401 © IWA Publishing 2021 Water Supply | 21.

1 | 2021

Comparison of crushed rock sand and natural river sand

as filter media for rapid filtration
Bruno Moreno Ramos da Silva, Rafael Kopschitz Xavier Bastos and
Pedro Kopschitz Xavier Bastos

ABSTRACT

The objective of this work was to evaluate the crushed rock sand (CRS) as a filter bed in rapid Bruno Moreno Ramos da Silva (corresponding
author)
filtration for water treatment. The experiments were carried out using pilot-scale filtration units: one Rafael Kopschitz Xavier Bastos
Department of Civil Engineering,
with a CRS filter bed and the other with natural river sand (NRS). Both filter media were prepared in Universidade Federal de Viçosa,
Av. P. H. Rolfs s/n, Viçosa, Minas Gerais,
accordance with typical standards for rapid sand filtration (particle size range and distribution, and Brasil
filter bed depth), and were further characterized in terms of chemical composition, particles and bulk E-mail: [email protected]

density, porosity, acid solubility and sphericity coefficient. Over four months, 14 filter runs using Pedro Kopschitz Xavier Bastos
Department of Civil Engineering,
filtration rates of 90, 180, 270 and 360 m3 m
2 d
1 were monitored and characterized in terms of run Universidade Federal de Juiz de Fora,
Campus Universitário, Rua José Lourenço Kelmer,
length, head loss increase along filter bed depth, turbidity removal along filter bed depth. Overall, the
s/n, Juiz de Fora, Minas Gerais,
performance of the CRS filter was similar to or even better than that of the NRS filter, producing Brasil

filtered water with turbidity lower than 0.50 NTU along the entire run, with head loss increasing rates
and run length similar to those of the NRS filter. It is concluded that CRS presents a high potential for
use as filter media for rapid filtration in water treatment, without technical or operational
disadvantages.
Key words | drinking water treatment, filter bed media, head loss, rapid filtration, turbidity

HIGHLIGHTS

• Crushed rock sand is a suitable filter bed media for rapid filtration.

• Similar turbidity removal and head loss increasing rate compared to natural river sand.

• High efficiency even under high filtration rates and high influent water turbidity.

doi: 10.2166/ws.2020.311

Downloaded from http://iwaponline.com/ws/article-pdf/21/1/401/840813/ws021010401.pdf

by guest
 402 B. M. R. da Silva et al. | Comparison of crushed rock and natural river sand as filter media Water Supply | 21.1 | 2021

GRAPHICAL ABSTRACT

INTRODUCTION

Drinking water treatment (DWT) plays a major role in effectively remove protozoa cysts (e.g. Giardia) and oocysts
waterborne diseases’ control. Currently, several techniques (e.g. Cryptosporidium) due to the similar characteristics pre-
are available for particle removal and disinfection, present- sented by (oo)cysts and colloidal or suspended particles,
ing different efficiencies, complexity and costs. Technology such as size (Emelko et al. ) and surface negative
selection for DWT relies mainly on raw water quality, charge (Bustamante et al. ), they can both be effectively
regulatory requirements, operational constraints, and econ- retained in the filter bed by also similar mechanisms (phys-
omical feasibility (Crittenden et al. ). Nowadays, ical-chemical mechanisms) (Betancourt & Rose ). Thus,
available technologies for water treatment are able to filtration may be the last barrier against disinfectant-resistant
grant drinking water production even from highly polluted organisms, such as Cryptosporidium, whose oocysts are
water sources. For instance, membrane filtration promotes extremely resistant to chlorine (the most frequently used dis-
the removal of micropollutants, besides being more effective infectant worldwide) (Tchobanoglous et al. ). However,
for the removal of particles and microorganisms compared due to financial and analytical constraints, monitoring pro-
to conventional systems (Mohammad et al. ). However, tozoa removal in a water treatment plant (WTP) on a
although membrane treatment technology costs have routine basis is rather difficult. Hence, among other surro-
decreased, high needs for capital, energy and specialized gates (like aerobic spore-forming bacteria; Headd &
personnel for operation are still limiting for developing Bradford ), filtered water turbidity has been used for
countries (Schäfer et al. ), where simpler technologies evaluating protozoa removal in DWT, and such an approach
are preferred. In Brazil, conventional DWT comprised of is present in many drinking water quality regulations
coagulation, flocculation, sedimentation, rapid filtration and (USEPA ; Ministry of Health ; Brasil ; Health
disinfection is, by far, the most used treatment technology. Canada ).
Granular media filtration is an important step in con- Rapid filtration is a combined result of physical and
ventional DWT, since: (i) it is the last step for water physical-chemical processes that take place as a dilute
clarification; (ii) it makes subsequent disinfection more water suspension passes through a granular medium
effective (as suspended particles can protect microorgan- (Hendricks ). Particles are transported near the surface
isms from the disinfectants’ action); and (iii) it can of the bed grains and are retained by attachment

Downloaded from http://iwaponline.com/ws/article-pdf/21/1/401/840813/ws021010401.pdf

by guest
 403 B. M. R. da Silva et al. | Comparison of crushed rock and natural river sand as filter media Water Supply | 21.1 | 2021

mechanisms, narrowing the pores of the filter media. Thus, Rao () had looked at using a crushed stone filter media
the interstitial velocity continuously increases during the prepared out of the waste dust from stone quarries.
filter run, increasing the head loss both over time and In Brazil, and elsewhere, there is a high demand for
through the filter bed depth. After some time of effective fil- sand for construction purposes. In big cities, sand has to
tration, breakthrough may occur due to either a decrease in be transported over long distances to meet the demand,
particle capture and/or an increase in detachment (Critten- hence increasing costs (Hofmann et al. ). In addition,
den et al. ). When the available head loss is exceeded extraction of NRS severely impacts the environment, caus-
and/or filtered water turbidity reaches a threshold value ing exposure of groundwater, deforestation and changes in
(generally imposed by drinking water standards) the filter the morphology of river channels as well as in its biotic com-
must be cleaned, usually by backwashing. However, since munities (Santo & Sánchez ). Thus, fine aggregates from
backwashing uses treated water, it is desirable to optimize rock mining have been studied as a substitute for sand for
treatment operation in order to increase the filter run. concrete production (Manasseh ; Manguriu et al.
The main factors that affect rapid filtration performance ). This material, often named artificial sand, industrial
are: influent water quality (e.g. turbidity, particle concen- sand or crushed rock sand (CRS) has less impurities, is
tration, particle size distribution, level of pre-treatment), more environmentally friendly and cheaper compared to
filtration rate, and the filter bed characteristics (e.g. porosity, NRS (Mundra et al. ). In this paper, the performance
depth, grains density, shape and zeta potential, size distri- of a rapid filter with gneissic CRS as filter bed was compared
bution – effective size and uniformity coefficient) to a conventional NRS filter in pilot-scale experiments.
(Crittenden et al. ). Depending on the coagulation pro-
cess, flocs with different sizes and destabilization levels
(floc strength) may be formed, affecting the adherence of METHODS
the particles to the filter bed and, consequently, turbidity
removal and increasing head loss rate. Filter media characterization
The material most commonly used in single medium
rapid filters is natural river sand (NRS) (silica sand) (Soyer NRS was obtained from a local WTP that had recently
et al. ), which is usually extracted from river beds and replaced the filter bed of its rapid filters. CRS was provided
is sieved according to the rapid filtration media require- by a mining company that produces aggregates for civil con-
ments for size distribution. Anthracite coal, garnet sand struction. Firstly, the raw materials (NRS and CRS) were
and ilmenite are examples of materials used in multimedia sieved in order to obtain the typical size distribution for
rapid filters (AWWA ). Depending on the availability, rapid sand filtration: effective size (d10) ¼ 0.45 mm and uni-
price and, most importantly, filtration performance, other formity coefficient (UC) ¼ 1.4. Both materials were then
materials may be employed instead of sand. According to washed with treated water and dried in an oven at 105  C
Uluatam (), limited availability of sand suitable for for 24 hours prior to physical and chemical characterization.
rapid filtration, along with the high cost of other materials For the chemical characterization, a 100 g sample of each
(e.g. anthracite, garnet sand), has led to a search for material was crushed to obtain particle sizes lower than
alternative materials as substitutes for sand in Turkey, and 0.075 mm. Chemical analyses were carried out by X-ray fluor-
successful experiences have been reported using perlite escence (XRF) using a Shimadzu equipment, model EDX-700.
(Uluatam , ), pumice (Farizoglu et al. ) and The following physical characteristics of both NRS and CRS
crushed glass (Soyer et al. ). Crushed glass has also grains were determined: particle density (ρs), bulk density
been evaluated by Rutledge & Gagnon () and Cescon (ρb), clean bed porosity (εo), acid solubility (S) and sphericity
et al. (), along with expanded clay in the latter. In coefficient (ψ). Particle density was determined using a water
Canada, as far back as 1995, the performance of a crushed displacement technique. Bulk density was obtained by measur-
quartz filter was compared to that of an anthracite-sand ing the mass needed to fill a 10 L-cylinder. Bed porosity was
dual-media filter (Suthaker et al. ). Earlier, in India, then calculated from the relationship between particle and

Downloaded from http://iwaponline.com/ws/article-pdf/21/1/401/840813/ws021010401.pdf

by guest
 404 B. M. R. da Silva et al. | Comparison of crushed rock and natural river sand as filter media Water Supply | 21.1 | 2021

bulk densities. Acid solubility was obtained by measuring the was more than 100 times the effective size of the filter
mass of each material after immersion of 100 g in HCl 1:1 media (0.45 mm), as recommended by Kawamura ()
(v/v) for 30 minutes (AWWA ). The sphericity coefficients in order to minimize sidewall effects. Through the filter
were determined by measuring clean bed head losses at differ- bed depth, eight piezometers were installed for head loss
ent filtration rates, using the Ergun equation as described by measuring: one at the support layer-bed interface; the next
Soyer & Akgiray (), and Soyer et al. (). two at 10 and 20 cm above the support layer, respectively;
and the remainder at 5 cm intervals. The water inlet was
Pilot filtration units made by a 32 mm pipe connected to a channel placed
20 cm above the filter bed, which was also used for collecting
Two identical pilot-scale filters were built with acrylic col- the backwash water. The flow rate was controlled by globe
umns, 190 mm internal diameter and 265 cm height, one valves. In the lower part of the filters there was a water
with CRS and the other with NRS as filter bed media. At outlet pipe and the backwash water inlet. The water outlet
10 cm above the filter bottom, a perforated acrylic plate con- tap was placed above the filter layer to avoid negative pressure
taining 56 holes (4.76 mm) was placed as a false floor. Over inside the bed. The pilot filters received settled water from a
the plate, 30 cm of gravel (1.68–19 mm) was used as a sup- full-scale conventional WTP located next to the experimental
port layer. The filter bed of both materials was 45 cm setting (coagulation with alum, hydraulic flocculation, sedi-
height, with effective size (d10) of 0.45 mm and uniformity mentation, rapid sand filtration and chlorination). Figure 1
coefficient of 1.4. The filter internal diameter (190 mm) presents, schematically, the experimental setup.

Figure 1 | Schematic representation of the experimental setup.

Downloaded from http://iwaponline.com/ws/article-pdf/21/1/401/840813/ws021010401.pdf

by guest
 405 B. M. R. da Silva et al. | Comparison of crushed rock and natural river sand as filter media Water Supply | 21.1 | 2021

Filtration experiments three with 360 m3 m
2 d
1; (ii) two runs during the rainy
season (raw water turbidity in the range of 20–200 NTU;
Preliminary tests settled water turbidity ranging from 2.0 to 5.0 NTU), one
with a filtration rate of 270 m3 m
2 d
1 and the other with
Firstly, laboratory-scale experiments were carried out to 360 m3 m
2 d
1 (the short duration of the rainy season
verify whether the passage of treated water from the full- explains the limited number of trials; the choice for high
scale WTP through the CRS bed would negatively affect values of filtration rate was influenced by the promising
the filtered water quality, e.g. altering the water pH. Also, results obtained over the dry season). All water quality ana-
the raw CRS contained a high fraction of fine aggregates lyses were performed according to Standard Methods
(<0.075 mm) which might be leached and deteriorate the procedures (APHA et al. ).
filter water effluent. The preliminary tests were carried out
using a 2 L-glass measuring cylinder containing a water Filtration efficiency evaluation
outlet, a false floor, 10 cm of support layer and 20 cm of
CRS filter bed. Two 10 min-trials were undertaken, during NRS and CRS filters were compared regarding turbidity
which the influent and effluent waters were monitored removal, head loss increasing rate and run length, under
every minute for pH, conductivity and turbidity. the same operating conditions of filtration rates and influent
water quality.
Pilot tests

The two pilot filters were operated in parallel. Settled water RESULTS AND DISCUSSION
from the full-scale WTP was conducted to the pilot filters by
gravity and the flow rate was adjusted according to the Filter media characterization
desired filtration rates. Filters were operated with constant
flow rate and variable water level above the filter bed. Table 1 shows the chemical composition of the two filter
During the filter runs, head loss and the filtrate turbidity bed media. The NRS was composed of more than 90%
were monitored through the bed depth in each filter every silica, 5% aluminium oxides and smaller amounts of cal-
two hours. Turbidity was analysed using a HACH 2100 cium, potassium, iron, sulphur and selenium oxides.
AN turbidimeter. Filter runs ended either by limiting head Although in smaller amounts than in NRS, silica was also
(available head ≈ 180 cm) or breakthrough (filtrate turbidity the major compound of the CRS (62.5%). In turn, CRS
>0.50 NTU – Brazilian standard (Brasil )). Filters were
then backwashed with treated water for 10 minutes with a Table 1 | Chemical composition of the filter bed media

suitable flow rate in order to achieve 33% bed expansion. Fil-

Compound (%) CRS NRS
tration rates were selected taking into account the upper
threshold stated in the Brazilian standard for the design of SiO2 62.48 91.06
3
2
1 Al2O3 18.46 5.02
drinking water treatment plants - 180 m m d (ABNT
), half (90 m3 m
2 d
1), 1.5 times (270 m3 m
2 d
1) and CaO 6.97 0.10
3
2
1 K2O 4.96 1.32
twice (360 m m d ) that value. Originally, an equal
number of filter runs per filtration rate was intended to be car- Fe2O3 4.77 0.53

ried out over both the dry and rainy seasons, but field SO3 1.46 1.60

constraints ended dictating the experimental plan, with 14 TiO2 0.72 ND

runs in total, as follows: (i) 12 runs over the dry season Ca2O ND 0.211

(raw water turbidity <20 NTU; settled water turbidity <1.0 Sn2O3 ND 0.156

2
1 Others 0.18 0.02
NTU), one of them with a filtration rate of 90 m m 3
d ,
four with 180 m3 m
2 d
1, four with 270 m3 m
2 d
1, and ND: not detected.

Downloaded from http://iwaponline.com/ws/article-pdf/21/1/401/840813/ws021010401.pdf

by guest
 406 B. M. R. da Silva et al. | Comparison of crushed rock and natural river sand as filter media Water Supply | 21.1 | 2021

showed higher amounts of alkaline materials, as lime (7%) rate were selected to be highlighted here, along with an over-
and potassium oxides, as well as aluminium (18.5%) and view of the results as a whole: Run 8 (dry season) and Run
iron oxides (4.8%). The higher iron oxides content gives 13 (rainy season), both operated with 360 m3 m
2 d
1.
CRS a dark-grey colour.
Table 2 shows the physical properties of the filter bed Filter Run 8: influent water turbidity between 1.0 and 1.1
media. The CRS particles’ density was slightly higher than NTU
that of NRS. Thus, it was necessary to apply higher back-
washing flow rates to CRS in order to achieve the same The results of Run 8 are shown in Figures 2–4. In both NRS
33% bed expansion as in NRS. The CRS filter was back- and CRS filters, the filtration run ended due to limiting head,

1
washed with a water velocity of 106 cm min , as opposed after 13 hours (i.e. there was no breakthrough). At the first

1
to 94 cm min in NRS, resulting in 13% more backwash 5 cm of the filter bed, the filtered water turbidity already pre-
water consumption. sented values lower than 0.50 NTU for both filters
Both materials presented similar sphericity coefficients. (Figure 2). The remaining 40 cm reduced the filtered water
The CRS coefficient was slightly lower than that of NRS, turbidity from values slightly higher than 0.40 NTU to
hence CRS porosity was slightly higher than that of NRS. values between 0.24 and 0.30 NTU in the CRS filter, and
The CRS acid solubility was five times higher than that of between 0.25 and 0.31 NTU in the NRS filter. This trend
NRS, probably due to the higher amounts of alkaline com- was observed over the entire run, meaning that filtration
pounds (lime, potassium oxides) and iron oxides present in took place, essentially, at the surface, leaving the bulk of
CRS. However, both materials presented acid solubility the filter beds unused.
values lower than 5%, being then classified as suitable for fil- Usually, rapid-sand filters work with design filtration
tration (AWWA ). rates between 5 and 15 m h
1 (Crittenden et al. ). At
high filtration rates, solids tend to penetrate deeper into
the filter bed and the rate of head loss accumulation may
Filtration experiments be slower because of the more efficient use of the filter
bed depth. However, according to the same authors, filter
The influent water used in the preliminary experiments had
the following characteristics (mean values): pH ¼ 6.3;
turbidity ¼ 0.48 NTU; electrical conductivity (EC) ¼ 83 μS
cm
1. The filter effluent was monitored during 10 minutes
over two trials and no marked changes were noticed in
the water quality: pH ¼ 6.2–6.5; turbidity ¼ 0.4–0.5 NTU;
EC ¼ 80–87 μS cm
1. Subsequently, as described in the
Material and Methods section, 14 filter runs were evaluated,
covering different influent water quality (turbidity) and fil-
tration rates. Worst-case scenarios in terms of filtration

Table 2 | Physical properties of the filter bed media

Characteristic CRS NRS

3
Particles density (g cm ) 2.72 2.64
Bulk density (g cm
3) 1.28 1.33
Porosity 0.53 0.50
Sphericity coefficient 0.77 0.82
Acid solubility (%) 1.16 0.22
Figure 2 | Filtered water turbidity during filtration Run 8: (a) CRS filter (b) NRS filter.

Downloaded from http://iwaponline.com/ws/article-pdf/21/1/401/840813/ws021010401.pdf

by guest
 407 B. M. R. da Silva et al. | Comparison of crushed rock and natural river sand as filter media Water Supply | 21.1 | 2021

effluent quality tends to get worse at filtration rates above

12.5 m h
1 (300 m3 m
2 d
1), specially with weak chemical
flocs, such as alum floc without polymer. Filtration rates are
often subject to regulatory limits and the current Brazilian
standard for rapid down flow sand filtration is 180
m3 m
2 d
1 (or 7.5 m h
1). Despite the use of a filtration
rate as high as 360 m3 m
2 d
1 (or 15 m h
1) in Run 8, the
head loss increase was mainly due to particle accumulation
in the first 5 cm of the filter bed, as presented in Figure 3,
and no filtered water quality deterioration was observed.
Both filters presented a similar behaviour during the run, Figure 4 | Influent (settled) and effluent (filtered) water turbidity and total head loss
with slightly higher head loss values being noted in the during filtration Run8. Ts ¼ settled water turbidity; Tf ¼ filtered water turbidity
(CRS); Tf2 ¼ filtered water turbidity (NRS); hf ¼ total head loss (CRS); hf2 ¼ total
NRS filter. head loss (NRS).

Figure 4 summarizes the results of Run 8. Both settled

(1.0–1.1 NTU) and filtered water turbidity (0.2–0.3 NTU)
varied only narrowly through the filter run in both filters. turbidity was higher than 0.50 NTU at the beginning, but
Moreover, NRS and CRS filters presented a rather similar after two hours CRS and NRS filters already produced fil-
behaviour, both in terms of effluent water quality and head tered water with, respectively, 0.38 NTU and 0.49 NTU.
loss increase. As the filtered water samples were collected every two
hours, it was not possible to identify clearly the filters ripen-
Filter Run 13: influent water turbidity between 2.4 and 5.0 ing times, but looking at the filtered water turbidity data, it
NTU seems that the CRS filter had a shorter ripening period com-
pared to the NRS filter. Afterwards (approximately from six
The results of Run 13 are shown in Figures 5–7. Again, in hours of operation onwards) CRS and NRS filters steadily
both cases (NRS and CRS) the filtration run ended due to produced water with average turbidity values of 0.29 NTU
limiting head, after 20 hours. In spite of the higher influent and 0.32 NTU, respectively. In Figure 5, the gradual satur-
water turbidity, the run length of both NRS and CRS filters ation of the first layers of the filters over time is
was longer than those of the dry season trials with the same noticeable; i.e. the increasing clogging of the first layers

2
1
filtration rate of 360 m m 3
d . Figure 5 shows the filtered until both the influent and filtered water turbidity values
water turbidity along the filtration run for CRS and NRS fil- were virtually the same. After 10 hours, in both filters, fil-
ters. Differently from all the other runs, filtered water tered water turbidity at 5 cm deep was the same as that of

Figure 3 | Head loss increase along the filter bed depth and over time, filtration Run 8: CRS filter (left) and NRS filter (right).

Downloaded from http://iwaponline.com/ws/article-pdf/21/1/401/840813/ws021010401.pdf

by guest
 408 B. M. R. da Silva et al. | Comparison of crushed rock and natural river sand as filter media Water Supply | 21.1 | 2021

Figure 7 | Influent (settled) and effluent (filtered) water turbidity and total head loss
during filtration Run 13; Ts ¼ settled water turbidity; Tf ¼ filtered water tur-
bidity (CRS); Tf2 ¼ filtered water turbidity (NRS); hf ¼ total head loss (CRS);
hf2 ¼ total head loss (NRS).

Results of Run 13 are summarized in Figure 7. The influ-

ent water presented relatively high turbidity values and wide
variations during the filtration run. Even so, both filters
showed an increasing efficiency over the first six hours
(mainly in the first two hours of filter ripening), producing
Figure 5 | Filtered water turbidity during filtration Run 13: (a) CRS filter, (b) NRS filter.
filtered water with turbidity around 0.30 NTU from there
onwards. The NRS and CRS filters presented a similar be-
the influent water; the same happened in NRS filter at 10 cm haviour during Run 13 regarding filtered water turbidity
after 14 hours. Thus, depth filtration seems to have hap- values and head loss increasing rates. However, differently
pened more clearly during this filtration run. from Run 8, particles removal took place at deeper layers
Figure 6 illustrates the head loss development along resulting in a more efficient use of the filter bed and, conse-
filter Run 13. Differently from the dry season trials, particles quently, in longer filter runs.
removed at deeper layers seem to have contributed to the
head loss increasing along the filter bed depth: the head Overview of the filtration runs results
loss curves lose their parallelism in relation to the initial
head loss curve (t ¼ 0) at gradually deeper layers of the Table 3 summarizes the main results of all the 14 filtration
filter along the filtration run. runs. Run 5 was conducted with 90 m3 m
2 d
1, but due to

Figure 6 | Head loss increase through the filter bed depth and over time, filtration Run 13: CRS filter (left), NRS filter (right).

Downloaded from http://iwaponline.com/ws/article-pdf/21/1/401/840813/ws021010401.pdf

by guest
 409 B. M. R. da Silva et al. | Comparison of crushed rock and natural river sand as filter media Water Supply | 21.1 | 2021

operational problems, this filtration rate was not tested any- Table 3 shows that increasing the filtration rate from 180
more. Only three runs ended due to breakthrough (filtered to 360 m3 m
2 d
1 did not affect the filtered water quality
water turbidity >0.50 NTU), all of them during the dry but reduced the average run length from 33.5 h to 13.7 h.
season trials. During Run 6 (360 m3 m
2 d
1), breakthrough However, during the rainy season the run length was
(filtered water turbidity ¼ 0.51 NTU) occurred in both filters, around 20 h for both the 270 and 360 m3 m
2d
1 filtration
at different moments of the run, but with the same head loss rates; that is, six hours longer than the dry season runs, in

2
1
values. At a lower filtration rate (180 m m 3
d ) (Run 9), spite of the higher influent water turbidity during the rainy
breakthrough (filtered water turbidity ¼ 0.56 NTU) occurred season. This behaviour was observed for both filter media
only in the NRS filter. (CRS and NRS).
The current Brazilian standard for filtered water turbid- Over the rainy season (alum dose ¼ 16–18 mg/L; pH ¼
ity is 0.50 NTU (for rapid filters) in 95% of collected samples 6.7–7.0), sweep coagulation mechanisms may have taken
(Brasil ). In addition, the filtered water turbidity must not place during the water pre-treatment in the full-scale WTP,
exceed 1.0 NTU in any sample. Filtered water turbidity combined with or even prevailing over charge neutralization
remained below 0.30 NTU (a rather strict standard, like (Ghernaout & Ghernaout ). Hence, coagulation may
those of the USA (USEPA ) and Canada (Health have produced more fragile flocs during the rainy season,
Canada ) drinking-water regulations)) in eight out of when raw water turbidity was higher than in the dry
the 13 CRS filtration runs which ended due to limiting season. In effect, in all the dry season trials (alum dose ¼
head loss, and below 0.50 NTU (the Brazilian standard) in 8–12 mg/L; pH ¼ 6.3–6.9), regardless of the filtration rate,
the other five runs. Also, it is worth noticing that in the filtration seems to have occurred mainly at the filter bed
only CRS filtration run that ended because of breakthrough upper layers, probably as a result of higher floc strength.
(Run 6), the Brazilian standard of 0.50 NTU was only Conversely, during the rainy season depth filtration seems
slightly exceeded (0.51 NTU). to have more clearly taken place. Thus, due to a deeper

Table 3 | Main results of the filtration runs

Filtered water turbidity (NTU)

Run length Total head loss Influent water
(hours) (cm) turbidity (NTU) CRS NRS

2
1
Run Filtration rate (m3 m d ) CRS NRS CRS NRS Min. Max. Min. Max Min. Max

1 270 16 16 155.2 157.0 0.60 1.20 0.16 0.30 0.19 0.25

2 270 12 12 153.4 155.6 0.59 0.76 0.17 0.25 0.18 0.25
3 270 15 15 173.9 174.0 0.51 0.74 0.12 0.16 0.13 0.17
4 270 16.5 16.7 173.8 174.0 0.57 0.88 0.14 0.17 0.14 0.17
5 90 86 86 156.9 156.7 0.74 1.00 0.14 0.20 0.14 0.20
6 360 13 12 149.3 150.8 1.10 1.30 0.29 0.51 0.28 0.51
7 360 14.3 14.3 172.7 174.7 1.00 1.50 0.31 0.39 0.31 0.42
8 360 13 13 171.9 174.6 0.96 1.10 0.24 0.29 0.25 0.30
9 180 35 22 174.1 122.2 0.78 1.10 0.25 0.42 0.29 0.56
10 180 31 31 173.8 174.4 0.80 1.10 0.26 0.38 0.27 0.43
11 180 30 31 173.3 174.5 0.62 1.10 0.17 0.28 0.21 0.31
12 180 38 36.5 174.1 174.5 0.51 0.88 0.13 0.27 0.14 0.27
13 360 20 20 172.9 174.5 2.40 5.00 0.25 0.38 0.29 0.49
14 270 20.5 20.5 176.0 173.5 2.10 4.40 0.27 0.48 0.30 0.48

Downloaded from http://iwaponline.com/ws/article-pdf/21/1/401/840813/ws021010401.pdf

by guest
 410 B. M. R. da Silva et al. | Comparison of crushed rock and natural river sand as filter media Water Supply | 21.1 | 2021

penetration of the particles into the filter bed, the head loss • Filtration run length, turbidity removal and head loss
accumulation rate was slower during the rainy season; con- accumulation were similar in both the CRS and the natu-
sequently, the respective filter runs were longer, revealing a ral sand river (NRS) filters.
more efficient use of the filter bed during filtration runs with • The CRS filter was able to steadily produce filtered water
higher influent water turbidity. turbidity lower than 0.50 NTU, and often below 0.30
As previously discussed, the main factors that affect NTU, even with high filtration rates and relatively high
rapid filtration performance are the influent water quality, influent water turbidity.
filtration rate and filter bed characteristics. In this study,
In conclusion, under the same operating conditions (simi-
the filters were fed with the same influent water and the
lar filter bed media characteristics, the same filtration
same filtration rate; moreover, some bed characteristics
rates and the same influent water quality), CRS filter effi-
(e.g. size distribution and bed length) were practically iden-
ciency (run length, turbidity removal and head loss
tical. Thus, despite other differences between the filter bed
accumulation) was shown to be at least as good as that of
materials (chemical composition, sphericity coefficients
the NRS filter.
and porosity), the NRS and CRS filters presented similar fil-
tration performances in terms of run length, filtered water
turbidity and head loss increasing rate. Hence, the same fil-
ACKNOWLEDGEMENTS
tration mechanisms supposedly took place in both filter
media, mainly surface capture, known as the principal
This research did not receive any specific grant from funding
mechanism of in-depth filtration (Tobiason et al. ).
agencies in the public, commercial, or not-for-profit sectors.
Given that the grain shape of the filter media also affects
We thank ‘Petra Agregados’ mining company for providing
particle capture and storage, and that angular particles are
the raw crushed rock sand. We also thank the Brazilian
preferable to rounded particles, the crushed stone surface
agency CAPES for providing an MSc scholarship.
roughness may have favoured absorption characteristics,
even better than the more rounded natural sand (Rao ;
Tobiason et al. ). Summing up, given the suggestions
presented herein of high treatment efficiency, the use of DATA AVAILABILITY STATEMENT
CRS as filter bed for rapid filtration in drinking water treat-
ment should be encouraged. Nevertheless, further studies All relevant data are included in the paper or its Supplemen-
should be undertaken in order to evaluate the filtration effi- tary Information.
ciency based on other water quality indicators (e.g. colour
and surrogates for pathogens such as bacteriophages and
aerobic spore-forming bacteria), as well as with varied REFERENCES
filter bed characteristics (e.g. depth and size distribution),
operating conditions and filtration techniques, such as ABNT. Associação Brasileira de Normas Técnicas (Brazilian
Association of Technical Norms)  NBR 12216. Projeto de
direct filtration, slow sand filtration or upflow filtration.
Estação de Tratamento de água Para Abastecimento Publico.
Rio de Janeiro, p. 18.
American Water Works Association – AWWA  AWWA B100-
96: Standard for Filtering Material. AWWA, Denver, p. 36.
APHA, AWWA & WEF – American Public Health Association,
CONCLUSIONS American Water Works Association & Water Environment
Federation  Standard Methods for the Examination of
• Crushed rock sand (CRS) did not present harmful com- Water and Wastewater, 21st edn. American Public Health
Association, Washington.
pounds and the CRS filter bed did not negatively affect
Betancourt, W. Q. & Rose, J. B.  Drinking water treatment
the quality of the filtered water (turbidity, pH and electri- processes for removal of Cryptosporidium and Giardia.
cal conductivity). Veterinary Parasitology 126 (1), 219–234.

Downloaded from http://iwaponline.com/ws/article-pdf/21/1/401/840813/ws021010401.pdf

by guest
 411 B. M. R. da Silva et al. | Comparison of crushed rock and natural river sand as filter media Water Supply | 21.1 | 2021

Brasil. Ministério da Saúde  (Ministry of Health, Regulation n Ministry of Health  Drinking Water Standards for New
2914, Guidelines for drinking water quality) Portaria n 2914, Zealand. 2005 (Revised 2008). Ministry of Health,
de 12 de Dezembro de 2011. Dispõe sobre os procedimentos Wellington, New Zealand.
de controle e de vigilância da qualidade da água para Mohammad, A. W., Teow, Y. H., Ang, W. L., Chung, Y. T., Oatley-
consumo humano e seu padrão de potabilidade. Diário Radcliffe, D. L. & Hilal, N.  Nanofiltration membranes
Oficial da União, Brasília, n 239, pp. 39–46. review: recent advances and future prospects. Desalination
Bustamante, H. A., Shanker, S. R., Pashley, R. M. & Karaman, 356, 226–254.
M. E.  Interaction between Cryptosporidium oocysts and Mundra, S., Sindhi, P. R., Chandwani, V., Nagar, R. & Agrawal, V.
water treatment coagulants. Water Research 35 (13),  Crushed rock sand – an economical and ecological
3179–3189. alternative to natural sand to optimize concrete mix.
Cescon, A., Jiang, J. Q., Haffey, M., Moore, G. & Callaghan, K. Perspectives in Science 8, 345–347.
 Assessment of recycled glass and expanded clay in a Rao, D. R. J.  Evolving high rate filter and use of crushed stone
dual media configuration for drinking water treatment. as filter media. Journal of Institution of Engineers 61, 92–96.
Separation Science and Technology 51 (14), 2455–2464. Rutledge, S. O. & Gagnon, G. A.  Comparing crushed
Crittenden, J. C., Trussell, R. R., Hand, D. W., Howe, K. J. & recycled glass to silica sand for dual media filtration. Journal
Tchobanoglous, G.  MWH’s Water Treatment: Principles of Environmental. Engineering and Science 1, 349–358.
and Design, 3rd edn. John Wiley & Sons, Inc., Hoboken, Santo, E. & Sánchez, L.  GIS applied to determine
New Jersey, p. 1920. environmental impact indicators made by sand mining in a
Emelko, M. B., Huck, P. M. & Coffey, B. M.  A review of floodplain in southeastern Brazil. Environmental Geology
cryptosporidium removal by granular media filtration. 41 (6), 628–637.
Journal of American Water Works Association 97 (12), Schäfer, A. I., Hughes, G. & Richards, B. S.  Renewable
101–115. energy powered membrane technology: a leapfrog approach
Farizoglu, B., Nuhoglu, A., Yildiz, E. & Keskinler, B.  to rural water treatment in developing countries? Renewable
The performance of pumice as a filter bed material under and Sustainable Energy Reviews 40, 542–556.
rapid filtration conditions. Filtration & Separation 40 (3), Soyer, E. & Akgiray, O.  A new simple equation for the
41–47. prediction of filter expansion during backwashing. Journal of
Ghernaout, D. & Ghernaout, B.  Sweep flocculation as a Water Supply: Research and Technology-Aqua 58 (5),
second form of charge neutralisation – a review. Desalination 336–345.
and Water Treatment 44 (1–3), 15–28. Soyer, E., Akgiray, Ö., Eldem, N. Ö. & Saatçı, A. M.  Crushed
Headd, B. & Bradford, S. A.  Use of aerobic spores as a recycled glass as a filter medium and comparison with silica
surrogate for Cryptosporidium oocysts in drinking water sand. Clean–Soil, Air, Water 38 (10), 927–935.
supplies. Water Research 90 (1), 185–202. Suthaker, S., Smith, D. W. & Stanley, S. J.  Filtrate turbidity
Health Canada  Guidelines for Canadian Drinking Water improvement of existing mono-media filters based on a pilot
Quality – Summary Table. Water and Air Quality Bureau, plant study. Canadian Journal of Civil Engineering 22 (6),
Healthy Environments and Consumer Safety Branch, Health 1051–1062.
Canada, Ottawa, Ontario. Tchobanoglous, G., Burton, F. L., Stensel, H. D. & Metcalf &
Hendricks, D. W.  Water Treatment Unit Processes: Physical Eddy  Wastewater Engineering: Treatment and Reuse.
and Chemical. CRC Press, Boca Raton, FL, p. 1266. McGraw-Hill, Boston, MA.
Hofmann, M., Büchi, A., Hoppe, A., Hornung, J., Karfunkel, J. & Tobiason, J. E., Cleasby, J. L., Logsdon, G. S. & O’Melia, C. R. 
Pagung, R.  Resources for a growing city – sand Granular Media Filtration, Chapt. 10. In: Water Quality &
extraction north of Belo Horizonte (Brazil). Neues Jahrbuch Treatment (Edzwald, J. K., ed.), 6th edn. AWWA &
Geologie Paläontologie Abhandlungen 253 (1), 61–78. McGrawHill, Denver, CO.
Kawamura, S.  Integrated Design and Operation of Uluatam, S. S.  Assessing perlite as a sand substitute in
Water Treatment Facilities, 2nd edn. John Wiley & Sons, filtration. Journal of American Water Works Association
New York, NY. 83 (6), 70–71.
Manasseh, J.  Use of crushed granite fine as replacement to Uluatam, S. S.  Laboratory evaluation of perlite bed filtration
river sand in concrete production. Leonardo Electronic in water treatment. International Journal of Environmental
Journal of Practices and Technologies 17, 85–96. Studies 42 (1), 1–10.
Manguriu, G. N., Karugu, C. K., Oyawa, W. O., Abuodha, S. O. & USEPA – United States Environmental Protection Agency 
Mulu, P. U.  Partial replacement of natural river sand National Primary Drinking Water. Regulations: Long Term 2
with crushed rock sand in concrete production. Global Enhanced Surface Water Treatment Rule. Final Rule. Federal
Engineers & Technologists Review 3 (4), 7–14. Register – Part II – 40CFR, Parts 9, 141 and 142.

First received 10 April 2020; accepted in revised form 29 October 2020. Available online 11 November 2020

Downloaded from http://iwaponline.com/ws/article-pdf/21/1/401/840813/ws021010401.pdf

by guest