Developments of High Rate Dissolved Air Flotation for Drinking Water

Treatment

James K. Edzwald
Department of Civil and Environmental Engineering
University of Massachusetts
Amherst, MA 01003 USA

ABSTRACT

In recent years there have been large increases in the hydraulic loading rates used to design
dissolved air flotation (DAF) facilities for drinking water applications. High rate DAF processes are
now available at loading rates of 20 to 40 m h-1. These high rate systems have a smaller plant
footprint compared to conventional systems. The paper examines bubble and floc-bubble rise
rates, and how the simple theory relating these rise rates to the hydraulic loading of the DAF
separation zone is inadequate. It is shown that in some high rate systems, the flow through the
separation zone is stratified and must be accounted for in relating separation zone efficiency to
the hydraulic loading and bubble or floc-bubble rise rates. An idealized and simple model is used
to explain why DAF tanks can be designed with hydraulic loadings > 20 m h-1.

KEYWORDS: bubble rise rates, drinking water, dissolved air flotation, floc-bubble rise rates,
high rate flotation, separation zone

INTRODUCTION

Dissolved air flotation (DAF) has been used to treat drinking water supplies for over 45 years. The
process was first used in Scandinavia and South Africa, and over the last 25 years has seen
increasing and widespread use around the world. It is a better clarification process than
sedimentation for treating reservoir supplies and supplies containing natural color. In short, it is
more efficient than settling for removing low density particles either initially in the raw water or
produced through precipitation from metal coagulants.
In the early years, the design and operation of dissolved air flotation plants were based
largely on experience and pilot plant data. Our understanding of DAF has been greatly improved
through development of fundamental principles and models. Some selected contributions are
those of Edzwald and co-workers (1990, 1995), the Tambo group including Fukushi and Matsui
(1986, 1995, 1998a, 1998b), and Han (2002). Haarhoff and Edzwald (2004) have summarized
the state of knowledge regarding modelling the contact and separation zones of the DAF process
and have identified areas that need additional research. The fundamental research conducted
over the last 25 years serves as a cornerstone for technological developments in DAF. DAF
plants until about 1990 were designed with long pretreatment flocculation times and the DAF tank
was sized with low hydraulic loading rates of about 5 to 10 m h-1. In the last 15 years, there has
been a trend of reducing the flocculation detention time and increasing the hydraulic loading rate.
The purpose of this paper is to describe the developments of high rate DAF processes.
The scope of the paper follows. First, a brief description of a DAF water plant is presented. Next,
the properties and concentrations of particles, floc particles or simply “flocs”, and bubbles are
summarized. Third, the theoretical rise velocities of bubbles and floc-bubble aggregates are
presented. Fourth, some pilot data are summarized showing that DAF processes are feasible with
hydraulic loadings greater than the theoretical rise velocities. Some high rate DAF processes now
available are briefly described. Fifth, a modification of the theory is made that relates rise
velocities to the DAF tank hydraulic loading for the removals of bubbles and aggregates.
DESCRIPTION OF DISSOLVED AIR FLOTATION

Figure 1 is a process schematic diagram for a DAF drinking water plant. Coagulation and
flocculation are required pretreatment processes. Coagulation is a chemical addition step. The
coagulants used most often in water treatment are aluminum and ferric salts. Coagulation has
two purposes: to destabilize particles present in the raw water and to convert dissolved natural
organic matter into particles. The latter purpose, over the last 25 years, has become an essential
part of water treatment used to remove as much dissolved organic carbon as feasible through
coagulation. The chemistry of coagulation involves selecting proper dosing and pH conditions to
produce particles – originally in the raw water and new particles produced through metal
coagulant precipitation – that can forms flocs in the flocculation tank, and ultimately to attach
bubbles to flocs in the DAF tank. The goal of coagulation chemistry then, is to produce particles
with little or no net electrical surface charge and with a relatively hydrophobic character. The
chemistry of coagulation for sedimentation and DAF plants is similar, so that coagulant dosages
and pH conditions are identical to forming flocs. An exception is that for some water types,
especially those of low to moderate turbidities, sedimentation plants may require higher coagulant
dosages so as to increase flocculation kinetics and produce flocs large enough to settle rapidly.

Saturator Recycle After

Coagulant + DAF or Filters
pH Control
Qr Qr

Q Rapid Filtration Q
Mixing Floc Tank DAF Tank

Pretreatment Clarification Filtration

Recycle
(Qr)

Q
Q

Separation Zone Contact Zone

Contact
Zone

Figure 1. Process schematic of a DAF water plant.

The flocculation step involves mixing of the water to induce collisions among particles and
precipitated particles from metal coagulants yielding flocs. The sizes of flocs in the influent to the
DAF tank affects greatly the collisions among flocs and bubbles in the DAF tank contact zone and
subsequently, the rise velocities of flocs with attached bubbles in the DAF tank separation zone.
The flocs with attached bubbles are called floc-bubble aggregates. Flocculation is used in both
sedimentation and DAF plants, but their objectives differ. When settling follows flocculation, the
goal is to produce large flocs capable of settling at a fairly rapid rate. Flocs with sizes of 100 μm
and greater are required to produce reasonable settling rates. Proper floc sizes for DAF differs,
especially in treating reservoir supplies with low density particles and little mineral turbidity. For
treating these type supplies, the optimum floc size for the DAF contact zone is 10s of µm
determined from modelling, laboratory, and pilot plant data (Edzwald et al. 1990; Edzwald and

2
Wingler 1990; Edzwald et al. 1992; Edzwald 1995). It has been demonstrated through extensive
pilot plant data that flocculation times of 5-10 min are adequate (Valade et al. 1996; Edzwald et al.
1999). Over the last 10 to 15 years, many full-scale plants around the world have been placed
into operation with short flocculation times. A large DAF plant (1.1x106 m3 d-1) with a flocculation
time of only 5 min is under construction for New York City.
The DAF tank is divided into two sections with different functions as illustrated at the
bottom of Figure 1. There is a section at the front end that is baffled, called the contact zone,
where floc particles are introduced and contacted with air bubbles. Here, collisions occur among
bubbles and floc particles. If the floc particles are prepared properly via coagulation with respect
to their surface chemistry, then bubbles colliding with the flocs may attach yielding floc-bubble-
aggregates. The water carrying the suspension of bubbles, flocs, and aggregates flows to the
second section of the tank, called the separation zone. Here, bubbles not attached to flocs and
floc-bubble-aggregates may rise to the surface of the tank. The float layer at the surface of the
tank consists of a mixture of bubbles and floc particles attached to bubbles. In drinking water
applications, this froth is called the float. Over time, this float layer is concentrated producing a
sludge that is collected and removed from the tank. Clarified water, often referred to as the
subnatant, is withdrawn from the bottom of the tank. In a standard type DAF water plant, granular
media filtration follows DAF. In some applications, DAF is placed above the filters.
Air is dissolved in a recycle flow by adding air under pressure in a saturator. The recycle
rate (R) is defined by Eq. (1), where Q is the plant through-put flow and Qr is the recycle flow.
Typical recycle rates are 8 to 12%. Saturator pressures are between 400 and 600 kPa. The
recycle flow is injected through nozzles or special valves at the bottom entrance of the contact
zone. Microbubbles of 10 to 100 μm are produced, and give a milky appearance to the water;
hence the name white water.

Qr (1)
R=
Q

Some introductory comments are made here about the hydraulic loading of DAF tanks.
Most often, the nominal hydraulic loading (vnom-hl) is used to describe DAF. It is defined by Eq. (2).
Note that this definition ignores the recycle flow and uses the gross plan tank area (A) of the
contact and separation zones.

Q (2)
v nom − hl =
A

The removal of bubbles and floc-bubble aggregates occurs in the separation zone. The
separation of bubbles and aggregates depends on their rise velocities relative to the hydraulic
loading of the separation zone; they must be greater. Therefore, the separation zone
performance should be described with the hydraulic loading (vsz-hl) using the sum of the plant
throughput flow and the recycle flow divided by the separation zone plan area (Asz) as presented
by Eq. (3).

Q + Qr (3)
v sz − hl =
Asz

A major part of this paper addresses the fundamentals of rise velocities, and modification
of the traditional Hazen theory approach relating hydraulic loadings to rise velocities to explain
developments of high rate DAF processes.

3
PARTICLE AND BUBBLE SUSPENSIONS

Some comparisons are made in this section among the concentrations of raw water particles,
flocs entering the DAF tank, and bubble concentrations in the DAF tank contact zone. These
concentrations are presented on both a number and volume concentration basis. The
applications considered are in treating reservoir type water supplies. This means that the raw
water mineral turbidity is low and not important compared to low density particles present in the
reservoir. These particles are mostly organic in nature and are attributed to algae, bacteria,
protozoa and the like.
The suspended solids concentration on a mass basis is also considered low, which is true
for most reservoir supplies. Expressed on a volume basis (Φraw), two cases are considered of 1
and 10 ppm. The raw water particle concentrations (nraw) are calculated from Eq. (4) for two
particle size cases of 1 and 10 µm. The results are summarized in Table 1. These calculations
show that particle number concentrations can vary over a wide range, and depends significantly
on the raw water particle size. Since actually there would be a distribution of particle sizes, the 1
µm size case yields numbers most likely far exceeding what occurs. It is more likely that the
number concentration of raw particles (nraw) lies between the 1 and 10 µm size cases. Actual
measurements using electronic particle counters are limited usually to sizes greater than 2 µm.
Some particle counters can detect sizes of 1 µm, and show a large number of particles at this
size. In short, nraw of at least 106 and often exceeding 107 particles per L are reasonable values.
The number concentration changes with coagulation and flocculation, and is of interest with
respect to DAF.

Φ raw (4)
n raw =
π (d raw )3
6

Table 1. Raw water before alum addition and floc particle concentrations after alum addition
(assumed alum adds10 ppm of particles)
Raw Water
Φraw of 1 ppm Φraw of 10 ppm
Particle diameter 1 10 1 10
(µm)
nraw (# L-1) 1.9x109 1.9x106 1.9x1010 1.9x107
Flocs
Φfloc of 11 ppm Φfloc of 21 ppm
Floc diameter 5 10 50 5 10 50
(µm)
nfloc (# L-1) 1.7x108 2.1x107 1.7x105 3.2x108 4.0x107 3.2x105

Alum coagulation is usually practiced at dosing and pH conditions called sweep floc
meaning that the dissolved aluminum added exceeds the solubility of Al(OH)3(s). Precipitation
occurs producing additional solids. For alum dosing (as Al) of 2.5 to 5 mg L-1, the solids added
on a volume basis are in the range of 6.5 to 13 ppm. A value of 10 ppm is used for illustrative
purposes. The floc volume concentrations are now 11 and 21 ppm as indicated in Table 1.
Flocculation decreases the particle size, so after flocculation we consider floc sizes of 5, 10, and
50 µm. The calculations for floc number concentrations (nfloc) are summarized at the bottom of
Table 1. If the particles are dominated with small flocs of say 5 µm, then the floc particle
concentrations are high exceeding 108 flocs per L. If the flocculation process grows flocs into
mean sizes exceeding 10’s of µm, then the floc concentrations decrease to 106 flocs per L. Of
course, particles do not flocculate into a perfect sphere, which is assumed in the simple
calculations here. They form different shapes and have an internal porous structure, or fractal

4
geometry. Haarhoff and Edzwald (2001; 2004) have examined in detail the effects of flocculation
on floc density and floc number concentration for a range of fractal dimensions. While, floc
density is highly dependent on the extent of flocculation (floc size compared to primary particle
size) and the fractal dimension, the number concentration is not affected much so that the simple
calculations presented in Table 1 suffice.
The mass concentration of air released in the contact zone (Cb) is calculated from Eq. (5),
which is obtained from a mass balance accounting for the efficiency (e) of dissolving and
delivering air, the air in the recycle flow (Cr), the saturation air concentration (Cs) in water in the
contact zone (101.3 kPa is used), the recycle rate (R), and k which accounts for any air deficit
concentration in the influent flocculated water to the DAF tank. The two main variables for
controlling Cb are the saturator pressure, which affects Cr, and the recycle rate (R). Cb is
controlled primarily in practice through R, since for a particular plant the saturator pressure is not
changed very much. The bubble volume (Φb) and number (nb) concentrations are calculated from
Eqs. (6) and (7).

Cb =
[e(C r − C s )R − k ] (5)
1+ R

Cb (6)
Φb =
ρb

Φb (7)
nb =
π (d b ) 3

Bubble suspension concentrations are summarized in Table 2 for a common range of

recycle rates and typical saturator design conditions. A mean bubble size of 60 µm is assumed
for the contact zone (Haarhoff and Edzwald 2004). It is instructive to compare these against floc
volume and number concentrations entering DAF from Table 1. A comparison is made for a
recycle ratio of 10%. The results are presented in Table 3 for three mean floc sizes. The results
show that bubble volumes are much greater than floc volumes. This provides for the potential
reduction in the density of floc-bubble aggregates assuming sufficient bubble attachment to flocs
occurs. For flotation to occur, the bubble numbers must exceed the floc concentrations – i.e.,
there has to be at least one bubble attached to every floc. The calculations in Table 3 show that if
the flocs are too small, then there may be an insufficient number of bubbles. Increasing flocs to
sizes greater than 10 µm insures bubbles exceed floc particles. Haarhoff and Edzwald (2004)
examined this problem in greater detail accounting for fractal floc geometry, and also concluded
that to have an excess air bubbles, particles should grow in flocculation to average floc sizes of at
least 10 µm.

Table 2. Bubble suspension concentrations. Conditions: saturator pressure 500 kPa, air
dissolution and delivery efficiency 90%, 20 oC, db 60 µm.
R Cb Φb nb
(%) (mg L-1) (ppm) (# L-1)
8 7.8 6500 5.8x107
10 9.6 8000 7.1x107
12 11.3 9400 8.3x107

5
Table 3. Ratios of bubble to floc volume and number concentrations for recycle ratio of 10% and
floc volume of 21 ppm.
Floc diameter Φb/ Φfloc nb/nfloc
(µm)
5 380 0.2
10 380 2
50 380 22

BUBBLE AND FLOC-BUBBLE AGGREGATE RISE VELOCITIES

Bubbles

The classical Stokes equation (Eq. (8)) is used to calculate rise velocities (vb) of free bubbles –
i.e., not attached to flocs – where g is the gravitation constant, ρw and ρb are the water and air
bubble densities, db is the bubble diameter, and µw is the water viscosity. Bubble rise velocity
depends greatly on bubble diameter, the density difference between water and air, and the water
viscosity. The calculations in Table 4 show that rise velocities from 0.1 to almost 20 m h-1,
depending on bubble size and water temperature. Bubble sizes in the DAF contact zone typically
have a range of 10 to 100 µm. If small bubbles are carried into the separation zone and do not
grow, then their rise rates would be low. However, there is evidence that the bubbles in the
separation zone are larger than in the contact zone (Haarhoff and Edzwald 2004; Leppinen and
Dalziel 2004). A bubble diameter of 100 µm is a good estimate for the separation zone. From
Table 4, it is shown that these bubbles rise at about 10 to 20 m h-1 depending on water
temperature.

g (ρ w − ρ b )d b2 (8)
vb =
18μ w

Table 4. Bubble rise velocities at 4 and 20 oC.

vb (m h-1)
Bubble diameter 4 oC 20 oC
(µm)
10 0.1 0.2
20 0.5 0.8
60 4.5 7.0
100 12.5 19.5

Floc-Bubble Aggregates

The rise velocity of floc-bubble aggregates is calculated using the following equations, where dfb
is the aggregate equivalent diameter produced from N bubbles of size db attached to a floc
particle of size dp. ρfb, and vfb are the floc-bubble aggregate density and aggregate rise velocity.
K accounts for floc shape and drag force. For spheres, K is 24 (Haarhoff and Edzwald 2004).
Here, it is assumed that K is 24 for flocs smaller than 40 µm, and increases gradually to 45 at a
floc size at or above 170 µm.

(
d fb = d 3f + Nd b3 )
1/ 3 (9)

6
 ρ f d 3f + Nρ b d b3 (10)
ρ fb =
d 3f + Nd b3

4 g (ρ w − ρ fb )d 2fb (11)
v fb =
3 Kμ w

Figure 2 presents the density difference (ρw – ρfb) as a function of particle or floc size for
two cases of bubble attachment: 1) one bubble attached per floc and 2) half the maximum
number (Nmax) of bubbles that can possibly attach. Nmax depends on the floc and bubble areas
according to Tambo et al. (1986). For flocs less than 70 µm, only one bubble is allowed to attach
per floc. The number increases gradually to 6 bubbles per floc for flocs of 200 µm. The density
difference represents conceptually the driving force for flotation. For flocs to rise, it must exceed 0
and its maximum value (998.23-1.19) is approximately 1000 kg m-3. The maximum value occurs
when the bubble size dominates compared to the floc particle size so the ρfb is about equal to ρb,
and so the driving force is approximately equal to that of free bubbles. For small flocs of 50 μm
or less, the density difference is close to the maximum value, and the driving force is large.
Density differences decrease for single bubble attachment to flocs greater than 50 μm and
approach zero for 200 μm flocs. Multiple bubble attachment is required for flocs > 50 μm, and
especially for flocs > 100 μm to maintain a density difference of ~ 500 kg m-3.

1200

1000
Density Difference (kg m )
-3

800

600

400

1 Bubble Attached
Half Nmax Bubbles Attached
200

0
1 10 100

Particle or Floc Size(μm)

Figure 2. Density difference (ρw – ρfb) versus particle size for 1 bubble attachment and multiple
bubble attachment (1/2 Nmax) (Conditions: db of 100 µm; ρf of 1100 kg m-3, 20 oC)

The floc-bubble aggregate rise velocity (vfb) is addressed next, and the calculations
displayed in Figure 3 for the same two cases of 1 bubble attached per floc and half the maximum

7
number of possible bubbles per floc. Aggregate rise velocities are at about their maximum value
of 20 m h-1 for flocs of 50 μm or less with just one bubble attached. As flocs increase in size the
rise velocity decreases approaching zero for 200 μm flocs with 1 bubble attached because the
driving force is about zero (Figure 2). Multiple bubble attachment is required for flocs > 100 μm to
achieve rise velocities of 10 to 20 m h-1. To achieve an aggregate rise velocity > 20 m h-1 for 200
μm flocs requires at least 6 bubbles attached per floc.

20
Rise Velocity (m h )
-1

15

1 Bubble Attached
10 Half Nmax Bubbles Attached

0
1 10 100

Particle or Floc Size (μm)

Figure 3. Rise velocity of floc-bubble aggregates versus particle size for 1 bubble attachment and
multiple bubble attachment (1/2 Nmax) (Conditions: db of 100 µm; ρf of 1100 kg m-3, 20 oC)

Since flocs of 200 μm and greater would most likely settle as they leave the flocculation
tank, we assess flocs less than this size in evaluating the rise rates of aggregates. What we learn,
is that rise rates of free bubbles of 100 μm are about 20 m h-1, and floc-bubble aggregates have
about the same rise velocity for flocs < 50 μm with one bubble attached per floc. For larger flocs
(100-200 μm), rise velocities of 10 to 20 m h-1 are achieved with multiple bubble attachment.
These calculations were made at 20 oC; for colder water temperatures the rise velocities are less.

DAF PERFORMANCE AT HIGH HYDRAULIC LOADINGS

Hazen Theory of Performance

The separation zone of the DAF tank (see Figure 1) allows for gravity removal of free bubbles
and floc-bubble aggregates. In practice, separation zone performance and design are based on
the Hazen theory that the rise velocities of the bubbles (vb) and aggregates (vfb) must exceed the
separation zone hydraulic loading according to Eqs. (12) and (13).

Q + Qr (12)
v b ≥ v sz − hl =
Asz

8
 Q + Qr (13)
v fb ≥ v sz − hl =
Asz

The Hazen theory limits the design of the separation zone to hydraulic loadings of 10 to
20 m h-1 as shown previously by the rise rates in Table 4 and Figure 3. Prior to the 1990’s, DAF
tanks were designed at 5 to 10 m h-1 based on nominal hydraulic loadings. Since that time, DAF
tanks are being designed at high hydraulic loadings, even greater than 20 m h-1 and performing
well. This means that the simple Hazen theory is inadequate in predicting separation zone
performance. I will revisit this matter below, but first some pilot data are presented illustrating
DAF performance at high hydraulic loadings.

DAF Pilot Data at High Hydraulic Loadings

Edzwald et al. (1999) demonstrated through pilot studies that DAF can operate at high hydraulic
loadings. One set of data from their studies is presented in Figure 4. These data show DAF
effluent turbidities for hydraulic loadings of 21, 29, and 40 m h-1. Two turbidity measurements
were made to distinguish between particles and bubbles (bubble carryover) in the DAF effluent.
The data show excellent DAF performance for particles with turbidities < 0.4 NTU at all hydraulic
loadings. At 21 m h-1 there is little bubble carryover, but the amount of bubbles in the effluent
increased at higher hydraulic loadings, particularly at 41 m h-1. The bubble carryover, however,
did not affect granular media filtration performance following DAF. Filtered water turbidities were
< 0.1 NTU and head loss was not affected as unit filter run volumes were about 400 m3/ m2,
indicating good filter run times and production. Edzwald et al. (1999) also examined methods to
reduce bubble carryover. One method was to provide surfaces (tubes) in the tank to increase the
separation zone area for removal of bubbles. Another way is to improve the hydraulic flow
patterns through the tank. Examples are given in the next section.

It is important to note that bubble carryover from DAF does not cause air binding in filters.
Air binding in filters is caused by a pressure reduction within the filter causing dissolved air to
precipitate within the filter media. This results in large air bubbles that occupy the pore space
between grains causing air binding. Most bubbles that may be in the DAF effluent from high rate
processes never enter the filter box in full-scale plants. They are removed as water moves in
channels and over weirs on the way to the filters. Within the filter box, most that remain will rise to
top of the water column.

Developments in DAF Equipment

In the last 5-10 years companies have improved DAF equipment so DAF can be designed and
operated at 20 to 40 m h-1 with excellent turbidity and particle removals and little or no bubble
carryover. One system was developed by Rictor Oy (Finland), and it has been used by several
drinking water plants. One plant of note is at Tampere (Finland), which was designed at about 40
m h-1. The Rictor technology is available for use around the world by license to Infilco Degrémont.
It is known by the trade name of AquaDAF™. In the USA there are plants in New York and
California designed at least at 30 m h-1. A schematic diagram is presented in the top of Figure 5.
The DAF tank operates at high rate with improved hydraulic flow through the tank compared to
many conventional tanks. A key development is the orifice plate floor at the bottom of the DAF
tank that produces better flow distribution in the separation zone and at the outlet through the
plate floor.

9
 2.0

Particle Turb
Bubble Turb

1.5

Turbidity (NTU)

1.0

0.5

0.0
21 29 40

DAF Hydraulic Loading (m h-1)

Figure 4. DAF effluent turbidity due to particles and bubbles as a function of hydraulic loading
(Pilot plant data for Hemlocks Reservoir, Fairfield, CT (USA), water temperature 12-17 oC)

Figure 5. High rate DAF systems. Top figure: AquaDAF™ Infilco Degrémont (courtesy of Bob
Raczko, United Water), note orifice plate floor; Bottom figure: DAFRapide® (courtesy of Purac
Ltd.), note tubes or plates.

10
 Amato et al. (2001) and Dahlquist and Göransson (2004) have reported on the
development and application of a high rate DAF process, called DAFRapide®, which was
developed by Purac Ltd. (UK) and Purac AB (Sweden). The DAFRapide®system is designed for
high hydraulic loadings upto about 40 m h-1, as depicted at the bottom of Figure 5. Internal tubes
or plates are placed in the DAF tank to improve the hydraulic flow distribution and to increase the
separation zone area for improved removal of floc-bubble aggregates and free bubbles. Another
feature of DAF Rapide® is that the DAF tank can perform well in drinking water applications
utilizing short flocculation times of about 5 min. Finally, in the past year, Leopold Company has
introduced a high rate DAF process known as Clari-DAF™.

MODIFICATION OF SEPARATION ZONE THEORY

The conventional Hazen based theory simply states that the free bubble and bubble-floc
aggregate rise velocities must be equal or greater than the separation zone hydraulic loading as
shown by Eqs. (12) and (13). Hence conventional DAF tanks were initially designed at 5-10 m h-1,
and then increased in the 1990s to 10-15 m h-1. These loadings agree with rise velocities of
bubbles and aggregates presented above and summarized in Table 4 and Figure 3.

The ability of the high rate DAF systems, that operate at rates 20 to 40 m h-1, to perform
well has to do with the hydraulics of flow through the separation zone. Hazen theory assumes
simple plug flow downward through the separation zone so that the clarification area and footprint
area are the same, but this does not occur. DAF tanks, especially with higher hydraulic loadings
exhibit stratified flow in which the flow moves horizontally along near the top of the separation
zone to the end and then returns in a horizontal flow structure in a layer below the surface (Kiuru
2000; Lundh et al. 2000) before proceeding to the exit at the tank bottom. A simple and idealized
illustration of this stratified flow is shown in Figure 6. This depicted flow pattern effectively triples
the clarification separation area, thereby reducing the equivalent Hazen hydraulic loading for
separation of free bubbles and aggregates, but allowing for a small separation zone footprint area
and a high footprint based hydraulic loading. Although the flow patterns through DAF tanks varies
with hydraulic loading, recycle rate, and tank geometry, this simple analysis explains why DAF
tanks with hydraulic loadings > 20 m h-1 can be designed. Stratified flow effectively increases the
area for clarification, and is analogous to inserting trays (floors) in a sedimentation tank. The
AquaDAF™ process described above insures good stratified flow. The DAF Rapide® process
described above contains tubes or plates to directly increase the separation zone area.

Figure 6. Idealized stratified flow in high rate DAF processes.

11
CONCLUSIONS

It is instructive to examine the bubble and floc particle concentrations in the contact zone of the
DAF tank where bubbles and flocs undergo collisions and attachment to form floc-bubble
aggregates. Bubble concentrations in terms of volume and number concentrations are about
8000 ppm and 7x107 bubbles per L, respectively. Floc concentrations are highly dependent on
the initial primary particle concentration and size, and on the extent of flocculation. For the
reservoir water applications examined in this paper for DAF, it is important to form flocs with sizes
exceeding 10’s of µm so floc particle number concentrations are 106 flocs per L or less. As a
result bubble volume and numbers exceed floc volumes and numbers, and thus insures adequate
opportunities to form floc-bubble aggregates with density less than water.

The removal of bubbles and floc-bubble aggregates occurs in the separation zone of the
DAF tank. Calculated bubble rise rates for 100 µm bubbles are 10 to 20 m h-1 depending on
water temperature. Aggregate rise velocities are at about their maximum value of 20 m h-1 for
flocs of 50 μm or less with just one bubble attached. As flocs increase in size the aggregate rise
velocity decreases unless there is multiple attachment. Multiple bubble attachment is required for
flocs > 100 μm to achieve rise velocities of 10 to 20 m h-1.

Conventional DAF tanks were initially designed at hydraulic loadings of 5-10 m h-1, and
then increased in the 1990s to 10-15 m h-1. These hydraulic loadings are in accordance with the
predicted rise velocities of bubbles and aggregates, thus linking rise velocities through the Hazen
theory to DAF tank separation zone hydraulic loadings. Since the 1990s, DAF processes have
been designed at higher hydraulic loading rates. In the last several years, high rate DAF systems
have been introduced with hydraulic loading rates of 20 to 40 m h-1. The simple Hazen theory is
inadequate for predicting the removals of bubbles and aggregates for these high rate systems. It
is shown that in some high rate systems, the flow through the separation zone is stratified and
must be accounted for in relating separation zone efficiency to the hydraulic loading and bubble
or floc-bubble aggregate rise rates. An idealized and simple model is used to explain why DAF
tanks can be designed with hydraulic loadings > 20 m h-1.

ACKNOWLEDGMENT

I am indebted to my former graduate students who have conducted research on dissolved air
flotation over the past 20 years. I thank M. Kelley, J. Malley, A. Paralkar, D. Pernitsky, J. Plummer,
K. Berger, K. Boudreau, S. Bullock, D. Bunker, M. Janay, S. Olson, L. Parento, W. Parmenter, C.
Tamulonis, C. Udden, M. Valade, J. Walsh, B. Wingler, and C. Yu.

REFERENCES

Amato, T., Edzwald, J.K., Tobiason, J.E., Dahlquist, J., & Hedberg, T. 2001 An Integrated
Approach to Dissolved Air Flotation. Wat. Sci. Technol. 43 (8), 19-26.

Dahlquist, J. & Göransson, K. 2004. evolution of a High Rate Dissolved Air Flotation Process –
From Idea to Full-Scale Applicatio, Chapter in Chemical Water and Wastewater Treatment, H.
Hahn, E. Hoffmann, and Ødegaard, Eds., IWA Publishing, London.

Edzwald, J.K., Malley, J.P., & Yu, C. 1990. A Conceptual Model for Dissolved Air Flotation in
Water Treatment. Wat. Suppl. 8, 141-150.

Edzwald, J.K., & Wingler. B.J. 1990. Chemical and Physical Aspects of Dissolved-air-flotation
for the Removal of Algae. J. Wat. Suppl.: Res. & Technol. - AQUA. 39 (1), 24-35.

12
Edzwald, J.K., Walsh, J.P., Kaminsky, G.S., & Dunn, H.J. 1992. Flocculation and Air
Requirements for Dissolved Air Flotation. J. Am. Wat. Wks. Assoc. 84 (3), 92-100.

Edzwald, J.K. 1995 Principles and Applications of Dissolved Air Flotation. Wat. Sci. & Technol.
31 (3-4), 1-23.

Edzwald, J.K., Tobiason, J.E., Amato, T., & Maggi, L.J. 1999 Integrating High Rate Dissolved Air
Flotation Technology into Plant Design. J. Am. Wat. Wks. Assoc. 91 (12), 41-53.

Fukushi K., Tambo N. & Matsui Y. 1995 A Kinetic Model for Dissolved Air Flotation in Water and
Wastewater Treatment Wat. Sci. & Technol. 31 (3-4), 37-48.

Fukushi, K., Matsui, Y., & Tambo, N. 1998b. Dissolved Air Flotation: Experiments and Kinetic
Analysis. J. Wat. Suppl.: Res. & Technol. - AQUA. 47 (2), 76-86.

Haarhoff, J., & Edzwald, J.K. 2001 Modelling of Floc-Bubble Aggregate Rise Rates in Dissolved
Air Flotation. Wat. Sci. & Technol. 43 (8), 175-184.

Haarhoff, J., & Edzwald, J.K. 2004 Dissolved Air Flotation Modelling: Insights and Shortcomins.
J. Wat. Suppl.: Res. & Technol. - AQUA. 53 (3), 127-150.

Han, M. 2002. Modeling of DAF: The Effect of Particle and Bubble Characteristics. J. Wat. Suppl.:
Res. & Technol. - AQUA. 51 (1), 27-34.

Kiuru, H.J., 2000. Proc. 4th Inter. Conf. on Dissolved Air Flotation in Water and Wastewater
Treatment, Helsinki.

Leppinen, D.M. & Dalziel, S.B. 2004. Bubble Size Distribution in Dissolved Air Flotation Tanks, J.
Wat. Suppl.: Res. & Technol. Aqua, 53 () 531-.543.

Lundh, M., Jönsson, L., & Dahlquist, J. 2000. Experimental Studies of the Fluid Dynamics in the
Separation Zone in Dissolved Air Flotation. Wat. Res. 34 (1), 21-30.

Matsui, Y, Fukushi, K., & Tambo, N. 1998a. Modeling, Simulation, and Operational Parameters
of Dissolved Air Flotation. J. Wat. Suppl.: Res. & Technol. - AQUA. 47 (1), 9-20.

Tambo, N., Matsui, Y., & Fukushi, K. 1986 A Kinetic Study of Dissolved Air Flotation. World
Congress of Chemical Engineering, Tokyo, 200-203.

Valade, M.T., Edzwald, J.K., Tobiason, J.E., Dahlquist, J., Hedberg, T. & Amato, T. 1996.
Particle Removal by Flotation and Filtration: Pretreatment Effects. J. Amer. Wat. Wks. Assoc. 88
(12), 35-47.

13