Estimations of copper roof runoff rates in the United States
-
Upload
ray-arnold -
Category
Documents
-
view
214 -
download
1
Transcript of Estimations of copper roof runoff rates in the United States
![Page 1: Estimations of copper roof runoff rates in the United States](https://reader035.fdocuments.us/reader035/viewer/2022081212/575004c01a28ab1148a07a96/html5/thumbnails/1.jpg)
Estimations of Copper Roof Runoff Rates in the United StatesRay Arnold*
Copper Development Association, 260 Madison Avenue, New York, New York 10016, USA
(Received 6 January 2005; Accepted 26 April 2005)
EDITOR’S NOTE:Additional supporting data are found in an appendix available on the online edition of IEAM volume (1), Number (4).
ABSTRACTCopper (Cu) concentrations in waterways of the United States are of widespread concern. Presently, 692 waterway
segments around the United States are listed by the U.S. Environmental Protection Agency (USEPA) as having unacceptably
high copper concentrations. As part of their water quality management strategy, the USEPA is mandated to understand
and manage sources and impacts of nonpoint releases of chemicals of concern. One potential nonpoint source of Cu is the
runoff of precipitation falling onto Cu used in external architecture (e.g., roofing). However, few studies of Cu roof runoff
have been published. This article is intended to provide estimations of Cu runoff rates and concentrations across the
United States. Copper runoff rates and concentrations are predicted at 179 locations with a recently developed model. The
average and range (in parentheses) of annual Cu loading rates, based on roof area; Cu export rates, based on amount of
precipitation; and Cu concentrations for the United States are 2.12 (1.05–4.85) g Cu/m2/y; 2.72 (0.69–16.48) mg Cu/m2/
mm; and 2.72 (0.69–16.48) mg Cu/L as total Cu, respectively. Statistics are presented that describe site-specific data
distributions for use in probabilistic exposure and risk assessments. The effects of air quality as well as the potential fate
and risks of Cu from roof runoff are discussed.
Keywords: Copper Risk Roof runoff Modeling Stormwater
INTRODUCTIONCopper (Cu) is one of the most important metals to man,
ranking 3rd in world consumption behind aluminum and steel
(Jolly 2000). Man has used Cu and its alloys for approx-
imately 9,000 y (Landner and Lindestrom 1999). Copper
sheeting has been used for centuries in architectural applica-
tions (Sundberg 1998) in both commercial and residential
buildings because it is fire-retardant (historically important),
durable, and attractive, forming a highly valued patina as it
ages. Some applications of copper are in external cladding
(e.g., roofs, flashing, gutters, facades, and other decorative
ornamentation). Supply of strip, sheet, and plate Cu products,
a portion of which is used in exterior architectural applica-
tions, averaged 168.3 million kg/y in the United States from
2001 to 2004 (Copper Development Association 2005).
Exterior use exposes copper to weathering forces such as
wind and all forms of meteorological precipitation. This leads
to dissolution of copper from product surfaces and its
introduction into local watersheds. The amount of Cu
dissolved and transported from exposed copper surfaces is a
function of atmospheric chemistry, precipitation rates, and
roof orientation. The potential for the exposure of organisms
in local watersheds is a function of the amount of Cu dissolved
and transported from the exposed surfaces to the watershed.
However, characteristics of a watershed affect exposure and,
thus, the risk of copper in runoff. These characteristics include
the amount of Cu used in architectural applications in the
watershed, as well as the dilution and assimilation capacity of
a wide variety of natural and human-made substrates that
transform, sequester, and dilute Cu before or soon after runoff
enters the watershed.
In most cases, limited input of copper to the environmentcan be benign or even beneficial. Unlike synthetic environ-mental contaminants, Cu is an essential micronutrient tohumans, aquatic organisms, and other terrestrial organisms(Richter 1978; Fernandes and Henriques 1991; Gabel et al.1994; Culotta et al. 1995; Harris and Gitlin 1996; Linder andHazegh-Azam 1996; Olivares and Uauy 1996; Slekar et al.1996; Dallinger et al. 1997; Uauy et al. 1998). Organismshave evolved in the presence of Cu and developed both a needfor and a means to regulate Cu. It is common practice toincrease Cu levels in soils through the application of fertilizersto lawns, pastures, and crops and to supplement Cu in foodsof both humans and livestock for beneficial reasons. However,Cu is recognized as being toxic to sensitive organisms in waterat concentrations in the low lg/L range.
Understanding the risk potential of Cu in roof runoff iscomplicated, site specific, and beyond the scope of this study.However, quantifying copper in runoff as it leaves rooftops isan essential 1st step in determining potential exposure andrisk to organisms in watersheds. Copper runoff data are usedalong with estimations of the quantity of Cu roofing in awatershed to estimate the percentage of contribution fromCu roofing to Cu loads measured in stormwater discharged tolocal waterways. This is critical for identifying the actual causeof unacceptable Cu concentrations and designing the bestmanagement practices to reduce Cu loadings to local water-ways because other sources of Cu (e.g., brake pads, fertilizers,pesticides, industrial site runoff) can also contribute sub-stantial loads of Cu to stormwater runoff.
Spatial and temporal variations in atmospheric conditionsaffect the rate of dissolution of Cu in roof runoff. ElementalCu (Cu0), the form found in Cu products, is highly insolublein water. However, Cu naturally forms moderately solublecorrosion products when exposed to the weather. The
* To whom correspondence may be addressed [email protected]
Integrated Environmental Assessment and Management — Volume 1, Number 4—pp. 333–342� 2005 SETAC 333
Orig
inalRese
arch
Review
![Page 2: Estimations of copper roof runoff rates in the United States](https://reader035.fdocuments.us/reader035/viewer/2022081212/575004c01a28ab1148a07a96/html5/thumbnails/2.jpg)
corrosion products vary in composition and are directlyaffected by the chemistry of the atmosphere and precipita-tion. Chemistry of the atmosphere and precipitation variesgeographically and, thus, corrosion products of Cu applica-tions can vary geographically. However, in general, the copperoxide, cuprite (Cu2O), and the copper hydroxysulphates,brochantite [Cu4SO4(OH)6] and posnjakite [Cu4SO4(OH)6-
H2O], are common corrosion products of Cu claddings(Graedel 1987; Kratschemer et al. 1997). Near marineenvironments, chloride may be present in the atmosphere insufficient amounts to cause a portion of the corrosion productto be the Cu chloride atacamite [Cu2Cl(OH)3] (Graedel1987). Cuprite layers begin forming within seconds to hoursof exposure. Posnjakite and atacamite take weeks to monthsto begin forming, and brochantite may take years. Cuprite isgenerally found closest to the Cu metal, with the othercorrosion layers closer to the surface (Graedel 1987;Kratschemer et al. 1997). Although a variety of atmosphericchemistries may affect copper dissolution on rooftops,Odnevall Wallinder et al. (2004) have concluded that rainacidity predominately determines the dissolution process, thatchloride and sulfate only negligibly increase dissolution, andthat nitrate has a small inhibitory effect.
The conversion of insoluble elemental Cu into moderatelysoluble Cu compounds and subsequent dissolution is slow,accounting for Cu’s durability. Although dissolution of Cu isrelatively minor from a material-life perspective, this processdoes create the potential for Cu exposure to organisms. Thus,it is important to understand the amount of soluble Cu inrunoff of architectural applications because large quantities ofCu materials are produced and used in architecturalapplications. Worldwide only a few studies of Cu roof runoffhave been published, and only 3 studies have documentedrunoff rates at 9 locations in the United States. Moreover, theU.S. Environmental Protection Agency (USEPA) total max-imum daily load and nonpoint source control programs(USEPA 1991,1993) make it necessary to determine sourcesand assess the potential impacts of both point and nonpointsources of chemicals of concern within potentially impairedwatersheds. At the time of the preparation of this article, theUSEPA had listed 692 water body segments in the UnitedStates as impaired due to Cu. Since 1996, 142 total maximumdaily loads assessments limiting Cu discharge to attainacceptable Cu concentrations in listed water bodies had beenconducted and approved (USEPA 2005).
Whereas data for point sources are often available, data fornonpoint sources are often lacking. Monitoring of individualbuildings is expensive and unlikely to occur on a wide scale.Thus, there is a need for estimates of Cu runoff rates fromCu-covered structures in watersheds across the United States.However, until recently a precise model to predict Cu runoffrates did not exist. Odnevall Wallinder et al. (2004) haverecently developed such a model. This model provides aconsistent method to make estimates of Cu roof runoff andcan be applied locally or across large-scale geographic areas.The objective of this study was to use the model with rainfallmonitoring data to predict runoff rates (g Cu/m2 of roof/y),export rates (mg Cu/mm of precipitation/m2 of roof), andconcentrations (mg Cu/L) of Cu in roof runoff at locationsthroughout the United States. Additional estimations aremade to quantify the effect of acid rain on copper-loadingrates in an effort to evaluate the benefits of air qualitymanagement efforts on reducing Cu dissolution.
METHODS AND MATERIALS
The model
Experimental data from parallel field and laboratoryinvestigations of pure and naturally patinated (0–145 y old)Cu were used by Odnevall Wallinder et al. (2004) to developthe model. In addition, Odnevall Wallinder et al. (2004)compiled data from 10 publications containing 27 data sets,representing most of the existing laboratory and field studiesmeasuring runoff rates from external Cu structures subjectedto various environmental factors. Five of the data sets werenot used by Odnevall Wallinder et al. (2004) because the pHof the precipitation was not measured. Based on the data set, asingle model was derived to estimate total Cu (i.e., dissolvedand particulate Cu) runoff rates. It was determined byOdnevall Wallinder et al. (2004, equation 8) that annualprecipitation rate, precipitation pH, and angle of roofinclination were critical variables in the model
PðV ;h; pHÞ ¼ ð1:04þ 0:96V � 10�0:62pHÞ � cosðhÞcos458
� �ð1Þ
where P(V, h, pH) is the predicted Cu runoff rate (g Cu/m2/y); Vis the annual rate of precipitation (mm/y); pH is the pH of theprecipitation; and h is the angle of the surface (8). Theexpression cos(h)/cos458 corrects the prediction of runoff ratesfor surfaces that deviate from 458. The addition of the constant(1.04) corrects the model runoff rate to account for the 1st-flush effect, where metal runoff rates are higher during theinitial portion of a runoff event, as noted in previous studies byHe (2000), as well as Boulanger and Nikolaidis (2001, 2003a,2003b). Odnevall Wallinder et al. (2004) reported that 70% ofthe field runoff rate data fall within 630% of the modelprediction. Simply stated, a greater amount of precipitation,lower precipitation pH, and smaller roof angle (i.e., a flatterroof) contribute to higher Cu runoff.
The ranges for each variable of data used to develop themodel are as follows:
1. Annual precipitation rate: 396 to 3,203 mm/y2. Precipitation pH: 3.9 to 5.83. Angle of surface: 208 to 708
Precipitation data
Precipitation rate, pH, and longitude and latitude datafor 177 locations throughout the lower 48 states plus Alaskaand Puerto Rico from January 1994 to December 2000 wereextracted from the U.S. Geological Survey National Atmos-pheric Deposition Program (NADP) database (USGS 2002).Data for 2 additional locations, Hawaii and the U.S. VirginIslands, were extracted but differed from the other locations.Data for Hawaii are from 1987 to 1993 (data collection inHawaii ceased after 1993). Data for the U.S. Virgin Islands arefrom 1998 to 2000 (data collection began in 1998). SomeNADP data fell outside of the range of data used to developthe model. Data for all 179 locations ranged as follows:precipitation pH 4.17 to 6.59 and annual precipitation rate64.6 to 4,303.5 mm/y. Cu runoff rates and concentrationswere estimated for all 179 locations. Locations where data felloutside the range used to develop the model are noted andshould be used with caution.
Estimation of Cu runoff rates, export rates,and concentrations
Three derivations of the model developed by OdnevallWallinder et al. (2004) were used with the data from the
334 Integr Environ Assess Manag 1, 2005—R. Arnold
![Page 3: Estimations of copper roof runoff rates in the United States](https://reader035.fdocuments.us/reader035/viewer/2022081212/575004c01a28ab1148a07a96/html5/thumbnails/3.jpg)
NADP to predict Cu runoff rates and concentrations. Aderivation of the equation 8 presented by Odnevall Wallinderet al. (2004) was used to estimate average annual mass ofcopper in runoff per m2 of Cu surface area,
PRRðV; pHÞ ¼ 1:04þ 0:96V � 10�0:62pH ð2Þ
where PRR(V,pH) is the predicted runoff rate (g Cu/m2/y); V isthe annual rate of precipitation (mm/y); and pH is the pH ofthe precipitation. This version of the model assumes allsurfaces are at a 458 angle, a common angle used for roofs andthe average angle assumed for the Cu roof runoff rateestimates that follow. To correct the average Cu roof runoffrate at a given location or building within a location that has aslope between 208 and 708, simply determine the average Curunoff rate by using Equation 2 and multiply by cos(h)/cos458, where h is the angle of the surface (8). The effect ofroof angle deviations from 458 is illustrated in Figure 1.
Another manipulation of the model allows the estimationof annual mass of Cu in runoff per m2 of roof per amount ofprecipitation. This is useful for site-specific estimations ofpotential export rates when annual precipitation deviatesfrom the average annual precipitation at a given location andis calculated as
PPRRðV; pHÞ ¼1:04þ 0:96V � 10�0:62pH
V� 1; 000 ð3Þ
where PPRR(V,pH) is the predicted precipitation runoff rate(mg Cu/m2/mm); V is the annual rate of precipitation (mm/y);and pH is the pH of the precipitation. Multiplying theequation by 1,000 (mg/g) is performed simply to convertgrams to milligrams of copper. A similar manipulation ofEquation 2 allows for site-specific estimations of averageannual Cu concentration of the runoff. Dividing the model bythe rate of precipitation and multiplying by 1,000 (mm/m),the units become g/m3 or mg/L and is calculated as
PRCðV; pHÞ ¼1:04þ 0:96V � 10�0:62pH
V� 1; 000 ð4Þ
where PRC(V,pH) is the predicted runoff concentration(mg Cu/L); V is the annual rate of precipitation (mm/y);and pH is the pH of the precipitation. Estimations of the massof Cu per unit of precipitation and concentrations in therunoff are equal. Annual Cu roof runoff rates based upon roofarea (i.e., g Cu/m2/y) and Cu roof runoff rates based uponamount of precipitation (i.e., mg Cu/m2/mm), here forward,
will be referred to as Cu roof loading and export rates,respectively.
RESULTS AND DISCUSSION
Effects of model modifications on predictability
The utility of using Equation 2 without the correctionfactor for roof angles that differ from 458 is demonstratedusing data compiled by Odnevall Wallinder et al. (2004)(Figure 2). The correlation between measured and predictedCu roof loading rates remains highly statistically significant(least squares: r2 ¼ 0.78, F ¼ 71.8, p , 0.001; Measuredloading rate¼0.98 3 Predicted loading rateþ0.09). PredictedCu loading rates deviate from measured Cu runoff rates by,35% for 77% of the data.
Deviation results are similar to those corrected for roof angleand reported by Odnevall Wallinder et al. (2004). Thissimilarity is not surprising because 17 of the 22 field data setsused by Odnevall Wallinder et al. (2004) were collected fromroof surfaces with angles of 458. Nevertheless, the need tocorrect for roof angle is supported by empirical data and shouldbe used when available. Moreover, the ability of Equation 2 topredict measured runoff should be evaluated further, as databecome available, and should be adjusted as appropriate.
Copper loading rates, export rates, and concentrationsfor the United States
Yearly data, summary statistics, and cumulative frequencyplots for estimates of Cu loading rates, export rates, andconcentrations at each of the 179 locations are provided in theAppendix. These data may prove helpful in site-specificexposure and risk assessments. Summary statistics for theUnited States are provided in Table 1. Mean annual copperloading rate, export rate, and concentration estimates fromeach location are illustrated in Figures 3 and 4.
Summary statistics were calculated 2 ways. The 1st methoduses only the 7-y average loading rate, export rate, or concen-tration at 177 locations (N ¼ 177). The 2nd method usesyearly average loading rates, export rates, or concentrationsat 177 locations (N¼ 1,239) (i.e., 177 locations 3 7 y). Datafrom Hawaii and the U.S. Virgin Islands were excludedbecause the time period of precipitation information does notcorrespond exactly with the other 177 locations.
The mean, standard deviation, and range of predicted aver-age annual loading rates for the United States from 1994 to
Figure 1. Plot of correction factors (correction factor¼ cosine of roof angle in(8)/ cosine of 458) used to correct Cu roof runoff rates for roof angles thatdiffer from 458.
Figure 2. Plot of measured versus predicted Cu runoff loading rates usingEquation 2 and data from Odnevall Wallinder et al. (2004). Linear regressionof measured and predicted Cu runoff rates is highly statistically significant(least squares, F ¼ 78.1, p , 0.001).
Copper Roof Runoff Rates—Integr Environ Assess Manag 1, 2005 335
![Page 4: Estimations of copper roof runoff rates in the United States](https://reader035.fdocuments.us/reader035/viewer/2022081212/575004c01a28ab1148a07a96/html5/thumbnails/4.jpg)
2000 (N¼ 177) were 2.12, 0.74, and 1.12 to 3.80 g Cu/m2/y,
respectively. The average, standard deviation, and range of
annual predicted loading rates calculated based on the
1,239 location-y of data from 177 locations and 7 consecutive
y, 1994–2000 (N¼1,239) were 2.12, 0.77, and 1.05 to 4.85 g
Cu/m2/y, respectively.
The mean, standard deviation, and range of runoff concen-
tration and predicted export rates for the United States based
on precipitation at 177 locations from 1994 to 2000 (N¼177)
were 2.72, 1.13, and 0.84 to 8.27 mg Cu/L or mg Cu/m2/mm,
respectively. The average, standard deviation, and range of
predicted concentration and export rate based on the 1,239
location-y of data from 177 locations and 7 consecutive y, 1994
to 2000 (N¼1,239) were 2.72, 1.26, and 0.69 to 16.48 mg Cu/
L or mg Cu/m2/mm, respectively.
Estimated loading rates .2 g Cu/m2/y occurred predom-
inately in the eastern half of the United States with only a few
exceptions (Figure 3). Estimated loading rates .3 g Cu/m2/y
occurred primarily in the eastern United States along the
various mountain ranges from eastern Tennessee through
New Hampshire. Most locations west of the Mississippi River
had loading rate estimates of ,2 g Cu/m2/y with only a few
exceptions. Estimated Cu roof runoff concentrations and
export rates were highest at extremely arid locations in thewestern United States (Figure 4).
Data for pH and annual precipitation rates at somelocations were outside of the range of data used to developthe model and are identified in the Appendix. Modelestimations for those locations should be used with cautionbecause confidence decreases with the increasing number andmagnitude of excursions. Data at 4 locations and a total of 13yearly averages were higher than the pH range of the model.There were no excursions resulting from low pH data. Data at3 locations and 7 yearly averages were higher than theprecipitation rate range of model verification. Data at 45locations and 195 yearly averages were lower than theprecipitation rate range of model verification.
Potential impacts of improving air quality
Precipitation pH has a profound effect on the dissolutionrates of copper and other metals and, thus, the importance ofair quality management should not be overlooked as a meansof improving water quality in a watershed. Acid precipitation(considered in this discussion as precipitation with pH , 5.6)has the potential to substantially increase metal dissolutionand increase metal loading to watersheds. To demonstrate thepotential impacts of air quality and to stress the potential
Table 1. Summary statistics for predicted annual copper roof runoff rates in the United States, based on precipitation datacollected at 177 locations for 7 y, 1994–2000ab
Parameter
Yearly loadingrate
(g Cu/m2/y)
Average locationloading rate(g Cu/m2/y)
Yearly export rateand concentration
(mg Cu/m2/mm or mg Cu/L)
Average locationexport rate and concentration(mg Cu/m2/mm or mg Cu/L)
No. of values 1,239 177 1,239 177
Minimum 1.05 1.12 0.69 0.84
25th percentile 1.39 1.38 2.02 2.07
Median 2.05 2.16 2.48 2.51
75th percentile 2.70 2.69 3.02 3.03
Maximum 4.85 3.80 16.48 8.27
Mean 2.12 2.12 2.72 2.72
Standard deviation 0.77 0.74 1.26 1.13
Standard error 0.02 0.06 0.04 0.09
Coefficientof variation 0.36 0.35
0.46 0.42
Lower 95% CL 2.07 2.01 2.65 2.55
Upper 95% CL 2.16 2.23 2.79 2.88
Kurtosis �0.59 �1.08 19.21 7.15
Skewness 0.53 0.18 3.16 2.20
Normality K-Sdistance 0.12 0.13
0.15 0.14
p Value ,0.001 0.006 ,0.001 0.002
Pass normalitytest (0.05)
No No No No
Geometric mean 1.98 1.99 2.51 2.54a Data for Hawaii and U.S. Virgin Islands are not included.b Cu¼ total copper (i.e., dissolved plus particulate copper); CL¼ confidence limit.
336 Integr Environ Assess Manag 1, 2005—R. Arnold
![Page 5: Estimations of copper roof runoff rates in the United States](https://reader035.fdocuments.us/reader035/viewer/2022081212/575004c01a28ab1148a07a96/html5/thumbnails/5.jpg)
benefits of air quality management on water quality, Curunoff rates and concentrations are estimated based onambient precipitation quantities at each location and bysubstituting a constant pH of 5.6 for the ambient pH.Resulting estimates of Cu in roof runoff across the UnitedStates are illustrated in Figures 5 and 6. This analysisdemonstrates that substantial reductions in Cu mobilizationcould be expected with improvement of air quality and canhave profound implications for management of other metalsin stormwater runoff. The average percentage of reduction inCu from roof runoff across the United States is estimated tobe approximately 31% with a maximum of approximately61% (Figure 7).
Considerations when conducting source, exposure,and risk assessments
Determining the relative contribution of different Cusources to a watershed is very watershed-specific and can bedifficult. Contributions from point source discharges such aspublicly owned treatment works are relatively easy toestimate because contaminants are routinely monitored andreported. However, determining contributions from nonpointsources such as runoff from roadways, roofs, and fields can bedifficult and highly uncertain. It is beyond the scope of thisstudy to estimate the relative contribution of sheet copperfrom architectural applications for all watersheds. However,it is possible and perhaps useful to provide a comparison of avariety of Cu sources to a single water body as an example.
Estimates of annual nonpoint Cu loadings to San FranciscoBay, California, USA, and their relative uncertainty weremade by the Clean Estuary Partnership (2004a) and aresummarized in Table 2. An additional 11,020 kg Cu/y isdischarged by industrial and publicly owned treatment works(Clean Estuary Partnership 2004b). In the report by theClean Estuary Partnership (2004a), it is assumed that all Culeaving the rooftops and gutters enters San Francisco Bay.Using these loading estimates, architectural sheet Cuapplications contribute about ,5% of the loading to SanFrancisco Bay and are the 6th largest nonpoint sourceanalyzed.
When conducting exposure and risk assessments usingroof runoff data, it is important to understand the pathwaythat the runoff travels before reaching a waterway. Materialsthat the stormwater contacts before entering the watershedcan alter the concentration or modify the bioavailability ofthe copper in the stormwater. The scenario producing thehighest probability of unacceptable risk is that of Cu roofrunoff entering directly into a water body without contactingother surfaces. In site-specific cases where unacceptable riskdue to Cu roof runoff is demonstrated, there exist othermanagement options that can be considered such ascommercially available coatings or filtration systems.
Designing buildings that allow Cu roof runoff to dischargedirectly into a waterway should be avoided unless mixing israpid and dilution is extensive (approximately 3 orders ofmagnitude). It has been shown that 60 to 90% of the total Cu
Figure 3. Estimations of average yearly Cu roof runoff loading rates (g Cu/m2/y) for 7 y (1994–2000) at 177 locations throughout the United States.Estimations for 2 additional locations differ from the others. Estimations for Hawaii are based on a different 7 y (1987–1993) because of the cessation in1993 of data collection. U.S. Virgin Island estimates are based on 3 y (1998–2000) because of the lack of any other data.
Copper Roof Runoff Rates—Integr Environ Assess Manag 1, 2005 337
![Page 6: Estimations of copper roof runoff rates in the United States](https://reader035.fdocuments.us/reader035/viewer/2022081212/575004c01a28ab1148a07a96/html5/thumbnails/6.jpg)
concentration in roof runoff is in a bioavailable form as it
leaves the roof (Karlen 2001). However, Cu roof runoff from
commercial buildings more often falls onto hard surfaces or
enters some type of conduit (e.g., sewer pipe) before reaching
a waterway. Residential roofs often discharge to lawns. In
some cases, roof runoff can even be directed to stormwater
treatment facilities before being discharged to a waterway.
Materials such as soils, concrete, and iron can rapidly and
extensively transform and sequester bioavailable forms of Cu
in runoff before reaching the watershed (Sunda and Guillard
1976; Sunda and Hansen 1979; Cabiniss and Shuman 1988;
Cantrell et al. 1995; Sundberg 1998; Shoakes and Moeller
1999; Boulanger and Nikolaidis 2001, 2003a, 2003b; Bertling
et al. 2002, 2003). One runoff study of a 9-y-old Cu roof and
16 storm events documents an average 45% reduction in the
concentration of Cu after traveling through 46 m of an iron
and concrete conduit system (Boulanger and Nikolaidis
2001). Data from Perkins et al. (2005) indicate that Cu is
removed from water containing 2.5 mg dissolved Cu/L at a
rate of approximately 3.1% per linear meter of new concrete
conduit. That study also indicated that Cu did not readily
dissolve back at detectable levels (minimum detection level of
,7 lg Cu/L) into control water that was subsequently passed
through the conduit. A study by Sundberg (1998) of Cu
runoff onto concrete gravel placed in the drip line of a
gutterless building indicates that approximately 50% of the
Cu reacted with the concrete and was removed from the
runoff. A study of soil retention of Cu demonstrates 85% and
72% removal of 3.5 mg Cu/L as dissolved CuCl2 and
organically complexed Cu, respectively, in 20-cm-long
columns of undisturbed soil and 100% removal in columns
of disturbed soil (Camobreco et al. 1996).
Boulanger and Nikolaidis (2001, 2003a, 2003b) developed a
model framework for the risk assessment of Cu roof runoff in
a watershed by incorporating water quality characterization,
watershed land use (including copper roofing), hydraulic data,
toxicity measurements, chemical speciation modeling, and a
probabilistic modeling technique. The speciation mode of the
biotic ligand model (Di Toro et al. 2001; Paquin et al. 2002)
was used to confirm Cu speciation measurements. However,
the toxicity mode of the biotic ligand model could be used in
place of toxicity testing to assess the bioavailability of Cu. In
the study by Boulanger and Nikolaidis (2001, 2003a, 2003b),
runoff from a 1,800-m2 Cu roof was monitored to assess
impacts on the receiving stream in a 0.465-km2 developed
urban watershed. The runoff flowed directly into a storm-
water conduit system and eventually into a small stream.
Because of removal within the conduits, dilution, and trans-
formation reactions in transit, Cu concentrations were
reduced on average from 3.63 mg Cu/L (total) and 1.73
mg Cu/L (ionic) to 0.046 mg Cu/L (total) and ,0.00005 mg
Cu/L (ionic) before reaching the stream. Thus, concentra-
Figure 4. Estimations of average Cu roof runoff export rates (mg Cu/m2/mm) and Cu roof runoff concentrations (mg Cu/L) for 7 y (1994–2000) at 177
locations throughout the United States. Estimations for 2 additional locations differ from the others. Estimations for Hawaii are based on a different 7 y
(1987–1993) because of the cessation in 1993 of data collection. U.S. Virgin Island estimates are based on 3 y (1998–2000) because of the lack of any other
data.
338 Integr Environ Assess Manag 1, 2005—R. Arnold
![Page 7: Estimations of copper roof runoff rates in the United States](https://reader035.fdocuments.us/reader035/viewer/2022081212/575004c01a28ab1148a07a96/html5/thumbnails/7.jpg)
tions of Cu were reduced to approximately 1/80th and 1/34,600th of the original concentrations.
This emphasizes how Cu concentrations and runoff ratescan be modified in a simple conduit system and whytransformation and sequestering processes must be consid-ered in exposure and risk assessments. Further studies areneeded to better quantify the degree that conduits modifyCu concentrations and the duration that they are effectivebecause eventually Cu binding sites may become exhausted.However, the substantial reduction of Cu documented in thestudy by Boulanger and Nikolaidis (2001) suggests that suchsystems can have active binding sites for years. Properquantification of removal rates under controlled conditionswill likely aid in making realistic, defensible, and informedwatershed management decisions.
Model improvements
Examination of the model and resulting predictions suggestthe need for additional roof runoff studies, especially in aridregions. Variation coefficients of predicted annual loadingrates at each site range from approximately 4 to 20%between predicted loading rates of 1.5 to 3.7 g Cu/m2/y.However, the range of the coefficient of variation decreasessubstantially at locations where predicted loading rates areapproximately ,1.5 g Cu/m2/y (Figure 8). This likelyindicates the rate at which the addition of the 1st-flushconstant (1.04 g Cu/m2/y, Eqn. 2) begins to dominate the
loading rate estimate, and thus, relative variability from yearto year begins to decrease.
The lower bound of the model for annual precipitation rate(400 mm/y) is too high to be applied with completeconfidence to a substantial amount of arid areas of the UnitedStates. In addition, there may be a need to refine the 1st-flushrate constant (1.04 g Cu/m2/y, Eqn. 2). For example, as theannual rate of precipitation approaches 0 mm/y, the runoffrate estimate is explained by the 1st-flush constant. Themodel ultimately predicts runoff of 1.04 g Cu/m2/y whenannual precipitation is 0 mm/y, an obvious impossibility. Thehighest estimates of runoff concentrations were at locationswith extremely low precipitation rate. Thus, overestimationof runoff rates and concentrations can be expected for drierclimates, but this is yet to be proven.
As mentioned previously, 17 of the 22 data points used todevelop the model were from 458 slopes, thus more data areneeded for roofs with angles other than 458. Lastly, it ishypothesized that accounting for roof length (i.e., distancefrom roof peak to roof drip edge), which affects the contacttime of precipitation with the copper, may further explainvariability between measured and predicted roof runoff rates.Studies are underway to test this hypothesis.
CONCLUSIONSEstimates of Cu loading rates, export rates, and concen-
trations of roof runoff across much of the United States are
Figure 5. Estimations of average yearly Cu roof runoff loading rates (g Cu/m2/y) at a constant pH of 5.6 and based on 7 y (1994–2000) of rainfall data at 177
locations throughout the United States. Estimations for 2 additional locations differ from the others. Estimations for Hawaii are based on a different 7 y (1987–
1993) because of the cessation in 1993 of data collection. U.S. Virgin Island estimates are based on 3 y (1998–2000) because of the lack of any other data.
Copper Roof Runoff Rates—Integr Environ Assess Manag 1, 2005 339
![Page 8: Estimations of copper roof runoff rates in the United States](https://reader035.fdocuments.us/reader035/viewer/2022081212/575004c01a28ab1148a07a96/html5/thumbnails/8.jpg)
now available. Loading rates .2 g Cu/m2/y occur predom-
inately east of the Mississippi River with few exceptions.
Loading rates .3 g Cu/m2/y occur primarily in the eastern
United States along the various mountain ranges from eastern
Tennessee through New Hampshire. Most locations west of
the Mississippi River have loading rate estimates of ,2 g Cu/
m2/y with few exceptions. Mean Cu concentrations and
export rates range from 0.84 to 8.27 mg Cu/L and mg Cu/m2/
mm, respectively, and are highest in the extremely aridwestern regions of the United States.
Improvement in air quality can lead to decreased Cumobilization. If air management programs are effective ineliminating acid precipitation (i.e., pH , 5.6) in the UnitedStates, it is estimated that copper runoff will be reduced by anaverage of 31%.
Further refinement of the model is needed. The modelshould be corrected as data become available, and new
Figure 6. Estimations of average Cu roof runoff export rates (mg Cu/m2/mm) and Cu roof runoff concentrations (mg Cu/L) at a constant pH of 5.6 and basedon 7 y (1994–2000) of rainfall data at 177 locations throughout the United States. Estimations for 2 additional locations differ from the others. Estimationsfor Hawaii are based on a different 7 y (1987–1993) because of the cessation in 1993 of data collection. U.S. Virgin Island estimates are based on 3 y (1998–2000) because of the lack of any other data.
Figure 7. Cumulative frequency plot of estimated percent reductions in runoffrates for all locations and dates (N¼ 179), assuming a constant precipitationpH of 5.6. Percent reduction¼ (Predicted runoff at measured pH – Predictedrunoff at pH 5.6)/(Predicted runoff at measured pH) � 100.
Figure 8. Coefficients of variation of predicted annual Cu loading rates as afunction of average predicted annual copper loading rates at 177 locationsthroughout the United States. Data for Hawaii and the U.S. Virgin Islands areomitted.
340 Integr Environ Assess Manag 1, 2005—R. Arnold
![Page 9: Estimations of copper roof runoff rates in the United States](https://reader035.fdocuments.us/reader035/viewer/2022081212/575004c01a28ab1148a07a96/html5/thumbnails/9.jpg)
estimates for the United States should be made. Correction ofthe 1st-flush constant to improve estimates for arid regions isa significant need. Incorporation of roof length into the modelmay improve the predictability of the model and provide theability to better predict Cu runoff for individual buildings.Nevertheless, the estimates provided here should prove usefulfor watershed management activities and as a surrogate for Curunoff studies.
The data presented here are important in assessing Culoadings to waterways, as well as exposure and risk of Curoofing runoff to aquatic organisms in watersheds. Theseestimates represent total Cu concentrations and loading andexport rates of runoff at the point it leaves the roof.Estimated concentrations of Cu in the runoff are highenough to suggest that direct discharge from a Cu roof intoa water body should be avoided whenever possible.However, the data should be used with caution. Studieshave shown that various surfaces that runoff can contactbefore entering aquatic environments readily sequester orchange the chemical speciation of Cu and thus reduce itsbioavailability. Therefore, these data should be used withcaution to estimate exposure or risk to aquatic organisms ina watershed. Rather, they should 1st be corrected for theeffects of landscapes, conduit materials, and best manage-ment practices (e.g., use of filtration systems to remove Cu).
Acknowledgement—This work was funded by the CopperDevelopment Association and the International CopperAssociation, New York, New York, USA.
REFERENCESBertling S, Odnevall Wallinder I, Leygraf C. 2002. The capacity of limestone to
immobilize copper in runoff water: A laboratory investigation. In: Morcillo M,
Costa JM, editors. Proceedings of the 15th International Corrosion Congress;
2002 Sept 22–Sept 27; Granada, Spain. Houston, TX: The International
Corrosion Council.
Bertling S, Odnevall Wallinder I, Leygraf C, Berggren D. 2002. Immobilization of
copper in runoff water from roofing materials by limestone, soil and concrete.
In: Morcillo M, Costa JM, editors. Proceedings of the 15th International
Corrosion Congress; 2002 Sept 22–Sept 27; Granada, Spain. Houston, TX: The
International Corrosion Council.
Boulanger B, Nikolaidis NP. 2001. Contribution of copper-based architectural
material to copper concentrations and toxicity in storm water runoff. New York
(NY), USA: Copper Development Association Inc. Contract report 6090-0002.
Boulanger B, Nikolaidis NP. 2003a. Mobility and aquatic toxicity of copper in an
urban watershed. Journal of the American Water Resources Association
39:325–336.
Boulanger B, Nikolaidis NP. 2003b. Modeling framework for managing copper
runoff in urban watersheds. Journal of the American Water Resources
Association 39:337–346.
Cabiniss SE, Shuman MS. 1988. Copper binding by dissolved organic matter: I.
Suwannee River fulvic acid equilibria. Geochim et Cosmochim Acta 52:185–
193.
Camobreco VC, Richards BK, Steenhuis TS, Peverly JH, McBride MB. 1996.
Movement of heavy metals through undisturbed and homogenized soil
columns. Soil Sci 161:740–750.
Cantrell K, Kaplan D, Wietsma T. 1995. Zero-valent iron for the in situ remediation
of selected metals in groundwater. J Hazard Mater 42:201–212.
Clean Estuary Partnership. 2004a. Copper sources in urban runoff and shoreline
activities. Oakland (CA), USA.
Clean Estuary Partnership. 2004b. North of Dumbarton Bridge copper and nickel
site-specific objectives state implementation policy justification report. Oak-
land (CA), USA: Clean Estuary Partnership.
Copper Development Association. 2005. Annual data 2005. Copper supply and
consumption, 1984–2004. www.copper.org/resources/market_data/pdfs/
annual-data/2005.pdf. Accessed 10 August 2005.
Culotta VC, Joh HD, Lin SJ, Slekar KH, Strain J. 1995. A physiological role of
Saccharomyces cerevisiae copper/zinc superoxide dismutase in copper
buffering. J Biol Chem 270:29991–29997.
Dallinger R, Berger B, Hunziker P, Kagi JHR. 1997. Metallothionein in snail
cadmium and copper metabolism. Nature 388:237–238.
Di Toro DM, Allen HE, Bergman HL, Meyer JS, Paquin PR, Santore RC. 2001. Biotic
ligand model of the acute toxicity of metals, I: Technical basis. Environ Toxicol
Chem 20:2383–2396.
Fernandes JC, Henriques FS. 1991. Biochemical, physiological, and structural
effects of excess copper in plants. Bot Rev 57:246–273.
Gabel CM, Bittinger MA, Maier RJ. 1994. Cytochrome aa3 gene regulation in
members of the family Rhizobiaceae: Comparison of copper and oxygen
effects in Bradyrhizobium japonicum and Rhizobium tropici. Appl Environ
Microbiol 60:141–148.
Graedel TE. 1987. Copper patinas formed in the atmosphere, II: A qualitative
assessment of mechanisms. Corrosion Science 27:721–740.
Harris ZL, Gitlin JD. 1996. Genetic and molecular basis for copper toxicity. Am J
Clin Nutr 63:836S–841S.
He W. 2000. Corrosion rates and runoff rates of copper and zinc as roofing
materials – A combined field and laboratory study [Licentiate thesis].
Stockholm, Sweden: Royal Institute of Technology.
Jolly JL. 2000. The U.S. copper-base scrap industry and its by-products: An
overview. New York (NY), USA: Copper Development Association Inc. Contract
report A1039–02/00.
Table 2. Nonpoint sources, estimates of annual copper (Cu) loadings and uncertainty descriptors for San Francisco Bay,California, USA (data from CEP 2005)
Nonpoint Cu source Load estimate kg Cu/y Uncertainty
Marine antifouling coatings 9,072 Moderate–high
Vehicle brake pads .4,536 High
Copper pesticides ,4,536 High
Air deposition 3,992 Low–moderate
Soil erosion 3,175 Moderate
Copper roofs and gutters 2,032 Moderate–high
Copper algaecides applied to surface water 1,814 High
Industrial use 1,497 Moderate
Domestic water discharge to storm drains 1,361 Moderate–high
Vehicle fluid leaks 272 Moderate–high
Copper Roof Runoff Rates—Integr Environ Assess Manag 1, 2005 341
![Page 10: Estimations of copper roof runoff rates in the United States](https://reader035.fdocuments.us/reader035/viewer/2022081212/575004c01a28ab1148a07a96/html5/thumbnails/10.jpg)
Karlen C. 2001. Atmospheric corrosion of copper and zinc-based materials: Runoff
rates, chemical speciation and ecotoxicity [licentiate thesis]. Stockholm,
Sweden: Royal Institute of Technology.
Kratschemer A, Odnevall Wallinder I, Leygraf C. 1997. The evolution of outdoor
copper patina. Corrosion Science 44:425–450.
Landner L, Lindestrom L. 1999. Copper in society and in the environment: An
account of the facts on fluxes, amounts, and effects of copper in Sweden, 2nd
ed. Vasteras, Sweden: Swedish Environmental Research Group (MFG).
Linder MC, Hazegh-Azam M. 1996. Copper biochemistry and molecular biology.
Am J Clin Nutr 63:797S–811S.
Odnevall Wallinder I, Bertling S, Zhang X, Leygraf C. 2004. Predictive models of
copper runoff from external structures. J Environ Monit 6:704–712.
Olivares M, Uauy R. 1996. Copper as an essential nutrient. Am J Clin Nutr
63:791S–796S.
Paquin PR, Gorsuch JW, Apte SC, Batley GE, Bowles KC, Campbell PGC, Delos CG,
Di Toro DM, Dwyer RL, Galvez F, Gensemer RW, Goss GG, Hogstrand C,
Janssen CR, McGeer JC, Naddy RB, Playle RC, Santore RC, Schneider U,
Stubblefield WA, Wood CM, Wu KB. 2002. The biotic ligand model: A
historical overview. Comp Biochem Physiol 133C:3–36.
Perkins C, Nadim F, Arnold WR. 2005. Effects of PVC, cast iron, and concrete
conduit on concentrations of copper in stormwater. Urban Water Journal
(forthcoming).
Richter G. 1978. Plant metabolism: Physiology and biochemistry of primary
metabolism. Stuttgart, Germany: Georg Thieme.
Shoakes T, Moeller G. 1999. Removal of dissolved heavy metals from acid rock
drainage using iron metal. Environ Sci Technol 33:282–287.
Slekar KH, Kosman DJ, Cullotta VC. 1996. The yeast copper/zinc superoxide
dismutase and the pentose phosphate pathway play overlapping roles in
oxidative stress protection. J Biol Chem 271:28831–28836.
Sunda WG, Guillard RR. 1976. The relationship between cupric ion activity and the
toxicity of copper to phytoplankton. J Mar Res 34:511–529.
Sunda WG, Hansen PJ. 1979. Chemical speciation of copper in river water. In:
Jennne EA, editor. Chemical modeling in aquatic systems. ACS Symposium
Series 93. Washington (DC), USA: American Chemical Society. p 147–188.
Sundberg R. 1998. The fate of copper released from the Vasa shipyard museum.
Metall, 52. Jahrgang 4:230–231.
Uauy R, Olivares M, Gonzales M. 1998. Essentiality of copper in humans. Am J Clin
Nutr 67:952S–959S.
[USEPA] U.S. Environmental Protection Agency. 1991. Guidance for water quality-
based decisions: The TMDL process. Washington DC. EPA/440/4-91-001.
[USEPA] U.S. Environmental Protection Agency. 1993. Guidance specifying
management measures for sources of nonpoint pollution in coastal waters.
Washington DC. EPA-840-B-93–001c.
[USEPA] U.S. Environmental Protection Agency. 2005. Top 100 impairments.
oaspub.epa.gov/waters/national_rept.control#TOP_IMP. Accessed 1 Novem-
ber 2004.
[USGS] United States Geological Survey. 2002. NADP National Atmospheric
Deposition Program database. Pacific Grove (CA), USA.
342 Integr Environ Assess Manag 1, 2005—R. Arnold