 0-D model was developed to project the seasonal cycle of loading and
flushing of chloride as well as the long-term accumulation of chloride in urban lakes
receiving runoff from roads to determine steady state concentrations under different
loading conditions. The model was calibrated using five years (2004-2008) of monthly
salinity profiles from 7 lakes in the Minneapolis/St. Paul Twin Cities Metropolitan Area
of Minnesota, USA, four model parameters and an initial concentration. Three of the
seven lakes appear headed towards year-round volume averaged chloride concentrations
above the 230 mg/L chronic standard for impairment to aquatic habitat. The two lakes
with the lowest projected equilibrium concentrations of chloride have already reached
equilibrium. It is projected that one lake will take up to 40 years to reach equilibrium.
If road salt application rates are reduced in future winters, it is projected that the lakes
will respond with noticeably lower Cl- concentrations within 5 to 10 years. If road salt
applications are discontinued altogether, chloride concentrations are projected to reduce
to natural levels within 10 to 30 years in all seven lakes.
78
4.2 Introduction
Rising chloride concentration are present in many lakes located in urban environments
or near major roadways due to the application of road salt in the watershed [56, 87, 51,
108, 109, 97, 71]. Snowmelt runoff flows into these lakes through storm sewers, small
streams, overland flow and interflow. Road salt applications in rural areas can affect
lakes a few hundred meters away [6]. In urban environments runoff into the lakes is
extended by the presence of impervious surfaces and storm sewers providing a direct
path to the surface waters.
With increased chloride concentrations comes damage to the aquatic life. Ele-
vated chloride concentrations decrease the biodiversity of diatoms in lakes resulting
in halophilic taxa to increase in relative abundance [110]. Tadpoles of woodland frogs
have significantly lower survivorship, decreased time to metamorphosis, reduced weight
and activity, and increased physical abnormalities with increased salt concentrations
pointing to effects that could be affecting amphibians as a whole [9]. Flathead minnows
are affected at chronic concentrations as low as 298 mg/L [111]. Other aquatic species
have impacts at chronic levels ranging between 194 mg/L for sensitive species such as
various daphnia taxa, to 327 mg/L for the snail, Physa gyrina, to 561 mg/L for cad-
dis flies, Anaobolia nervosa and lemnephilis stigma, to 1,036 mg/L for bluegill sunfish
[59]. Not only are surface water communities affected, but also the biodiversity in lake
benthic sediments [109].
Over 317,000 tons of NaCl is being applied to the roads in the Minneapolis/St. Paul
Twin Cities Metropolitan Area (TCMA) with over 70% of the chloride being retained
and not flushed through runoff into the Mississippi River [112]. One of the sinks for
chloride in the watershed are lakes with seasonal chloride stratification and rising vol-
umetric chloride concentrations [71]. Background chloride concentrations before urban
development were around 3-10 mg/L [14]. From 1970 to 2000 in the TCMA significant
changes in lake chloride concentrations were detected using diatom assemblages with a
strong correlation to the percent of urban developments in each of the watersheds stud-
ied [13]. Similar increasing trends in chloride concentrations (an average of 1.5mg/L
per year and a range between 0.1 to 15 mg/L per year) were detected in 38 lakes of the
Twin Cities Metropolitan Area. That trend was correlated with the amount of rock salt
79
purchases by the state of Minnesota [71]. Median concentrations in the 38 lakes in 2005
were 87 mg/L well above the 3-10 mg/L observed during predevelopment times [71].
The objective of this study is to investigate chloride concentration trends in lakes
from a mechanistic point of view and determine if a 0-D lake chloride model can be
used to project seasonal salinity fluctuations and future lake chloride concentrations
based on recently acquired year round monitoring data. While chemical stratification
is common in lakes receiving runoff with high concentrations of salts our main interest
is not in the spatial distribution of chloride in lakes, but in the total amount of chloride
contained in each lake, and the variation of this chloride content over time. A water
quality model simulating inflow, accumulation and flushing of Cl- was created. The
model formulation, validation and model projections will be described. Lakes located
in the TCMA were used to simulate seasonal salinity cycles, long-term trends, ultimate
equilibrium chloride levels, and responses to reduced road salt application rates using
this model.
4.3 Lake Chloride Model Fomulation
4.3.1 Zero-dimensional model formulation
A model simulating the volume-weighted average chloride concentration C(t) in a lake
was formulated and used. The total amount of chloride (kg) in a lake is the volume-
weighted average concentration C(t) multiplied by the volume of the lake. The behavior
of C(t) in a lake can be reproduced by a 0-dimensional (0-D) model simulating inflow,
accumulation and flushing of Cl- over time. 0-D models have been used extensively
and very successfully to simulate phosphorus management scenarios for eutrophication
control of lakes [113, 114].
4.3.2 Daily time scale model
Chloride is a highly soluble substance that is not easily removed by chemical reactions
or biological processes once it is solution. In the model Cl- is therefore treated as a
conservative substance and the lake is treated as being fully mixed at all times. It is
assumed that chloride is added at a constant rate to the lake during the months when
80
snowmelt runoff occurs and that flushing occurs during the open water season at a
constant flowrate. The equations used (Eq. 4.1 and 4.2) for this model represent a fully
mixed continuous flow reactor for a conservative material (chloride) that is added at a
constant rate during one period and flushed at a constant rate during another.
Loading phase :dC
dt=M/D
V(4.1)
Flushing phase :dC
dt= −K ∗ C (4.2)
where M is the mass (g) of chloride added to the lake in the winter. D is the number
of days during which chloride is added to the lake, V is the lake volume (m3), K is
the flushing rate coefficient equal to 1/T, where T = V/Q is a hydraulic residence time
(days), and C is the volume-weighted average chloride concentration (mg/L). Solving
the differential equations (4.1) and (4.2) for these two phases with their respective initial
conditions results in equations (4.3) and (4.4).
@ t = tmin, C = Cmin : C(t) =M/(tmax − tmin)
V(t− tmin) + Cmin (4.3)
@ t = tmax, C = Cmax : C(t) = Cmaxe− 1
T(t−tmax) (4.4)
where t (days) is the day number in a particular year (t = 1 represents Jan 1 and
t = 365 represents Dec 31), tmax is the day number when the maximum concentration
(Cmax) is reached after all of the loading has occurred and tmin is the day number when
the minimum concentrations (Cmin) is reached after all of the flushing has occurred.
The other two variables M (g) and T (days) are mass and residence time, respectively,
as defined previously.
Equations (4.3) and (4.4) describe the seasonally cyclic nature of the Cl- concentra-
tion in the lake, starting at a minimum concentration in the late fall, accumulating to a
maximum concentration in the late winter and then flushing the salt out to a minimum
concentration in the following fall. By setting an initial value for Cmin, solving equation
(4.3) until the day equals tmax the value for Cmax can be obtained for equation (4.4).
This equation can then be solved until the day equals tmin thus obtaining the new Cmin
81
value to be used in equation (4.3). This cycle is repeated describing a time series (with
a daily time step) of chloride concentrations in the lake based on the mass of chloride
M that is entering the lake in a cold season, the lakes material residence time T, and
the dates when loading and flushing begin, tmin and tmax, respectively.
Two informative values in equations (4.3) and (4.4) are the minimum and maximum
concentrations. They tell the lowest and highest chloride values that can be reached in
a lake for a particular year. These parameters can be used to assess the stresses on the
aquatic life within a lake due to chloride, both the maximum concentration the aquatic
life will have to live in and the sustained concentration that will be felt throughout the
entire year. Equation (4.3) and (4.4) can be adjusted to an annual time step and solved
for the maximum and minimum concentrations in each year.
Setting C = Cmax and t = tmax in equation (4.3), setting C = Cmin and t = tmin in
equation (4.4), and introducing Tf = (tmin - tmax ) (the number of days in a year that
flushing occurs), allows the conversion of equations (4.3) and (4.4) to equations (4.5)
and (4.6) which define the maximum and minimum concentrations that can be obtained
in a particular year i.
Cmax(i) =M
V+ Cmin(i−1) (4.5)
Cmin(i) = Cmax(i)e−
TfT (4.6)
Equations (4.5) and (4.6) can be used to determine how the maximum and minimum
concentrations in a lake change from year to year based on the mass of chloride M(g)
entering the lake, the volume of the lake V(m3), and the length of a lakes flushing period
Tf = (tmin - tmax) relative to the hydraulic residence time T, i.e. Tf /T.
If the annual salt loading of a lake and the summertime flushing remain the same
year after year, each of the lakes will eventually come to a point where the amount of
salt entering that lake during the snowmelt runoff is equal to the mass of salt leaving the
lake due to flushing. At this point the minimum and maximum concentrations reached
from year to year will be the same. When this equilibrium is reached Cmin(i) will equal
Cmin(i−1) and equations (4.5) and (4.6) can be solved for Cmax(i) and Cmin(i) that are
then defined as the equilibrium concentrations Cmax(eq) and Cmin(eq) in equations (4.7)
82
and (4.8).
Cmax(eq) =M/V
1− e−Tf /T(4.7)
Cmin(eq) =M/V
eTf /T − 1(4.8)
The long-term effect of salt applications within a lakes watershed can be determined
from equations (4.7) and (4.8) if the total amount of salt M (g) entering the lake in a
winter season and the lakes hydraulic residence time (T ) are known and are constant.
Residence time (T) is related to lake volume (V) and water outflow rate (Q) from a
lake as T =V/Q. Since volumes of individual lakes in the TCMA are fairly constant
in the long term, except under severe drought conditions, the hydrology of a lakes
watershed has to remain fairly unchanged to assure a constant residence time T. How
the equilibrium concentrations would change based on reductions or increases in salt
applications in a lakes watershed can also be studied using the above equations.
4.3.3 Annual time scale model
The lake can also be modeled at an annual time scale. How the annual average con-
centration in the lake is changing from year to year can be determined from equation
(4.9).
VdC
dt= M∗ −QC (4.9)
where V (m3) is the volume of the lake, M∗(g/yr) is the total mass (rate)of chloride
added to the lake each year, Q is the total outflow from the lake and C is the current
annual average concentration in the lake. This equation can be solved with the initial
condition @ t = 0, C = Co. The solution is in equations (4.10 and 4.11).
ln
[M∗
V T − Co
M∗
V T − C
]=
t
T(4.10)
C =M∗
VT −
(M∗
VT − Co
)e−t/T (4.11)
83
where T = V/Q is again the residence time. At equilibrium, where dC/dt =0, the
annual average concentration would be:
CE =M∗
VT (4.12)
Equation 4.11 can be solved for the time necessary to reach x% of the equilibrium
concentration. (x = 100% requires an infinite amount of time, and x = 97% is a
considered a reasonable approximation). By substituting equation 4.12 and solving
equation 4.11 when C = .97 * CE @ t = tE equation 4.13 is created. This equation can
be used to find an estimate of the time it will take for a lake to reach an equilibrium
chloride concentration.
tE = ln
[CE − Co
0.03CE
]T (4.13)
4.3.4 Model assumptions
The five-parameter zero-dimensional model simplifies the actual lake processes consid-
erably, yet will be shown to capture the principal behavior of the time-variable chloride
concentrations. It is appropriate to briefly describe the actual inflow, mixing and out-
flow processes that produce the observed flow-weighted chloride concentrations, in order
to highlight the model simplifications.
The first minor shortcoming of the model is that the water budget is ignored, and a
constant lake volume is assumed. During the loading phase (tmin to tmax ) only dissolved
material (chloride), but no water is added to the lake. It is assumed that there is a net
inflow of chloride into the system during this phase and that any exported chloride is
outweighed by the mass entering the lake. The water flow rates into and out of the lake
are assumed to be equal. The outflow of water from the lake during the flushing phase
is also not explicitly accounted for although it is assumed to be equal to any inflow,
and the lake water budget is again ignored. A volumetric water outflow rate is hidden
in the residence time T of the chemical, which is equal to the hydraulic residence time,
i.e. the ratio of lake volume V divided by the water outflow rate, because the chemical
is conservative. During this flushing phase a net export of chloride is assumed, and
any delayed inflow of chloride through subsurface flows into the lake is assumed to be
84
outweighed by the flushing of chloride from the system.
The second assumption is that concentrations in the model are volume-weighted
averages for the entire lake, as if the lake were completely mixed all the time. How-
ever, strong chloride concentration gradients have been observed in TCMA lakes in
spring. The deepest lake waters typically have the highest chloride concentrations be-
cause snowmelt runoff containing dissolved sodium chloride flow to the bottom of a
lake as a density current [93, 71]. Since lake outflows are typically from the surface,
the saline water at the bottom cannot be flushed out until the lake has become fully
mixed vertically. Chloride-laden salt water at the bottom of a lake can also penetrate
by convective mixing into the pore system of lake sediments [71]. Some of this salt can
re-enter the water column after lake-mixing events. The net effect is an extension of the
residence time T of chloride in a lake when compared to the water residence time.
These assumptions would point towards the need of a 1-D model that accounts
for vertical stratification. However, a 0-D model is a very useful tool and sufficient
for the analysis of seasonal and long-term trends in overall lake salinity. Phosphorus
management in lakes for the control of eutrophication has benefited from 0-D modeling
since the 1970s. 0-D phosphorus models have been applied to lakes as large as the
Laurentian Great Lakes [113, 114].
Our 0-D lake salinity model has five parameters: the mass of chloride from road salt
(NaCl) entering the lake (M), the material residence time (T), the day (tmin) when the
average concentration in the lake has reached a minimum and new salt begins to enter
the lake, the day (tmax) when the concentration in the lake has reached a maximum and
flushing starts to dominate over salt accumulation, and the initial concentration (Co)
before the first year simulated. To obtain values of the five parameters for seven lakes
in the TCMA, the model was fitted to data (16 to 33 data points per lake) collected
over up to 5 years.
4.4 Lake data collection and model calibration
Field data from 13 lakes of the Twin Cities Metropolitan were available for periods
of up to five years [93, 71]. The model was applied to seven of the lakes with the
longest records and the characteristics given in Table 4.1 [93, 71]. Lakes were sampled
85
every 4-6 weeks between 2/15/2004 and 10/20/2008. All of the lakes were natural lakes
with inflows coming from streams, storm sewers or overland flow. Vertical profiles of
specific conductance were measured in the water column every 0.5 meters at the deepest
location in each lake using a YSI Model 63 probe [91]. A relationship between specific
conductance and chloride was previously determined [71], and the measured specific
conductance values were converted to chloride concentrations using that relationship.
This relationship existed because chloride and sodium ions were the dominant drivers
in changes in salinity. All other ions remained relatively constant throughout the year.
Table 4.1: Lakes modeledMaximum Lake Impervious
Lake Lake Lake Surface Watershed watershedname Volume Depth Area area area
Volume weighted average chloride concentrations were determined for each of the
sampling dates [93]. The four model parameters (M, T, tmin tmax) were estimated
by fitting the model to the data. An initial concentration (Co) set at time tmin of
the first year of data was also determined. Least square errors between measurements
and model results were used to determine the best-fit model. A program was created
to cycle through permutations of the four parameters plus the initial concentration
until a combination was found that had the lowest root mean square error (Eq. 4.14)
between the model values and the measured values using the entire time series of the
measurements.
RMSE =
√(Cmodel(t)− CObserved(t))2
N − 1(4.14)
Numerical solutions of equations (4.3) and (4.4) with a time step of 1 day were used
for this simulation. Cmin and Cmax were calculated by applying equations (4.3) and
86
(4.4) sequentially. Once the best-fit parameters had been determined, simulations were
extended in time to project how chloride concentrations may change in the future, and
what maximum concentrations are expected if current loading and flushing conditions
continue. Estimations were made for the equilibrium concentrations using equations
(4.7) and (4.8) along with the time required to reach concentrations within 97% of
equilibrium if conditions were to remain constant. Finally simulations were run to
determine how changes in the amount of NaCl entering the lakes would influence the
final equilibrium concentrations of chloride.
4.5 Results
The best-fit model parameters for the seven simulated lakes are given in Table 4.2. tmin,
tmax, M and T and the initial concentration Cmin,o at the beginning of the simulation,
i.e. in the fall of the first simulated year (i = 1), is shown. Model simulations and
data are displayed in Figure 4.1 for all seven of the lakes. Identical patterns of chloride
accumulation in the cold season and flushing in the summer are apparent. With the
best-fit parameters, the model was used to project future equilibrium concentrations in
the seven lakes under different salt loading conditions.
Table 4.2: Model parameters. Cmin,o is the chloride concentration at the beginning ofthe simulation in fall (tmin) of the first simulated year (i=1). RMSE (%) is the ratioof root mean square error to average concentration over the record length. N is thenumber of data points used for model fitting.
tmin tmax M T Cmin,o RMSE N(mo/day) (mo/day) t/yr (yrs) (mg/L) (%)
Figure 4.2 illustrates how the chloride concentrations change with time, and how
87
21
Figure 4.1: Modeled and observed chloride concentrations in seven lakes of the TwinCities metropolitan area, Minnesota.
88
equilibrium is reached under four different chloride (salt) loading scenarios. The sce-
narios are: (a) current conditions continue in the future (1M), (b) loading is increased
by 50% (1.5M), (c) loading is decreased by 50% (0.5M) and (d) salt applications are
stopped and loading is reduced to 0 (0M). Figures 4.1 and 4.2 were created by using
the best-fit parameters for M, T, tmin and tmax (Table 4.2) in equations (4.3) and (4.4).
Cmax and Cmin for every year (i) were obtained from equations (4.5) and (4.6).
Equilibrium concentrations were determined by solving equations (4.7) and (4.8)
with the parameters listed in Table 4.2 and are given in Table 4.3. Another important
parameter is the time it will take to reach equilibrium. To obtain this value the con-
centrations were examined at an annual time scale. An initial concentration (Co) was
determined by averaging the simulated data from equation 4.3 and 4.4 between 1/1/2008
and 12/31/2008. The annual average equilibrium concentrations were also calculated
using the parameters in Table 4.2 and equation 4.12. These values were then inserted
into equation 4.13 to determine how long it would take for the lake to reach a concentra-
tion within 3% of the equilibrium concentration. Values obtained were between 0 and
41 years. If the equilibrium concentration was lower than the initial concentrations it
was assumed that the lake had already reached equilibrium and the time to equilibrium
was set to 0.
Table 4.3: Maximum (Cmax(eq)) and minimum (Cmin(eq) ) chloride concentrations atequilibrium determined by equations (4.7) and (4.8). C2008 is the average annual chlorideconcentration in the lake between 1/1/2008 and 1/1/2009 and is used as Co in equations(4.13) to find the time required to reach equilibrium (tE).
The observed seasonal patterns of lake chloride concentrations matched the model
results with an RMSE between 2.8 and 9.5%. On average, accumulation of chloride
89
22
Figure 4.2: Projected future chloride concentrations under four road salt loading condi-tions. Current loading (1M), 50% increase in loading (1.5M), 50% decrease in loading(0.5M) and zero loading (0M).
90
in the lakes started on Nov 14, and ended about 3 months later on Feb 24. Flushing
was evident during the remaining nine months in spring, summer and fall. Annual
fluctuations between maximum and minimum chloride concentrations are from 18 to 23
mg/L in five of the lakes, 55 mg/L in Parkers Lake and 109 mg/L in Sweeney Lake.
In five of the lakes (Cedar, McCarron, Parkers, Sweeney, and Bryant) volume-
weighted average chloride concentrations have been rising, both in the simulations and
the data. This shows that the lakes have not reached equilibrium with the amount of
salt applied in their watersheds annually. Two of the lakes have slight decreasing trends
(Tanners and Ryan) indicating that chloride concentrations had reached equilibrium
and fluctuations are possibly occurring due variations in salt application rates from
year to year. Current concentrations in Ryan and Tanners Lakes are slightly above (by
8 and 10 mg/L) their volume-weighted projected equilibrium Cl- concentrations of 95
mg/L and 140 mg/L, respectively. Two lakes (Cedar and McCarron) are currently 17
and 33 mg/L below their equilibrium concentrations of 128 and 164 mg/L, respectively.
Three lakes (Parkers, Bryant and Sweeney) are currently 70 to136 mg/L below their
equilibrium concentrations of 248, 250 and 328 mg/L, respectively. With current road
salt application rates, the two lakes (Tanner and Ryan) with the lowest equilibrium con-
centrations have already reached equilibrium, four will reach it in about 5 to 15 years,
and Bryant Lake with the second highest equilibrium concentration will take more than
40 years to reach it.
Three equilibrium concentrations for current road salt application rate are above
the 230 mg/L chronic level for impairment to aquatic life. Even though the other four
lakes are not projected to reach such high concentrations throughout the water column,
the standards may be exceeded at the bottom of the lake in winter and spring due to
salinity stratification. In Tanners Lake, for example, current volume-weighted chloride
concentrations are on average 150 mg/l, but concentrations as high as 400 mg/L have
been measured near the bottom of the lake after snowmelt events [71].
It is also of interest to examine how model parameters relate to watershed and
lake characteristics. Annual chloride loading (M ) of the seven lakes obtained from the
model fit are between 7 and 104 t/yr. Watershed areas range from 77 ha (Ryan) to
1512 ha (Sweeney) of which 24 to 37% are impervious surfaces. Road salt is applied
on impervious areas, and chloride loading rates per unit impervious watershed area are
91
given in Table 4.4 for each of the seven lakes. The annual salt application rates on
the impervious areas vary from 186 kg/ha to 850 kg/ha with an average of 445 kg/ha.
This is much lower then the average application rates in the major streamsheds each
of the lakes are located in of 2200 kg/ha impervious area [112]. The flushing rates are
really small, ranging from 2.3 to 20 L/s. This corresponds to only 60 to 370mm per
year with an average of 204mm per year of runoff from the impervious areas. This
is much lower than the typical annual precipitation of about 800 mm/yr in the Twin
Cities metropolitan area.
Table 4.4: Chloride loading rates (M) per impervious watershed area (Aimp) , flushingflow rate (Q = V/T ), and flushing rate (Q) per impervious watershed area, determinedby fitting of model parameters to measured lake chloride concentrations.
The maximum chloride concentration in a lake due to road salt applications depends
on three parameters: the annual mass of chloride entering the lake M, the material
residence time T and the lake volume V (Equations 4.7 and 4.8). Changes in any of
these parameters would influence chloride concentrations in the lake. For a lake with the
same residence time and volume, doubling the entering chloride mass M would cause a
doubling of the equilibrium concentrations. Likewise if the mass entering the lake were
cut in half the equilibrium concentration would be cut in half. These changes in the
mass of chloride entering the lake are shown in Figure 4.2.
92
The residence time T has a large influence on the final equilibrium concentrations in
the lake. In Lake McCarron the entering mass/volume (M/V) ratio is 20 mg/L while in
Bryant Lake that ratio is 18 mg/L. Even though the amount of chloride entering Lake
McCarron per lake volume every year is larger, the equilibrium concentration is much
lower, i.e. 174 mg/L and 259 mg/L for Lake McCarron and Bryant Lake respectively.
The equilibrium concentration is proportional to residence time (equation 4.12).
Lake volume has an inverse relationship with the equilibrium concentrations. If two
lakes had the same residence times and the same mass of chloride entering the lake, but
one lake had twice the volume, the equilibrium concentrations in the larger lake would
be half as high as in the smaller lake (equation 4.12).
In summary, changes in any of the three model parameters ( M, V and T) in the
future would influence the final equilibrium concentrations. Therefore, projections of
lake concentrations have been made with the assumption that current conditions will
continue in the future.
4.6.2 Model projections
While increasing chloride concentrations in 38 Twin Cities area lakes were observed
between 1984 to 2005 [71], if the application rates and lake parameters remain constant
the chloride concentration in these lakes would level off in the future eventually reaching
equilibrium. The assumption of constant application rates and lake parameters may
be correct for inner city lakes where development and road expansion have reached
a maximum, however, expanding roadways and increasing population densities in the
suburbs may push road salt application rates in some lake watersheds higher.
Climate, population, urban development and road salt application rates are dynamic
variables that can influence future chloride concentrations in urban lakes. Increased
precipitation in Minnesota is expected under climate change scenarios [115, 116]. More
days with precipitation and increased intensity of rainfall events in Minnesota are pre-
dicted resulting in more runoff reaching the lakes. The elevated runoff volume would
enhance the flushing of lakes causing lake residence times and equilibrium concentra-
tions to decrease. However, increased snowfall amounts and more snowfall events are
also projected causing higher road salt application rates and increased lake equilibrium
concentrations.
93
In addition, changes in populations and/or impervious surfaces in the lake water-
sheds would influence the final equilibrium concentrations. In the Twin Cities metropoli-
tan area the population is expected in rise by 30% from 2.8 million in 2008 to 3.6 million
by 2030 [117]. With this expansion come more roads and wider highways elevating the
amount of salt applied in the lake watersheds and exposing new lakes to chloride loads.
Impervious surfaces would also cover larger areas leading to higher flowrates entering
into the lakes, possibly decreasing the lakes residence times, but also increasing the flow
of salt into the lakes.
Another factor influencing the final equilibrium concentrations in the urban lakes is
training and awareness of best management practices (BMPs) designed to reduce the
amount of salt applied in a watershed. With added awareness of the environmental
impact of road salt applications, BMPs on how to more effectively apply road salt or
reduce its impact on water bodies, and alternative deicers that do not contain chloride,
decreases in road salt application rates could be expected. If practices were changed,
chloride concentrations in the lakes would quickly decrease since the hydraulic residence
time (T) in the Twin Cities urban lakes is typically small (3 to 14 years). It is uncertain
how much of the salt is being stored in the sediments or how much is also being stored
in the groundwater feeding the lakes, however rapid changes would still be expected if
the mass of salt entering the lakes were dramatically reduced. As shown in Figure 2,
if road salt applications were eliminated it would take only 10 to 30 year for chloride
concentrations to reach the level of predevelopment values of 3-10 mg/L [14]. Even
if chloride applications were reduced by only 50%, concentrations in all seven lakes
would be reduced to levels below the chronic standard (Figure 4.2). Current stormwater
management in the Twin Cities favors the routing of surface runoff to lakes and wetlands.
If the runoff is from rainfall, this practice may be favorable, although water quality
has to be considered. If the runoff is snowmelt water, the practice has unfavorable
consequences.
4.7 Conclusions
A model was successfully used to simulate the chloride seasonal cycle of loading and
flushing as well as the long-term accumulation of chloride in urban lakes affected by
94
road salt applications. With monthly data from several years, the 5-parameter model
was used to determine long-term equilibrium concentrations in a lake if hydrologic con-
ditions and salt loading rates remain constant as well as simulations based on either
reduction or increased loading. Seven lakes in the Twin Cities metropolitan of Min-
nesota area were modeled. Using five years (2004-2008) of monthly salinity profiles,
four model parameters and an initial concentration were determined and subsequently
used to project future chloride concentration in each of the lakes under several road salt
application scenarios. The following average model parameters were determined from
the lake data: an annual chloride loading rate of 445kg per ha of paved surface in the
watershed, a hydraulic residence time of 6.8 yrs, and an average flushing rate of 10 L/s
in the seven lakes studied. From the simulation it was determined that 3 of the 7 lakes
will reach chloride concentrations year round throughout the water column above the
230 mg/L standard for chronic impairment to biota under current salt loads. If road salt
applications were stopped or the high salinity water from snowmelt events was diverted
from the lake concentrations could be reduced to predevelopment concentrations of 3-10
mg/L within 10-30 years.
Aknowledgements
Funding for the project, especially the data collection, was provided by the Local Road
Research Board, St. Paul, MN and the James L. Record Fund, University of Minnesota.
The University of Minnesota provided a Doctoral Dissertation Fellowship for the senior
author.
Chapter 5
Road salt impact on vertical lake
mixing
Eric V. Novotny and Heinz G. Stefan
St. Anthony Falls Laboratory,
Department of Civil Engineering, University of Minnesota
Minneapolis, Minnesota 55414
95
96
5.1 Abstract
Runoff from roadways on which road salt (NaCl) has been applied for driving safety
in winter can form a saline water layer at the bottom of a lake, pond, reservoir or
river impoundment. Natural vertical mixing of such lentic surface water bodies can be
hindered by this benthic saline layer. To study the formation and disappearance of the
saline layer temperature and specific conductance profiles were measured intermittently
over two years (2007, 2008) in eight urban lakes of the northern temperate region and
recorded at high frequency during one year (2009) in one lake. A deterministic dynamic
1-D lake temperature and salinity model was developed and used to simulate the summer
stratification and mixing dynamics in Tanners Lake, Oakdale, Minnesota. Erosion of the
saline layer in the spring occurred in only one of the three years examined (2007). In the
other two years (2008 and 2009), the saline layer persisted throughout the summer, and
was destroyed only by fall turnover and mixing between the epilimnion and hypolimnion
when thermal stratification was at a minimum. Density stratification was dominated
by salinity after ice-out, but was quickly overtaken by temperature stratification as the
epilimnion warmed. Inclusion of the lake number in the calculation of the hypolimnetic
eddy diffusion parameter made the mixing in the hypolimnion stronger when the lake
stratification became unstable in the fall and spring and weaker in the summer after
thermal stratification had formed. It was demonstrated that the saline benthic layer
prevents dissolved oxygen from reaching the lake sediments. Overall the results show
how salinity from road salt applications can influence water quality and natural mixing
in urban lakes.
97
5.2 Introduction
Sodium and chloride concentrations have been on the rise in lakes, streams and ground-
water in northern regions where road deicing salt (NaCl) is applied [85, 65, 3, 88, 23, 118,
71, 86, 112, 51, 97, 5]. Lakes receive sodium and chloride from roadways by snowmelt
runoff directly through overland flow, streams and storm sewers within the watershed,
or indirectly through the soil and groundwater. Salt concentrations in urban streams
and drainage systems are sufficiently high during the winter months to cause density
currents and chemical stratification in the receiving water bodies, especially lakes, deten-
tion ponds and reservoirs. This phenomenon occurred synoptically with above freezing
air temperatures and snowmelt runoff in a Minneapolis lake. The saline water flowed to
the deepest part of a small lake where it remained until spring [54]. The density change
caused by dissolved NaCl is small, but significant in relation to temperature-induced
density changes. For example, a temperature change from 4 to 5oC produces the same
specific gravity change as a NaCl concentration of 10 mg/L [55].
A saline water layer can cause a lake to become permanently density stratified. In
such a meromictic lake the bottom waters never mix with the surface waters. The con-
sequences of a meromictic lakes often include dissolved oxygen depletion in the benthic
saline layer and high concentrations of phosphate, ammonia, and hydrogen sulfide at
the sediment water interface. Small lakes and deep lakes are more vulnerable to be-
coming meromictic than large lakes and shallow lakes [6]. The formation of meromictic
conditions due to road salt applications has been reported for a few individual small
lakes or ponds [87, 88]. No meromictic lakes due to road salt application have been
found in the Twin Cities metropolitan area of Minnesota, but monomictic behavior has
been observed [71]. In these monomictic lakes full mixing is prevented during the spring
overturn period, but not during the fall. The formation of meromixis or even monomixis
is important because it can have serious ecological consequences for a lake.
In addition to monomictic behavior, rising overall concentrations and seasonal cy-
cles of sodium and chloride ion concentrations have been observed in lakes of the Twin
Cities metropolitan area [71]. A 0-D model was previously developed and used to
98
project long-term salinity trends and seasonal salinity cycles [86]. The 0-D model cap-
tured volumetrically-averaged salinity concentrations, but a 1-D model lake stratifica-
tion model is needed to capture the dynamics of vertical salinity stratification in a lake
from the high concentrations at the bottom to the low salinity in the surface waters.
The main objective of this paper was to demonstrate the influence of saline water
inflow from road salt applications on the stratification dynamics of a freshwater lake.
This includes (1) the formation of a benthic saline layer, (2) the effect on the vertical
mixing mechanics of a lake including the disruption of natural turnover events, (3)
the dissipation of the saline layer during the summer and fall seasons, and (4) the
consequences of the saline layer for oxygen transfer dynamics throughout the water
column. To pursue these objectives a combined field study and modeling approach was
adopted. The field study had two components: (1) measuring monthly temperature and
salinity profiles over a period of several years and (2) continuous sensing and recording
of vertical temperature and salinity profiles. Both research components were conducted
in a lake located near a major interstate highway in the Twin Cities metropolitan area.
The simulation study also had two components: (1) simulate lake stratification and
vertical mixing in a lake during the open water season following a winter in which a
saline benthic layer had formed and (2) simulate vertical dissolved oxygen transfer in a
lake with a benthic saline layer had formed.
In the deterministic model simulations, principles and governing equations previ-
ously used in 1-D lake temperature simulation models were applied and extended. The
extended model was based on the MINLAKE hydrothermal model, which has been
used successfully to simulate lake temperature stratification in many individual lakes
[119, 120, 121, 122]. The model was extended to include density gradients due to salinity
and the Lake Number [123] in the calculation of effective hypolimnetic diffusion coeffi-
cients. To illustrate the consequences of the benthic saline layer on lake water quality,
the dissolved oxygen profiles in a lake without and with a saline layer were simulated.
99
5.3 Methods
5.3.1 Field Investigation: Data Collection/Sampling Site
Data were collected in Tanners Lake in the eastern Twin Cities metropolitan area (Oak-
dale, MN). Tanners Lake is located next to Interstate Highway I-94 and a high volume
county road (Figure 5.1a). The lake receives runoff from a 214 ha watershed area of
which 33% is covered by impervious surfaces [71]. The lake has a surface area of 30 ha
(0.30 km2), littoral area of 11 ha (0.11 km2) and a maximum depth of 14 m [124]. The
depth area profile for Tanners Lake is shown in Figure 5.1b.
TannersLake
´ 0.5Kilometers
§̈¦I-94
")120
(a)
!"!!#
$"!!#
%"!!#
&'"!!#
&("!!#
!"!!# &!"!!# '!"!!# )!"!!# $!"!!#
!"#$%&'()&
*+",&'%,)&
(b)
Figure 5.1: (a) Tanners Lake location in Oakdale, Minnesota and (b) Bathymetry ofTanners Lake
Data were collected in Tanners Lakes in two phases. In the years 2007 and 2008
specific conductivity/temperature profiles were measured in Tanners Lake, and eight
other lakes in the Twin Cities metropolitan area, at 4- to 6-week intervals [71]. These
measurements were made in the water column every 0.5 meters at approximately the
100
deepest location in each lake using a YSI Model 63 probe [91]. Supplemental data
for Tanners Lake collected by the Ramsey-Washington Watershed District in Tanners
Lake were obtained from the Minnesota Pollution Control Agency Environmental Data
Access website (http://www.pca.state.mn.us/data/edaWater/index.cfm). The District
takes profiles every two weeks between May and October.
In November 2008 a buoy system connected to a chain of sensors was installed in
Tanners Lake to monitor continuously temperature and specific conductance (Figure
5.2). Fourteen Sensorex CS150TC probes, installed at depth intervals from 0.5 m to
1.5 m with a higher concentration of probes in the bottom half of the lake, measured
temperature and specific conductance every two minutes. These sensors were connected
to a Campbell Scientific 32- channel relay multiplexor, which was connected to a Camp-
bell scientific CR-10X data logger. A Garmin GPS device was attached to the buoy
to record any movement from its original position at the end of the ice cover period.
Remote data access was installed for easy data retrieval from the laboratory.
The Sensorex probes measure voltage across resistors, which is converted to conduc-
tivity or temperature. The conductivity and temperature values were used to find the
specific conductance value at 25 oC. The probes were calibrated before being placed in
Tanners Lake. While the probes were operating in the lake, a verification/recalibration
was conducted. Temperature and conductivity profiles in the lake next to the buoy were
measured independently using a handheld YSI Model 63 probe. Profiles were taken on
11/19/2008, 2/24/2009 and 5/11/2009. These profiles were used to estimate creep in
the values recorded by the Sensorex probes. Temperature probes did not need to be
recalibrated.
5.3.2 Model Formulation: Simulation of Summer Stratification in a
Lake with a Benthic Saline Layer
Basic heat and salinity transfer equations
The 1-D temperature and salinity model is based on the following diffusion equations.
A∂T
∂t=
∂
∂z
(KzA
∂T
∂z
)+
Hn
ρwCp(5.1)
A∂C
∂t=
∂
∂z
(KzA
∂C
∂z
)(5.2)
101
Antenna Raven XTV CDMA Digital Cellular Modem
CR 10X Data Logger
AM16/32B Delay Multiplexer
Sensorex CS150 TC Temp/ Conductivity Probes
GPS16-HVS Geographical position receiver
SC932A interface
GPS516 – HVS RJ45 Cable
COAXSM-L cable (6ft)
Figure 5.2: Schematic of data acquisition and transmission system (”Buoy set-up”)to continuously record specific conductance and temperature profiles in Tanners Lake.Buoy has connection for remote data access.
102
where T(z, t) is the temperature of a water layer (oC) at time t (days) and depth z (m),
C is the total salinity expressed as specific conductance (µS/cm), A is the horizontal
area of the water layer (m2), Kz is the vertical effective diffusion coefficient (m2/d), Hn
is the strength of internal heat sources (kJ m−3 d−1), ρw is the density of the water
(kg/m3), Cp is the specific heat of the water (kJ/oC). Equations (5.1) and (5.2) include a
number of important assumptions: (1) Salinity is treated as conservative with no losses
of solute by chemical or biological processes. (2) changes in lake salinity are dominated
by the sodium and chloride ions derived from road salt. Previous studies [71] justify this
assumption. (3) Vertical transport of heat and of solute (salt) are analogous, i.e, the
same effective diffusion coefficient Kz can be used. (4) Because the salt layer is located
at the bottom of the lake and is associated with the colder heavier water during most
of the open water season double diffusion [125] can be ignored.
The boundary conditions include the heat flux across the lake surface between water
and air and the heat flux between water and sediments at the bottom of the lake. The
surface heat flux has to be calculated from daily weather data. The bottom heat flux
was ignored by imposing an adiabatic boundary condition at the bottom of the lake.
Computations progressed in daily time steps in the following sequence: The first
step was the computation of the surface heat flux using weather parameters and wa-
ter surface temperatures from the previous time step. This was followed by solving
equations (5.1) and (5.2) for vertical mixing by an effective diffusion coefficient (Kz).
Equations (5.1) and (5.2) were solved numerically in daily time steps and 0.5 m depth
increments using an explicit numerical scheme. Next, convective mixing was simulated
to remove density instabilities. If a layer had a greater density than the layer below
the two layers were mixed. Net convective mixing proceeded from the surface to the
bottom until the density profile was stabilized. Finally wind mixing was simulated to
determine the surface mixed layer depth.
Surface heat transfer
The air-water heat exchange at the lake surface can be estimated by equation (5.3)
[126].
Hn = Hsn +Han −Hc −He −Hbr (5.3)
where Hn is the strength of internal heat sources, Hsn is the difference between incoming
103
short-wave radiation and reflected short wave radiation, Han is the difference between
incoming long-wave atmospheric radiation and reflected atmospheric ration, Hc is the
heat loss from the water by conduction, He is the heat loss from the water body by
evaporation and Hbr is longwave back-radiation from the water to the atmosphere. The
net short wave radiation (Hsn) is defined by equations (5.4) to (5.6) [120].
Hsn(0) = (1− r)βHs (kJ m−2 day−1) (5.4)
Hsn(i) = Hsn(i− 1)exp(−µ∆z) (kJ m−2 day−1) (5.5)
µ = 1.84(ZSD)−1 (5.6)
Where Hs is the incoming solar radiation (kJ m−2 day−1), r is the reflection coefficient
defined by the albedo of the water surface (= 0.087; [127]), β is the surface absorption
factor (= 0.4; [128]), Hsn is the solar radiation at the top of each layer of water and µ
is the total shortwave radiation attenuation coefficient, which is related to secchi depth
ZSD by equation (5.6) [120]. The effect of long wave radiation was calculated using
Convective mixing occurs if the density of water is greater in a layer above another
layer. This condition is unstable and induces convective mixing between the two layers.
Mixing continues downward until the density of the water increases from lowest to
highest throughout the water column going from the water surface to the lake bottom.
On occasion, the water temperature of a layer in the lake can increase or decrease
past the temperature of maximum density (TMD) during the daily time step to the point
where the new density of the layer is lower than that of the water layer below. This
apparent results masks the fact that during the period of cooling or heating the layers
temperature would have reached the point of maximum density resulting in instability-
induced mixing of the water layer of higher density above with the lower density layer
107
below it. This convective mixing could be missed due to the length of the time step
in the computations. Therefore a routine was added to the computations to check if
water temperature increased or decreased past TMD; if it did, convective mixing was
induced until the temperature of the overlaying layer reached a temperature past TMD
or until the density of the lower layer was higher than the density of the water at TMD.
Using equations (5.22) through (5.25), a linear approximation between TMD and specific
conductivity (C) was developed (equation 5.26).
TMD = 3.981− 1.22 ∗ 10−4C (5.26)
where TMD is in (oC) and C is specific conductivity in (/muS/cm ). According to
equation (5.26) a 0.22 oC decrease in the temperature of maximum density occurs for
every 1000 mg/L ( 1800 µS/cm) of salinity increase.
Surface wind mixing
The depth of the wind-mixed layer from the lake surface was determined from an en-
ergy balance consideration [133]. The ratio R (equation 5.27) is essentially the energy
transferred from the wind to the lake surface relative to the lifting work required to
deepen the mixed layer.
R =ρwWSCTU
∗3A∆tVm∆ρg(Zm − Zg)
(5.27)
U∗ = 0.0343W√CZ (5.28)
where ∆t is the time interval of one day, A is the surface area affected by wind, ρw is
the density of water, Vm is the volume of the surface mixed layer, ∆ρ is the density
difference between the mixed layer and the layer immediately below the mixed layer,
Zm is the depth of the mixed layer, Zg is the center of gravity of the mixed layer, CZ
is a drag coefficient on the water surface given by CZ = 0.0005√W when W < 15 m/s
or CZ=0.0026 when W >= 15 m/s [133]. U* is the wind induced water shear velocity,
W is the wind speed (m/s) and WSCT is the wind sheltering coefficient. When R is
greater that 1.0 mixing between the layers occurs. When R reaches a value below 1.0
wind mixing has reached it maximum depth.
108
Model calibration
Four model parameters were calibrated by minimizing the error between model results
and lake data. These parameters are WSCT , WFCT , α and ZSD. They appear in equa-
tions (5.27), (5.10), (5.17) and (5.6) respectively. The parameters WSCT and WFCT are
functions of wind sheltering. They represent the effective wind speed reduction for the
convective or evaporative heat transfer at the water surface, and for wind mixing at the
lake surface, respectively. The parameter α is used to determine the maximum vertical
effective diffusivity in the hypolimnion. The Secchi depth (ZSD) measures water trans-
parency, and is used to determine the attenuation coefficient of short wave radiation in
the lake water.
Simulations were run to obtain the best fit between the observed and modeled data
by changing the four calibration parameters. The set of parameter values that gave
the best combination of NSC-values between observed and modeled temperature and
specific conductance profiles was retained. This calibration was conducted with data
from the ice-free periods of 2008. The calibrated model parameter values were used for
model validation with data from 2007.
Weather data used as model input for the years 2007 and 2008 were obtained from the
Minneapolis/St. Paul International Airport (courtesy of Dr. X. Fang). The weather
station is about 19.8 km from the lake site. The weather data provided were daily
average values of air temperature (oF), dew point temperature (oF), solar radiation
(Langleys), wind speed (mph), and cloud cover (%). The values were converted to (oC)
for the temperatures, kj/m2 for radiation and m/s for wind speed (Figure 5.3).
5.3.3 Model Simulation of Dissolved Oxygen Transfer in a Lake with
a Saline Benthic Layer
Simulations were made to show how dissolved oxygen concentration profiles would
change in the lake under the following scenarios 1) uniform salinity concentrations
throughout the water column 2) observed salinity concentrations for 2008. The weather
data for the year 2008 were used along with the initial temperature profile. The initial
salinity profile was used for scenario 2 and a constant salinity was used for scenario 1.
A simplified dissolved oxygen model was used to simulate the vertical dissolved oxygen
109
0
10
20
30
Air
Tem
p (!
C)
-10
0
10
20
Dew
Poi
nt T
emp
(!C
)
5
10
Win
d sp
eed
(m
/s)
2,000
4,000
6,000
8,000
Sol
ar ra
diat
ion
(Kca
l/m2 )
04/18/07 07/27/07 11/04/070
0.5
1
Clo
ud C
over
(%
)
04/18/08 07/27/08 11/04/08
Figure 5.3: Weather parameters recorded at the Minneapolis/St. Paul InternationalAirport. and used as model input.
110
profile in the lake.∂O
∂t=
1A
∂
∂z
(AKz
∂O
∂z
)− SSOD (5.29)
where A, Kz and z are previously defined in equations 5.1 and 5.2, O is the dissolved
oxygen concentration (mg/L) and SSOD is the sediment oxygen demand (mg L−1 day−1).
All other sources and sinks of dissolved oxygen such as photosynthesis, plant respiration
and biochemical oxygen demand are ignored. Sediment oxygen demand (SSOD) is
defined by equations (5.30) and (5.31).
SSOD =Sb
A
∂A
∂z(5.30)
Sb = Sb20θT−20S (5.31)
where Sb is the sediment oxygen demand per unit area (g O m−2 day−1), A is the
horizontal lake area at a particular depth, and ∂A∂z is an estimate of the amount of
sediment surface area that is in contact with a particular water layer. A SSOD value
is calculated for each water depth. The Sb value is calculated using equation (5.31)
where Sb20 is an estimate of the sediment oxygen demand at 20 oC, θS is a temperature
adjustment coefficient and T is the temperature at the particular water depth. Sb20 is
estimated based on the trophic state of the lake. Tanners Lake is a mesotrophic lake. A
value of 1 g O m−2 day−1 was used for the value of Sb20 [134, 135]. Values for θS used
were 1.065 for T>10 oC and 1.130 for T <= 10 oC [136].
As a boundary condition, the DO concentration in the top layer of the lake was
always set to the saturation concentration of dissolved oxygen in water (equations (5.32)
and (5.33)) [136].
ln(CSO) =− 139.34411 +1.575701 ∗ 105
T− 6.642308 ∗ 107
T 2(5.32)
+1.2438 ∗ 1010
T 3− 8.621949 ∗ 1011
T 4
CS = CSO ∗ (1.0− 0.000035∆H) (5.33)
where T is the water temperature CSO is the saturation concentration at sea level and
CS is the dissolved oxygen saturation concentration adjusted for the elevation of the lake
111
above sea level (∆H (ft)). ∆H was set to 293.6 m [124]. The initial concentration profile
was set to 0 throughout the water column except for the surface layer, which was set to
the saturation concentration. In the computation of the DO profiles convective mixing
due to density instability and wind mixing in the surface layer were also accounted
for in separate computational steps, the same as in the transfer of salinity and heat
(temperature).
5.4 Results
5.4.1 Saline Layer Formation: Continuous lake monitoring results
The continuous record of water temperature and specific conductance profiles collected
in Tanners Lake from Nov 28 2008 to July 31 2009 showed the formation of a saline
water layer on the bottom of an urban lake in winter and the dissipation in spring
and summer. The conductivity probes experienced significant creep while in the lake.
Profiles were taken with the YSI Model 63 probe on Nov 19 2008, Feb 24 2009 and May
11 2009 and compared with daily average concentrations recorded by the Buoy (Table
5.1). ). On a plot of YSI data vs. buoy data a slope was determined representing the
daily creep for each of the buoy sensors. This daily creep value for individual probes
was subtracted from the recorded value for individual probes to produce the corrected
specific conductance values (Figure 5.4).
The water temperature and salinity record displayed the influence that runoff con-
taining road salt can have on a lake (Figure 5.5). In late fall of 2008 the salinity, defined
as the (specific conductance of the water,) was uniform throughout the water column.
As the winter progressed a saline water layer began to formed at the bottom of the lake.
The accumulation of saline water continued throughout the winter months resulting in
the formation of a saline layer approximately 5 meters thick above the bottom sediments
by April 2009.
The thickness of the saline benthic layers increased abruptly at certain times coin-
cident with daily average air temperatures above 0 oC (Figure 6 points A, B and C).
When the average daily air temperature reached above freezing causing the existing
snowpack to melt, the specific conductance (salinity) of the bottom waters in Tanners
Lake increased and/or the thickness of the saline layer expanded.
112
Table 5.1: Comparison of specific conductance recorded by the Buoy system to valuesmeasured with the YSI Model 63 probe. A slope representing the daily creep of theprobes was calculated using the difference between the values from the buoy system andthe values from a YSI Model 63 probe.Depth 11/19/08 2/24/08 5/11/08 Difference Slope
Figure 5.4: Specific conductance recorded at depths of 1.22 m and 12.19 m . Solid linesrepresent raw data collected by the probes. Dashed lines represent corrected data afterrecalibration.
Figure 5.5: Isopleths of recorded (measured) specific conductance (top) and isotherms(bottom) in a depth vs. time plot. Values recorded in Tanners Lake from 28 Nov 2008to 31 July 2009.
114
600800
1000120014001600
Sur
face
wat
erC
ondu
ctan
ce
(!S
/cm
)
8001,0001,2001,4001,600
Ben
thic
wat
erC
ondu
ctan
ce(!
S/c
m)
Nov Jan Mar May0
0.5
1
Pre
cipi
tatio
nW
ater
Equ
ival
ent
(inch
es)
5
10
Sno
w D
epth
(inch
es)
0
20
40
Dai
ly A
vera
geA
ir Te
mpe
ratu
re("
C)
A B C D
Figure 5.6: From top to bottom: Specific conductance at the surface of the lake (seriesdepths from surface 1.2, 2.1, 3.0. 6.1 meters), Specific conductance at the bottom of thelake (series depths from surface in order of highest observed specific conductance valuesto lowest 13.4, 12.8, 12.2, 11.6, 11.0, 10.4, 9.1), daily average air temperature, dailyaverage snow depth, and precipitation water equivalent. Points A, B and C representstimes when accumulation of saline water occurred at the bottom of the lake. Point Drepresents when a fresh water intrusion was observed in the lakes surface water. Weatherdata were collected by the Minnesota Climatology working group.
115
The accumulation of salt laden (snowmelt) runoff water at the bottom of the lake
continued until ice-out. Ice-out occurred between March 31 and April 2, based on ice-out
dates of similar sized lakes located near Tanners Lake (Kohlman, Gervais, and Harriet;
[137]).
Right before ice-out a water intrusion occurred in the surface waters of the lake
(Figure 5.6 point D). During this time a layer with low specific conductance and tem-
peratures that reached 7 oC formed below the ice surface (Figure 5.4). The water for
these intrusions came from a few snowfalls that quickly melted and rainfall that oc-
curred from March 23 to April 1. The air temperatures during this period were above
freezing and likely most of the road salt had already been washed away allowing a warm
and less saline layer to form under the ice cover.
Shortly after the ice layer on the lake had melted, the lake began to mix (spring
overturn). Only the top 10 meters of the 14 m deep lake mixed after ice-out, resulting
in a saline water layer of about 4 m thickness at the bottom of the lake. The density
of this saline layer was strong enough to resist the convective and wind mixing events
right after ice out, i.e. the lake had become monomictic.
After ice-out the salinity of the benthic layers increased even though road salt ap-
plications had stopped. Shortly after the erosion of the saline layer began. This erosion
continued, but was not strong enough to remove the saline layer. By July 31st Tanners
Lake still has a saline benthic layer.
5.4.2 Lake Stratification: Model calibration results
The best-fit model parameters obtained by model calibration are given in Table 5.2.
WSTR and WFCT are wind-sheltering coefficients. Typical values for wind sheltering
coefficients range from 0.1 to 1 [120]. The value of 0.25 for Tanners Lake is low pointing
towards a well-sheltered lake.
Table 5.2: Best fit parameters from model calibrationWSTR 0.25WFCT 0.44Zsd (m) 3.00α (m2/day) 0.19
116
A maximum hypolimnetic effective diffusivity (Kzmax) was also determined by cali-
bration. The best-fit value for this parameter was 0.19 m2/day.
The Secchi depth determined by model calibration was 3 m. It matched the mea-
sured average Secchi depth in 2007 and 2008 in Tanners Lake of 3 meters with a range
from 1.8 to 5 m [124].
The temperature and specific conductance profiles obtained from the calibrated
model for the year 2008 were comparable to the observed values (Figure 5.7). Com-
parisons between observed and modeled values at three depths including the waters
surface, the mid-depth of the lake and the lake bottom depth are given in Figure 5.8.
Values obtained were consistent for temperature and salinity at the three depths. Dif-
ferences between the two years were significant. In 2007, values in the middle of the
lake decreased rapidly and remained even with the surface concentrations shortly after
ice out. The deep layer also eroded quickly, compared to the values in 2008, resulting
in the removal of the chemocline shortly after ice out for most of the water column. In
2008 the chemocline persisted until fall.
The average RMSE values for the temperature profiles in 2007 and 2008 were 1.06
and 1.10 oC respectively. The average RMSE values for the specific conductance profiles
in 2007 and 2008 were 57 and 60 µS/cm respectively. The temperature profiles for
both 2007 and 2008 compare well with the observed values throughout the year 5.9.
Calibration was only conducted with the 2008 data, yet the RMSE values for the 2007
temperature series are comparable to the 2008 values. The RMSE values for the specific
conductance are also consistent throughout the year 5.9. For the 2008 simulation, RMSE
values hover between 50 and 60 µS/cm. In 2007 values around 50 µS/cm are observed
during the first 5 months of the simulation, but increase beyond that point. In the later
months of the simulation concentrations in the surface waters of the observed dataset
are reduced in 2007. This is likely caused by freshwater inflows from rainfall. Rainwater
inflows are not included in the model resulting in an increase in RMSE values in the
epilimnion.
5.4.3 Lake Stratification and Vertical Mixing: Model simulation re-
sults
Temperature and specific conductance profiles
117
0
5
10
07-May-2008 30-May-20080
5
10
30-Jul-2008
Dep
th (m
)
15-Aug-2008
0 10 20
0
5
10
1525-Sep-2008
800 1,200 1,600 0 10 2023-Oct-2008
800 1,200 1,600
Temperature !C Conductance "S/cm Temperature !C Conductance "S/cm
Figure 5.7: Vertical profiles of measured (dashed line) and simulated (solid line) watertemperatures and specific conductance in Tanners Lake in 2008.
118
4/22/07 6/11/07 7/31/07 9/19/07 11/7/07
600
800
1000
1200
1400
1600
1800
Con
duct
ivity
(!S
/cm
)
0 m7 m14 m0 m14 m7 m
4/22/07 6/11/07 7/31/07 9/19/07 11/7/070
5
10
15
20
25
30
Tem
pera
ture
(!C
)
4/22/08 6/11/08 7/31/08 9/19/08 11/7/08
600
800
1000
1200
1400
1600
1800
Con
duct
ivity
(!S
/cm
)
4/22/08 6/11/08 7/31/08 9/19/08 11/7/080
5
10
15
20
25
30
Tem
pera
ture
(!C
)
Figure 5.8: Time series of specific conductance (left) and water temperatures (right) atthe lake surface (0m), at mid-depth (7m) and at the bottom (14m) of Tanners Lake,measured and simulated for 2007 (top) and 2008 (bottom).
Figure 5.9: Root mean square error (RMSE) between modeled and observed watertemperature and specific conductance profiles in 2007 and 2008.
Model simulations were made for the ice-free periods of 2007 and 2008. The initial
conditions used in the model for temperature and specific conductance were obtained
from field measurements using an YSI Model 63 probe (Figure 5.10). Initial values were
acquired as close to the ice-out date as possible on 1 April 2007 and 22 April 2008.
Estimates of ice-out dates in Tanners Lake are March 26-27, 2007 and April 1821, 2008
respectively. These ice out dates were determined from similar sized lakes located near
Tanners Lake (Kohlman, Gervais, and Harriet; [137]).
The two initial conductivity profiles are approximately similar but the two initial
temperature profiles are different. The temperature of the surface waters in 2008 were
much warmer than in 2007, but the deeper layers were colder in 2008 than in 2007.
In 2008 the saline layer persisted throughout the summer and diluted gradually by
effective diffusion into the overlying water. By fall the density of the bottom layer was
reduced due to heating and gradual mixing with hypolimnetic water, allowing for com-
plete mixing (Figure 5.11). In 2007 the chemocline eroded quickly after ice-out resulting
in an almost completely mixed water column (Figure 5.3).
Stratification stability
Density gradients caused by temperature differences and those caused by salinity were
120
!"
#"
$"
%"
&"
'!"
'#"
'$"
'%"
!" (!!" '!!!" '(!!" #!!!"
!"#$%&'()&
*#"+,-,+&+./01+$2/+"&'µ*3+()&
!" #" $" %" &" '!"
4"(#"52$15"&'.6)&
$)')!*"
$)##)!&"
Figure 5.10: Measured specific conductance (dashed line) and temperature (solid line)profiles used as initial conditions for simulations of the open water periods 2007 and2008.
calculated separately to evaluate the contribution of the salinity to lake stratification
(Figure 5.12).
In April, after ice-out, the density stratification in the lake was mostly caused by
salinity. On April 1 2007 all of the density stratification in the lake was from the salinity.
The largest salinity induced density gradient was located at 13 m and was equal to 0.17
Kg/m4. On April 22 2008 temperature stratification was present up to a depth of 4
m and the strongest density gradient (equall to 0.15 Kg/m4) caused by temperature
stratification was at a depth of 4m. Salinity stratification began at a depth of 8 m.
The strongest salinity induced density gradient (also equal to 0.15 Kg/m4) was at 10 m
depth .
121Depth (m)
4/01
/07
6/01
/07
8/01
/07
10/0
1/07
0 2 4 6 8 10 12 14
Specific Conductance (!S/cm)
600
700
800
900
1000
1100
1200
1300
1400
1500
1600
Depth (m)
4/01
/08
6/01
/08
8/01
/08
10/0
1/08
0 2 4 6 8 10 12 14
Specific Conductance (!S/cm)
600
700
800
900
1000
1100
1200
1300
1400
1500
1600
Depth (m)
4/01
/07
6/01
/07
8/01
/07
10/0
1/07
0 2 4 6 8 10 12 14
Temperature ( !C )
0510152025
Depth (m)
4/01
/08
6/01
/08
8/01
/08
10/0
1/08
0 2 4 6 8 10 12 14
Temperature ( !C )
0510152025
Fig
ure
5.11
:Is
ople
ths
ofsi
mul
ated
spec
ific
cond
ucta
nce
(top
)an
dis
othe
rms
(bot
tom
)in
ade
pth
vsti
me
plot
duri
ngth
eic
e-fr
eepe
riod
sof
2007
(lef
t)an
d20
08(r
ight
).
122
In May the density gradients were dominated by temperature. In 2007 a slight chem-
ical stratification was present in the bottom 11 meters, but almost all salinity induced
density gradients had disappeared. Temperature-induced density gradient reached 0.7
Kg/m4. In 2008 a more pronounces salinity layer was present in May, but the density
gradients caused by the salinity only reached 0.06 kg/m4 while temperature stratifica-
tion caused a maximum density gradient of 0.5 kg/m4.
In July the temperature stratification in both 2007 and 2008 caused density gradients
to reach up to 1.4 kg/m4. Salinity stratification was non-existent in July of 2007, but
it was still present in July of 2008 starting at about 8 meters.
In October of 2008 salinity stratification had been reduced to a maximum value of
0.03 kg/m4 at 7 m. Temperature stratification was reduced to about 0.4 kg/m4 for both
2007 and 2008.
Hypolimnetic effective diffusivity
In order to get the variable results between 2007 and 2008 the Lake Number had to
be included in the calculation of the hypolimnetic effective diffusivity. With out the
inclusion of this parameter either to much mixing occurred in the summer months when
temperature stratification was strong or to little mixing occurred in the early spring
and late fall months when temperature stratification was reduced.
Without the Lake Number the hypolimnetic Kz values would be constant throughout
the year year if no localized density stratification was present at that particular depth.
The Lake number changes with strength of the overall lake stratification as well as
with wind speed. In summer when thermal stratification was strong the Lake number
increased causing the maximum hypolimnetic eddy diffusion (Kz) to decrease (Figure
13). In fall and spring when density gradients were absent or weak mixing was increased
in the hypolimnion. When wind speeds are high and more wind energy was applied at
the lake surface Kz values increased
5.4.4 Dissolved Oxygen Modeling Results
Simulations were run using the 2008 temperature and salinity profiles to determine how
dissolved oxygen (DO) profiles in Tanners Lake would respond to the presence of a
benthic saline layer. Under the first scenario with no vertical salinity gradient (Figure
123
!!
"!
#!
$!
%!
&!!
&"!
&#!
&$!
!! !'"! !'#! !'$! !'%! &! &'"! &'#!
!"#$%&'()&
"#("*)'+,-(.)&
"#("$!
"#("$!
!"#$%&%'()*
!+,-.)*
/0102334*!!
"!
#!
$!
%!
&!!
&"!
&#!
&$!
!! !'"! !'#! !'$! !'%! &! &'"! &'#!
!"#$%&'()&
"#("*)'+,-(.)&
/02202335*
!!
"!
#!
$!
%!
&!!
&"!
&#!
&$!
!! !'"! !'#! !'$! !'%! &! &'"! &'#!
!"#$%&'()&
601502334*!!
"!
#!
$!
%!
&!!
&"!
&#!
&$!
!! !'"! !'#! !'$! !'%! &! &'"! &'#!
!"#$%&'()&
601502335*
!!
"!
#!
$!
%!
&!!
&"!
&#!
&$!
!! !'"! !'#! !'$! !'%! &! &'"! &'#!
!"#$%&'()&
401502334*!!
"!
#!
$!
%!
&!!
&"!
&#!
&$!
!! !'"! !'#! !'$! !'%! &! &'"! &'#!
!"#$%&'()&
401502335*
!!
"!
#!
$!
%!
&!!
&"!
&#!
&$!
!! !'"! !'#! !'$! !'%! &! &'"! &'#!
!"#$%&'()&
1301502334*!!
"!
#!
$!
%!
&!!
&"!
&#!
&$!
!! !'"! !'#! !'$! !'%! &! &'"! &'#!
!"#$%&'()&
1301502335*
Figure 5.12: Profiles of density gradients caused by salinity (solid line) and temperature(dotted line).
Figure 5.13: Simulated effective hypolimnetic diffusion coefficients near the lake bed(depth of 14m) plotted vs time.
5.14)the lake is dimictic, i.e. it turns over in fall and spring and the entire water column
is oxygenated on both occasions. In midsummer the lake bottom goes anoxic for about
100 days from July 10 to Oct 20.
Under the second scenario, using identical weather conditions, but introducing a
saline layer at the bottom of the lake produced a very different result. The benthic
saline layer reduced vertical mixing especially near the lake bed. In spring the benthic
saline layer prevented the complete convective mixing of the entire water column. In
fall, when the saline layer was diluted, vertical lake mixing was delayed, but eventually
the saline layer was eroded and the surface waters mixed with the benthic waters late
in November. At that time, DO was transport from the overlying water to the lake
sediment. Overall DO did not reach the lake bottom during the entire open-water
period from April to November (Figure 13 bottom)
Vertical dissolved oxygen profiles were very similar in late summer for the two simu-
lated scenarios, with and without a benthic saline layer, when temperature stratification
was at a maximum.
125D
etpt
h (m
)
6/1/08 8/1/08 10/1/08
0
2
4
6
8
10
12
14
Dissolved O
xygen (mg/L)
0
1
2
3
4
5
6
7
8
9
(a)
Dep
th (m
)
6/1/08 8/1/08 10/1/08
0
2
4
6
8
10
12
14
Dissolved O
xygen (mg/L)
0
1
2
3
4
5
6
7
8
9
(b)
Figure 5.14: Isopleths of simulated dissolved oxygen concentrations in a depth vs. timeplot without a saline layer (top) and with a saline layer (bottom) for the 2008 testscenario.
126
5.5 Discussion
5.5.1 Interpretation of Measurements (2007-2009) and Simulation Re-
sults (2007-2008)
Equations developed for the MINLAKE temperature model [133][120]were used to sim-
ulate temperature and salinity profiles in a stratified lake during the open-water season.
A conversion of salinity (specific conductance) to density and an effective hypolimnetic
diffusion coefficient linked to a Lake Number were introduced to accurately simulate
temperature and salinity profiles in Tanners Lake, a lake in the Twin Cities metropoli-
tan area of Minnesota. The dynamic lake model used daily weather parameters and
initial temperature and salinity profiles measured after ice-out as inputs. Daily sim-
ulated data and a few observed profiles for 2007 and 2008 and continuously recorded
profiles for 2008/2009 were compared. The interpretation of the three years of data and
two years of simulations focused on the saline layer.
In 2007 the saline layer eroded away a few weeks after ice-out resulting in an almost
completely mixed water column throughout the summer. In 2008 and 2009 this did
not occur; instead the saline benthic layer persisted throughout the summer. The main
cause for this significant and consequential difference appears to be a 9-day cooling
period between April 4 and April 12, 2007 right after ice-out where air temperatures
and dew point temperature were below 0 oC 5.3. The prolonged stratification instability
allowed for the effective hypolimnetic diffusion caused by wind mixing to erode away
the saline layer before temperature stratification was able to reduce the mixing between
the hypolimnion and epilimnion.
In 2008 and 2009 a steady warming of the surface waters followed ice-out. This warm-
ing quickly created a thermocline, which reduced the mixing between the hypolimnion
and the surface waters effectively limiting the erosion of the saline layer.
Convective mixing appeared to be a minimal component in the erosion of the saline
layer. Convective mixing alone was not strong enough to mix the surface waters with
the salinity layer in 2007, 2008 or 2009. In 2007 the density increase due to the heating
of the water to TMD (just below 4 oC) from 3.5 oC was only 0.002 kg/m3. At this same
time the density increase due to the change in salinity between the surface waters and
the hypolimnion was 0.42 kg/m3. In 2008 when the hypolimnetic water temperatures
127
were much cooler the difference in density caused by heating the water from 3 oC to TMD
amounted to 0.007 kg/m3 compared to the density difference of .47 kg/m3 between the
surface waters and the hypolimnion caused by salinity. The large differences between
the added density from salinity and the added density from heating the surface waters
to TMD eliminated the complete convective mixing of the water column. This means
that the major driving force for the erosion of the saline layer during the spring and fall
was wind energy.
The density stratification right after ice-out was dominated by salinity stratification.
Throughout the summer, however, when the temperature of the surface water heated
up, the density difference across the thermocline far outweighed the density gradients
caused by the chemocline. Consequently, it is the formation of the thermocline that
prevents the mixing of the saline layer with the surface waters during the summer. If
only the chemocline was present the density stratification caused by the salinity gradient
would not be sufficient to withstand wind mixing from top to bottom.
The data recorded continuously in 2009 at 2-minute intervals clearly showed the
accumulation of a saline water layer at the bottom of Tanners Lake. This layer grew
during the winter months, when melting events occurred, until ice-out (Figure 5.5).
The accumulation of salt water at the bottom of the lake suggests that the snowmelt
runoff plunged upon entering the lake to form a density current that flow to the lake
bottom. This flow pattern was documented using only temperature probes in Ryan
Lake in Minnesota [54].
Right after ice-out a final salinity increase was recorded in the deepest part of Tan-
ners Lake (Figures 5.5 and 5.66). At least two different processes could account for this:
(1) mixing between the sediment layer and the benthic water could have uprooted salt
stored in the sediments [138, 44, 71] or (2) arrival of a delayed snowmelt runoff traveling
through extended pathways.
Shortly after this final input of saline water, the erosion of the saline benthic layer
began. This erosion was not enough to completely mix the lake before a thermocline
formed in the lake reducing the vertical transport of mass, momentum and energy from
the lake surface to the lake bottom.
128
5.5.2 Effects of Benthic Saline Layer Formation on Lake Water Quality
The intermittent or persistent presence of a benthic saline layer at the bottom of an
urban lake can have a number of consequences for lake water quality. The saline layer
prevents natural convection and mixing of the surface waters with the benthic water.
This prevents oxygen from reaching the botom of the lake lengthening the anoxic periods
of the lake sediments [61]. This effect was clearly shown in the simulated DO profiles
(Figure 5.14 bottom). The saline layer reduced mixing from convection as well as
wind mixing of the hypolimnion, preventing oxygen from reaching the sediments in the
spring right after ice out. When the salt layer was removed oxygen was able to travel
throughout the water column due to convective and wind mixing (Figure 5.14 top).
DO remained in the hypolimnion until July when the oxygen demand of the sediments
finally reduced DO levels to those observed when a saline layer was present (Figure 5.14
bottom). The saline layer increased the anoxic period in the lower depths of the lake
by 3 months.
Biochemical (mostly microbial) DO uptake at the sediment water interface removes
DO from the water column, and diffusive transport from above usually replenishes it.
These processes were represented in the model. Lake bottom waters and lake sediments
become anaerobic when insufficient DO was supplied by convection or wind mixing from
the water above. Starting with zero DO at the lake bottom due to the presence of a
saline layer that prevents lake overturn, will cause continuing anoxia in summer on the
lake bed.
Under winter conditions, which were not modeled, there is no oxygen source in the
lake water (no surface aeration due to the ice cover, and no photosynthesis due to a
snow cover on the ice). Benthic DO continues, however, and in order to maintain aerobic
conditions near the lake, fall turnover and oxygen replenishment near the lake bed are
crucial.
A long anoxic period in the hypolimnion can facilitate the release of phosphorus and
metals from the sediments. Enhanced internal phosphorus release from the anaerobic
sediment [139, 140] increases cultural eutrophication. Increased chloride concentrations
have also been observed to increase the bioavailability of metals such as cadmium, lead,
chromium mercury among others [40, 37, 39, 4, 34]. By increasing the contact time
of the saline layer containing high concentrations of chloride with the sediments, the
129
release of metals from these sediments is increased.
Finally, the increased contact time of benthic organisms with elevated chloride con-
centrations can affect the biodiversity of the lake [59, 110]. Macro-invertebrates and
bottom feeding organisms would have to adjust to the increased length of the anoxic
period at the lake bed as well as increased chloride concentrations. All of these con-
sequences of the saline benthic layer formation are detrimental to water quality and
aquatic life in the lake.
5.6 Summary and Conclusions
Two years (2007, 2008) of intermittently measured temperature and specific conduc-
tance profiles and one year (2009) of continuously monitored data were used to show
the formation of a benthic saline layer in winter, and its effect on summer stratification
and mixing dynamic in Tanners Lake, Oakdale, Minnesota.
Erosion of the saline layer in the spring occurred in only one of the three years studied
(2007). In the other two years (2008 and 2009), the saline layer persisted throughout
the summer, and was destroyed only by fall turnover and mixing between the epilimnion
and hypolimnion when thermo stratification was at a minimum.
Simulations showed that from ice-out in April 2008 until November 2009 the saline
layer was able to prevent dissolved oxygen from reaching the lake sediments. Without
the saline layer the lake sediments would have been anoxic only from the beginning
of July until the end of October 2008. With the addition of a saline layer the anoxic
period experienced by the lake sediments persisted throughout the spring and summer.
Simulations also showed that in the fall turnover of the lake was delayed, when the
saline layer was present, until the end of November.
The results of this study provide information on the formation, the mixing and the
consequences of a benthic saline layer in a northern temperate urban lake that receives
snowmelt runoff from roads on which salt (NaCl) has been applied to increase driving
safety. Specific results are:
(1) The formation of the benthic saline layer has been documented in detail. The
record consists of specific conductance profiles recorded at 2-minute intervals from 28
Nov 2008 until 31 July 2009 in Tanners Lake in Oakdale, Minnesota. The formation of
130
the saline benthic layer is episodic, and follows the air temperature pattern.
(2) Natural convective mixing in spring or fall (spring and fall turnover) is not
intensive enough to remove the saline layer after ice out. Mixing of the hypolimnion by
wind energy applied at the water surface is needed to completely erode the saline layer.
(3) The formation of monomixis (one complete lake turnover per year) or meromixis
(no complete lake turnover per year) in a northern temperate lake due to road salt
application in the watershed appears to be contingent on both the strength (salt con-
centration) of the saline layer and the timing of the seasonal thermocline formation after
ice out.
(4) Formation of a seasonal thermocline effectively reduces the transport of wind
energy from the epilimnion to the hypolimnion, and hinders the erosion of the saline
layer. Lakes number was used in the model simulations to capture this effect.
Overall the presence of a saline benthic layer due to runoff containing road salt
(NaCl) was shown to disrupt the natural mixing mechanics of a dimictic lake. This
disruption has significant consequences for a lakes water quality and ecology.
Aknowledgements
We acknowledge the Minnesota Local Road Research Board (LRRB) and the University
of Minnesota Doctoral Dissertation Fellowship Program for providing support to com-
plete this research. We also thank Ben Erickson and Chris Ellis from the St. Anthony
Falls Laboratory for their assistance in designing and implementing the Buoy system.
References
[1] J. Marsalek. Road salts in urban stormwater: An emerging issue in stormwater
management in cold climates. Water Science and Technology, 48(9):61–70, 2003.
[2] United Stated Geological Survey. Salt statistics and information, 2007.
[3] R. B. Jackson and E. G. Jobbagy. From icy roads to salty streams. Proceedings of
the National Academy of Sciences, 102(41):14487–14488, October 11, 2005 2005.
[4] V. Novotny, D. W. Smith, D. A. Kuemmel, J. Mastriano, and Al Bartosova. Urban
and highway snowmelt: Minimizing the impact on receiving water. Technical
Report Pr. 94-IRM-2, Water Environment Research Foundation, Alexandria, VA,
1999.
[5] E. L Thunqvist. Regional increase of mean annual chloride concnetration in water
due to the application of deicing salt. Science of the Total Environment, 325:29–37,
2004.
[6] Environment Canada Health Canada. Priority substances list assessment report
Bryant LakeC = Measured Conductivity (υS/cm)T = Temperature (oC)SC = calculated specific conductance (υS/cm)Depth
(m) C T SC C T SC C T SC C T SC C T SC C T SC C T SC C T SC0 274 0.8 510 280 0.8 521 419 14 530 510 23 530 524 27.4 501 441 16.4 528 320 6.5 495 310 1 572
Cedar LakeC = Measured Conductivity (υS/cm)T = Temperature (oC)SC = calculated specific conductance (υS/cm)Depth
(m) C T SC C T SC C T SC C T SC C T SC C T SC C T SC C T SC0 307 1.1 564 249 0 477 442 13.6 565 550 21.7 587 506 26.5 492 423 15.9 512 347 6.3 540 317 0 607
Lake GervaisC = Measured Conductivity (υS/cm)T = Temperature (oC)SC = calculated specific conductance (υS/cm)Depth
(m) C T SC C T SC C T SC C T SC C T SC C T SC C T SC C T SC0 700 26.4 682 539 16.4 645 451 8.8 653 390 0.2 741 405 1.1 745 432 5.9 680 625 17.1 736 747 24.1 742
9/5/07 COTTAGE_GR 315 2379/19/07 COTTAGE_GR 365 26610/3/07 COTTAGE_GR 343 25710/17/07 COTTAGE_GR 343 26210/31/07 COTTAGE_GR 348 26311/14/07 COTTAGE_GR 330 26311/28/07 COTTAGE_GR 368 26312/12/07 COTTAGE_GR 389 28112/26/07 COTTAGE_GR 376 2761/9/08 COTTAGE_GR 376 2551/23/08 COTTAGE_GR 335 2472/6/08 COTTAGE_GR 351 2722/20/08 COTTAGE_GR 327 2653/5/08 COTTAGE_GR 322 2503/19/08 COTTAGE_GR 372 2654/2/08 COTTAGE_GR 372 2634/16/08 COTTAGE_GR 350 2474/30/08 COTTAGE_GR 336 2515/14/08 COTTAGE_GR 330 2385/28/08 COTTAGE_GR 336 2456/11/08 COTTAGE_GR 302 2166/25/08 COTTAGE_GR 356 2277/9/08 COTTAGE_GR 356 2187/23/08 COTTAGE_GR 337 54.68/6/08 COTTAGE_GR 334 2458/20/08 COTTAGE_GR 376 2459/3/08 COTTAGE_GR 342 2419/17/08 COTTAGE_GR 387 25910/1/08 COTTAGE_GR 388 26910/15/08 COTTAGE_GR 366 26210/29/08 COTTAGE_GR 336 24111/12/08 COTTAGE_GR 370 2676/20/07 METRO 246 1897/3/07 METRO 209 1667/11/07 METRO 248 1877/24/07 METRO 223 1737/25/07 METRO 232 1807/26/07 METRO 238 1877/27/07 METRO 234 1967/30/07 METRO 217 1877/31/07 METRO 218 1858/1/07 METRO 232 1958/2/07 METRO 239 2028/3/07 METRO 234 1968/6/07 METRO 206 1708/7/07 METRO 220 1718/8/07 METRO 228 1888/22/07 METRO 211 1809/5/07 METRO 200 180
9/19/07 METRO 223 19410/3/07 METRO 226 18310/17/07 METRO 225 17510/31/07 METRO 251 19411/14/07 METRO 232 19211/28/07 METRO 255 18712/12/07 METRO 268 20512/26/07 METRO 260 1961/9/08 METRO 266 2071/23/08 METRO 270 1702/6/08 METRO 268 2072/20/08 METRO 248 1983/5/08 METRO 272 1893/19/08 METRO 278 1944/2/08 METRO 297 2244/16/08 METRO 248 1844/30/08 METRO 248 1745/14/08 METRO 263 1725/28/08 METRO 232 1586/11/08 METRO 240 1756/25/08 METRO 244 1747/9/08 METRO 240 6337/23/08 METRO 242 1678/6/08 METRO 248 1838/20/08 METRO 258 1869/3/08 METRO 208 1629/17/08 METRO 270 18710/1/08 METRO 269 20010/15/08 METRO 230 18610/29/08 METRO 232 19611/12/08 METRO 244 1996/20/07 SENECA 298 4507/3/07 SENECA 274 3907/25/07 SENECA 274 4208/1/07 SENECA 273 4248/8/07 SENECA 296 4828/22/07 SENECA 235 3989/5/07 SENECA 249 3749/19/07 SENECA 291 46210/3/07 SENECA 263 39310/17/07 SENECA 265 48610/31/07 SENECA 257 39211/14/07 SENECA 278 34811/28/07 SENECA 278 44112/12/07 SENECA 294 39212/26/07 SENECA 312 4631/9/08 SENECA 292 3851/23/08 SENECA 276 4252/6/08 SENECA 294 452