2779 © IWA Publishing 2012 Water Science & Technology | 66.

12 | 2012

Pilot plant study of alternative filter media for rapid

gravity filtration
P. D. Davies and A. D. Wheatley

ABSTRACT

Sand has been the main filter media used in rapid gravity filtration since its introduction. P. D. Davies (corresponding author)
A. D. Wheatley
The dominance of sand has been due to its low cost and availability. Extensive experience School of Civil and Building Engineering,
Loughborough University,
has led to sand filters with a dependable and predictable performance. Sand remains the Leicestershire,
LE11 3TU
preferred filter medium but usually with a larger sized anthracite capping to reduce the onset E-mail: P.D.Davies@lboro.ac.uk
of head loss. Other approved filter media are now commercially available and this paper
compares sand with recycled glass, Filtralite® and slate at pilot scale. The results have
reaffirmed the basic importance of particle size on head loss and turbidity performance rather
than surface activity or specific surface area. The results did suggest, however, that particle shape
and packing exerted a stronger influence on performance than previously acknowledged. These
could be used to improve the design and the contribution to sustainability made by rapid gravity
filters.
Key words | deep bed filtration, Filtralite®, rapid gravity filtration, recycled glass, slate

INTRODUCTION

Within a water treatment works the highest operating costs
are typically chemicals and power required mainly for
pumping (WATER UK ). Improvements in rapid gravity
filtration (RGF) could provide savings through a reduction
in frequency or duration of backwashing. Better filter per-
formance could also reduce chemical usage by being able
to guarantee removal of a greater proportion of solids for
the same run-time.
Research into alternative filter media has been limited
with the emphasis on turbidity or solids removal. Materials
tested included Filtralite® by Mitrouli et al. (, )
and Saltnes et al. (), pumice (Morgeli & Ives ; Fari-
zoglu et al. ; Ghebremichael ) and recycled glass
(Evans et al. ; Rutledge & Gagnon ). The fundamen-
tal effects of particle size on filtration performance and
headloss are well known but not often accounted for in the
reported research (Rutledge & Gagnon ), while particle
shape is often overlooked entirely. Ives () has discussed
particle shape and its influence on turbidity removal and
highlighted the problem caused by reduced permeability of
the filter bed caused by low sphericity media (<0.6).
Filter media were chosen for the study based on BS
standard tests and initial laboratory-scale experiments
(Ives ; Stevenson ). The laboratory experiments
used a wider range of materials selected based on fundamen-
tal characteristics which were predicted to contribute to
performance. The sustainable supply of any new materials
was also important and although pumice has been tested
before because of its high surface area and light weight,
supply is limited to a few areas of the world – Italy and
Turkey are the largest suppliers (Farizoglu et al. ). The
pumice results also showed poor durability. Other materials
tested were types of slag and limestone with surface polarity,
but the slags leached alkalinity and the limestone failed the
acid test.
Final choices were Filtralite, which has a large surface
area that manufacturers claim improves filter performance,
and glass, which has a very smooth surface texture, shown
by scanning electron microscopy and as a recycled material
has better sustainability than sand which requires quarrying
(Davies & Wheatley ). Our studies report on the effect of
a particle shape on turbidity removal and head loss perform-
ance using the standard sand and the approved alternative
Filtralite, recycled glass, which is seeking approval, and also
a non-approved material – slate – as an extreme plate-shaped
example.

doi: 10.2166/wst.2012.517
 2780 P. D. Davies & A. D. Wheatley | Alternative filter media: pilot study
Filtralite and glass are marketed as replacement media
in either mono- or dual-media filter designs. The results of
the first series of experiments reported in this paper are
for the mono-media; a second series of tests are to be
reported in a second paper with dual media using the rec-
ommended two sizes of Filtralite, two sizes of glass,
anthracite and sand and mono sand. The work formed
part of a thesis and more results are available to compare
with the previous laboratory work (Davies & Wheatley ).
MATERIALS AND METHODS
Site description
Treatment processes
The pilot plant was located at a water treatment works with
a maximum throughput of 30,000 m3/day, although during
the time of testing reported the throughput was around
16,000 m3/day (January 2012) due to seasonal reduced
demand. The works operate a treatment process that has
been developed over a number of years in sympathy
with improved standards. Raw water is drawn from three
sources – boreholes, reservoir and river water stored in a
bankside balancing reservoir; these are blended together
prior to treatment at the entry to works.
The raw water is coagulated with ferric sulphate at a
concentration adjusted to match the incoming raw water
turbidity; it then enters a first stage ozone contact tank
before four parallel hopper bottomed clarifiers. The clarified
water is then second stage ozone contacted before entering
five rapid gravity filters. Four of the filters are original and
have a surface area of 24 m2 and the fifth newest filter has
a surface area of 30.7 m2. The filters were designed for use
with an activated carbon filter media but this has been chan-
ged to a dual-media configuration of anthracite/sand to a
total depth of around 0.9 m. During the pilot study the filters
were being operated at a typical runtime of 48 h before
backwashing.
Table 1 | Results of specific surface area (SSA) and particle size analysis for the media
SSA (m2/g) Sphericity
Sand 0.2261 0.89
Glass 0.0558 0.73
Filtralite 1.4685 0.78
Slate 0.5964 0.49
Water Science & Technology | 66.12 | 2012
Clarified water quality
During the study clarified turbidity measured by the works’
online monitoring was in the range 0.15–0.19 NTU. The
pilot plant feed water was also manually grab sampled and
the results were in the same range (0.15–0.20 NTU). There
were, however, two spikes in turbidity which were thought
to be due to accumulated flocculated solids observed in
the pilot plant inlet pipework being dislodged by changes
in flow rate through the filter columns.
Filter media characterisation
The basic physical characteristics of the media that influence
performance have previously been reported (Davies &
Wheatley ). The details of these key media properties
are given in Table 1. The earlier laboratory trial demon-
strated that sphericity was an important performance
criterion. It was shown that the more angular the media
the better the turbidity removal and headloss performance.
Ideally all media would be identically sized but this was
not possible due to manufacturer constraints. The smallest
available Filtralite is twice the size of other media used
but the increased surface area is claimed to offset this. Filtra-
lite is also the most widely used alternative to sand and the
Utility wished to gain some operating data prior to a
planned refurbishment. Slate is also larger in one dimension
but porosity is also affected by the uniformity coefficient and
packing. There is the likelihood that smaller voids will
become filled and rapidly blocked. The combination of
low uniformity coefficient and lower sphericity of the glass
compared with sand leads to a greater bed porosity than
for the sand.
Pilot plant
The four pilot plant filter columns were 150 mm internal
diameter clear Perspex tube and 1.65 m in height. The
filter media depth was 600 mm to be similar to the works’
filters and allow a comparison with the full-scale results.
Porosity d10 (mm) d60 (mm) Uniformity coefficient (U)
0.362 0.59 0.75 1.27
0.370 0.76 0.92 1.21
0.480 0.77 1.40 1.82
0.434 0.58 1.00 1.72
 2781 P. D. Davies & A. D. Wheatley | Alternative filter media: pilot study
Each column was supplied by peristaltic pumps, which were
chosen to limit damage to any incoming fragile flocs, which
were observed in the feed water.
Gravity feed would have been preferable but this was
not practical if a constant flow rate was to be guaranteed
irrespective of the flow rate into the main works or if the
pilot plant was to be transportable. The turbidimeters also
required a de-bubbler (located at the same height as the
inlet at the top of the filter) which reduced the available
head to the filters. About 2–3 m of head is required to
match the maximum reached in the full-scale filters. Pump-
ing was also necessary to pressurise the filters to achieve this
head loss. A pressure relief valve was installed at the top of
each filter column to release water when the pressure
exceeded 0.3 bar or 3 m head of water. The filters were
thus operating at a constant flow (controlled by the peristal-
tic pumps) and variable head configuration.
Data collection and analysis
Turbidity and headloss were used to assess performance in
both the laboratory and the pilot plant; these are the stan-
dard methods for operating the full-scale filters. Turbidity
was monitored continuously by an Aquascat P (Sigrist,
Ely), operating at a range of 0–3 NTU, the range expected
during operation and ripening. Headloss was monitored by
a SITRANS P DS III (Siemens, Manchester) as total
across the whole filter bed in mmH2O. Flowrate was moni-
tored by a SITRANS F M MAGFLO MAG6000 (Siemens,
Manchester) which was also used to ensure there was no
Figure 1 | Average filtered water turbidity according to flow rate.
Water Science & Technology | 66.12 | 2012
significant change in the flow rate during the operation of
the filters; the variable peristaltic pumps were used to set
the initial flow rate to the required value.
A laptop installed with SquirrelView (Grant Instru-
ments, Cambridge) was used to collect stored results from
a Squirrel 2040 data logger (Grant Instruments, Cambridge).
Readings for each of the monitored sensors were taken
every 10 seconds by the data logger, but the data collected
were converted to an averaged reading for a 1 min (turbid-
ity/flowrate) or 2 min for head loss, which reduced the
noise in the raw results and provided a more manageable
volume of data.
RESULTS AND DISCUSSION
Turbidity
The average turbidity results, after ripening, for the test runs
for the different hydraulic loading rates are shown in
Figure 1; each run was carried out in at least triplicate and
stopped when the head loss reached 3 m or the normal
back wash point. The feed or clarified turbidity was taken
from the works’ online monitoring as an average for the
period but manual grab samples were taken periodically to
confirm the turbidity values and also the flow rates through
the pilot filters.
The tests were carried out over a period of 6 weeks and
feed turbidity results were in the range 0.18–0.28 NTU.
Observations suggested that the majority of the mass of
 2782 P. D. Davies & A. D. Wheatley | Alternative filter media: pilot study
solids was floc carry over, not turbidity. All the filter media
therefore achieved below the normal limit of <0.1 NTU and
the standard deviation in NTU was <0.01 NTU. The turbid-
ity results follow a similar pattern irrespective of the media.
The consistent difference of Filtralite compared with
other media is due to its larger particle size giving the
filter bed a more open pore structure and therefore increas-
ing the distances that suspended particles must be
transported to the surface for attachment to occur. The
other media begin to show an increase in turbidity at the
highest flows possibly as a result of pore occlusion.
Headloss
Figure 2 shows representative filter run times and influence
on turbidity at a typical UK flowrate terminated by a 2 m
head loss (the standard deviation in NTU across the dupli-
cates was <0.01 and omitted). The ripening times between
the media is attributed to the initial beneficial effects of par-
ticle accumulation reducing porosity. The smaller the media
the faster the ripening except for the slate which could be
due to the particle shape. The highly angular particle
shape, in the case of slate, leading to a more tortuous path
through the media, was suggested by previous laboratory
results and predicted by Ives ().
Figure 3 corroborates this idea showing that initial head
loss is linked to porosity as well as flow rate. The slate has a
lower head loss compared with the similar porosity Filtralite
which is also attributed to the contribution of particle shape
Figure 2 | Example of turbidity results from a test run at a flowrate of 7.4 m/h.
Water Science & Technology | 66.12 | 2012
(Table 1) that helps to reduce the energy loss as the flow
moves through the heterogeneous filter bed.
Filtralite, although similar to sand in sphericity, has
increased particle size which enlarges the pore size and
leads to lower initial and accumulated headloss (Figure 4).
Similarity between the glass and the slate is attributed to
the similarity of d60 and larger pore size.
With a headloss limit set at 2 m, the results indicate
increases of between two and three times the run times of
the sand with a 25–30% increase in size and reduction in
sphericity. This would lead to operational savings if it can
be maintained over the lifetime of the filter due to the
reduced washing frequency. The increased runtime, how-
ever, will lead to a greater mass of solids to be removed in
the backwash and further work would be needed to deter-
mine if this could be achieved.
Discussion
Traditionally it was believed that a rounded sand filter
media was best for use in rapid gravity filters (Ives ).
Although the reasons for this were not explicitly stated, it
is likely this was intended to avoid the formation of preferen-
tial flow paths. Ives () states that an angular filter media
would achieve better removal due to the more tortuous flow
path improving some of the transportation mechanisms
which move suspended particles to the media grain surface
for attachment. Ives () does not analyse the transpor-
tation mechanisms but it is believed an increase in inertia
 2783 P. D. Davies & A. D. Wheatley | Alternative filter media: pilot study
Figure 3 | Summary of initial headloss results against flow rate.
Figure 4 | Example headloss curves for flow rate of 7.4 m/h.
(Ives , ) would be the primary effect increasing the
forces available for the suspended particles to break free
of flow-streams. Ives also suggested a limit of 0.6 sphericity
below which the permeability of the filter bed would be
too low and the initial head loss too high.
Slate used here has a sphericity of 0.49, which falls
below this limit set by Ives () and should be poorly per-
forming with a high initial head loss and resistance to flow.
This is not shown by either our pilot plant or laboratory
results. In fact as the media becomes more angular the initial
Water Science & Technology | 66.12 | 2012
headloss is reduced, caused by the variable packing of the
angular media creating areas of both low and high per-
meability. Variable permeability creates preferential flow
paths through the bed carrying solids to high porosity
zones where their attachment has less effect on headloss
accumulation. The glass results are a closer match to the
slate more than the sphericity alone would suggest, probably
due to the low surface area and surface roughness.
Filtralite does not have an angular shape and its head-
loss performance can be attributed to the greater particle
 2784 P. D. Davies & A. D. Wheatley | Alternative filter media: pilot study
size leading to large open pores and a high permeability
through the bed. This is corroborated by the turbidity
removal of the Filtralite being consistently poorer than the
other three media. This allows the majority of suspended
particles to penetrate deeper into the filter bed before they
are captured and for a greater proportion to escape. The
media still performs to achieve the <0.1 NTU limit and is
a suitable filter media for low turbidity clarified water such
as this example water works where Filtralite could increase
filter run times by up to 3.5 times that of sand.
Packing variations between pilot and full scale are a
possibility that will alter the results and which needs further
work and highlights the need for better filter optimisation
from plant to plant. Research on multi-media configurations
to further the understanding of filtration is also suggested.
CONCLUSIONS
The results presented are in accordance with the accepted
theory that smaller particle sizes and lower sphericity filter
media improve turbidity removal. It was suggested by
improving the inertia transportation of more, smaller sus-
pended particles to the media surface. If there is good
quality clarified water then turbidity improvements are not
significant compared with the angularity of normal sand.
Improvements in head loss performance from larger less
spherical media were also shown by the results, caused by
the more varied bed packing of angular filter media and
heterogeneous flow paths through the bed. This could
reduce the frequency of backwashing.
The work suggests improvements to design procedures
to match filter media to the anticipated raw water that are
specific to a works. Filtralite, which has approval for use
and is commercially guaranteed, could improve filter run
times without affecting water quality. Multimedia filters or
increasing the particle size of other media which could be
separated after backwashing should also give the same
benefits.
ACKNOWLEDGEMENTS
The authors gratefully acknowledge the financial and techni-
cal support of Severn Trent Water Ltd for providing the pilot
First received 1 May 2012; accepted in revised form 20 July 2012
Water Science & Technology | 66.12 | 2012
plant and allowing this work to be conducted on one of their
operating treatment works.
REFERENCES
Davies, P. & Wheatley, A.  Alternative filter media in rapid
gravity filtration of potable water. Proceedings of IWA Young
Water Professionals UK Conference 2011, Edinburgh, UK.
Evans, G., Dennis, P., Cousins, M. & Campbell, R.  Use of
recycled crushed glass as a filtration medium in municipal
potable water treatment plants. Water Science and
Technology: Water Supply 2, 9–16.
Farizoglu, B., Nuhoglu, A., Yildiz, E. & Keskinler, B.  The
performance of pumice as a filter bed material under rapid
filtration conditions. Filtration þ Separation 40, 41–47.
Ghebremichael, K. A.  Moringa Seed and Pumice as
Alternative Natural Materials for Drinking Water Treatment.
Ph.D. KA Royal Institute of Technology.
Ives, K. J.  Review paper rapid filtration. Water Research 4,
201–223.
Ives, K. J.  Specifications for granular filter media. Effluent and
Water Treatment Journal 15, 296–305.
Ives, K. J.  Filtration of clay suspensions through [quartz] sand.
Clay Minerals 22, 49–61.
Ives, K. J.  Testing of filter media. Aqua 39, 144–151.
Mitrouli, S. T., Karabelas, A. J., Yiantsios, S. G. & Kjolseth, P. A.
 New granular materials for dual-media filtration of
seawater: Pilot testing. Separation and Purification
Technology 65, 147–155.
Mitrouli, S. T., Yiantsios, S. G., Karabelas, A. J., Mitrakasa, M.,
Follesdal, M. & Kjolseth, P. A.  Pretreatment or
desalination of seawater from an open intake by dual-media
filtration: Pilot testing and comparison of two different
media. Desalination 222, 24–37.
Morgeli, B. & Ives, K. J.  New media or effluent filtration.
Water Research 13, 1001–1007.
Rutledge, S. O. & Gagnon, G. A.  Comparing crushed
recycled glass to silica sand for dual media filtration. Journal
of Environmental Engineering 1, 349–358.
Saltnes, T., Eikebrokk, B. & Odegaard, H.  Contact filtration
of humic waters: performance of an expanded clay aggregate
filter (Filtralite) compared to a dual anthracite/sand filter.
Water Science and Technology: Water Supply 2, 17–23.
Stevenson, D. G.  The specification of filtering materials for
rapid-gravity filtration. Water and Environmental
Management 8, 527–533.
WATER UK  Sustainability Indicators 2009/10. Water UK,
London. Available from: http://www.water.org.uk/home/
news/press-releases/sustainability-indicators-09-10/
sustainability-2010-final.pdf