Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 0 4 9 e2 0 5 9

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

The effect of roofing material on the quality of harvested


rainwater

Carolina B. Mendez a, J. Brandon Klenzendorf a,1, Brigit R. Afshar a,2, Mark T. Simmons b,
Michael E. Barrett a, Kerry A. Kinney a, Mary Jo Kirisits a,*
a
Department of Civil, Architectural, and Environmental Engineering, The University of Texas at Austin, 1 University Station C1786,
Austin, TX 78712, USA
b
Lady Bird Johnson Wildflower Center, The University of Texas at Austin, 4801 La Crosse Avenue, Austin, TX 78739, USA

article info abstract

Article history: Due to decreases in the availability and quality of traditional water resources, harvested
Received 7 November 2010 rainwater is increasingly used for potable and non-potable purposes. In this study, we
Received in revised form examined the effect of conventional roofing materials (i.e., asphalt fiberglass shingle,
15 December 2010 Galvalume metal, and concrete tile) and alternative roofing materials (i.e., cool and green) on
Accepted 16 December 2010 the quality of harvested rainwater. Results from pilot-scale and full-scale roofs demonstrated
Available online 22 December 2010 that rainwater harvested from any of these roofing materials would require treatment if the
consumer wanted to meet United States Environmental Protection Agency primary and
Keywords: secondary drinking water standards or non-potable water reuse guidelines; at a minimum,
Rainwater harvesting first-flush diversion, filtration, and disinfection are recommended. Metal roofs are commonly
Roof-runoff recommended for rainwater harvesting applications, and this study showed that rainwater
Water quality harvested from metal roofs tends to have lower concentrations of fecal indicator bacteria as
First-flush compared to other roofing materials. However, concrete tile and cool roofs produced harvested
Roofing material rainwater quality similar to that from the metal roofs, indicating that these roofing materials
Potable also are suitable for rainwater harvesting applications. Although the shingle and green roofs
produced water quality comparable in many respects to that from the other roofing materials,
their dissolved organic carbon concentrations were very high (approximately one order of
magnitude higher than what is typical for a finished drinking water in the United States), which
might lead to high concentrations of disinfection byproducts after chlorination. Furthermore
the concentrations of some metals (e.g., arsenic) in rainwater harvested from the green roof
suggest that the quality of commercial growing media should be carefully examined if the
harvested rainwater is being considered for domestic use. Hence, roofing material is an
important consideration when designing a rainwater catchment.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction harvesting is undergoing a surge in popularity in the United


States, particularly in locations where traditional, high-quality
Roof-based rainwater harvesting is the capture of rain- freshwater supplies are absent or where consumers desire to
water from a roof for potable or non-potable use. Rainwater contribute to sustainability. The United States Environmental

* Corresponding author. Tel.: þ1 512 232 7120; fax: þ1 512 471 0592.
E-mail address: kirisits@mail.utexas.edu (M.J. Kirisits).
1
Present address: Geosyntec Consultants, 3600 Bee Caves Road, Suite 101, Austin, TX 78746, USA.
2
Present address: CH2M Hill, 12377 Merit Drive, Ste. 1000, Dallas, TX 75251, USA.
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2010.12.015
2050 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 0 4 9 e2 0 5 9

Nomenclature mL milliliter
mm millimeter
cm centimeter
NTU nephelometric turbidity unit
CFU colony forming unit
 % percent
degree angle
 PAH polycyclic aromatic hydrocarbon
C degrees Celsius
PVC polyvinyl chloride
DOC dissolved organic carbon
SMCL secondary maximum contaminant level
FC fecal coliform
SVOC semi-volatile organic compound
h hour
a significance level
km kilometer
m2 square meter
L liter
TC total coliform
MCL maximum contaminant level
TDS total dissolved solids
mg/L microgram per liter
TSS total suspended solids
mS/cm microsiemens per centimeter
USEPA United States Environmental Protection Agency
mg-N/L milligram of nitrogen per liter
VOC volatile organic compound
mg/L milligram per liter

Protection Agency (USEPA) does not regulate the water quality neither has been widely considered for rainwater harvesting.
of residential rainwater harvesting systems in the U.S., but To our knowledge, only one study has analyzed harvested
some state and local agencies do issue voluntary water quality rainwater quality from a green roof (Nicholson et al., 2009),
guidelines for residential rainwater harvesting systems (e.g., and it was examined in the context of non-potable use.
Texas Rainwater Harvesting Evaluation Committee, 2006). The The objective of the current study was to thoroughly
quality of harvested rainwater is of great importance because it examine the effect of roofing material on the quality of har-
is increasingly used for domestic purposes. vested rainwater (untreated, except for the use of a first-flush
Several types of chemical contaminants have been found in diverter) from the standpoint of suitability for domestic use.
harvested rainwater including heavy metals (Förster, 1999; Lee Five co-located pilot-scale roofs were used to compare har-
et al., 2010), polycyclic aromatic hydrocarbons (PAHs) (Förster, vested rainwater quality among conventional roofing mate-
1998, 1999), pesticides (Zobrist et al., 2000), and herbicides rials (i.e., asphalt fiberglass shingle, Galvalume metal, and
(Bucheli et al., 1998). Microorganisms also are present in roof- concrete tile) and alternative roofing materials (i.e., cool and
runoff, and fecal indicator bacteria and potentially pathogenic green). This is the first study to extensively compare harvested
bacteria and protozoa have been detected (Ahmed et al., 2008). rainwater quality (20 water quality parameters) among these
The type of roofing material used for the catchment can conventional and alternative roofing materials. In addition, the
affect the quality of harvested rainwater. Nicholson et al. quality of the rainwater harvested from the asphalt fiberglass
(2009) compared harvested rainwater quality among six roof shingle and Galvalume roofing materials was compared
types: galvanized metal, cedar shake, asphalt shingle, two between the pilot-scale roofs and full-scale residential roofs.
types of treated wood, and green (i.e., vegetated). The galva-
nized metal, asphalt shingle, and green roofs neutralized the
acidic rainwater to a greater extent than did the other roofing 2. Materials and methods
materials. The treated woods yielded the highest copper
concentrations (mg/L range), and the galvanized metal yiel- 2.1. Study sites
ded the highest zinc concentrations (mg/L range), as com-
pared to the mg/L concentrations of these metals from the This study was conducted in Austin, Texas, where the mean
other roofing materials. Van Metre and Mahler (2003) found annual rainfall is 856 mm (National Oceanic and Atmospheric
galvanized metal roofs to be a source of particulate zinc and Administration, 2010). The pilot-scale roofs were constructed
cadmium and asphalt shingle roofs to be a source of partic- at the Lady Bird Johnson Wildflower Center (Austin, TX), and
ulate lead and potentially mercury. Kingett Mitchell Ltd. the full-scale residential roofs were located within 42 km of
(2003) found higher zinc concentrations in rainwater har- the pilot site.
vested from painted galvanized iron roofs that showed
evidence of weathering as compared to those in excellent 2.2. Pilot-scale roofs
condition. Despins et al. (2009) found that harvested rain-
water quality from steel roofs was superior to that from The conventional pilot-scale roofs, constructed in 2009, were
asphalt shingle roofs, particularly with respect to turbidity, made of asphalt fiberglass shingle (GAF-Elk, Wayne, NJ),
total organic carbon, and color. Galvalume (55% aluminum-zinc alloy coated steel, United
Most studies to date have focused on examining conven- States Steel Corporation, Pittsburgh, PA), or concrete tile
tional roofing materials, such as galvanized metal, for rain- (MonierLifetile, Irvine, CA) (Fig. 1A). They were angled 18.4
water harvesting. Green roofs are increasingly being installed from the horizontal, with a footprint area of 2.8 m2. The alter-
for their stormwater retention characteristics, and both green native pilot-scale roofs, constructed in 2007 as described by
and cool roofs (made of reflective roofing material) are being Simmons et al. (2008), were an unfertilized (type E) green roof
installed for their energy conservation potential; however, (Fig. 1B) and an acrylic-surfaced, 2-ply atactic polypropylene
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 0 4 9 e2 0 5 9 2051

Fig. 1 e Pilot-scale roofs (A) from left to right: asphalt fiberglass shingle, metal, concrete tile, (B) green, and (C) cool.

modified bituminous membrane cool roof (Fig. 1C); these low- sampler, first-flush bottle, and first and second tanks were
slope roofs were angled 1.2 from the horizontal, with a foot- analyzed in triplicate for pH, conductivity, total coliform
print area of 3.4 m2. All of the pilot-scale roofs faced north. (TC), fecal coliform (FC), turbidity, total suspended solids
(TSS), dissolved organic carbon (DOC), and nine metals
2.3. Full-scale residential roofs (dissolved þ particulate): aluminum, arsenic, cadmium, chr-
omium, copper, iron, lead, selenium, and zinc. Nitrate and
Three full-scale residential roofs were sampled: a 12-year-old nitrite were measured once for each sample. For the fourth
Galvalume roof (22 from the horizontal, footprint area of rain event, samples from the ambient sampler and first-flush
4.3 m2), a 5-year-old asphalt fiberglass shingle roof (23 from bottle were analyzed for pesticides and PAHs.
the horizontal, footprint area of 4.3 m2) and another 5-year-old Rainwater harvested from the full-scale roofs was sampled
asphalt fiberglass shingle roof (18 from the horizontal, foot- during three rain events in 2009: February 9, February 11, and
print area of 5.3 m2). All of the full-scale roofs faced north. March 11. For all three rain events, the ambient sampler, first-
flush bottle, and first and second tanks were analyzed in tripli-
cate for TC, FC, TSS, DOC, lead, and zinc. The pH, conductivity,
2.4. Rainwater sampling devices
turbidity, nitrate, and nitrite were measured once for each
sample. For the first rain event, samples from the ambient
A sampling insert (7.6-cm inner-diameter potable-quality
sampler and first-flush bottle were analyzed for a suite of volatile
polyvinyl chloride (PVC) pipe cut lengthwise in half) was placed
and semi-volatile organic compounds (VOCs and SVOCs).
in the gutters of the conventional pilot-scale and full-scale
At the end of each rain event, samples were transported to
roofs. The sampling insert drained to a passive collection
the laboratory for analysis. The pH, conductivity, TC, FC,
system consisting of a 2-L glass bottle to collect the “first-flush”
turbidity, and TSS were analyzed immediately upon arrival at
and two 10-L polypropylene tanks in series (Fig. 2). The first-
the laboratory. The pH was measured with a combination
flush volume was designed according to the Texas Water
electrode and meter (Model 920A, Orion Research Inc., Boston,
Development Board guidelines, which state that the first-flush
should divert a minimum of 38 L for every 93 m2 of collection
area (Texas Water Development Board, 2005). For the alterna-
tive pilot-scale roofs, the drainage system described in
Simmons et al. (2008) emptied to the aforementioned passive
collection system. The first-flush bottle and tanks in the passive
collection system filled sequentially during a rain event, with
excess rainfall exiting the system through an overflow spout
(Fig. 2). A 46-cm inner-diameter polyethylene funnel attached to
a 10-L polypropylene tank was used as an ambient sampler to
collect rainwater without roof exposure; the cover on the funnel
was removed up to 24 h prior to a rain event, so some atmo-
spheric deposition was possible. Between rain events, the
sampling devices were washed and rinsed, and the collection
bottle and tanks were autoclaved. Rainfall was measured with
a rain gauge, and the sampled events had rainfalls between
25 mm and 38 mm at the pilot-scale and full-scale sites.

2.5. Sampling and analytical methods

Rainwater harvested from the pilot-scale roofs was sampled


during four rain events in 2009: April 18, June 11, July 23, and
September 11. For the first three rain events, the ambient Fig. 2 e Schematic of the sampling device.
2052 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 0 4 9 e2 0 5 9

MA). Conductivity was measured with a conductivity meter data could be characterized by a normal distribution. A
(Model CDM 230, Radiometer Analytical, Lyon, France). TC was significance level of a ¼ 0.1 was used.
measured according to method 9222B (American Public Health Since the consumer often diverts the first-flush water from
Association, 1998) with m-Endo broth (Millipore, Billerica, MA), use, the discussion in this study generally focuses on the
and FC was measured according to method 9222D (American quality of the rainwater harvested after the first-flush (i.e.,
Public Health Association, 1998) with m-FC agar (Difco, tanks 1 and 2 in Fig. 2). The quality of the rainwater harvested
Detroit, MI). Turbidity was measured with a turbidimeter (Model after the first-flush was compared to the quality of the first-
2100A, Hach, Loveland, CO). TSS was measured according to flush, to the quality of rainwater captured by the ambient
method 2540B (American Public Health Association, 1998). sampler, to the U.S. drinking water standards, and to the U.S.
Samples for the remaining analytes were preserved for non-potable urban water reuse guidelines. Finally, the quality
later analysis. Samples for nitrate and nitrite analysis were of the rainwater harvested after the first-flush was compared
stored at 4  C and analyzed within 48 h with Test 0 N Tube among the roofing materials.
Reactor/Cuvette Tubes (Hach, Loveland, CO); nitrate was
measured with the NitraVerX reagent set, and nitrite was
measured with the NitriVer3 reagent set according to the 3. Results and discussion
manufacturer’s instructions. Samples for DOC analysis were
stored at 4  C after addition of reagent grade phosphoric acid; 3.1. pH and conductivity e pilot-scale roofs
DOC was measured with an Apollo 9000 Combustion Analyzer
(Tekmar-Dohrmann, Cincinnati, OH). Samples for metal The average pH of rainfall collected in the ambient sampler
analysis were preserved with trace metal grade nitric acid. was 6.1, and all of the harvested rainwater samples from the
For the pilot-scale study, metal concentrations were mea- pilot-scale roofs were in the near-neutral range (Fig. 3A). For
sured by inductively coupled plasma-mass spectrometry all of the pilot-scale roofs except the metal roof, the pH of the
(Model 7500ce, Agilent Technologies, Inc., Tokyo, Japan). For rainwater harvested after the first-flush was significantly
the full-scale study, metal concentrations were measured by higher than that in the ambient sampler ( p-values < 0.048).
graphite furnace atomic absorption spectroscopy (Model Similarly, other studies have reported an increase in pH from
AAnalyst 600, PerkineElmer, Waltham, MA). ambient rainwater to harvested rainwater (Kingett Mitchell
At the end of the February 9 and September 11, 2009 rain Ltd., 2003). The pH of the rainwater harvested after the first-
events, samples were delivered to DHL Analytical (Round flush for the tile roof was significantly higher than those of the
Rock, TX) for the analysis of organic compounds. Pesticides other pilot-scale roofs ( p-values < 0.021). This was expected
were measured according to method 8081A (USEPA, 1996a); due to the alkaline nature of concrete and is consistent with
SVOCs and PAHs were measured according to method 8270C other studies (Kingett Mitchell Ltd., 2003). The average pH of
(USEPA, 1996b); VOCs were measured according to method the rainwater harvested after the first-flush for all pilot-scale
8260B (USEPA, 1996c). roofs met the USEPA secondary drinking water standard range
of 6.5e8.5 (Fig. 3A) and the non-potable urban water reuse
2.6. Statistical analysis guideline range of 6.0e9.0 (USEPA, 2004).
The average conductivity of rainfall collected in the amb-
The nonparametric Mann-Whitney test was used to make ient sampler was 34 mS/cm (Fig. 3B), which is in the typical
comparisons among the data because of the small number of range for ambient rainwater (Yaziz et al., 1989; Lee et al., 2010).
rainfall events and the inability to show that the water quality For the green roof only, the conductivity of the rainwater

Fig. 3 e Average pH (panel A) and conductivity (panel B) for the pilot-scale events: ( ) Quality of the first-flush, ( )
Quality after the first-flush (average of tank 1 and tank 2),  USEPA secondary drinking water standard range for pH
(6.5e8.5), - - - Ambient sampler. One standard deviation is shown.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 0 4 9 e2 0 5 9 2053

harvested after the first-flush was significantly higher than after the first-flush was statistically indistinguishable from
that in the ambient sampler ( p-value ¼ 0.012). The conduc- that in the ambient sampler ( p-values > 0.179). According to
tivity of the rainwater harvested after the first-flush for the the USEPA primary drinking water standards, if two consec-
green roof also was significantly higher than those for the utive samples have TC and at least one of them has FC (or
other pilot-scale roofs ( p-values < 0.013); this was expected Escherichia coli), then the system has an acute maximum
due to mineral dissolution as the rainwater passed through contaminant level (MCL) violation; all of the pilot-scale roofs
the growing media of the green roof. For the shingle, metal, violate this standard. Only the rainwater harvested after the
and cool pilot-scale roofs, the conductivity of the first-flush first-flush from the metal roof meets the non-potable urban
was significantly higher than that of the rainwater harvested water reuse guidelines (USEPA, 2004), which state that FC
after the first-flush ( p-values < 0.024). should not exceed 14 colony forming units (CFU)/100 mL in
Conductivity and total dissolved solids (TDS) are corre- any sample.
lated, although a specific correlation was not prepared in this Our data are consistent with the literature, which shows
study. Using the correlation of Singh and Kalra (1975) between detectable TC and FC in harvested rainwater (Yaziz et al., 1989;
conductivity and TDS in mountain streams, the average, Simmons et al., 2001). The use of these traditional indicator
estimated TDS of rainwater harvested after the first-flush for bacteria might cause the underestimation of the risk of
all pilot-scale roofs met the USEPA secondary maximum waterborne disease to the rainwater consumer (Lye, 2002;
contaminant level (SMCL) for TDS, which is 500 mg/L. Ahmed et al., 2008). Nonetheless, based on the TC and FC
concentrations in the rainwater harvested after the first-flush,
3.2. TC and FC e pilot-scale roofs most of the pilot-scale roofs produced water that did not meet
potable and non-potable guidelines. Thus treatment is needed
The harvested rainwater was assessed using TC (Fig. 4A) and if the guidelines are to be met.
FC (Fig. 4B) as indicators of microbial quality. Due to filtration of the rainwater through the substrate
For the shingle, tile, and cool pilot-scale roofs, the TC layer, we expected the pilot-scale green roof to consistently
concentration of the first-flush was significantly higher than yield the lowest concentrations of indicator bacteria in har-
that of the rainwater harvested after the first-flush ( p- vested rainwater, but this was not always the case. For
values < 0.024), but the TC concentration from the metal and instance, while the first two rain events showed no detectable
green pilot-scale roofs did not change significantly from the FC in the rainwater harvested from the green roof, the third
first-flush to the subsequent tanks ( p-values ¼ 0.131 and 0.179, rain event showed a spike in FC (330 CFU/100 mL) in the
respectively). For all of the pilot-scale roofs, the TC concentra- rainwater harvested after the first-flush. The third rain event
tion of the rainwater harvested after the first-flush was statis- also showed a spike in TC concentrations from the green roof
tically indistinguishable from that in the ambient sampler but no spikes in TSS or turbidity (data not shown). The other
( p-values > 0.131). According to the USEPA primary drinking pilot-scale roofs did not have similar contaminant spikes in
water standards, no more than five percent of samples in the third rain event; thus, the reason for the spike in indicator
a month are allowed to be positive for TC; all of the pilot-scale bacteria concentrations in this event from the green roof is
roofs violate this standard because they showed measurable TC unclear at this time. On a per-event basis, the green roof most
in the first-flush and subsequent tanks for all rain events. often showed the best water quality with respect to indicator
For the shingle, metal, tile, and cool pilot-scale roofs, the bacteria among the roofing materials, but the potential for
FC concentration of the first-flush was significantly higher spikes in indicator bacteria concentrations from the green
than that of the rainwater harvested after the first-flush roof should be investigated further.
( p-values < 0.083). For all of the pilot-scale roofs except the For the pilot-scale metal roof, the FC concentration of the
metal roof, the FC concentration of the rainwater harvested rainwater harvested after the first-flush was significantly

Fig. 4 e Geometric mean of TC (panel A) and FC (panel B) for the pilot-scale events: ( ) Quality of the first-flush, ( ) Quality
after the first-flush (average of tank 1 and tank 2), - - - Ambient sampler. One geometric standard deviation is shown.
2054 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 0 4 9 e2 0 5 9

lower than that in the ambient sampler ( p-value ¼ 0.083). The must never be above 1 nephelometric turbidity unit (NTU),
TC and FC concentrations of the first-flush from the metal roof and 95% of samples in one month must be less than or equal
were significantly lower than those for the cool, shingle, and to 0.3 NTU. According to the non-potable urban water reuse
tile roofs ( p-values < 0.001). Similarly, Yaziz et al. (1989) guidelines (USEPA, 2004), the average turbidity in a 24-h period
showed that the concentration of indicator bacteria in rain- should be less than or equal to 2 NTU and turbidity should not
water harvested from metal roofs was lower than that from exceed 5 NTU at any time. All samples of rainwater harvested
concrete tile roofs. One possible reason for this trend is that after the first-flush from the pilot-scale roofs violate these
low emissivity materials, like metals, have higher surface standards.
temperatures in sunlight (Bretz et al., 1998), which might The TSS data showed very similar trends to the turbidity
have inactivated a fraction of the indicator bacteria. Another data. For example, the TSS of the rainwater harvested after
possible reason for this trend is that heavy metal exposure the first-flush for the green roof was lower than those for the
can decrease cellular viability (Teitzel and Parsek, 2003), so other pilot-scale roofs ( p-values < 0.047); thus, the green roof
contact with the metal roofing material might have inacti- produced the highest quality rainwater with respect to TSS.
vated a fraction of the indicator bacteria. No primary or secondary drinking water standards exist for
TSS. The non-potable urban water reuse guidelines state that
3.3. Turbidity and TSS concentrations e pilot-scale roofs TSS should not exceed 5 mg/L (USEPA, 2004); the rainwater
harvested after the first-flush for all pilot-scale roofs often
Turbidity (Fig. 5A) and TSS concentrations (Fig. 5B) were exceeded this standard.
measured for the pilot-scale roofs. Based on the turbidity and TSS concentrations in the
For all of the pilot-scale roofs, the turbidity of the first-flush rainwater harvested after the first-flush, all of the pilot-scale
was significantly higher than that of the rainwater harvested roofs produced water that did not meet potable and non-
after the first-flush ( p-values < 0.083). The turbidity of the potable guidelines. Thus, treatment is needed if the guidelines
first-flush from the metal and cool pilot-scale roofs was are to be met.
significantly higher than those for the shingle, tile, and green
pilot-scale roofing materials ( p-values < 0.001). This was 3.4. Nitrate and nitrite concentrations e pilot-scale roofs
expected because the metal and cool roofs visually had the
smoothest surfaces, and Egodawatta et al. (2009) noted that Nitrate (Fig. 6A) and nitrite (Fig. 6B) concentrations were
particulate matter washes off from smoother surfaces more measured for the pilot-scale roofs.
rapidly than it does from rougher surfaces. For the tile, cool, For the shingle, metal, and tile pilot-scale roofs, the average
and green pilot-scale roofs, the turbidity of the rainwater nitrate concentration of the first-flush was significantly higher
harvested after the first-flush was statistically indistinguish- than that of the rainwater harvested after the first-flush
able from that in the ambient sampler ( p-values > 0.131). The ( p-values < 0.082). For all of the pilot-scale roofs, the average
turbidity of the rainwater harvested after the first-flush for the nitrate concentration of the rainwater harvested after the first-
green roof was lower than that for the shingle, metal, and tile flush was statistically indistinguishable from that in the
pilot-scale roofs but indistinguishable from that of the cool ambient sampler ( p-values > 0.25). Elevated nitrate levels were
pilot-scale roof ( p-values < 0.078); thus, the green roof not expected in the rainwater harvested from the green roof
produced the highest quality rainwater with respect to because no fertilizers were applied to the green roof, and plants
turbidity. and soil biota can consume nitrate. All samples of the rain-
According to the USEPA primary drinking water standards, water harvested after the first-flush for all pilot-scale roofs met
for systems using conventional or direct filtration, turbidity the USEPA MCL for nitrate (10 mg-N/L), consistent with the

Fig. 5 e Average turbidity (panel A) and TSS concentrations (panel B) for the pilot-scale events: ( ) Quality of the first-flush,
( ) Quality after the first-flush (average of tank 1 and tank 2),  USEPA filtered system guideline (1 NTU), - - - Ambient
sampler. One standard deviation is shown.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 0 4 9 e2 0 5 9 2055

Fig. 6 e Average nitrate (panel A) and nitrite (panel B) concentrations for the pilot-scale events: ( ) Quality of the first-flush,
( ) Quality after the first-flush (average of tank 1 and tank 2),  USEPA MCLs for nitrate (10 mg-N/L) and nitrite (1 mg-N/L),
- - - Ambient sampler. One standard deviation is shown.

findings of Nicholson et al. (2009) for asphalt fiberglass shingle, concentrations in the rainwater harvested after the first-
metal, and green roofs. flush as compared to all of the other pilot-scale roofs ( p-
The average concentration of nitrate in the first-flush from values < 0.047). Given the nature of the roofing materials, the
all of the pilot-scale roofs was greater than the average nitrate shingle and green pilot-scale roofs were expected to be sources
concentration in the ambient sampler (Fig. 6A), which is likely of organic matter. Asphalt fiberglass shingles are noted for
due to dry deposition of nitrate onto the roof surfaces (Förster, producing colored water (Despins et al., 2009). Berndtsson et al.
1998). The nitrate concentrations in the first-flush increased (2009) hypothesized that the elevated DOC concentrations
from 4 to 8 antecedent dry days and then either reached observed in green roof-runoff come from the soil organic matter
a plateau or continued to increase from 8 to 14 antecedent dry and from decay of the vegetation. DOC also might be produced
days (Fig. 7), supporting the hypothesis that dry deposition by soil bacteria colonizing the green roof. The DOC concentra-
contributes to nitrate levels in harvested rainwater. tions in the rainwater harvested after the first-flush for the
For all of the pilot-scale roofs, the average nitrite concen- shingle and green roofs were elevated as compared to the DOC
tration of the first-flush was significantly higher than that of concentration of a typical, finished drinking water in the U.S.
the rainwater harvested after the first-flush ( p-values < 0.083). (w2.5e3.5 mg/L).
For the metal, cool, and green pilot-scale roofs, the average Although DOC does not have an MCL, it is a precursor for
nitrite concentration of the rainwater harvested after the first- regulated disinfection byproducts (e.g., trihalomethanes and
flush was statistically indistinguishable from that in the haloacetic acids). To reduce disinfection byproduct formation,
ambient sampler ( p-values > 0.25), and the shingle and tile the Stage 1 Disinfectants/Disinfection Byproducts Rule
pilot-scale roofs produced nitrite concentrations in the rain- mandates that water systems that treat with disinfectants
water harvested after the first-flush that were just slightly other than ultraviolet light must reduce the total organic
higher than ambient ( p-values < 0.083). All samples of the carbon of the water by 15e50%, where the percentage is based
rainwater harvested after the first-flush for all pilot-scale roofs
met the USEPA MCL for nitrite (1 mg-N/L).

3.5. DOC concentrations e pilot-scale roofs

DOC concentrations were measured for the pilot-scale roofs


(Fig. 8). For the metal and cool roofs, the average DOC concen-
tration of the first-flush was significantly higher than that of the
rainwater harvested after the first-flush ( p-values < 0.048). For
the metal, tile, and cool pilot-scale roofs, the average DOC
concentration of the rainwater harvested after the first-flush
was statistically indistinguishable from that in the ambient
sampler ( p-values > 0.33). Conversely, for the shingle and green
pilot-scale roofs, the average DOC concentration in the rain-
water harvested after the first-flush was significantly higher
than that in the ambient sampler ( p-values < 0.012). In fact,
the shingle pilot-scale roof showed a significant increase in Fig. 7 e Effect of antecedent dry days on average nitrate
DOC concentration from rainwater in the first-flush tank to concentrations in first-flush samples from the pilot-scale
the rainwater harvested after the first-flush ( p-value < 0.012). roofs: dCd Shingle, dBd Metal, d;d Tile, d6d Cool,
The green roof showed statistically significant higher DOC d-d Green,  USEPA MCL for nitrate (10 mg-N/L).
2056 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 0 4 9 e2 0 5 9

harvested from the green roof are hypothesized to have come


from the growing media, which suggests that the quality of
commercial growing media must be carefully examined if the
harvested rainwater is being considered for domestic use.
Third, the rainwater harvested after the first-flush from the
shingle roof had copper concentrations that were significantly
higher than that in the ambient sampler ( p-value ¼ 0.012) and
those from all of the other pilot-scale roofs ( p-values < 0.001),
suggesting that the shingle roof is a source of copper. This is
consistent with the study of Nicholson et al. (2009), which
showed that rainwater harvested from asphalt fiberglass
shingle roofs contained more copper than did rainwater har-
vested from galvanized metal or green roofs. In other studies,
elevated copper in roof-runoff was attributed to atmospheric
deposition due to vehicles (Van Metre and Mahler, 2003) and
Fig. 8 e Average DOC concentrations for the pilot-scale copper fittings (Förster, 1999), but those sources were not
events: ( ) Quality of the first-flush, ( ) Quality after the present at our pilot-scale site.
first-flush (average of tank 1 and tank 2), - - - Ambient The concentration of each metal in the rainwater harvested
sampler. One standard deviation is shown. after the first-flush was compared to the appropriate USEPA
primary and secondary drinking water standards or action
on the total alkalinity and source water total organic carbon levels (Fig. 9). None of the pilot-scale roofs violated the primary
concentration. Thus, systems with high source water total standards or action levels for the rainwater harvested after the
organic carbon concentrations (>8 mg/L) must decrease the first-flush (i.e., arsenic, copper, and lead). However, aluminum
total organic carbon by 30e50%. Given the particularly high (Fig. 9A) and iron (Fig. 9D) concentrations in the rainwater
DOC concentrations in the rainwater harvested after the first- harvested after the first-flush often exceeded the secondary
flush from the shingle and green roofs, it is quite possible that standards. All of the pilot-scale roofs violated the aluminum
the concentrations of disinfection byproducts formed in these SMCL of 200 mg/L in the rainwater harvested after the first-flush.
waters would violate the primary drinking water standards. If Only the green pilot-scale roof did not violate the iron SMCL of
the water is used for non-potable purposes, the consumer also 300 mg/L in the rainwater harvested after the first-flush.
might come into contact with these disinfection byproducts by Zinc concentrations in the rainwater harvested after the
inhalation, since many of them are volatile (e.g., chloroform). first-flush from all roofing materials were at least one order of
magnitude below the zinc secondary drinking water standard
3.6. Metal concentrations e pilot-scale roofs (Fig. 9F). Previous studies have shown that galvanized metal
roofs (Kingett Mitchell Ltd., 2003; Nicholson et al., 2009) or
The average concentrations of cadmium, chromium, and painted galvanized metal roofs in less-than-excellent condition
selenium in the rainwater harvested after the first-flush for (Kingett Mitchell Ltd., 2003) yield high concentrations of zinc
each pilot-scale roof were at least an order of magnitude below (mg/L) in harvested rainwater. Conversely, painted galvanized
the corresponding primary drinking water standard (data not metal roofs in excellent condition and Zincalume (55%
shown); these metals are not discussed further. Data for the aluminum-zinc alloy coated steel, which is the same compo-
remaining metals (aluminum, arsenic, copper, iron, lead, and sition as Galvalume) roofs yield lower concentrations of zinc
zinc) are shown in Fig. 9. (mg/L) in harvested rainwater (Kingett Mitchell Ltd., 2003),
In general, the average concentration of each metal in the similar to our data from the Galvalume roof. This is consistent
first-flush was significantly higher than that in the rainwater with the work of Sullivan and Worsley (2002), who found that
harvested after the first-flush at a 10% significance level. The the presence of aluminum in alloy-coated steel roofs increases
notable exception to this is the green roof, where the metal the corrosion resistance as compared to uncoated galvanized
concentrations (except for lead) in the first-flush were statis- roofs, thereby decreasing the zinc concentration in roof-runoff.
tically indistinguishable from those in the rainwater har- In terms of primary standards and action levels for metals
vested after the first-flush at a 10% significance level. in drinking water, all of the pilot-scale roofs produced
In general, the average concentration of each metal in the acceptable quality water. The green pilot-scale roof produced
rainwater harvested after the first-flush was significantly the best quality water in terms of secondary standards, as its
less than or statistically indistinguishable from that in the aluminum and iron concentrations in the rainwater harvested
ambient sampler at a 10% significance level, with three after the first-flush were the lowest as compared to the other
notable exceptions. First, the rainwater harvested after the pilot-scale roofs.
first-flush from the tile and green roofs had arsenic and
lead concentrations that were significantly higher than those 3.7. Water quality ranges e comparing full-scale and
in the ambient sampler ( p-values < 0.083). Second, the rain- pilot-scale roofs
water harvested after the first-flush from the metal, tile, and
green roofs had zinc concentrations that were significantly Previous studies have shown that variation in roof-runoff
higher than those in the ambient sampler ( p-values < 0.083). quality is attributable to the age of the roof, presence of over-
The elevated arsenic, lead, and zinc found in the rainwater hanging vegetation, wind direction, and proximity to local
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 0 4 9 e2 0 5 9 2057

Fig. 9 e Average total aluminum (panel A), arsenic (panel B), copper (panel C), iron (panel D), lead (panel E), and zinc (panel F)
concentrations for the pilot-scale events: ( ) Quality of the first-flush, ( ) Quality after the first-flush (average of tank 1 and
tank 2),  USEPA primary or secondary drinking water standards or action levels: aluminum (200 mg/L), arsenic (10 mg/L),
copper (1300 mg/L), iron (300 mg/L), lead (15 mg/L), and zinc (5000 mg/L). - - - Ambient sampler. One standard deviation is
shown.

sources (Förster, 1999; Kingett Mitchell Ltd., 2003; Evans et al., shingle roofs at a concentration of 3 mg/L; benzyl alcohol also
2006). In contrast to the pilot-scale roofs, our full-scale roofs was detected in the first-flush from the same shingle roof at
were older, had overhanging vegetation, and were located a concentration of 0.20 mg/L. These compounds are both
closer to local emission sources (e.g., chimneys and traffic). unregulated in drinking water. Our lack of detection of PAHs
Despite these differences, the qualities of the rainwater in harvested rainwater is consistent with Van Metre and
harvested after the first-flush from the pilot-scale and full- Mahler (2003), who observed that metal and shingle roofing
scale roofs were remarkably similar (Table 1). Consistent with materials are not a source of PAHs but that PAHs come from
the pilot data, the rainwater harvested after the first-flush atmospheric deposition. Our lack of detection of pesticides in
from the full-scale metal roof had average TC and FC
concentrations that were significantly lower than those from
both full-scale shingle roofs ( p-values < 0.09). Also consistent
with the pilot data, the TC, FC, turbidity, and TSS in the
Table 1 e Water quality parameters
rainwater harvested after the first-flush for the full-scale roofs
(minimumemaximum) of the rainwater harvested after
do not meet the USEPA primary drinking water standards nor the first-flush for the pilot-scale and full-scale roofs.
the non-potable urban water reuse guidelines. Similar to the
Parameter Metal Shingle
pilot data, the average nitrate and nitrite concentrations in the
rainwater harvested after the first-flush for the full-scale roofs Pilot-scale Full-scale Pilot-scale Full-scale
were well below the USEPA MCLs. pH 6.0e6.8 5.4e6.3 6.7e6.9 5.8e6.5
Conductivity 9e56 18e60 18e57 20e102
(mS/cm)
3.8. VOCs, SVOCs, PAHs, and pesticides e pilot- and
TC (CFU/100 mL) 117e770 64e173 177e1367 102e353
full-scale roofs FC (CFU/100 mL) <1e8 37e127 9e87 73e253
Turbidity (NTU) 7e30 5e35 8e24 6e23
The first-flush samples and ambient rain samples from one TSS (mg/L) 20e87 10e50 12e54 20e150
rain event were analyzed for PAHs and pesticides for the pilot- Nitrate (mg-N/L) 0.0e2.0 0.4e4.1 0.0e1.8 0.3e4.7
scale roofs and for a larger suite of VOCs and SVOCs for the Nitrite (mg-N/L) 0.01e0.03 0.01e0.05 0.01e0.04 0.01e0.06
DOC (mg/L) 2e11 4e13 10e15 5e31
full-scale roofs. In the pilot-scale study, no pesticides or PAHs
Lead (mg/L) 0.3e2.3 2.1e5.8 0.4e1.2 0.7e8.6
were detected. In the full-scale study, 2,4-dinitrophenol was
Zinc (mg/L) 77e362 18e23 8e85 1e15
detected in the first-flush of the metal roof and one of the
2058 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 0 4 9 e2 0 5 9

harvested rainwater is likely related to surrounding land-use, of commercial growing media should be carefully examined
where major pesticide applications are absent. if the harvested rainwater is being considered for domestic
use.
 Concentrations of most of the tested water quality param-
4. Conclusions eters decreased as a result of diverting a first-flush of at least
38 L for every 93 m2 of collection area. While the depth of
Conventional roofing materials (i.e., asphalt fiberglass shingle, rainfall to be diverted is a subject of debate, a first-flush
Galvalume metal, and concrete tile) and alternative roofing device can provide some improvement in water quality.
materials (i.e., cool and green) were examined for their suit-
ability to harvest rainwater for domestic use. pH, conductivity,
TC, FC, turbidity, TSS, nitrite, nitrate, DOC, metals, VOCs, and
SVOCs were measured in harvested rainwater. The following Acknowledgments
conclusions were drawn:
We are grateful to several sources of funding for this work: the
 Rainwater harvested from any of the tested roofing mate- Texas Water Development Board (contract number
rials would require treatment if the consumer wanted to 0804830855), Texas AgriLife Research (subcontract number
meet the USEPA primary and secondary drinking water 570464), the National Science Foundation (Graduate Research
standards or the USEPA non-potable water reuse guidelines. Fellowship to C.B. Mendez), and the American Water Works
In particular, the rainwater harvested after a first-flush Association and CH2M Hill (Holly A. Cornell Scholarships to
consisting of a minimum of 38 L for every 93 m2 of collection B.R. Afshar and C.B. Mendez). We thank Dr. Sanjeev Kalaswad
area from the asphalt fiberglass shingle, metal, concrete and Mr. Jorge Arroyo from the Texas Water Development
tile, and cool roofs would need treatment for TC, FC, Board for their input on this project and Mr. David Phillips (Ja-
turbidity, aluminum, and iron to meet the drinking water Mar Roofing) for donating the materials and labor for
standards. The rainwater harvested after the first-flush construction of the pilot-scale roofs. We also thank Mr.
from the green roof would need treatment for TC, FC, Charles Perego, Dr. Steve Windhager, Dr. Cindy Menches, Dan
turbidity, and aluminum. Thus, at a minimum, first-flush Hemme, Jeffrey Stump, Brett Buff, and David Stump for their
diversion, filtration, and disinfection are recommended to assistance on the project.
meet current USEPA standards/guidelines. However, since
the quality of ambient rainwater varies by location, we
caution that rainwater harvested in other locations might references
exceed standards/guidelines for different water quality
parameters than those observed in this study.
 While metal roofs are commonly recommended for rain- Ahmed, W., Huygens, F., Goonetilleke, A., Gardner, T., 2008.
water harvesting applications, the metal roofs examined in Real-time PCR detection of pathogenic microorganisms in
this study did not produce clearly superior harvested rain- roof-harvested rainwater in southeast Queensland,
Australia. Applied and Environmental Microbiology 74 (17),
water quality as compared to the other roofing materials.
5490e5496.
Concrete tile and cool roofs also appear to be good candi-
American Public Health Association, 1998. Standard Methods for
dates for rainwater harvesting catchments, given the overall the Examination of Water and Wastewater, twentieth ed.
similarity in harvested rainwater quality among those three Washington, D.C.
materials. Berndtsson, J.C., Bengtsson, L., Jinno, K., 2009. Runoff water
 The DOC concentrations in rainwater harvested from the quality from intensive and extensive vegetated roofs.
shingle and green roofs are very high (i.e., approximately Ecological Engineering 35 (3), 369e380.
Bretz, S., Akbari, H., Rosenfeld, A., 1998. Practical issues for using
one order of magnitude higher than that of a typical,
solar-reflective materials to mitigate urban heat islands.
finished drinking water in the U.S.) and could lead to high Atmospheric Environment 32 (1), 95e101.
concentrations of disinfection byproducts. The formation of Bucheli, T.D., Müller, S.R., Voegelin, A., Schwarzenbach, R.P.,
trihalomethanes and haloacetic acids, two classes of regu- 1998. Bituminous roof sealing membranes as major
lated disinfection byproducts in drinking water, should be sources of the herbicide (R,S )-mecoprop in roof runoff
assessed after disinfecting harvested rainwater with chlo- waters: potential contamination of groundwater and surface
rine. The elevated DOC concentrations observed in the waters. Environmental Science and Technology 32 (22),
3465e3471.
rainwater harvested from the shingle and green roofs in our
Despins, C., Farahbakhsh, K., Leidl, C., 2009. Assessment of
short-term study suggest that other roofing materials (e.g., rainwater quality from rainwater harvesting systems in
metal, tile, and cool) might be preferable for rainwater Ontario, Canada. Journal of Water Supply: Research and
harvesting applications when chlorine is the disinfectant of Technology-AQUA 58 (2), 117e134.
choice. Egodawatta, P., Thomas, E., Goonetilleke, A., 2009. Understanding
 The rainwater harvested from the metal roof showed lower the physical processes of pollutant build-up and wash-off on
roof surfaces. Science of the Total Environment 407 (6),
concentrations of fecal indicator bacteria than did the other
1834e1841.
roofing materials. This might be due to the low emissivity of
Evans, C.A., Coombes, P.J., Dunstan, R.H., 2006. Wind, rain,
metal, resulting in higher surface temperatures on the roof. and bacteria: the effect of weather on the microbial
 The concentrations of some metals (e.g., arsenic) in rain- composition of roof-harvested rainwater. Water Research
water harvested from the green roof suggest that the quality 40 (1), 37e44.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 0 4 9 e2 0 5 9 2059

Förster, J., 1998. The influence of location and season on Singh, T., Kalra, Y.P., 1975. Specific conductance method for in
the concentrations of macroions and organic trace pollutants situ estimation of total dissolved solids. Journal of American
in roof runoff. Water Science and Technology 38 (10), 83e90. Water Works Association 67 (2), 99e100.
Förster, J., 1999. Variability of roof runoff quality. Water Science Sullivan, J.H., Worsley, D.A., 2002. Zinc runoff from galvanised
and Technology 39 (5), 137e144. steel materials exposed in industrial/marine environment.
Kingett Mitchell Ltd., 2003. A Study of Roof Runoff Quality in British Corrosion Journal 37 (4), 282e288.
Auckland New Zealand: Implications for Stormwater Teitzel, G.M., Parsek, M.R., 2003. Heavy metal resistance
Management. Auckland Regional Council, Auckland, of biofilm and planktonic Pseudomonas aeruginosa.
New Zealand. Applied and Environmental Microbiology 69 (4),
Lee, J.Y., Yang, J.S., Han, M., Choi, J., 2010. Comparison of the 2313e2320.
microbiological and chemical characterization of harvested Texas Rainwater Harvesting Evaluation Committee, 2006.
rainwater and reservoir water as alternative water resources. Rainwater Harvesting Potential and Guidelines for Texas.
Science of the Total Environment 408 (4), 896e905. Texas Water Development Board, Austin, Texas.
Lye, D.J., 2002. Health risks associated with consumption of The Texas Manual on Rainwater Harvesting, third ed., 2005.
untreated water from household roof catchment systems. Texas Water Development Board, Austin, Texas.
Journal of the American Water Resources Association 38 (5), USEPA, 1996a. Organochlorine Pesticides by Gas Chromatography
1301e1306. SW-846 Ch. 4.3.1, Washington, D.C.
National Oceanic and Atmospheric Administration, 2010. Austin USEPA, 1996b. Semivolatile Organic Compounds by Gas
Climate Records e NWS Austin/San Antonio, National Chromatography/Mass Spectrometry SW-846 Chapter 4.3.2,
Weather Service. National Oceanic and Atmospheric Washington, D.C.
Administration. http://www.srh.noaa.gov/ewx/?n¼ausclidata. USEPA, 1996c. Volatile Organic Compounds by Gas
htm (accessed July 2010). Chromatography/Mass Spectrometry SW-846 Chapter 4.3.2,
Nicholson, N., Clark, S.E., Long, B.V., Spicher, J., Steele, K.A., 2009. Washington, D.C.
Rainwater harvesting for non-potable use in gardens: USEPA, 2004. Guidelines for Water Reuse Washington, D.C. EPA/
a comparison of runoff water quality from green vs. traditional 625/R-04/108.
roofs. In: Proceedings of World Environmental and Water Van Metre, P.C., Mahler, B.J., 2003. The contribution of particles
Resources Congress 2009 e Great Rivers Kansas City, Missouri. washed from rooftops to contaminant loading to urban
Simmons, G., Hope, V., Lewis, G., Whitmore, J., Gao, W., 2001. streams. Chemosphere 52 (10), 1727e1741.
Contamination of potable roof-collected rainwater in Yaziz, M.I., Gunting, H., Sapari, N., Ghazali, A.W., 1989. Variations
Auckland, New Zealand. Water Research 35 (6), 1518e1524. in rainwater quality from roof catchments. Water Research 23
Simmons, M.T., Gardiner, B., Windhager, S., Tinsley, J., 2008. (6), 761e765.
Green roofs are not created equal: the hydrologic and thermal Zobrist, J., Müller, S.R., Ammann, A., Bucheli, T.D., Mottier, V.,
performance of six different extensive green roofs and Ochs, M., Schoenenberger, R., Eugster, J., Boller, M., 2000.
reflective and non-reflective roofs in a sub-tropical climate. Quality of roof runoff for groundwater infiltration. Water
Urban Ecosystems 11 (4), 339e348. Research 34 (5), 1455e1462.

You might also like