You are on page 1of 30

Accepted Manuscript

Estimating effective density of cohesive sediment using shape factors from


holographic images

Sun Min Choi, Jun Young Seo, Ho Kyung Ha, Guan-hong Lee

PII: S0272-7714(18)30094-5
DOI: 10.1016/j.ecss.2018.10.008
Reference: YECSS 5998

To appear in: Estuarine, Coastal and Shelf Science

Received Date: 5 February 2018


Revised Date: 8 October 2018
Accepted Date: 15 October 2018

Please cite this article as: Choi, S.M., Seo, J.Y., Ha, H.K., Lee, G.-h., Estimating effective density of
cohesive sediment using shape factors from holographic images, Estuarine, Coastal and Shelf Science
(2018), doi: https://doi.org/10.1016/j.ecss.2018.10.008.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
1 Estimating effective density of cohesive sediment using shape factors

2 from holographic images

4 Sun Min Choi1, Jun Young Seo1, Ho Kyung Ha1,*, Guan-hong Lee1

PT
5
1
6 Department of Ocean Sciences, Inha University, Incheon 22212, Republic of Korea

RI
7
*
8 Corresponding author

SC
9 Tel: +82-32-860-7702; Fax: +82-32-862-5236

U
10 E-mail addresses: hahk@inha.ac.kr; hokyung.ha@gmail.com (H.K. Ha)
AN
11
M
D
TE
C EP
AC

1
ACCEPTED MANUSCRIPT
12 Abstract

13 To reveal the influence of shape factors on the effective density (ρe) of cohesive

14 sediment, a series of laboratory experiments has been conducted under different

15 turbulence intensities using a submersible digital holographic camera (LISST-Holo).

16 Shape factors representing axis ratio, boundary roughness, and sturdiness of flocs

PT
17 exhibited characteristic variations with turbulence intensity. For turbulence intensities

RI
18 from 0.54 to 1.42 Pa, micro-flocs were broken into single grains, resulting in a

19 constant axis ratio, a 13% decrease in boundary roughness, and a 16% increase in

SC
20 sturdiness. Out of these shape factors, ρe was mainly derived from the concept of

21 sturdiness, which represents the fraction of primary particles within flocs. In fact, the

22
U
16% increase in sturdiness added 9% to ρe. When the flocs became sturdier, ρe had
AN
23 lower dependence on floc size due to the increase in the fraction of primary particles.
M

24 Furthermore, ρe as a function of floc size might be affected by the response of the

25 optical instrument. In the case of fine particles (<25 µm), not every suspended particle
D

26 was reconstructed from the holographic image, inducing an underestimation of ρe.


TE

27 This is because insufficient scattering from floc edges impacts the particle detection
EP

28 ability. Despite the uncertainties, LISST-Holo is well suited to obtain a series of data

29 for shape factors of flocs. These outcomes give insight into means for estimating ρe.
C

30
AC

31 Keywords: Cohesive sediment; Shape factor; Effective density; LISST-Holo; Image

32 analysis

33

2
ACCEPTED MANUSCRIPT
34 1. Introduction

35 Cohesive sediment particles in fluid tend to form aggregates known as flocs,

36 because of physical attraction by electrochemical forces. The flocs composed of fine

37 particles are described as having porous and irregular structures (Kim et al., 2001;

38 Jarvis et al., 2005a). They are de-flocculated or flocculated depending on the

PT
39 turbulence intensity (Spicer and Pratsinis, 1996; Fettweis, 2008). It is known that the

RI
40 turbulence intensity range of 0.6–1 Pa is known by a key zone for de-flocculation

41 from macro-flocs to micro-flocs (Manning et al., 2010). In this process, the floc size

SC
42 and floc shape are significantly varied, resulting in non-uniform shape factor and

effective density (ρe) defined as the difference between floc density and water density

U
43 AN
44 (Spicer and Pratsinis, 1996; Dyer and Manning, 1999; Schwarz et al., 2017). As the

45 flocs become larger, for instance, the rigid bond between fine particles becomes
M

46 looser and they have lower ρe. Such variation of ρe has influence on the settling

47 velocity, which is one of the essential parameters determining sediment transport and
D

48 fate (Fugate and Friedrichs, 2003; Stone and Krishnappan, 2003; Mikkelsen et al.,
TE

49 2006). To date, for this reason, substantial efforts have been devoted to understand the
EP

50 relationship between floc structure, floc size, and ρe. In most studies, ρe represents a

51 power-function relationship (ρe∝DF) with floc size (D) (e.g., Al Ani et al., 1991;
C

52 Khelifa and Hill, 2006). The exponent term (F) is called as fractal dimension, which
AC

53 describes floc structure (Kranenburg, 1994; Winterwerp, 1998; Vahedi and Gorczyca,

54 2011). It can be also calculated using image analysis (i.e., 1D, 2D, and 3D capacity

55 fractal dimensions) (Stone and Krishnappan, 2003; Maggi et al., 2007). However, the

56 linkage of ρe to floc structure in terms of floc shape relationships has been barely

57 studied.

58 Various factors (e.g., floc size, settling velocity, and concentration) are used to

3
ACCEPTED MANUSCRIPT
59 estimate ρe occurring in natural settings (Droppo et al., 2005). Even though floc shape

60 is a critical component of floc structure, it is often ignored because of the high

61 variability of floc shapes (Shen and Maa, 2016). Of the numerous methods, image

62 analysis is one of the most powerful approaches to measure the floc shape (Eisma et

63 al., 1990; Manning and Dyer, 1999; Van der Lee, 2000; Droppo et al., 2005; Graham

PT
64 and Nimmo-Smith, 2010; Markussen and Andersen, 2013; Many et al., 2016). In

RI
65 particular, image analysis using a submersible digital holographic camera (LISST-

66 Holo, Sequoia Scientific Inc.) can effectively characterize the floc shape without any

SC
67 physical disturbance. It is thus possible to consider the influence of floc shape on ρe.

68 In this study, therefore, a series of laboratory water tank experiments has been

69
U
conducted to estimate ρe using holographic images. The main objectives are (1) to
AN
70 reveal the influence of turbulence intensity on floc shapes and (2) to understand the
M

71 capability of using floc shapes in calculating ρe under different turbulence intensities.

72
D

73 2. Materials and methods


TE

74 2.1. Holographic camera

75 The LISST-Holo was used to measure floc size and shape. Each holographic
EP

76 image is formed from the 658-nm laser by combining beam scattered from particles
C

77 with beam transmitted directly. The camera captures a holographic image (.pgm),
AC

78 which contains information about floc size and shape. The holographic image is

79 composed of several focal planes (≤50), and records the interference patterns, which

80 contain information about both phases and amplitudes of the resulting waves

81 produced from diffraction and/or refraction of the two beams. It enables the

82 determination of floc size and position in the sample volume (1.86 cm3) (Graham and

83 Nimmo-Smith, 2010; Ha et al., 2015). An outstanding advantage of LISST-Holo is

4
ACCEPTED MANUSCRIPT
84 that the focus on flocs is not fixed at a certain depth, allowing all of the flocs within

85 the sample volume to be recorded in the holographic images. The recorded flocs are

86 analyzed into reconstructed images by Sequoia’s image analysis software, and the

87 information on floc size and shape is computed in digital format (1 pixel = 4.4 × 4.4

88 µm2) (Ha et al., 2015).

PT
89 The floc shape was quantitatively represented by shape factors in projected

RI
90 images. The representatives are the major and minor axis (a and b for half length,

91 respectively), perimeter (P), area (A), and convex area (Aconv) as shown in Fig. 1. A is

SC
92 the extent converted from pixels occupied by the projected floc area. Aconv is the inner

93 area of the convex hull including empty pixels. P is the perimeter surrounding the

94
U
projected floc area. Such basic factors facilitate the calculation of advanced shape
AN
95 factors of flocs: axis ratio (AR), circularity (CI), and solidity (S). AR (=a/b) indicates
M

96 how close the projected area of floc is to that of circle. CI (=4πA/P2) represents the

97 smoothness of the floc boundary on the assumption that P of primary particles are
D

98 smaller than that of the flocs (Keyvani and Strom, 2014). S (=A/Aconv) is a measure of
TE

99 the primary particle fraction within flocs. A larger S represents flocs characterized by

100 more sturdiness and less roughness (Olson, 2011). In the analysis of projected image
EP

101 (2D rendering image) from LISST-Holo, the basic assumption is that the “randomly”
C

102 detected floc might be representative of the floc in 3D (Jarvis et al., 2005b). In the
AC

103 case of hollow sphere-like floc, this assumption can have an exception that S would

104 represent a solid floc without pore spaces.

105

106 2.2. Data analysis

107 In estimating ρe (=ρf – ρw =Mf/Vf – Mw/Vw), the floc mass (Mf) can be

108 expressed as (Fettweis, 2008)

5
ACCEPTED MANUSCRIPT

= + = + − (1)

109 where subscripts f, p, and w denote floc, primary particle, and pore water,

110 respectively. Within the floc, Mp denotes the mass of primary particles, Mw the mass

111 of pore water, ρw the pore water density, and Vf the floc volume. ρp is the density of

PT
112 primary particles (2,583 kg m-3 in this paper), which was estimated from both the

113 normal density of quartz (2,650 kg m-3) and organic matter (1,300 kg m-3) derived

RI
114 from loss on ignition (Markussen and Andersen, 2013). Using S and Vf derived from

SC
115 LISST-Holo, Mp was calculated in such a way that the ratio of Vp to Vf determines

116 how many primary particles are included in the floc. Assumptions are that (1) primary

117
U
particles maintain a constant size (median size (d50)=12.78 µm) and (2) primary
AN
118 particles and flocs are ellipsoid. S can then be defined by the expression


= = = = (2)
M

where ri is radius of the ith primary particle, and n is the total number of primary
D

119
TE

120 particles contained within a floc. The definition of S leads to an expression for the

121 volume ratio as follows:

4
EP

×∑ ×$ ′ $′ 0.6
= 3 = = (3)
4 $ )
$
3
C

122 where c is axis length parallel to settling direction of flocs, and c′ is that of primary
AC

123 particles. c′ = 0.6rp was used based on the previous studies (e.g., Vlieghe et al., 2014;

124 Many et al., 2016). By multiplying Vf at both sides of Equation (3), Vp was derived,

125 and then Mp was expressed as

0.6
= = * + (4)
)

126 From Equations (1) and (4), Mf was calculated as

6
ACCEPTED MANUSCRIPT
0.6
=, − - + (5)
)

127 Finally, ρe was successfully calculated as

0.6
/ =, − - (6)
)

128 2.3. Experimental methods

PT
129 Prior to the start of each experiment, the hydrated sediment was fully mixed

RI
130 with tap water (0 psu and 7.8 pH) within a rectangular water tank under a high

131 turbulence intensity (3.04 Pa for 20 minutes) using four submersible pumps installed

SC
132 at the bottom (Fig. 2). The turbulence intensity was adjusted by varying the pumping

133 direction of outlet vents. A series of five tank experiments was carried out under

134
U
different turbulence intensities (0.07, 0.21, 0.40, 0.54, and 1.42 Pa). Each experiment
AN
135 was conducted continuously within 1.5 hours maintaining water temperature at
M

136 26.6±1.04°C and suspended sediment concentration (SSC) at 30.22 mg l-1. A 10-MHz

137 acoustic Doppler velocimeter (ADV, Vectrino I, Nortek), optical backscatter sensor
D

138 (OBS, Campbell), and LISST-Holo were installed (Fig. 2). All sampling volumes of
TE

139 the instruments were placed at 40 cm above the tank bottom. The ADV measured 3-D

140 water velocities with a sampling rate of 200 Hz. Turbulence intensities were
EP

141 calculated by applying Soulsby’s constant of 0.2 in the turbulence kinetic energy
C

142 method (Kim et al., 2000). The OBS measured turbidity (in NTU) with a sampling
AC

143 rate of 4 Hz using 850-nm wavelength. The turbidity measured by the OBS was

144 converted into SSC with a good correlation coefficient (r = 0.98). The LISST-Holo

145 captured holographic images with a sampling rate of 0.2 Hz. The analyzed

146 holographic images (430 images in average) provided the distributions of floc size

147 and shape. Using a syringe with a long tube, water samples were collected every 20

148 minutes to estimate SSC at the same sampling depth as the OBS.

7
ACCEPTED MANUSCRIPT
149

150 2.4. Sediment

151 The tested sediment was sampled from Dongmak tidal flat (37º 35' 28.08"N,

152 126º 27' 45.66"E) off Ganghwa Island, west coast of Korea. A portion of dried

153 sediment was weighed and burned for 3 hours at 550°C. The organic matter content

PT
154 measured by the loss on ignition was 5% by mass. The particle size distribution (PSD)

RI
155 of tested sediment was measured by a particle-sizing instrument in state of dispersion

156 (Malvern, Mastersizer 2000S). The d50 was about 12.78 µm. The sediment consisted

SC
157 of 6.78% of clay, 69.46% of silt, and 23.76% of sand size particles (Fig. 3).

158

159 3. Results
U
AN
160 The PSDs of flocs generally exhibited coarse-skewed distributions over the
M

161 entire range of turbulence intensities (Fig. 4a–e). While turbulence intensity was

162 increasing from 0.07 to 1.42 Pa, d50 was in the range of 37.70–42.06 µm, and the
D

163 fraction-increment position shifted towards smaller floc sizes. At 0.21, 0.40, and 1.42
TE

164 Pa, the fraction increments were up to 3.71, 7.01, and 10.78% in the size ranges of

165 100–133, 40–100, and 25–40 µm, respectively. These shifting trends indicate that a
EP

166 de-flocculation process was dominant beyond 0.21 Pa. Moreover, the floc shape
C

167 factors (AR, CI, and S) exhibited characteristic variations as the turbulence intensity
AC

168 was increased from 0.07 to 1.42 Pa (Fig. 4f). On average, AR had a relatively stable

169 value of about 0.55, whereas CI and S increased from 0.40 to 0.49 and from 0.56 to

170 0.64, respectively. Such a distinct increase in CI and S could be associated with the

171 large amount of fraction increment (10.78%) in the floc size range of 25–40 µm. At

172 1.42 Pa, most flocs can be categorized as single grains (<36 µm), playing a role as

173 building blocks of micro-flocs (36–133 µm) (Mikkelsen et al., 2006). Even with

8
ACCEPTED MANUSCRIPT
174 strong turbulence, single grains are hardly broken up into primary particles and are

175 characterized as circular, smooth, and sturdy structures (Lee et al., 2012). It is noted

176 that a primary particle in this study refers to an individual particle before flocculation,

177 and that the single grain (also known as flocculi; Lee et al., 2012) is categorized as

178 “floc” aggregated with several primary particles. As the de-flocculation from micro-

PT
179 flocs to single grains occurred, the ratio of A and P decreased, implying that the

RI
180 boundary roughness of single grains decreased. Above all, the reduction of Aconv was

181 determined by the decrease in P. These results reflected the fact that the increase in

SC
182 the number of single grains caused increases in CI and S. Representative holographic

183 images in Fig. 5 show the variation of shape factors. The single grains exhibited

184
U
smooth boundaries and sturdy structures. While floc size was decreasing with
AN
185 increasing turbulence intensity, the floc boundary became more abrasive, breaking to
M

186 single grains. At 0.54 Pa, the micro-floc was described as a complex boundary

187 (CI=0.30) with many empty pixels inside the floc (S=0.64), but at 1.42 Pa, the single
D

188 grain had a smoother boundary (CI=0.65) with a compacted floc (S=0.79).
TE

189 Furthermore, both CI and S at 1.42 Pa were always large, compared to those at 0.54

190 Pa.
EP

191 The variations of shape factors were subject to a combination of the


C

192 influences of floc size and turbulence intensity (Fig. 6). At both 0.54 and 1.42 Pa, AR
AC

193 distributions were in the range of 0.2 to 1.0, and exhibited a long tail feature to the

194 right (skewness: 0.24) (top row in Fig. 6). It represented shape diversity between

195 elliptical and circular flocs. Over the entire floc size range, AR was evenly scattered

196 maintaining an average of 0.56, indicating that most flocs retained a half squashed-

197 elliptical shape regardless of turbulence intensity.

198 CI distributions at both turbulence intensities were widely spread over the

9
ACCEPTED MANUSCRIPT
199 entire range of 0 to 1 (middle row in Fig. 6). CI was largely scattered in the range of

200 single grains, providing evidence for a variety of floc boundary roughness. In the

201 range of micro-flocs, it converged to 0.1 resulting in higher floc boundary roughness.

202 On the other hand, when turbulence intensity increased from 0.54 to 1.42 Pa, the

203 mode of the CI distributions shifted from 0.38 to 0.43. The change in CI with

PT
204 turbulence intensity could occur even with floc size maintained, but the roughness at

RI
205 1.42 Pa was as much as 13% lower on average, compared to that at 0.54 Pa. S

206 distributions exhibited patterns similar to those of CI and with a nearly normal

SC
207 distribution ranging from 0.4 to 1 (bottom row in Fig. 6). However, dependence of S

208 on floc size had a gentler slope than that of CI. While the turbulence intensity was

209
U
being increased, the mode of S distributions shifted from 0.63 to 0.73, resulting in
AN
210 increased sturdiness. This indicates that increased turbulence intensity caused flocs to
M

211 include a large fraction of primary particles, as much as 16% on average. In other

212 words, when the flocculation process occurred with a decrease in CI and S, the flocs
D

213 became rougher and less sturdy representing a lower fraction of primary particles
TE

214 within the flocs. This resulted from the fact that P and Aconv increased more rapidly

215 than A (Stone and Krishnappan, 2003). In the de-flocculation process with an increase
EP

216 in CI and S, smoother and more compacted flocs resulted from the rapid decrease of P
C

217 and Aconv rather than A representing a large fraction of primary particles within the
AC

218 flocs.

219

220 4. Discussion

221 The results presented above demonstrate that the variation of floc structure

222 under different turbulence intensities can be described through shape factors. Using

223 the fact that the variation of floc shape is concomitant with that of ρe, the influence of

10
ACCEPTED MANUSCRIPT
224 shape factors on ρe will be documented in this section.

225 At present, there are only a few ways to calculate ρe from floc shape.

226 Measuring floc shape generally requires a significant amount of time, and there is the

227 possibility of floc breakage in the process (Jarvis et al., 2005a; Koivuranta et al.,

228 2013). Such disadvantages can be overcome by LISST-Holo, and ρe was calculated

PT
229 properly as a function of shape factors through the Equation (6). The ρe calculated

RI
230 from LISST-Holo (ρe_Holo) follows an inverse power relationship with floc size in a

231 reasonable range of 60–500 kg m-3 (Fig. 7a). Overall, the scattering range of ρe_Holo

SC
232 decreased with increase in floc size at both 0.54 and 1.42 Pa. It was speculated that

U
233 frequent minor changes in floc shape attributed to the large degree of scatter in CI and
AN
234 S (middle and bottom rows in Fig. 6). Based on the relationship between ρe_Holo and

235 shape factors, such changes in single grain size could lead to a drastic increase in non-
M

236 linearity for ρe_Holo (Gibbs, 1985).

For average floc sizes, on the other hand, ρe_Holo at 1.42 Pa was larger than that
D

237

at 0.54 Pa by 9% (as much as 18.77 kg m-3). This implies that ρe_Holo was influenced
TE

238

239 by the variation of turbulence intensity. In this de-flocculation process, as mentioned


EP

240 above, Mp increased with decreasing proportion of pore water. The floc structure

241 became sturdy enough to survive when the turbulence intensity was higher, and thus
C

242 floc breakage no longer occurred easily (Shao et al., 2010).


AC

243 Such floc structure is able to be described as the slope of the regression line

244 between ρe_Holo and floc size in a log-log plot (Kranenburg, 1994; Dyer and Manning,

245 1999). Previous research exhibited various slopes of this regression line from -1.40 to

246 -0.42 in the range from 10 to 1000 kg m-3 (McCave, 1975; Fennessy et al., 1994; Hill

247 et al., 1998; Manning and Dyer, 1999; Sternberg, et al., 1999; Mikkelsen and Pejrup,

248 2001; Markussen and Andersen, 2013; Guo et al., 2017) (Fig. 7b). Sturdy and

11
ACCEPTED MANUSCRIPT
249 compact flocs are represented with gentle slopes, while steep slopes indicate fragile

250 and loose flocs. If the flocs are composed of only primary particles without pore

251 water, they are characterized as the sturdiest and most compact (Dyer and Manning,

252 1999). Regardless of floc size, ρe of the sturdiest flocs does not vary, so the slope

253 becomes zero. The slope, therefore, is inversely proportional to S, which indicates the

PT
254 fraction of primary particles within the flocs. In the case of this study, while varying

RI
255 from 0.54 Pa to 1.42 Pa, the regression slope became gentler from -1.375 to -1.312

256 because the 16% increase in S added 9% to ρe_Holo. This suggests that ρe_Holo has lower

SC
257 dependence on floc size with increase in fraction of primary particles within flocs.

258 The ρe_Holo as a function of floc size might be affected by the response of the

U
optical instrument (Fettweis, 2008). In order to reveal the uncertainty, ρe derived from
AN
259

260 OBS (ρe_OBS) was used to compare with the ρe_Holo (Hsu and Liu, 2010; Guo et al.,
M

261 2017). As shown in Fig. 7a, the ρe_OBS’s were 1053, 609, and 384 kg m-3 at 38, 45, and

262 53 µm of floc size, respectively. This suggests that ρe_Holo was an underestimate
D

263 compared to ρe_OBS, and the difference between them was larger toward the single
TE

264 grain size. The main reason for the discrepancy was the difference in sensing ranges
EP

265 for LISST-Holo and OBS. In case of LISST-Holo, the floc size range from 25 to 2500

266 µm is reliable for getting high-resolution images of all particles in reconstructed


C

267 images. As a practical size threshold for determining what is a particle, Aconv of 9
AC

268 pixels (equivalent circular diameter 15 µm) was recommended (Graham et al., 2012;

269 Many et al., 2016). Within the range of 15–25 µm in floc size, insufficient scattering

270 from edges of flocs has an impact on the ability of LISST-Holo to detect the particles

271 in a reconstructed image. For this reason, every suspended particle is not

272 reconstructed (Personal conversation with Tomas Leeuw of Sequoia, 2017). On the

273 other hand, the short wavelength of OBS allows suspended particles smaller than 25

12
ACCEPTED MANUSCRIPT
274 µm to be detected. They were included only in Mp, causing an underestimation of

275 ρe_Holo compared to ρe_OBS. Furthermore, as floc size decreased, a widening difference

276 between ρe_Holo and ρe_OBS was observed. This is because the OBS response is more

277 effective for floc sizes smaller than 25 µm (Wren et al., 2000; Ha et al., 2009).

278 Consequently, there are some uncertainties, as in all shape-factor

PT
279 measurement techniques. Despite the uncertainties, LISST-Holo is well suited to

RI
280 obtain a series of data for floc shape. Shape factors are useful for determining ρe, but

281 are still seldom quantified properly. In this study, shape factors are determined in

SC
282 association with characteristics of flocs and these outcomes give insight into means

for estimating ρe.

U
283
AN
284

285 5. Conclusions
M

286 Using the LISST-Holo, this study emphasized the capacity of shape factors to

287 represent behaviors of cohesive sediment. The following conclusions were drawn
D

288 from the laboratory experiments.


TE

289 (1) As turbulence intensity increased from 0.54 to 1.42 Pa, micro-flocs were de-
EP

290 flocculated into single grains, causing variation of shape factors: axis ratio

291 remained constant with half-squashed ellipse, boundary roughness decreased


C

292 by 13%, and sturdiness increased by 16%.


AC

293 (2) Such variation of floc shape being concomitant with that of ρe_Holo, ρe_Holo was

294 mainly derived from the concept of sturdiness. The 16% increase in sturdiness

295 added 9% to ρe_Holo due to the large fraction of primary particles within flocs.

296 (3) When the relationship between ρe_Holo and floc size was established, the

297 increase in fraction of primary particles within flocs caused lower dependence

298 of ρe_Holo on floc size.

13
ACCEPTED MANUSCRIPT
299 (4) The function relating ρe_Holo to floc size can be affected by response of the

300 optical instrument. For fine particles (<25 µm), insufficient scattering from

301 particle edges resulted in underestimation of ρe_Holo. This was because every

302 suspended particle was not necessarily reconstructed in the holographic

303 image.

PT
304

RI
305 Acknowledgements
306 This study was supported by Inha University Research Grant (58380).

U SC
AN
M
D
TE
C EP
AC

14
ACCEPTED MANUSCRIPT
307 References

308 Al Ani, S., Dyer, K.R., Huntley, D.A., 1991. Measurement of the influence of salinity

309 on floc density and strength. Geo-Marine Letters 11(3), 154–158.

310 Droppo, I.G., Nackaerts, K., Walling, D.E., Williams, N., 2005. Can flocs and water

311 stable soil aggregates be differentiated within fluvial systems? Catena 60(1), 1–

PT
312 18.

RI
313 Dyer, K.R., Manning, A.J., 1999. Observation of the size, settling velocity and

314 effective density of flocs, and their fractal dimensions. Journal of Sea Research

SC
315 41(1), 87–95.

316 Eisma, D., Schuhmacher, T., Boekel, H., Van Heerwaarden, J., Franken, H., Laan, M.,

317
U
Vaars, A., Eijgenraam, F., Kalf, J., 1990. A camera and image-analysis system for
AN
318 in situ observation of flocs in natural waters. Netherlands Journal of Sea
M

319 Research 27(1), 43–56.

320 Fennessy, M.J., Dyer, K.R., Huntley, D.A., 1994. INSSEV: an instrument to measure
D

321 the size and settling velocity of flocs in situ. Marine Geology 117(1), 107–117.
TE

322 Fettweis, M., 2008. Uncertainty of excess density and settling velocity of mud flocs

323 derived from in situ measurements. Estuarine, Coastal and Shelf Science 78(2),
EP

324 426–436.
C

325 Fugate, D.C., Friedrichs, C.T., 2003. Controls on suspended aggregate size in partially
AC

326 mixed estuaries. Estuarine, Coastal and Shelf Science 58(2), 389–404.

327 Gibbs, R.J., 1985. Settling velocity, diameter, and density for flocs of illite, kaolinite,

328 and montmorillonite. Journal of Sedimentary Research 55(1), 65–68.

329 Graham, G.W., Davies, E.J., Nimmo-Smith, W.A.M., Bowers, D.G., Braithwaite,

330 K.M., 2012. Interpreting LISST-100X measurements of particles with complex

331 shape using digital in-line holography. Journal of Geophysical Research: Oceans

15
ACCEPTED MANUSCRIPT
332 117(C5), 1–20.

333 Graham, G.W., Nimmo-Smith, W.A.M., 2010. The application of holography to the

334 analysis of size and settling velocity of suspended cohesive sediments.

335 Limnology and Oceanography: Methods 8(1), 1–15.

336 Guo, C., He, Q., Guo, L., Winterwerp, J.C., 2017. A study of in-situ sediment

PT
337 flocculation in the turbidity maxima of the Yangtze Estuary. Estuarine, Coastal

RI
338 and Shelf Science 191, 1–9.

339 Ha, H.K., Hsu, W.Y., Maa, J.Y., Shao, Y.Y., Holland, C.W., 2009. Using ADV

SC
340 backscatter strength for measuring suspended cohesive sediment concentration.

341 Continental Shelf Research 29(10), 1310–1316.

342
U
Ha, H.K., Kim, Y.H., Lee, H.J., Hwang, B., Joo, H.M., 2015. Under-ice measurements
AN
343 of suspended particulate matters using ADCP and LISST-Holo. Ocean Science
M

344 Journal 50(1), 97–108.

345 Hill, P.S., Syvitski, J.P., Cowan, E.A., Powell, R.D., 1998. In situ observations of floc
D

346 settling velocities in Glacier Bay, Alaska. Marine Geology 145(1–2), 85–94.
TE

347 Hsu, R.T., Liu, J.T., 2010. In-situ estimations of the density and porosity of flocs of

348 varying sizes in a submarine canyon. Marine Geology 276(1), 105–109.


EP

349 Jarvis, P., Jefferson, B., Gregory, J.O.H.N., Parsons, S.A., 2005b. A review of floc
C

350 strength and breakage. Water Research 39(14), 3121–3137.


AC

351 Jarvis, P., Jefferson, B., Parsons, S.A., 2005a. Measuring floc structural

352 characteristics. Reviews in Environmental Science and Biotechnology 4(1), 1–

353 18.

354 Keyvani, A., Strom, K., 2014. Influence of cycles of high and low turbulent shear on

355 the growth rate and equilibrium size of mud flocs. Marine Geology 354, 1–14.

356 Khelifa, A., Hill, P.S., 2006. Models for effective density and settling velocity of

16
ACCEPTED MANUSCRIPT
357 flocs. Journal of Hydraulic Research 44(3), 390–401.

358 Kim, S.C., Friedrichs, C.T., Maa, J.P.Y., Wright, L.D., 2000. Estimating bottom stress

359 in tidal boundary layer from acoustic Doppler velocimeter data. Journal of

360 Hydraulic Engineering 126(6), 399–406.

361 Kim, S.H., Moon, B.H., Lee, H.I., 2001. Effects of pH and dosage on pollutant

PT
362 removal and floc structure during coagulation. Microchemical Journal 68(2-3),

RI
363 197–203.

364 Koivuranta, E., Keskitalo, J., Haapala, A., Stoor, T., Sarén, M., Niinimäki, J., 2013.

SC
365 Optical monitoring of activated sludge flocs in bulking and non-bulking

366 conditions. Environmental Technology 34(5), 679–686.

367
U
Kranenburg, C., 1994. The fractal structure of cohesive sediment aggregates.
AN
368 Estuarine, Coastal and Shelf Science 39(6), 451–460.
M

369 Lee, B.J., Fettweis, M., Toorman, E., Molz, F.J., 2012. Multimodality of a particle

370 size distribution of cohesive suspended particulate matters in a coastal zone.


D

371 Journal of Geophysical Research: Oceans 117(C3).


TE

372 Maggi, F., Mietta, F., Winterwerp, J.C., 2007. Effect of variable fractal dimension on

373 the floc size distribution of suspended cohesive sediment. Journal of Hydrology
EP

374 343(1–2), 43–55.


C

375 Manning, A.J., Baugh, J.V., Spearman, J.R., Whitehouse, R.J., 2010. Flocculation
AC

376 settling characteristics of mud: sand mixtures. Ocean Dynamics 60(2), 237–253.

377 Manning, A.J., Dyer, K.R., 1999. A laboratory examination of floc characteristics with

378 regard to turbulent shearing. Marine Geology 160(1), 147–170.

379 Many, G., Bourrin, F., Durrieu de Madron, X., Pairaud, I., Gangloff, A., Doxaran, D.,

380 Ody, A., Verney, R., Menniti, C., Le Berre, D., Jacquet, M., 2016. Particle

381 assemblage characterization in the Rhone River ROFI. Journal of Marine

17
ACCEPTED MANUSCRIPT
382 Systems 157, 39–51.

383 Markussen, T.N., Andersen, T.J., 2013. A simple method for calculating in situ floc

384 settling velocities based on effective density functions. Marine Geology 344, 10–

385 18.

386 McCave, I.N., 1975. Vertical flux of particles in the ocean. Deep Sea Research and

PT
387 Oceanographic Abstracts 22(7), 491–502.

RI
388 Mikkelsen, O.A., Hill, P.S., Milligan, T.G., 2006. Single-grain, microfloc and

389 macrofloc volume variations observed with a LISST-100 and a digital floc

SC
390 camera. Journal of Sea Research 55(2), 87–102.

391 Mikkelsen, O.A., Pejrup, M., 2001. The use of a LISST-100 laser particle sizer for

392
U
in-situ estimates of floc size, density and settling velocity. Geo-Marine Letters
AN
393 20(4), 187–195.
M

394 Olson, E., 2011. Particle shape factors and their use in image analysis-Part 1: Theory.

395 Journal of GXP Compliance 15(3), 85.


D

396 Schwarz, C., Cox, T., Van Engeland, T., Van Oevelen, D., Van Belzen, J., Van de
TE

397 Koppel, J., Soetaert, K., Bouma, T.J., Meire, P., Temmerman, S., 2017. Field

398 estimates of floc dynamics and settling velocities in a tidal creek with significant
EP

399 along-channel gradients in velocity and SPM. Estuarine, Coastal and Shelf
C

400 Science 197, 221–235.


AC

401 Shao, Y., Yan, Y., Maa, J.P.Y., 2010. In situ measurements of settling velocity near

402 Baimao shoal in Changjiang estuary. Journal of Hydraulic Engineering 137(3),

403 372–380.

404 Shen, X., Maa, J.P.Y., 2016. A camera and image processing system for floc size

405 distributions of suspended particles. Marine Geology 376, 132–146.

406 Spicer, P.T., Pratsinis, S.E., 1996. Shear-induced flocculation: the evolution of floc

18
ACCEPTED MANUSCRIPT
407 structure and the shape of the size distribution at steady state. Water Research

408 30(5), 1049–1056.

409 Sternberg, R.W., Berhane, I., Ogston, A.S., 1999. Measurement of size and settling

410 velocity of suspended aggregates on the northern California continental shelf.

411 Marine Geology 154(1), 43–53.

PT
412 Stone, M., Krishnappan, B.G., 2003. Floc morphology and size distributions of

RI
413 cohesive sediment in steady-state flow. Water Research 37(11), 2739–2747.

414 Vahedi, A., Gorczyca, B., 2011. Application of fractal dimensions to study the

SC
415 structure of flocs formed in lime softening process. Water Research 45(2), 545–

416 556.

417
U
Van der Lee, W.T.B., 2000. Temporal variation of floc size and settling velocity in the
AN
418 Dollard estuary. Continental Shelf Research 20(12), 1495–1511.
M

419 Vlieghe, M., Coufort-Saudejaud, C., Frances, C., Liné, A., 2014. In situ

420 characterization of floc morphology by image analysis in a turbulent Taylor–


D

421 Couette reactor. AIChE Journal 60(7), 2389–2403.


TE

422 Winterwerp, J.C., 1998. A simple model for turbulence induced flocculation of

423 cohesive sediment. Journal of Hydraulic Research 36(3), 309–326.


EP

424 Wren, D.G., Barkdoll, B.D., Kuhnle, R.A., Derrow, R.W., 2000. Field techniques for
C

425 suspended-sediment measurement. Journal of Hydraulic Engineering 126(2), 97–


AC

426 104.

427

19
ACCEPTED MANUSCRIPT
428 Figure captions

429

430 Fig. 1. Concept diagram of floc shape factors derived from LISST-Holo. Red dots are

431 the grid points for image analysis. Numbers indicate magnitudes of parameters.

432

PT
433 Fig. 2. Schematic setup of tank experiments. Sampling volumes (see red areas) of

RI
434 ADV, OBS, and LISST-Holo are placed at the same level. Four submersible pumps

435 are located on the tank bottom, one in each corner. Arrows indicate the pumping

SC
436 directions.

437

438
U
Fig. 3. Particle size distribution of the sediment used in water tank experiments.
AN
439
M

440 Fig. 4. (a–e) Particle size distribution (PSD) of flocs under different turbulence

441 intensities. Red and blue shaded areas respectively represent the increment and
D

442 decrement in fraction between the PSD and its immediately previous PSD. Value
TE

443 within parentheses denotes the measured particle number. (f) Dependencies of shape

444 factors on the turbulence intensity, where AR, CI, and S are the axis ratio, circularity,
EP

445 and solidity, respectively.


C

446
AC

447 Fig. 5. Floc images processed by LISST-Holo at (a) 0.54 Pa, and (b) 1.42 Pa. Red

448 boxes (100 × 100 µm2) in each panel contain the particles of interest. Values within

449 parentheses indicate the shape factors (AR, CI, and S, respectively).

450

451 Fig. 6. Variations of AR (top row), CI (middle row) and S (bottom row). Red and blue

452 colors indicate 0.54 and 1.42 Pa, respectively. In left panels, markers with lines

20
ACCEPTED MANUSCRIPT
453 represent fractions of shape factors. The horizontal lines with values denote mode of

454 shape factor. In right panels, dots denote the raw data and markers with lines denote

455 mean values of each size class.

456

457 Fig. 7. (a) ρe_Holo derived from AR and S. Red and blue colors indicate 0.54 and 1.42

PT
458 Pa, respectively. Lines are fitted curves of power functions. Cross marks represent

RI
459 ρe_OBS. (b) Comparison of ρe_Holo with previous studies. Note that x- and y- axes are

460 plotted on logarithmic scales.

U SC
AN
M
D
TE
C EP
AC

21
ACCEPTED MANUSCRIPT

PT
RI
SC
1
1 U
AN
a
M

b
D

Major axis (μm) Perimeter (μm


TE

a=3 P = 22
EP

Minor axis (μm)


b=2
C
AC

Area (μm 2 ) Convex area (μ


A = 14 Aconv = 20.5
ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
EP
C
AC
ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
5
M

4
D
TE
Fraction (%)

3
EP

2
C
AC

0
Particle size (μm)
ACCEPTED MANUSCRIPT

PT
RI
20
(a) 0.07 Pa (b) 0.21 Pa (c)

SC
15 (n=115,489) (n=66,893)

10
5 U
AN
0
1 2 1
M

10 10 10
Floc size (μm) F
D
Fraction (%)

20
(d) 0.54 Pa (f)
TE

15 (n=143,368)

10
EP

5
C

0
AC

20
(e) 1.42 Pa
15 (n=12,874)

10
5
0
1 2
10 10 0.07 0.1 0.2 0.3 0.4 0.5
Floc size (μm) Turbulence inte
ACCEPTED MANUSCRIPT

PT
RI
SC
(a)
U
AN
0.54 Pa

43.73 μm
(0.36, 0.30, 0.64
M
D

34.92 μm
TE

(0.51, 0.40, 0.66)


EP

(b) 45.18 μm
C

(0.32, 0.44, 0
1.42 PaC
A

35.79 μm
(0.77, 0.65, 0.79) 100

Single grains Micro-flocs


ACCEPTED MANUSCRIPT

PT
RI
1 0.54 Pa

SC
0.8 1.42 Pa

0.58
0.6
AR

U
0.53
0.4
AN
0.2
M

0
1
D

0.8
Shape factors

TE

0.6
CI

0.43
EP

0.4 0.38

0.2
C

0
AC

1
0.8 0.73

0.6 0.63
S

0.4
0.2
0
0 5 10 15 20 25 101 10
2

Fraction (%) Floc size (μm


ACCEPTED MANUSCRIPT

T
IP
CR
(a) + ρe_OBS
1000 ρe_Holo 0.54 P
ρe_Holo 1.42 P

US
800 AN
600 ρe_Holo 4 -1.375
= (3.900 * 10 ) *D
ρ (kg m )
-3

2
r = 0.70
400
M

ρe_Holo 4
= (3.356 * 10 ) *D
-1.312
2
r = 0.65
200
D
e

TE
Cdensity,

3
20 60 100 140 180
10
EP

(b)
Effective
AC

2
10
This study
Markussen and Andersen (2013)
Fennessy et al. (1994)
Manning and Dyer (1999)
Mikkelsen and Pejrup (2001)
Guo et al. (2017)
1
10
20 40 60 100 140 1
Floc size, D (μm)
ACCEPTED MANUSCRIPT
Highlights

• Laboratory experiments were conducted to reveal the influence of shape


factors on effective density.
• Shape factors were determined by LISST-Holo to investigate the
characteristics of flocs.
• Outcomes give insight into means for estimating effective density.

PT
RI
U SC
AN
M
D
TE
C EP
AC

You might also like