You are on page 1of 4

Technical Note

Simplified Physically Based Model for


Estimating Effective Floc Density
Hoang-Ha Nguyen1 and Lloyd H. C. Chua2

Abstract: This paper applies dimensional analysis to propose an alternative model for estimating the effective density of flocs (Δρf ). The
Downloaded from ascelibrary.org by University of Southampton on 06/27/15. Copyright ASCE. For personal use only; all rights reserved.

model takes into account the effective density of the primary particles, in addition to the sizes of the floc and primary particles, and does not
consider the concept of self-similarity. The model contains three dimensionless products and two empirical parameters (αf and β f ), which
were calibrated by using data available in the literature. Values of αf ¼ 0:7 and β f ¼ 0:8 were obtained. The average value of the primary
particle size (Dp ) for the data used in the analysis, inferred from the new model, was found to vary from 0.05 μm to 100 μm with a mean value
of 2.5 μm. Good comparisons were obtained in comparing the estimated floc-settling velocity on the basis of the proposed model for effective
floc density with the measured value. DOI: 10.1061/(ASCE)HY.1943-7900.0000355. © 2011 American Society of Civil Engineers.
CE Database subject headings: Density; Dimensional analysis; Fractals; Particles; Sediment; Settling velocity.
Author keywords: Density; Dimensional analysis; Fractals; Particles; Cohesive sediment; Settling velocity.

Introduction F ¼ 3ðDf =Dp50 Þβ with β ¼ logðF c =3Þ= logðDf c =Dp50 Þ, Dp50


median size of primary particles, and F c and Df c characteristic
Floc density (ρf ) is one of key parameters affecting floc-settling values of F and Df , respectively. The effective density of flocs
velocity (ws ) in natural systems. As there is no direct method avail- was estimated from
able to measure ρf , ρf is usually estimated indirectly by using the
measured settling velocity and floc size (Df ), and the modified
Stokes’ law Δρf ¼ ðρs  ρw ÞðDp50 =Df Þ3F Ψ ð3Þ
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
4Δρf gDf where ψ ¼ 1 for monosized primary particles. The authors com-
ws ¼ ð1Þ
3ρw C D piled 26 published data sets covering both field and laboratory mea-
surements of ws as a function of Df . Based on this compiled data set
where Δρf ¼ ðρf -ρw Þ = floc effective density; g = gravitational and assuming ρs ¼ 2;300 kg=m3 , the authors used Eq. (1) with
acceleration, ρw = density of water; C D ¼ f ðR p Þ = drag coefficient; C D ¼ 24ð1 þ 0:15R 0:687
p Þ=R p to derive Δρf as a function of Df
R p ¼ ws Df =ν = (floc) particle Reynolds number; and ν = and then calibrated Eq. (3). The authors found that for the compiled
kinematic viscosity of water (Droppo et al. 2005; Khelifa and Hill data set, Dp50 varies from 0.05 μm to 20 μm, with a mean Dp50 of
2006). Floc density is also estimated by applying the concept of 1 μm, Df c ¼ 2;000 μm, F c ¼ 2:0, and ρs ¼ 2;300 kg=m3 . Also,
self-similar fractal flocs. Tambo and Watanabe (1979) derived incorporating Eq. (3) for Δρf , Khelifa and Hill (2006) verified
the expression for Δρf from volume and mass balance the use of Eq. (3) by comparing the settling velocity estimated by

Δρf ¼ αðρs  ρw ÞðDp =Df Þ3nf ð2Þ


1 ρ  ρw 3F DfF1
ws ¼ θg s Dp50 Ψ ð4Þ
where ρs and Dp = density and size of primary particles, res- 18 ρw ν 1 þ 0:15R 0:687
p
pectively; α ≈ 1:0, nf ≈ 2:1 and can be considered as the frac-
tal dimension of flocs (Kranenburg 1994). Khelifa and Hill (2006)
proposed an improvement to Eq. (2) by using, instead of a con- with the compiled data. In Eq. (4), θ is coefficient depending on the
stant nf , a variable three-dimensional (3D) fractal dimension shape of the floc. This model predicts that ws decreases at large floc
sizes (Df ≳ 2;000 μm for mean Dp50 ¼ 1 μm). Although fractal
1
Graduate Student, Division of Environmental and Water Resources analysis has been widely applied, Kranenburg (1994) and Khelifa
Engineering, School of Civil and Environmental Engineering, Nanyang and Hill (2006) argued that the self-similar fractal flocs may only
Technological Univ., 50 Nanyang Ave., Singapore 639798 (corresponding be an idealization because there is no true fractal entity in the real
author). E-mail: ng0003ha@ntu.edu.sg world; Maggi (2005) argued that real flocs display a complicated
2
Assistant Professor, Division of Environmental and Water Resources network of clusters of primary particles that cannot be simply mod-
Engineering, School of Civil and Environmental Engineering, Nanyang eled as fully self-similar structures.
Technological Univ., 50 Nanyang Ave., Singapore 639798. E-mail: A new model for Δρf is proposed in this paper. The modeling
chcchua@ntu.edu.sg
approach is based on dimensional analysis and does not consider
Note. This manuscript was submitted on September 2, 2009; approved
on October 28, 2010; published online on January 24, 2011. Discussion the concept of self-similarity of flocs. Nonlinear dependence of ρf
period open until January 1, 2012; separate discussions must be submitted is modeled as a function of not only Df and Dp but also the effective
for individual papers. This technical note is part of the Journal of Hydrau- density of the primary particles. The effects of flow turbulence and
lic Engineering, Vol. 137, No. 8, August 1, 2011. ©ASCE, ISSN 0733- sediment concentration, etc., on flocculation and ws , however, are
9429/2011/8-843–846/$25.00. beyond the scope of the present study.

JOURNAL OF HYDRAULIC ENGINEERING © ASCE / AUGUST 2011 / 843

J. Hydraul. Eng. 2011.137:843-846.


Proposed Model process where Δρf was calculated by Eq. (6) and compared with
the data from Khelifa and Hill (2006). The other parameter to be
Given that Δρf ¼ f ðDf ; Dp ; ρs ; ρw Þ and applying the Buckingham determined is Dp50 , which is a physical quantity and is constrained
Pi theorem (Hughes 1993), the dimensionless groups are by the size of primary particle in the experiments. This is different
π1 ¼ Δρf =ðρs  ρw Þ, π2 ¼ Df =Dp , and π3 ¼ ρs =ρw . Introducing from αf and β f , which are fitting parameters. Khelifa and Hill
g, ν, and Df and combining with π3 gives (2006) obtained Dp50 ¼ 0:05–20 μm for their model. Although
  1=3 the actual values of Dp50 for the data set computed by Khelifa
~ ¼ ρs
D  1 g=ν 2 Df ð5Þ and Hill (2006) may not be known, the field data of van Rijn
ρw (1993) suggests that the maximum value of Dp50 could be larger,
with a maximum value of 100 μm (see Table 1, reproduced from
where the inclusion of g and ν is to obtain a new dimensionless van Rijn 1993). Thus, setting a range for Dp50 ¼ 0:05 to 100 μm,
~  is similar to a floc particle parameter, which is a function
group. D the values of αf and β f that gave the best fit to the data from Khelifa
of Df and ρf . This parameter accounts for the joint contribution and Hill (2006) were found to be αf ¼ 0:7 and β f ¼ 0:8. The
of π3 and Df on π1 . It can be seen that D ~  decreases when Df
Downloaded from ascelibrary.org by University of Southampton on 06/27/15. Copyright ASCE. For personal use only; all rights reserved.

plot of Eq. (6) with Dp50 ranging from 0.05 to 100 μm and set-
and/or ρs decreases. By trial, it can be shown that a formula for ting αf ¼ 0:7 and β f ¼ 0:8 is shown in Fig. 2. A minimum value
Δρf =ðρs  ρw Þ ¼ f ðDf =Dp ; D ~  Þ can be expressed as
of Dp50 ¼ 0:05 μm was chosen to be consistent with the
   β 1:0 proposal by Khelifa and Hill (2006). Fig. 2 shows that the mean
Df f
Δρf ¼ ðρs  ρw Þ 1 þ ðD ~  Þαf 1 ð6Þ value of Dp50 is 2.5 μm, close to the Dp50 of kaolinite, which is
Dp50 3 μm, according to van Rijn (1993). This is in comparison to
the mean value of Dp50 ¼ 1 μm obtained by Khelifa and Hill
where αf and β f = empirical coefficients, which depend on physical
(2006). Note that in Fig. 2, Point A corresponds to ðDf ; Δρf Þ ¼
and chemical properties of the primary particles and surrounding
ð100 μm; 1;300 kg=m3 Þ, and at this point, ρf ¼ ρw þ Δρf ¼ ρs ¼
liquid. Comparing Eq. (6) with Eq. (2) shows that the exponent
2;300 kg=m3 , by definition, Dp ¼ Df ¼ 100 μm. As pointed out
β f is analogous to (3  nf ), which reflects the fractal property
~  Þαf is similar to α. Bache by an anonymous reviewer, the scatter in the data in Fig. 2 is large
of the floc, and the component ðD
and the proposed formula, as well as any other similar formula in
(2004) argued that α in Eq. (2) reflects the compaction of the pri-
the literature, should be used with caution because they represent
mary particles within the floc. The compaction should be charac-
just an “average order of magnitude” value for Δρf .
terized by the mass of the primary particles within a given floc and
Independent verification of Eq. (6) was carried out by compar-
is a function of ρs and Df . Therefore, in the present study, the com-
~  Þαf , which includes both ρs and Df , is analogous to α ing the model predictions with the data found in Camenen (2008).
ponent ðD
Camenen (2008) used Thorn’s (1982) settling velocity data for
and represents the compaction of flocs.
high-concentration mud obtained from the Severn estuary and
The dependence of Δρf on ρs is illustrated by a series of curves
Dankers et al.’s (2007) settling column data for kaolinite flocs.
shown in Fig. 1. It is observed from the figure that the curves
The data of Thorn (1982) and Dankers et al. (2007) are not included
move upward when ρs decreases. Thus, for a given value of Df ,
as part of the 26 data sets analyzed by Khelifa and Hill (2006).
when ρs is high, ρf is large (Point A). When ρs decreases, ρf
Camenen (2008) found that Df ≈ 295 μm, ρf ≈ 1;053 kg=m3 ,
becomes smaller (Point B), since Point B is founded on the curve
or Δρf ¼ 53 kg=m3 provided a good fit to Thorn’s (1982) data.
for smaller ρs, located above the curve for larger ρs. In the limit,
Similarly, Df ≈ 97 μm, ρf ≈ 1;066 kg=m3 , or Δρf ¼ 66 kg=m3
ðρf  ρw Þ=ðρs  ρw Þ ¼ 1 when ρs ¼ ρw (Point C). The effect of
~  on ðρf  ρw Þ=ðρs  ρw Þ can be seen in Fig. 1. In the figure,
D was found to fit well the data of Dankers et al. (2007). The pro-
the shape of the curves depends on D ~  ; ðρf  ρw Þ=ðρs  ρw Þ in- posed model was used to predict the Δρf observed in Thorn’s
creases when ρs and/or Df decreases. Conversely, ðρf  ρw Þ= (1982) and Danker et al.’s (2007) experiments. The following
ðρs  ρw Þ decreases when ρs and/or Df increases. values were assigned: αf ¼ 0:7, β f ¼ 0:8, ρs ¼ 2;300 kg=m3 ,
ρw ¼ 1;000 kg=m3 , ν ¼ 1 × 106 m2 =s. For the natural mud used
in Thorn (1982), Dp50 ¼ 20 μm was assumed. This is within the
Results and Discussion range of Dp50 suggested by van Rijn (1993) (see Table 1). For
the kaolinite used in Dankers et al. (2007), Dp50 ¼ 3 μm, which
Following the procedure by Khelifa and Hill (2006), the following is the average value of primary particle size suggested by van Rijn
values of ρs , ρw , and ν were assigned: ρs ¼ 2;300 kg=m3 , ρw ¼ (1993) for kaolinite (see Table 1), was assumed. The effective den-
1;000 kg=m3 and ν ¼ 1 × 106 m2 =s. Best fit values for αf , β f sity of flocs was calculated by using Eq. (6) and compared with the
and the estimation of Dp50 essentially followed a trial-and-error data provided by Camenen (2008). It was found that for mud in the
Severn estuary, the predicted value Δρf ¼ 52 kg=m3 is nearly
equal to the value of 53 kg=m3 suggested by Camenen (2008);

Table 1. Stokes Diameter of Deflocculated Material of Natural Muds


Found in the Netherlands and That of Kaolinite (van Rijn 1993,
reprinted with permission from Aqua Publications)
Dp50 Dp50
Mud type (μm) Mud type (μm)
Kaolinite 3 Maas (river) 25
Hollands Diep 1 (lake) 5 Brekens Harbor (estuary) 10
Hollands Diep 2 (lake) 10 Delfijn Harbor (estuary) 80
Ketelmeer (lake) 7 Loswal Noord (sea) 100
Fig. 1. Variation of floc density with the floc size Biesbosch (lake) 8

844 / JOURNAL OF HYDRAULIC ENGINEERING © ASCE / AUGUST 2011

J. Hydraul. Eng. 2011.137:843-846.


Downloaded from ascelibrary.org by University of Southampton on 06/27/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. Eq. (6) with αf ¼ 0:7, β f ¼ 0:8, ρs ¼ 2;300 kg=m3 , ρw ¼ 1;000 kg=m3 , ν ¼ 1 × 106 m2 =s, and (i) Dp ¼ 0:05 μm; (ii) Dp ¼ 2:5 μm;
(iii) Dp ¼ 100 μm [Note: original graph with d being primary particle size is reprinted from Khelifa and Hill (2006) with permission from the
International Association of Hydro-Environment Engineering and Research]

Fig. 3. Settling velocity of flocs with αf ¼ 0:7, β f ¼ 0:8, and (i) Dp ¼ 0:05 μm, (ii) Dp ¼ 2:5 μm, (iii) Dp ¼ 100 μm [Note: original graph with d
being primary particle size is reprinted from Khelifa and Hill (2006) with permission from the International Association of Hydro-Environment
Engineering and Research, and inset is reprinted from Alldredge and Gotschalk (1988) with permission from the American Society of Limnology
and Oceanography, Inc.]

JOURNAL OF HYDRAULIC ENGINEERING © ASCE / AUGUST 2011 / 845

J. Hydraul. Eng. 2011.137:843-846.


for kaolinite, the predicted value Δρf ¼ 50:2 kg=m3 is close to the F c , Df c =
characteristic values of F and Df ;
value of 66 kg=m3 suggested by Camenen (2008). Given the dis- g =
gravitational acceleration;
crepancies and scatter associated with data of this nature, the com- N =
number of primary particles in a floc;
parison suggests that the model is sufficiently robust. Rp =
floc particle Reynolds number;
The settling velocity data published in Khelifa and Hill (2006) αf , β f =
empirical coefficients;
were used to compare with the calculated ws by using Δρf Δρf =
effective density of flocs;
determined by the new model. The modified Stokes law, given ν =
kinematic viscosity of fluid;
in Eq. (1) where C D ¼ 24ð1 þ 0:15R 0:687 p Þ=R p , as used by π1 , π2 , π3 =
dimensionless groups;
Khelifa and Hill (2006), was adopted in the calculation of ws . ρf , ρs , ρw =
densities of flocs, primary particles, and water;
The following values were also assigned: αf ¼ 0:7, β f ¼ 0:8, and
ρs ¼ 2;300 kg=m3 , ρw ¼ 1;000 kg=m3 , and ν ¼ 1 × 106 m2 =s. ws = settling velocity of flocs.
The calculated ws are compared with ws estimated by Khelifa
and Hill (2006) in Fig. 3. For Df < 2;000 μm, the present model
Downloaded from ascelibrary.org by University of Southampton on 06/27/15. Copyright ASCE. For personal use only; all rights reserved.

shows that ws increases with Df . This is consistent with the model


by Khelifa and Hill (2006) and observations by Droppo (2001), References
who remarked that “as floc size increases, settling velocity in-
creases … as floc size seems to have a much stronger effect over Alldredge, A. L., and Gotschalk, C. (1988). “In situ settling behavior of
floc settling than density.” The increase in ws with Df for Df < marine snow.” Limnol. Oceanogr., 33(3), 339–351.
Bache, D. H. (2004). “Floc rupture and turbulence: A framework for analy-
2;000 μm is further supported by the data of Curran et al.
sis.” Chem. Eng. Sci., 59(12), 2521–2534.
(2007), Petticrew (2005), and Williams et al. (2008). For Camenen, B. (2008). “Settling velocity of sediments at high concentra-
Df ≳ 2;000 μm, the present model predicts that ws increases with tions.” Sediment and Ecohydraulics: INTERCOH 2005, T. Kusuda,
Df while Khelifa and Hill’s (2006) model shows that ws decreases H. Yamanishi, J. Spearman and J. Z. Gailani, eds., Elsevier, London,
with Df . However, it can be argued that the data points in Fig. 3 do 211–226.
not show a clear trend of ws decreasing with an increase in Df for Curran, K. J., et al. (2007). “Settling velocity, effective density, and mass
Df > 2;000 μm. The data points for Df ≳ 2;000 μm used in Fig. 3 composition of suspended sediment in a coastal bottom boundary layer,
were obtained from Diercks and Asper (1997) and Alldredge and Gulf of Lions, France.” Cont. Shelf Res., 27(10-11), 1408–1421.
Gotschalk (1988). In their paper, Diercks and Asper (1997) stated Dankers, P. J. T., Winterwerp, J. C., and Van Kesteren, W. G. M. (2007). “A
that their data (Df < 5:5 mm) “showed a wide variability” and “no preliminary study on the hindered settling of kaolinite flocs.” Estuarine
and Coastal Fine Sediment Dynamics: INTERCOH 2003, J. P.-Y. Maa,
statistically significant correlation of the aggregate settling speeds
L. P. Sanford, and D. H. Schoellhamer, eds., Elsevier, London, 227–
with their size ... was found.” On the contrary, Alldredge and 241.
Gotschalk’s (1988) data (Df < 25:5 mm) showed that ws increases Diercks, A. R., and Asper, V. L. (1997). “In situ settling speeds of marine
when Df increases (see Fig. 3 in Alldredge and Gotschalk 1988). snow aggregates below the mixed layer: Black Sea and Gulf of
Hence, it can be remarked here that, for Df ≳ 2;000 μm, the in- Mexico.” Deep Sea Res. Part I, 44(3), 385–398.
creasing trend in ws as proposed by the present model is supported Droppo, I. G. (2001). “Rethinking what constitutes suspended sediment.”
by the data of Alldredge and Gotschalk (1988). Hydrol. Processes, 15(9), 1551–1564.
Droppo, I. G., Leppard, G. G., Liss, S. N., and Milligan, T. G. (2005).
“Opportunities, needs, and strategic direction for research on floccula-
Conclusions tion in natural and engineered systems.” Flocculation in natural and
engineered environmental systems, I. G. Droppo, G. G. Leppard,
The following can be concluded from this study: S. N. Liss, and T. G. Milligan, eds., CRC Press, London, 407–421.
1. The value of Dp50 was found to vary from 0.05 μm to 100 μm Hughes, S. A. (1993). Physical models and laboratory techniques in
with a mean value of 2.5 μm, which is greater than the value of coastal engineering, World Scientific, Singapore.
1 μm suggested by Khelifa and Hill (2006) but is consistent Khelifa, A., and Hill, P. S. (2006). “Models for effective density and settling
velocity of flocs.” J. Hydraul. Res., 44(3), 390–401.
with the Dp50 of kaolinite.
Kranenburg, C. (1994). “The fractal structure of cohesive sediment aggre-
2. Independent verification of the new model by using the data of gates.” Estuarine Coastal Shelf Sci., 39(6), 451–460.
Thorn (1982) and Dankers et al. (2007) showed that the Maggi, F. (2005). “Flocculation dynamics of cohesive sediment.” Ph.D.
predicted Δρf are close to the values suggested by Camenen thesis, Delft Univ. of Technology, Netherlands.
(2008). Petticrew, E. L. (2005). “The composie nature of suspended and
3. The settling velocity estimated on the basis of the pro- gravel stored fine sediment in stream: A case study of O’Ne-eil Creek,
posed model for Δρf is consistent with the calculated ws based British Columbia, Canda.” Flocculation in natural and engineered
on Khelifa and Hill’s (2006) for Df < 2;000 μm and the environmental systems, I. G. Droppo, G. G. Leppard, S. N. Liss,
measured data by Alldredge and Gotschalk (1988) for and T. G. Milligan, eds., CRC Press, London, 1–21.
Df ≳ 2;000 μm. Tambo, N., and Watanabe, Y. (1979). “Physical characteristics of flocs—I.
The floc density function and aluminium floc.” Water Res., 13(5),
409–419.
Notation Thorn, M. F. C. (1982). “Physical process of siltation in tidal channels.”
Hydraulic modelling in maritime engineering: Proc., conf. organized
by the Institution of Civil Engineers, Thomas Telford, London, 65–73.
van Rijn, L. C. (1993). Principles of sediment transport in rivers, estuaries
The following symbols are used in this paper:
and coastal seas, Aqua Publications, Netherlands.
C D = drag coefficient; Williams, N. D., Walling, D. E., and Leeks, G. J. L. (2008). “An analysis of
Df , Dp (or d) = sizes of flocs and primary particles; the factors contributing to the settling potential of fine fluvial sediment.”
Dp50 = median size of primary particles; Hydrol. Processes, 22(20), 4153–4162.
D~  = dimensionless parameter; Winterwerp, J. C. (1999). “On the dynamics of high-concentrated mud sus-
F, nf = fractal dimensions; pensions.” Ph.D. thesis, Delft Univ. of Technology, Netherlands.

846 / JOURNAL OF HYDRAULIC ENGINEERING © ASCE / AUGUST 2011

J. Hydraul. Eng. 2011.137:843-846.

You might also like