Dynamics and sources of colloids in shallow groundwater in lowland wells and fracture flow in sloping farmland

Dynamics and sources of colloids in shallow groundwater in lowland wells and fracture flow in sloping farmland

Accepted Manuscript Dynamics and sources of colloids in shallow groundwater in lowland wells and fracture flow in sloping farmland Wei Zhang, Jian-Hua...

2MB Sizes 0 Downloads 22 Views

Accepted Manuscript Dynamics and sources of colloids in shallow groundwater in lowland wells and fracture flow in sloping farmland Wei Zhang, Jian-Hua Cheng, Qing-Song Xian, Jun-Fang Cui, Xiang-Yu Tang, Gen-Xu Wang PII:

S0043-1354(19)30225-8

DOI:

https://doi.org/10.1016/j.watres.2019.03.012

Reference:

WR 14498

To appear in:

Water Research

Received Date: 16 May 2018 Revised Date:

4 March 2019

Accepted Date: 12 March 2019

Please cite this article as: Zhang, W., Cheng, J.-H., Xian, Q.-S., Cui, J.-F., Tang, X.-Y., Wang, G.-X., Dynamics and sources of colloids in shallow groundwater in lowland wells and fracture flow in sloping farmland, Water Research (2019), doi: https://doi.org/10.1016/j.watres.2019.03.012. 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.

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT

Dynamics and sources of colloids in shallow groundwater in

2

lowland wells and fracture flow in sloping farmland

3

Wei Zhanga,b,c, Jian-Hua Chengb,c, Qing-Song Xianb,c, Jun-Fang Cuib,c,*, Xiang-Yu

4

Tangb,c, Gen-Xu Wangb,c

RI PT

1

5

b

Key Laboratory of Mountain Surface Processes and Ecological Regulation, Institute of Mountain Hazards and Environment, Chinese Academy of Sciences, Chengdu

9

610041, China

10

11

M AN U

University, Chongqing 400067, China

7

8

School of Tourism and Land Resource, Chongqing Technology and Business

SC

a

c

University of Chinese Academy of Sciences, Beijing 100049, China

TE D

6

* Corresponding author: Institute of Mountain Hazards and Environment, Chinese Academy of Sciences, No.

13

9, Block 4, Renminnanlu Road, Chengdu 610041, China. Tel.: +86 28 85213556; fax: +86 28 85222258.

15

AC C

14

EP

12

E-mail address: [email protected] (J.F. Cui).

16

Abstract

17

Field-scale studies of natural colloid mobilization and transport in finely fractured aquifer as well

18

as the source identification of groundwater colloids are of great importance to the safety of

19

shallow groundwater. In this study, the daily monitoring of fracture flow from a sloping farmland 1

ACCEPTED MANUSCRIPT plot and the biweekly monitoring of three lowland shallow wells within the same catchment were

21

carried out simultaneously in 2013. The effects of physicochemical perturbations on groundwater

22

colloid dynamics were explored in detail using partial redundancy analysis, structural equation

23

modeling, Pearson correlation and multi-linear regression analyses. The characterization and

24

source identification of groundwater colloids were addressed via multiple parameters. The daily

25

colloid concentration in the fracture flow varied between 0.54 and 31.90 mg/L (1.64 mg/L on

26

average). Unique periods of high colloid concentration (5.59 mg/L on average) occurred during

27

the initially generated flow following the dry season. In comparison, a narrower colloid

28

concentration range of 0.24-11.66 mg/L was observed in the lowland shallow wells, with a smaller

29

temporal variation than that of the fracture flow. A low percentage (2.4-7.0%) of colloids and a

30

high percentage (47.7-92.0%) of coarse particles (2-10 µm) were present in the lowland well water.

31

Hydraulic perturbation by rainwater infiltration in the sloping farmland was the dominant

32

mechanism for colloid mobilization in general; this effect retreated to secondary importance

33

behind chemical perturbations (pH, Mg2+ and DOC) at low flow discharges (<1.3 L/min). In

34

contrast, water chemistry (e.g., EC, cations and DOC concentrations) exhibited a major effect on

35

colloid dynamics in the water of the lowland wells, except for the extremely high-salinity water of

36

one well, in which water temperature showed a negative dominant influence on colloid stability.

37

The combined use of multiple parameters (e.g., mineral composition and organic matter, calcium

38

carbonate and δ13C contents) traced groundwater colloids to the shallow soil in the upper

AC C

EP

TE D

M AN U

SC

RI PT

20

2

ACCEPTED MANUSCRIPT farmlands. It is strongly advised that in finely fractured aquifers within agricultural catchments,

40

not only the small colloids but also the coarse particles in the size range of 2-10 µm should be

41

monitored in case of colloid-associated contamination from agricultural wastes e.g., N, P,

42

pesticides and/or heavy metals, especially at the early stages of the rainy seasons.

RI PT

39

43

Keywords: Fracture flow, groundwater, colloid transport, source identification

45

1. Introduction

46

Mass transfer in the vadose zone has been a major concern in transport studies of

47

different geospheres in general. In particular, the frequent temporal and spatial

48

variations in physical and chemical conditions in the vadose zone exert a great effect

49

on the fate (e.g., release, mobilization and transport) of water dispersible colloids.

50

Colloids include fine particles of several micrometers to macromolecules of several

51

nanometers and are ubiquitous in the subsurface (Liu et al., 2018). The release of

52

colloids from a solid surface is determined mechanically by the interactions of

53

repulsive and attractive forces between the colloid and the solid surface. Only when

54

the repulsive forces overcome the attractive forces can colloids be detached and

55

mobilized (Degueldre and Benedicto, 2012; Torkzaban et al., 2010; Zhang et al.,

56

2012). Mobilized colloids are susceptible to transport into groundwater, during which

57

colloid-facilitated transport of various strongly sorbing contaminants (e.g.,

AC C

EP

TE D

M AN U

SC

44

3

ACCEPTED MANUSCRIPT phosphorus, nitrogen, pharmaceuticals, pesticides, heavy metals and radionuclides)

59

has been frequently reported (Delwiche et al., 2014; Ge et al., 2018; Judy et al., 2018;

60

Missong et al., 2018; Mohanty et al., 2014; Tran et al., 2018; Xing et al., 2015). In

61

particular, pathogens, e.g., viruses and bacteria, either occur in groundwater in bio-

62

colloid forms (Zhang et al., 2012) as in most cases, or are associated with inorganic

63

colloids in wastewater (Walsge et al., 2010). Both forms have been demonstrated to

64

move through shallow aquifers with little retention and are present in groundwater

65

due to colloid-facilitated transport (Göppert and Goldscheider, 2008; Maciopinto et al.,

66

2008), which may lead to the outbreak of waterborne diseases. Therefore, it is of great

67

importance to explore subsurface colloid transport characteristics, which provides a

68

pre-requisite for experimental and modeling research on colloid-facilitated

69

contaminant transport.

70

Numerous studies concerning colloid mobilization and subsequent transport, as well

71

as the facilitated transport of various contaminants by colloids, have been carried out

72

in the laboratory, e.g., in repacked/intact columns and fractured cores (Ge et al., 2018;

73

Kuno et al., 2002; Liu et al., 2018; Mishurov et al., 2008; Mitropoulou et al., 2013;

74

Mohanty et al., 2014; Mondal and Sleep, 2012; Morales et al., 2011; Tran et al., 2018;

75

Walshe et al., 2010). The flow rate, pH, ionic strength, divalent cations, dissolved

76

organic matter, etc., of a solution have been widely reported to affect colloid

AC C

EP

TE D

M AN U

SC

RI PT

58

4

ACCEPTED MANUSCRIPT mobilization, transport and the association of contaminants to colloids in these studies.

78

However, the ideal or synthetic colloid tracers (e.g., nanoparticles or microspheres) of

79

specific sizes, shapes, densities and surface properties that were frequently used in

80

these studies do not resemble the various natural colloids present in groundwater. In

81

addition, colloid mobilization and transport in a column/core with a relatively

82

homogeneous structure or a steady state of flow may exhibit large discrepancies from

83

those observed in field scenarios. Seaman et al. (2007) demonstrated that colloid

84

mobilization and transport in an aquifer did not follow shear force predictions from

85

well-controlled column-scale studies. This discrepancy leads to the necessity for the

86

field-scale study of the mobilization and transport of naturally occurring colloids in

87

groundwater.

88

At the field scale, water flow and organic carbon were observed to have a profound

89

effect on colloid generation in groundwater (Kaplan et al., 1993). The released colloid

90

load in shallow aquifers was also found to increase with the pH of recharge water,

91

especially when it exceeded the pHpzc of the inorganic colloids (Rebacca et al. 2002).

92

The authors also reported that the decrease in ionic strength did not facilitate

93

significant colloid mobilization, which differs from that observed in many column-

94

scale studies (e.g., Kuno et al., 2002; Mitropoulou et al., 2013; Mondal and Sleep,

95

2012). Colloid mobilization and transport in response to natural rainfall and artificial

AC C

EP

TE D

M AN U

SC

RI PT

77

5

ACCEPTED MANUSCRIPT irrigation events were also observed in vadose zone lysimeters and were partially

97

attributed to the moving air-water interfaces in the unsaturated aquifer (Liu et al.,

98

2013). A similar quick response of natural colloid mobilization and transport in

99

fracture flow due to hydraulic scouring and the movement of the air-water interface

100

has been demonstrated in a fractured mudstone (Zhang et al., 2015, 2016) and in a

101

fractured karst aquifer (Pronk et al., 2009). In addition, the quick mobilization and

102

transport of colloids within the first several hours following flow infiltration as well

103

as the occurrences of lower colloid concentration peaks several days later in fractured

104

chalk have been reported (Weisbrod et al., 2002). These results imply the potential for

105

large-load and long-distance transport of colloids in groundwater originating from

106

macroporous or fractured media where preferential flow pathways prevail. However,

107

colloid mobilization in shallow aquifers, for example, induced by water injection,

108

showed significant spatial and temporal heterogeneities due to the local hydrogeologic

109

conditions (Seaman et al., 2007). Moreover, the complex interplay between physical

110

perturbations of rainfall infiltrations and chemical reactions (e.g., water pH and ionic

111

strength) contributed to colloid load in well water in karst aquifers (Shevenell and

112

McCarthy, 2002). However, the relative importance of physical and chemical

113

perturbation has not yet been identified.

114

Nevertheless, the existing field-scale studies of natural colloid mobilization and

AC C

EP

TE D

M AN U

SC

RI PT

96

6

ACCEPTED MANUSCRIPT transport in the subsurface were based on single events (e.g., natural or artificial

116

rainfall, irrigation, groundwater recharge). The long tails and erratic pulses of colloid

117

concentrations observed during the drainage process in these field-scale studies

118

further indicate the importance of the investigation of colloid mobilization and

119

transport over a longer time scale. In addition, the re-mobilization or re-suspension of

120

the clogged/coagulated colloids and the increased load of small colloids caused by the

121

breakdown of large aggregates may also occur due to the flow regime and the

122

resultant variations in water chemistry induced over time, especially under fluctuating

123

natural weather conditions (Göppert and Goldscheider, 2008). Continuous field-scale

124

investigations of natural colloids and groundwater/flow dynamics in response to

125

various rainfall events in different geological media are necessary to better understand

126

the temporal and spatial variations of colloid transport in a large field. Various

127

geological media (e.g., soils, weathered rocks, sediments, till deposits, epikarst) could

128

act as the sources of natural colloids present in groundwater, but contribute different

129

colloid loads (Albarran et al., 2014; Degueldre and Cloet, 2016; McCarthy and

130

McKay, 2004; Pronk et al., 2009; Schiperski et al., 2016). In addition, field studies

131

have indicated that colloid transport in response to water infiltration showed site-

132

specific characteristics (e.g., breakthrough time, recovery, peak colloid concentration)

133

(Göppert and Goldscheider, 2008; Zhang et al., 2016). Therefore, it is also important

AC C

EP

TE D

M AN U

SC

RI PT

115

7

ACCEPTED MANUSCRIPT to trace the sources of natural colloids to more accurately evaluate the potential of

135

colloid-facilitated contaminant transport. Particle size distribution (PSD) was used for

136

the source identification of autochthonous particles inside an aquifer and

137

allochthonous particles derived from the land surface in karst groundwater (Pronk et

138

al., 2007). Morphological and mineralogical evidence is also useful for tracking the

139

potential sources of suspended sediments in surface water (Chanudet and Filella, 2008;

140

Mills et al., 2017) and shallow groundwater (Filella et al., 2009; Herman et al., 2007).

141

However, the feasibility of these technologies for groundwater colloid source

142

identification needs to be tested, and other technologies are still inadequate. 13C stable

143

isotope analysis of organic matter provides information about carbon transfer between

144

different geomedia with distinct carbon isotope signatures (Engelmann et al., 2018).

145

This parameter has been increasingly applied in recent decades in the source

146

identification of sediment from soils of different land uses within a catchment (Gibbs,

147

2008; Guan et al., 2017). However, no application of

148

identification has been conducted.

149

The vast (160, 000 km2) hilly region of central Sichuan in the upper reaches of the

150

Yangtze River is characterized by thin purple soil cover and underlying finely

151

fractured mudrock. Frequent occurrences of preferential flow have been demonstrated

152

in finely fractured aquifer in this region (Zhao et al., 2013); as a result, the mobilized

13

C in colloid source

AC C

EP

TE D

M AN U

SC

RI PT

134

8

ACCEPTED MANUSCRIPT colloids in the subsurface were enriched in the fracture flow (Zhang et al., 2015, 2016)

154

and could be transported to lowland drinking wells.

155

In this study, a sloping farmland plot (0.15 hm2) located in a relatively high part of the

156

Jieliu catchment and three lowland shallow wells along the streamline to the

157

catchment outlet were selected. Fracture flow from the hillslope and well water from

158

the lowland wells were investigated simultaneously for a year to investigate the

159

mechanisms regarding the mobilization and transport of natural colloids. Daily and

160

biweekly monitoring were carried out for the fracture flow and the well water,

161

respectively. The relationship between the dynamics of colloid concentration and the

162

dynamics of physiochemical perturbations in the shallow groundwater was analyzed

163

and discussed thoroughly using partial redundancy analysis (pRDA), structural

164

equation modeling (SEM), Pearson correlation and multi-linear regression analysis. In

165

addition, the characterization and source identification of groundwater colloids were

166

carried out by multi-parameter comparisons between the colloids and nearby

167

geological samples (soils and weathered mudrocks from the upper farmlands and/or

168

forests). To the best of our knowledge, this is the first study to quantitatively

169

characterize the interplay of physical and chemical factors in natural colloid dynamics

170

for different shallow groundwater patterns (e.g., fracture flow and lowland well water)

171

by a combination of pRDA and SEM. In addition, exploring the reliability and

AC C

EP

TE D

M AN U

SC

RI PT

153

9

ACCEPTED MANUSCRIPT 13

validity of coupling

C stable isotope with other chemical parameters (e.g., mineral

173

composition, organic matter and calcium carbonate) to identify groundwater colloid

174

sources is expected to provide new insights regarding colloid source identification.

175

The in-situ investigation results presented in this study may provide theoretical and

176

modeling information regarding natural colloid mobilization and transport in large

177

fields (e.g., catchment-scale) and may also help protect shallow groundwater from

178

potential contamination by colloids and colloid-associated agricultural wastes, in not

179

only finely fractured media but also karst aquifers elsewhere.

180

2. Materials and methods

181

2.1 Experimental site

182

The study site is located in the small Jieliu catchment (Fig. 1), which has an area of

183

0.35 km2 and is located in hilly central Sichuan Province, Southwest China. The

184

annual mean temperature and rainfall from 1981 to 2006 in this area were 17.3 °C and

185

826 mm, respectively. The cumulative rainfall from May to September accounted for

186

85% of the annual precipitation. Sloping farmland is the major land use type (a

187

proportion of 44%) in the catchment, with an average slope of approximately 6°. The

188

typical soil in the catchment is a readily erodible purple soil with abundant

189

macropores, which could contribute over 87% of the flow (Wang et al., 2015). The

190

soil thickness on the slopes varies from 25 cm to 60 cm. Mudrock with visible fine

AC C

EP

TE D

M AN U

SC

RI PT

172

10

ACCEPTED MANUSCRIPT fractures lies beneath the shallow purple soil and overlies an impermeable sandstone.

192

Other land use types of the catchment include forests (woods) (35%), paddy fields

193

(11%) and residential areas (10%). Considering the application of agricultural

194

chemicals to farmlands and their potential rapid transport via preferential flow in

195

dissolved and colloid-associated forms, the subsurface transport of colloids from a

196

sloping farmland plot on the upper slope and the dynamics of colloid concentrations

197

in three lowland shallow wells were investigated in this study.

198

2.2 Methods

199

2.2.1 Monitoring of the sloping farmland plot

200

A farmland plot (0.15 hm2) with an average slope of 6° was constructed on the upper

201

slope in the catchment (Fig. 1). Details on the experimental plot can be found in

202

Zhang et al. (2015). Briefly, rainfall was recorded automatically at 15-min intervals

203

by a tipping bucket. A conflux groove was constructed in a trench at the lower end of

204

the profile of the plot to collect the fracture flow from the interface of the fractured

205

mudrock and the impermeable sandstone. Fracture flow discharge was measured by a

206

customized tipping bucket gauge at 15-min intervals. Almost no fracture flow occurs

207

during the dry season from early December to the following April. During the rainy

208

season from May to November, fracture flow is the dominant flow type (Zhao et al.,

209

2016).

AC C

EP

TE D

M AN U

SC

RI PT

191

11

ACCEPTED MANUSCRIPT In this system, fracture flow was sampled daily with a syringe pump at the outlet of

211

the conflux groove as the flow occurred at the beginning of the rainy season in 2013.

212

In cases of large rainfall events, sampling of the fracture flow was carried out twice

213

per day at 08:00 am and 20:00 pm. The whole sampling period lasted for eight months

214

from May through December until fracture flow ceased. All the flow samples were

215

stored in glass bottles prior to analysis. After the flow sampling period, surface

216

sediment (≤5 cm) in a small settling pond (200 cm long, 100 cm wide and 150 cm

217

deep) receiving fracture flow was collected manually in January 2014. The sediment

218

was dispersed in the fracture flow water and sonicated in a water bath sonicator (KQ-

219

3000VDE, Huqin Equipement Co., Ltd., Shanghai, China) for 30 min before settling.

220

The suspension with colloids (<2 µm) was recovered by the pipette method according

221

to Stokes’ law and was subsequently oven dried (60 °C). In addition, soil and rock

222

samples were collected in the plot. The sampling depth was 0-15 cm (plough layer)

223

for the soil samples and 5 cm below the interface between the soil layer and the

224

weathered mudrock for the rock samples. Three upper locations near the trench were

225

sampled. The samples were air-dried and sieved through 150 µm for analysis.

226

2.2.2 Monitoring of the well system

227

Five wells were constructed and are used as the only sources of drinking water for the

228

local residents. In the lower part of the catchment, three of these wells were selected

AC C

EP

TE D

M AN U

SC

RI PT

210

12

ACCEPTED MANUSCRIPT and monitored for water and colloid dynamics (Fig. 1). Two wells (i.e., the

230

Zhaoxingqiang well and Zhangfei well) that exhibit different perturbations were

231

chosen. The Zhaoxingqiang well is located at the footslope, and the water is extracted

232

only via bailing. The Zhangfei well is located in the valley close to the outlet of the

233

catchment, and the water is extracted via pump withdrawal. In addition, a third well

234

(the Xinjing well) that experiences no human perturbations and that is located in the

235

woods on the mid-slope below the sloping farmland plot was also selected as a

236

monitoring well. Given the much lower fluctuations of the well water compared with

237

the fluctuations of fracture flow in the sloping farmland plot, water was sampled

238

biweekly from the three wells. The water level (recorded as the depth below the

239

ground) and temperature were measured manually while sampling. Water samples

240

from the wells were collected by a low flow rate syringe pump at 30-50 cm below the

241

water surface. The sampling period lasted for one year starting in January 2013. After

242

the last water sampling, surface sediments (≤5 cm) at the bottoms of the Xinjing and

243

Zhaoxingqiang wells were collected through core sampling in January 2014. The

244

collected sediment samples were dispersed in the well water, and colloid fractions

245

were recovered using the same method that was described for the sloping farmland

246

plot. Similarly, soil and rock samples from the farmland and/or forestland at the mid-

247

slope and foot slope close to each well (approximately 5 m away) were collected,

AC C

EP

TE D

M AN U

SC

RI PT

229

13

ACCEPTED MANUSCRIPT each with three replicates. The soil sampling depths were 0-15 cm and 35-40 cm for

249

the farmland and 0-10 cm for the forest. The sampling depth for the rock layer was 5

250

cm below the soil-rock interface. The samples were treated in the same way as the soil

251

and rock samples collected from the sloping farmland plot.

252

2.2.3 Sample analysis

253

Fracture flow and well water samples were analyzed for water chemistry and colloid

254

concentrations. pH, electrical conductivity (EC) and colloid concentration were

255

determined within 12 h after sampling. pH and EC were measured by a pH and EC

256

meter (SensION+MM150, Hach Company, Loveland, CO., USA). The PSD of the

257

groundwater samples was measured by a laser scattering PSD analyzer (LA950,

258

Horiba, Ltd., Kyoto, Japan); detailed information on this process can be found in

259

Zhang et al. (2016). Colloid numbers in the groundwater samples were calculated by

260

Eq. 1 (Supporting Information). Colloid absorbency in all the water samples was

261

determined for colloid concentration by a spectrophotometer (Tu-1810, Purkinje

262

General Instrument Co., Beijing, China) at a wavelength of 400 nm using a calibration

263

curve of the colloid mass concentration (mg/L) against absorbance. Additional

264

absorbance and PSD measurements were carried out on the water samples after

265

ultrasound treatment (2 min, 100 W) in a water bath sonicator. Then, the water

266

samples were filtered through 0.45-µm filters and used for the determination of the

AC C

EP

TE D

M AN U

SC

RI PT

248

14

ACCEPTED MANUSCRIPT dissolved organic carbon (DOC) and divalent cation (Ca2+ and Mg2+) concentrations.

268

The DOC concentration was measured by a continuous flow analyzer (Auto Analyzer

269

3, SEAL Analytical, Norderstedt, Germany). Cation concentrations were determined

270

by ion chromatography (ICS-900, Dionex, Sunnyvale, CA, USA).

271

The colloids that were separated from the surface sediments collected at the fracture

272

flow outlet and well bottoms, as well as the soil and rock samples from the two

273

systems, were analyzed for organic matter (OC) content, carbonate content, mineral

274

composition and δ13C content. The OC and carbonate contents were measured by the

275

potassium dichromate method and titration method, respectively. The mineral

276

composition was determined by X-ray diffraction (D/max-2500, Rigaku Corporation,

277

Japan). The δ13C content was determined by an isotope ratio mass spectrometer

278

(IRMS Delta V Plus, Thermo Scientific, USA) coupled to an elemental analyzer (1112

279

Flash EA, Thermo Scientific, USA).

280

2.3 Statistical analysis

281

Partial redundancy analysis (pRDA), performed by variation partitioning analysis

282

(VPA) using “vegan” in R, was conducted to identify the quantitative variation in the

283

contribution of flow discharge/well depth and the physicochemical groundwater

284

parameters to colloid dynamics. pRDA has already been used in water quality studies

285

by other researchers (Buffam et al., 2016; Heier et al., 2010). Moreover, structural

AC C

EP

TE D

M AN U

SC

RI PT

267

15

ACCEPTED MANUSCRIPT equation modeling (SEM), performed by “lavaan” in R (Rosseel, 2012), was

287

conducted to further test the relationships between colloid dynamics in the fracture

288

flow and the rainfall hydraulics and flow chemistries. Multiple goodness-of-fit criteria,

289

including the χ2 test (P>0.05), the goodness-of-fit index (GFI>0.90) and the root mean

290

square error of approximation (RMSEA<0.08), were used in the SEM. In addition,

291

Pearson correlation and multi-linear regression (stepwise method) analyses were used

292

in combination to evaluate the relationship between colloid concentration dynamics

293

and flow/water physicochemical properties. Moreover, one-way ANOVA (S-N-K test

294

for post hoc multiple comparisons, P<0.05) was applied in the significance difference

295

tests for the OC, carbonate and δ13C contents of the environmental samples. The

296

software package SPSS 17.0 and Origin 8.0 for Windows were used for these

297

statistical analyses.

298

3. Results and discussion

299

3.1 Dynamics of fracture flow and well water

300

In 2013, fracture flow did not occur until the onset of a rain event on 1 May

301

(rainfall=17.6 mm, Imax=5.6 mm/h) (Fig. 2). Afterwards, fracture flow lasted

302

throughout the course of the rainy season and ceased in December. Generally, distinct

303

peaks of fracture flow discharge occurred following the perturbations caused by rain

304

events. Specifically, sharp peaks of fracture flow discharge occurred during heavy

AC C

EP

TE D

M AN U

SC

RI PT

286

16

ACCEPTED MANUSCRIPT storms only (e.g., on 18 July, 22 July and 19 September). Descriptive statistics of the

306

hydraulic and physicochemical parameters of the fracture flow are presented in Table

307

1. The fracture flow discharge showed the largest variation (CV=1.87), between 0.089

308

and 67.22 L/min, and the corresponding daily flow depth varied between 0.004 and

309

14.64 mm/day (CV=1.31). The well-developed preferential pathways (i.e., soil

310

macropores and mudrock fractures) facilitated rapid rainwater infiltration and

311

subsurface flow recession, resulting in a substantial variation in flow discharge within

312

a short time in response to a rain event. The fracture flow was chemically neutral to

313

alkaline. Flow pH showed little variation (CV=0.05) between 6.75 and 8.81, which

314

could be ascribed to the buffering capacity of the calcareous purple soil in this region.

315

The EC of the fracture flow varied between 441 and 769 µS/cm. The mixing of the

316

dilute rainwater (EC<30 µS/cm) with the mobile soil pore water was responsible for

317

the decrease in fracture flow EC. Divalent cation (Ca2+ and Mg2+) concentrations in

318

the fracture flow showed moderate variations. The average Ca2+ and Mg2+

319

concentrations were 83.36 mg/L and 15.70 mg/L, respectively, and were in a similar

320

range as those of the fracture flow from the karst limestone aquifers (Atteia et al.,

321

1998; Yang et al., 2013). The DOC concentrations of the fracture flow showed greater

322

variation (CV=0.46) between 0.28 and 3.82 mg/L. Higher DOC concentrations were

323

generally observed in the summer (from June to September), when the intensive dry-

AC C

EP

TE D

M AN U

SC

RI PT

305

17

ACCEPTED MANUSCRIPT wet cycles facilitated the release of soil organic carbon into the soil pore water and

325

subsequently appeared in the fracture flow.

326

The water levels in the Xinjing and Zhaoxingqiang wells gradually decreased to the

327

lowest levels at the end of the dry season (April 2013) (Figs. S1-S2, Supporting

328

Information). In contrast, the water level in the Zhangfei well, which is located near

329

the catchment outlet, showed obvious fluctuations (Fig. S3) that could have been

330

caused by the frequent pumping of drinking water. With the arrival of the rainy season

331

in early May, the water levels of the three wells gradually increased and exhibited the

332

first peaks in late June, while other distinct peaks also occurred after heavy storms. In

333

comparison to the distinct peaks of fracture flow discharge observed in the sloping

334

farmland plot, the three wells exhibited fewer water level peaks but with much longer

335

durations, indicating the slower and lagged response of the shallow well water to rain

336

events.

337

The water temperature of the Xinjing and Zhaoxingqiang wells varied between 10.1

338

and 25.1 °C (Table S1), with an obvious peak occurring in August (Figs. S1-S2). The

339

water temperature of the Zhangfei well remained at a plateau of approximately 19 °C

340

from early August to early November (Fig. S3), which is much lower than those of the

341

other two wells. The smallest variation in water temperature in this well (CV=0.07,

342

Table S1) was due to its largest depth below ground (>4 m), as compared to the depths

AC C

EP

TE D

M AN U

SC

RI PT

324

18

ACCEPTED MANUSCRIPT of the other two wells, so that the deeper groundwater was less affected by air

344

temperature. The pH of the well water was in the same range as that of the fracture

345

flow, but the EC of the well water was significantly higher (P<0.01) than that of the

346

fracture flow. Both the EC and the major divalent cation concentrations (Ca2+ and

347

Mg2+) of water from the Zhangfei well were significantly higher than those from the

348

other two wells (P<0.01), probably due to the release of cations from thicker soil and

349

rock layers into the groundwater. The DOC concentration of the Xinjing well water

350

increased rapidly in early May and reached a maximum (11.8 mg/L) in mid-June (Fig.

351

S1). The increase in the DOC concentration was mainly ascribed to the early leaching

352

of the organic carbon pool in the shallow soil, which was supplied by decaying wood

353

litter that had accumulated in the dry season. A lower average DOC concentration was

354

observed in the Zhaoxingqiang well (Fig. S2), and the lowest level was found in the

355

Zhangfei well (Fig. S3), indicating the retention of DOC by the soil during its

356

downward transport.

357

3.2 Dynamics of colloids in shallow groundwater

358

The daily colloid concentration of the fracture flow varied from 0.54 to 31.90 mg/L,

359

with an average concentration of 1.64 mg/L (Table 1). A high colloid concentration

360

lasted approximately 1 week, with an average concentration of 5.59 mg/L observed

361

after the commencement of the fracture flow on 1 May (Fig. 2). This phenomenon

AC C

EP

TE D

M AN U

SC

RI PT

343

19

ACCEPTED MANUSCRIPT was unique and specific to the very beginning of the rainy season following a long dry

363

period, during which the differential capillary stresses might lead to soil matrix and/or

364

soil pores weakening and further colloid generation (Michel et al., 2010). The

365

abundant mobilized colloids were continuously transported to the fracture flow due to

366

the advancement of the wetting front of the newly infiltrated rainwater, which was

367

responsible for the clustering of high colloid concentration. This finding is consistent

368

with other findings from column studies (Majdalini et al., 2008; Mohanty et al.,

369

2015b). Afterwards, the colloid concentration generally stayed at a lower base level

370

(1.15 mg/L) during the non-rainy days, except for a few peaks that were several or

371

even tens of times the base level in response to heavy storms. The highest colloid

372

concentration was 31.90 mg/L and was observed during the storm on 19 September

373

(rainfall=92.2 mm, Fig. 2). The second highest colloid concentration was 27.50 mg/L

374

and was observed during the storm on 22 July (rainfall=120.1 mm, Fig. 2). The

375

durations of elevated colloid levels (less than one day) were shorter than those of

376

elevated flow discharges (several days), which is consistent with the reported faster

377

increases in colloid concentration compared to those of the subsurface flow discharge

378

during rain events in a number of field studies (El-farhan et al., 2000; McKay et al.,

379

2000; Schiperski et al., 2016; Toran et al., 2006; Zhang et al., 2016).

380

In the three lowland wells, the colloid mass concentration varied between 0.20 and

AC C

EP

TE D

M AN U

SC

RI PT

362

20

ACCEPTED MANUSCRIPT 11.66 mg/L, and the corresponding number concentration varied between 1.62×107

382

and 1.07×109 colloids/L (Table 2). The colloid populations observed in this study were

383

1-3 orders of magnitude lower than those in the granitic groundwater from the

384

Grimsel Test Site (Degueldre et al., 1989) and in coastal groundwater (Rani and

385

Sasidhar, 2011). Colloid concentrations of the groundwater in the wells varied in a

386

much smaller range and exhibited much lower peak concentrations and less temporal

387

variations than those of the fracture flow from the sloping farmland. In addition,

388

decreasing trend of the average colloid concentration from 2.72 mg/L in the Xinjing

389

well to 1.42 mg/L in the Zhangfei well close to the catchment outlet was observed,

390

which could be attributed partially to the increased retention of subsurface colloid

391

transported along the streamline. A similar colloid concentration decline with distance

392

in the down-gradient wells was previously reported (McKay et al., 2000). However,

393

the temporal dynamics of the colloid concentration showed different patterns among

394

the three wells. In the Xinjing well which was not subjected to human activity, the

395

colloid mass concentration increased gradually from less than 2.00 mg/L at the end of

396

the dry season to a peak of 7.62 mg/L in mid-June, followed by a rapid decrease to

397

below 2.00 mg/L (Fig. S1). This observation is similar to observations from the

398

sloping farmland plot in which colloid concentration increased to a plateau in

399

response to rain events at the early rainy season. However, this is not the case for the

AC C

EP

TE D

M AN U

SC

RI PT

381

21

ACCEPTED MANUSCRIPT other two wells. In the Zhaoxingqiang well, a distinct colloid concentration peak

401

(11.66 mg/L) occurred on 2 July in response to the previous consecutive storms (30

402

June to 1 July, cumulative rainfall of 203.2 mm) (Fig. S2). A much smaller response

403

of the colloid concentration to these storms was observed in the Zhangfei well (Fig.

404

S3). However, a secondary colloid concentration peak in the Zhaoxingqiang well and

405

major peaks in the Zhangfei well appeared during the dry season (by the end of April),

406

when water levels were very low and continuously declined in both wells. These

407

peaks were mainly the result of perturbations related to water extraction. The frequent

408

bailing from the Zhaoxingqiang well and pump withdrawal of water from the

409

Zhangfei well at low water levels facilitated the re-suspension of previously deposited

410

colloids at the well bottoms. Similarly, Rani and Sasidhar (2011) reported that the

411

physical perturbations of water extraction led to the artifact of increased colloid

412

concentrations in the wells. In addition, sediment re-suspension has been reported to

413

contribute up to 55% of the natural colloids suspended in the water column in a lake

414

(Xu et al., 2018).

415

The effects of the ultrasound treatment of the water bath on the colloid concentration

416

of the groundwater samples are presented in Figs. S4-S5. The Cus/C ratios were all

417

higher than 1:1 (the dotted line), regardless of the fracture flow samples or well water

418

samples, indicating that the colloids in the shallow groundwater existed as aggregates.

AC C

EP

TE D

M AN U

SC

RI PT

400

22

ACCEPTED MANUSCRIPT In the present study, the high electrical conductivities and particularly high Ca2+

420

concentrations (57.04-107.03 mg/L for the fracture flow and 77.00-207.03 mg/L for

421

the well water) that far exceeded the critical coagulation concentration (CCC) of

422

purple soil (24 mg/L) and mudrock (16 mg/L) particles (Zhang et al., 2016) would

423

have extensively neutralized the negative charges in the diffuse electric double layer

424

(EDL) of the colloids. The resulting compression of the EDL further promoted the

425

colloid coagulation. Similar findings of divalent cation-induced colloid coagulation

426

were reported by other studies (Kuno et al., 2002; Shiyan et al., 2014; Xu et al., 2018).

427

The PSD of fracture flow samples from the hillslope has been explored in our

428

previous single-event based studies (Zhang et al., 2015, 2016). Suspended particles in

429

the fracture flow ranging in size from submicron to more than 100 µm were observed

430

while the colloid (<2 µm) presence was lower than 1%. However, the PSD of the well

431

water exhibited a different pattern from that of the fracture flow, as shown in Fig. 3. A

432

smaller size range from submicron up to 40 µm, a slightly higher colloid percentage

433

(2.4-7.0%) and smaller median sizes (4.2-10.0 µm) were observed (Table 3). Low

434

percentages of colloid in the fracture flow and lowland wells were ascribed to colloid

435

diffusion into dead-end voids, such as the soil matrix or fracture apertures, during the

436

subsurface transport process. This matched the observations from other studies, e.g.,

437

Alonso et al. (2007), Mondal and Sleep (2012), Zvikelsky and Weisbrod (2006). The

AC C

EP

TE D

M AN U

SC

RI PT

419

23

ACCEPTED MANUSCRIPT much smaller maximum size of suspended particles and the higher cumulative particle

439

percentage from colloids to coarse particles (2-10 µm) in the well water relative to the

440

fracture flow implies a greater retardation of large particles (>10 µm) during their

441

transport in the catchment. Gravity sedimentation acts as the main cause of

442

immobilization of large particles, especially in horizontal flow conduits and/or when

443

the flow rate decreases. However, the gravity effect of colloids exhibited an

444

insignificant influence on colloid mobilization and transport in finely fractured aquifer

445

in the present study, considering the intense driving force of macropore flow and/or

446

fracture flow that enabled the transport and enrichment of coarse particles in the well

447

water. Monhanty et al. (2015a) also reported the irrelevance of gravity sedimentation

448

on colloid transport, as the 1.8-µm colloids recovered to a greater extent than the 0.5-

449

µm colloids during transport in a macroporous soil. The environmental implications

450

of the PSD results for regions characterized by finely fractured aquifers are that

451

particle investigation of the groundwater in these areas should include but not be

452

limited to colloids; coarse particles up to 10 µm should also be taken into account.

453

Apparent shifts towards smaller groundwater PSDs following ultrasound treatment

454

are also shown in Fig. 3. The colloid percentage increased by a factor of 3.0, 2.5 and

455

13.8 in the Xinjing well, Zhaoxingqiang well and Zhangfei well, respectively.

456

Breakage of the coagulated colloids, especially in the high-salinity Zhangfei well, led

AC C

EP

TE D

M AN U

SC

RI PT

438

24

ACCEPTED MANUSCRIPT to the increase in the colloid presence following the ultrasound treatment.

458

3.3 Physicochemical perturbations of colloid dynamics in the fracture flow and

459

shallow groundwater

460

Due to the large temporal variation in the fracture flow discharge (CV=1.87, Table 1),

461

it is reasonable to separate the flow discharge into high and low ranges according to

462

the median flow discharge (1.3 L/min) for further detailed analyses via pRDA,

463

Pearson correlation and multi-linear regression analysis.

464

According to the pRDA, a total of 67.84% and 84.28% of the variations in colloid

465

dynamics in the fracture flow could be explained for the general cases and the high

466

flow discharge cases, respectively, to which flow discharge contributed more than 45%

467

(Fig. 4). This finding indicates the dominance of flow discharge, instead of flow

468

chemistry, on colloid mobilization and transport in the fracture flow from the hillslope.

469

SEM was employed to explore the routes and extensions of the physicochemical

470

factors on colloid dynamics in the fracture flow. Our model explained 60% of the

471

variance in colloid dynamics in the fracture flow (Fig. 5). Flow discharge can impact

472

colloid dynamics directly (λ=0.70, P<0.001) and can also impose an indirect effect on

473

colloid dynamics by affecting flow chemistry (λ=-0.25, P<0.01). In contrast, flow

474

chemistry showed a half less importance of direct impact (λ=0.31, P<0.001) on

475

colloid dynamics. This is in accordance with pRDA in which flow discharge acted as

AC C

EP

TE D

M AN U

SC

RI PT

457

25

ACCEPTED MANUSCRIPT the primary and positive factor influencing colloid dynamics.

477

In addition, the colloid concentration was significantly positively correlated with the

478

fracture flow discharge, with a Pearson correlation coefficient of 0.756, which is

479

much greater than that of the chemical parameters (Table 4). Similarly, flow discharge

480

showed the largest influence (53.27%) on colloid mobilization and transport dynamics

481

among all the physicochemical parameters according to the multi-linear regression

482

analysis (Table 4). For the fracture flow samples taken at high discharges (>1.3

483

L/min), an increased correlation (0.885) of colloid dynamics with flow discharge and

484

a slightly larger effect (58.09%) of flow discharge on colloid transport was observed.

485

This result agrees with the results of the pRDA and SEM analyses. The fracture flow-

486

derived shear force, which corresponds to flow discharge and possibly coupled with

487

the air-water interface especially following the dry period, is the main mechanism for

488

colloid scoring, mobilization and transport in the fracture flow. Hydraulic shearing

489

has also been reported as the dominant force controlling colloid dynamics in fractured

490

geologic media elsewhere in response to rainfall (Mohanty et al., 2015a; Shevenell

491

and McCarthy, 2002; Weisbrod et al., 2002).

492

In contrast, for the low flow discharge cases, only 20.73% of the variation in colloid

493

dynamics in the fracture flow was explained and was due nearly exclusively to flow

494

chemistry (Fig. 4). The colloid concentration showed insignificant correlations with

AC C

EP

TE D

M AN U

SC

RI PT

476

26

ACCEPTED MANUSCRIPT flow discharge in this situation (Table 4). The influence of the fracture flow discharge

496

decreased by half more (from >50% to 21.67%, Table 4) and became the secondary

497

factor for colloid mobilization and transport, unlike in the two aforementioned

498

situations. Therefore, the physical perturbation of hydraulic infiltration exhibits a

499

primary effect on colloid mobilization and transport dynamics in the fracture flow

500

from the hillslope in general. In particular, this effect retreats to secondary importance

501

behind flow chemistry (pH, Mg2+ and DOC) when the fracture flow discharge is low

502

(<1.3 L/min). Increased hydraulic retention of fracture flow occurs at low flow

503

discharges, facilitating the interactions of colloids with groundwater. The pH (7.21-

504

8.81, 8.10 on average) of fracture flow at low discharges was significantly higher

505

(P<0.05) than that of the general fracture flow (6.75-8.81, 7.96 on average) and flow

506

at high discharges (6.75-8.74, 7.82 on average). According to the calculations from

507

Zhou et al. (2011), the surface potential (ζ-potential) of soil colloids increased from -

508

49 mV to -68 mV (more negative) when the solution pH increased from 7.0 to 9.0.

509

Therefore, the significantly increased flow pH at low discharges leads to the increase

510

in electrostatic repulsion between colloids and the solid surfaces and further favors

511

colloid mobilization into fracture flow.

512

The depth variations in the well water, which were directly affected by subsurface

513

flow discharge, offered almost no explanations for colloid dynamics in the lowland

AC C

EP

TE D

M AN U

SC

RI PT

495

27

ACCEPTED MANUSCRIPT wells (Fig. 4). Water chemistry accounted for 21.73-70.11% of the explained

515

variations (24.35-77.03%), indicating the dominant role of water chemistry on colloid

516

dynamics in the lowland wells. With respect to water chemistry, EC and Mg2+ in most

517

cases, as well as Ca2+ in certain cases (Tables 4-5), showed a significant negative

518

effect on colloid dynamics in both the well water and fracture flow. This phenomenon

519

agrees with the DLVO theory. With the remarkable decrease in EC or divalent cation

520

concentrations in groundwater, for example due to the incorporation of dilute

521

rainwater, irrigation and/or recharge water, an increased repulsive energy barrier could

522

develop between the dispersed colloids and the solid surfaces (soil matrix and/or

523

pore/fracture surfaces) and thus favor the dispersion and mobilization of colloids

524

(Ryan and Gschwend, 1994; Zhang et al., 2012). Many researchers have reported a

525

negative correlation between the stability of colloids and solution ionic strength in

526

column-scale studies (Albarran et al., 2014; Masciopinto and Visino, 2017; Morales et

527

al., 2011; Zhang et al., 2017) and in stream water in a small watershed (Mills et al.,

528

2017).

529

In contrast, the DOC exhibits a general positive correlation with groundwater colloid

530

concentration and a dominant effect on colloid dynamics in the Xinjing well. The

531

sorption of DOC molecules to natural colloids has been widely recognized, leading to

532

the modified surface negative charges of colloids in most cases (Xu et al., 2018). This

AC C

EP

TE D

M AN U

SC

RI PT

514

28

ACCEPTED MANUSCRIPT modification would promote the electric repulsions between colloids and solid

534

surfaces and further increase the chances of colloid mobilization. Other researchers

535

have reported the promotion of colloid stability by dissolved organic matter in natural

536

groundwater as well as improved colloid transport in the presence of DOC (Cheng

537

and Sailers, 2015; Morales et al., 2011; Shiyan et al., 2014). Moreover, the adsorbed

538

DOC could also block the deposition sites, especially at a high flow rate (Yang et al.,

539

2015), leading to the enhanced mobility of colloids in groundwater. However, no

540

significant correlations between the colloid concentration and water chemistries were

541

found for the Zhangfei well. Instead, there was a distinct negative correlation between

542

the water temperature and the colloid concentration in this well (Table 5), a trend that

543

was not observed in the fracture flow or the other two wells. This observation implies

544

that, provided the highest EC and cation concentrations which are favorable for

545

colloid coagulation in this well, the small increases in water temperature (3.3°C) from

546

May to October could lead to the enhanced Brownian motion of the coagulated

547

colloids. The improved collision of the inter-coagulated colloids potentially promotes

548

the formation of large particles that could easily settle to the well bottom due to

549

gravity sedimentation, resulting in fewer colloids present in the well water. This

550

hypothesis is in agreement with a laboratory study in which a decrease of the stability

551

of the latex colloidal system with increasing of solution temperature was observed

AC C

EP

TE D

M AN U

SC

RI PT

533

29

ACCEPTED MANUSCRIPT (Garcia-Garcia et al., 2006). This relationship does not exist for the other two wells or

553

the fracture flow, attributable to the much lower salinities in these cases.

554

3.4 Groundwater colloid characteristics and source identification

555

Mineral composition and calcium carbonate, organic matter (OC) and δ13C abundance

556

were used to characterize groundwater colloids and soil and rock samples from upper

557

geological media for the identification of colloid sources.

558

A number of minerals (quartz, plagioclase, feldspar, calcite, hematite, montmorillonite,

559

illite, chlorite, etc.) were detected in all three categories of solids but in different

560

proportions (Fig. S6). Quartz and illite were the two major minerals in the soil and

561

underlying mudrock in the sloping farmland plot, while quartz was the dominant

562

mineral in both the soil and mudrock samples collected from the farmlands and

563

forestlands near the wells. At both the sloping farmland site and the well sites, the

564

lowest content of quartz and the highest content of illite were observed in the

565

groundwater colloids and not the soil or mudrock samples. This observation is

566

consistent with the findings of Lei et al. (2004), who also found a higher proportion of

567

illite in the mineral colloids than in the purple soil itself. This is also the case for the

568

colloids in the subsurface water from the adit of an abandoned mine (Filella et al.,

569

2009). In addition, an increase in the percentage of illite in colloids in a river was

570

ascribed to the high load of soil colloids in the watershed during the snow melting

AC C

EP

TE D

M AN U

SC

RI PT

552

30

ACCEPTED MANUSCRIPT period (Chanudet and Filella, 2008).

572

As presented in Fig. S7, groundwater colloids in both the fracture flow and well water

573

had a significantly (P<0.05) higher calcium carbonate content (>25%) than the soils

574

or mudrocks of the nearby geological media. The enriched carbonate content in the

575

groundwater colloids could be explained by the precipitation of calcium carbonates

576

facilitating the physical inter- and intra-cementing of colloids during the process of

577

colloid dispersion and generation (Zhao et al., 2012). In addition, the hematite could

578

also act as the cementing material. As shown in the inset graphs in Fig. S6,

579

groundwater colloids had the highest proportion of hematite, whose presence can

580

facilitate the chemical bonding of organic substances to mineral particles (Schulten

581

and Leinweber, 2000). Therefore, groundwater colloid complexation by dissolved

582

organic matter potentially increases the OC content of colloids.

583

In the sloping farmland plot, the OC content of the colloids (0.91%±0.06%) in the

584

fracture flow showed an insignificant difference (P>0.05) to that of the soil

585

(0.71%±0.13%) while both of them were significantly higher (P<0.05) than that of the

586

rock (0.2%±0.02%) (Fig. S7). Groundwater colloids from both the Xinjing well and

587

Zhaoxingqiang well had significantly higher (P<0.05) OC contents (1.02%±0.11%

588

and 0.96%±0.08%, respectively) than the rocks (<0.21%) of the upper farmlands

589

or forestlands (Fig. S7). These findings might exclude the possibility that fractured

AC C

EP

TE D

M AN U

SC

RI PT

571

31

ACCEPTED MANUSCRIPT rocks in the hillslope or the upper farmlands and forestlands of the lowland wells are

591

the source of the groundwater colloids. In addition, the OC content of the

592

groundwater colloids was slightly higher than that of the shallow soil of the upper

593

farmlands. Considering the increased OC content of colloids due to the complexation

594

of dissolved organic matter in the groundwater, shallow soil from the upper farmland,

595

rather than the forest soil, may act as the major source of groundwater colloids in the

596

present study.

597

In addition to the above parameters, δ13C was also used to help identify the source of

598

the groundwater colloids. In the sloping farmland plot, the δ13C abundance of the

599

colloids (-15.13‰±0.24‰) in the fracture flow was not significantly different

600

(P>0.05) from that of the soil (-14.45‰±0.46‰), but both of them were significantly

601

lower (P<0.05) than that of the underlying mudrock (-9.24‰±0.27‰) (Fig. 6). This

602

result indicates that the colloids in the fracture flow originate from the shallow soil of

603

the sloping farmland, which agrees with the results of the source identification based

604

on the mineral composition, organic matter and calcium carbonate. In the Xinjing well,

605

the δ13C abundance of colloids (-14.61‰±0.18‰) was slightly lower (P>0.05) than

606

that of the shallow soil (-14.34‰±0.33‰) from the upper farmland. These results

607

were markedly different from those of the other geological samples. In the

608

Zhaoxingqiang well, groundwater colloids showed a similar δ13C abundance (-

AC C

EP

TE D

M AN U

SC

RI PT

590

32

ACCEPTED MANUSCRIPT 14.61‰±0.18‰) as that of the shallow soil (-15.27‰±0.07) and underlying mudrock

610

(-14.90‰±0.43‰) of the upper farmland, while all of these samples had markedly

611

lower (P<0.05) δ13C abundances than those of the deep farmland soil and forest soil

612

and rocks. This phenomenon implies little contributions from nearby forestlands to

613

the groundwater colloids in the well. As mudrock has been demonstrated to be an

614

unlikely source of the groundwater colloids, the δ13C abundances also suggest that

615

shallow farmland soil is the major source of colloids present in the lowland wells.

616

This is the first attempt to identify groundwater colloid origination by using δ13C.

617

However, the incorporation of δ13C in colloid source tracing should be combined with

618

other characteristic parameters, e.g., mineral composition and calcium carbonate and

619

organic matter content, for better reliability.

620

4. Conclusion

621

Our study provides quantitative evidence indicating that different mechanisms

622

dominate colloid dynamics in different groundwater patterns. Hydraulic perturbations

623

of rainwater infiltration exhibit a large and direct effect on colloid mobilization and

624

transport via the fracture flow from the upper hillslope. Subsequently, the variations in

625

physicochemical properties of flow water along subsurface transport pathways exert a

626

major effect on colloid dynamics in the lowland wells. These understandings of

627

colloid mobilization and transport mechanisms are crucial to prioritizing the

AC C

EP

TE D

M AN U

SC

RI PT

609

33

ACCEPTED MANUSCRIPT monitoring of flow/water parameters, which may govern colloid dynamics in shallow

629

groundwater. In addition to the high concentrations of colloids (maximums of 31.90

630

mg/L and 11.66 mg/L in the facture flow and well water, respectively), which exist in

631

coagulated forms, a high percentage (47.7-92.0%) of coarse particles (2-10 µm)

632

indicate the importance of including coarse particles in the monitoring scheme of

633

shallow groundwater, especially in finely fractured aquifers and/or karst aquifers.

634

Moreover, the new approach proposed in this study for source identification of

635

groundwater colloids by simultaneous analyses of δ13C and mineral composition,

636

organic matter and carbonate content in groundwater colloids and the upper

637

geological samples would be useful for colloid-facilitated contaminant transport

638

studies in large catchments with spatially varying geological media.

639

Acknowledgements

640

This study was supported by the National Natural Science Foundation of China (Grant

641

Nos. 41790431 and 41601539), the 135 Strategic Program of the Institute of

642

Mountain Hazards and Environment (Grant No. SDS-135-1702) and the Key

643

Research Program of Frontier Sciences, Chinese Academy of Sciences (QYZDJ-

644

SSW-DQC006).

645

References:

646

Albarran, N., Degueldre, C., Missana, T., Alonso, U., García-Gutiérrez, M., López, T., 2014. Size

AC C

EP

TE D

M AN U

SC

RI PT

628

34

ACCEPTED MANUSCRIPT 647

distribution analysis of colloid generated from compacted bentonite in low ionic strength

648

aqueous solutions. Appl. Clay Sci. 95, 284-293.

651 652 653 654 655 656 657 658

rock of a deep geological repository. Phys. Chem. Earth 32, 469-476. Atteia, O., Perret, D., Adatte, T., Kozel, R., Rossi, P., 1998. Characterization of natural colloids from a river and spring in a karstic basin. Environ. Geol. 34, 257-269. Buffam, I., Mitchell, M.E., Durtsche, R.D., 2016. Environmental drivers of seasonal variation in green roof runoff water quality. Ecol. Eng. 91, 506-514.

RI PT

650

Alonso, U., Missana, T., Patelli, A., Rigato, V., 2007. Bentonite colloid diffusion through the host

Chanudet, V., Filella, V., 2008. Size and composition of inorganic colloids in a peri-alpine, glacial flour-rich lake. Geochim. Cosmochim. Ac. 72, 1466-1479.

Cheng, T., Sailer, J.E., 2015. Effects of dissolved organic matter on the co-transport of mineral colloids and sorptive contaminants. J. Contam. Hydrol. 177-178, 148-157.

SC

649

Degueldre, C., Baeyens, B., Goerlich, W., Riga, J., Verbist, J., Stadelmann, P., 1989. Colloids in

660

water from a subsurface fracture in granitic rock, Grimsel Test Site, Switzerland.

661

Geochim. Cosmochim. Ac. 53, 603-610.

663 664 665 666 667

Degueldre, C., Benedicto, A., 2012. Colloid generation during water flow transients. Appl. Geochem. 27, 1220-1225.

Degueldre, C., Cloet, V., 2016. Pore water colloid properties in argillaceous sedimentary rocks. Sci. Total Environ. 569, 423-433.

Delwiche, K.B., Lehmann, J., Walter, M.T., 2014. Atrazine leaching from biochar-amended soils. Chemosphere 95, 346-352.

TE D

662

M AN U

659

668

El-farhan, Y.H., Denovio, N.M., Herman, J.S., Hornberger, G.M., 2000. Mobilization and

669

transport of soil particles during infiltration experiments in an agricultural field,

670

Shenandoah Valley, Virginia. Environ. Sci. Technol. 34, 3555-3559. Engelmann, P., Barbieri, C., Augustin, A., Ketzer, J., Rodrigues, L., 2018. Environmental

672

monitoring of a landfill area through the application of carbon stable isotopes, chemical

673

parameters and multivariate analysis. Waste Manage. 76, 591-605.

675 676 677 678

Filella, M., Chanudet, V., Philippo, S., Quentel, F., 2009. Particle size and mineralogical

AC C

674

EP

671

composition of inorganic colloids in waters draining the adit of an abandoned mine,

Goesdorf, Luxembourg. Appl. Geochem. 24, 52-61.

Garcia-Garcia, S., Jonsson, M., Wold, S., 2006. Temperature effect on the stability of bentonite colloids in water. J. Colloid Interf. Sci. 298, 694-705.

679

Ge, M., Wang, D., Yang, J., 2018. Co-transport of U(VI) and akaganeite colloids in water-

680

saturated porous media: Role of U (IV) concentration, pH and ionic strength. Water Res.

681

147, 359-361.

682 683 684

Göppert, N., Goldscheider, N., 2008. Solute and colloid transport in karst conduits under low-and high-flow conditions. Ground Water 46, 61-68. Gibbs, M.M., 2008. Identifying source soils in contemporary estuarine sediments: A new

35

ACCEPTED MANUSCRIPT 685

compound specific isotope method. Estuar. Coast 31, 344-359.

686

Guan, Z., Tang, X.Y., Yang, J.E., Ok, Y.S., Xu, Z.H., Nishimura, T., Reid, B.J., 2017. A review of

687

source tracking techniques for fine sediment within a catchment. Environ. Geochem. Hlth.

688

39, 1221-1243. Heier, L.S., Meland, S., Ljones, M., Salbu, B., Stromseng, A.E., 2010. Short-term temporal

690

variations in speciation of Pb, Cu, Zn and Sb in a shooting range runoff stream. Sci. Total

691

Environ. 408, 2409-2417.

692 693

RI PT

689

Herman, E.K., Tancredi, J.H., Toran, L., White, W.B., 2007. Mineralogy of suspended sediment in three karst springs. Hydrogeol. J. 15, 255-266.

Judy, J.D., Kirby, J.K., Farrell, M., McLaughlin, M.J., Wilkinson, S.N., Bartley, R., Bertsch, P.M.,

695

2018. Colloidal nitrogen is an important and highly-mobile form of nitrogen discharging

696

into the Freat Barrier Reef lagoon. Sci. Rep-UK 8, doi: 10.1038/s41598-018-31115-z.

698

Kaplan, D.I., Bertsch, P.M., Adriano, D.C., Miller, W.P., 1993. Soil-borne mobile colloids as influenced by water flow and organic carbon. Environ. Sci. Technol. 27, 1193-1200.

M AN U

697

SC

694

699

Kuno, Y., Kamei, G., Ohtani, H., 2002. Natural colloids in groundwater from a bentonite mine -

700

Correlation between colloid generation and groundwater chemistry. Mat. Res. Soc. Symp.

701

Proc. 713, 841-848.

702 703

Lei, J., He, Y., Yu, A., 2004. Relationship between inorganic colloid and fertility of purple soil. J. Sichuan Agr. Univ. 22, 233-238 (in Chinese with an English abstract). Liu, F., Xu, B.L., He, Y., Brookes, P.C., Tang, C.X., Xu, J.M., 2018. Differences in transport

705

behavior of natural soil colloids of contrasting sizes from nanometer to micron and the

706

environmental implications. Sci. Total Environ. 634, 802-810.

707 708

TE D

704

Majdalani, S.M., Di-Pietro, M.E., Angulo-Jaramilli, R., 2008. Effects of wetting and drying cycles on in situ soil particle mobilization. Eur. J. Soil Sci. 59, 147-155. Masciopinto, C., La Mantia, R., Chrysikopoulos, C.V., 2008. Fate and transport of pathogens in a

710

fractured aquifer in the Salento area, Italy. Water Resour. Res. 44, doi:10.1029/2006WR

711

005643.

713 714

Masciopinto, C., Visino, F., 2017. Strong release of viruses in fracture flow in response to a

AC C

712

EP

709

perturbation in ionic strength: Filtration/retention tests and modeling. Water Res. 126,

240-251.

715

Mitropoulou, P.N., Syngouna, V.I., Chrysikopoulos, C.V., 2013. Transport of colloids in

716

unsaturated packed columns: Role of ionic strength and sand grain size. Chem. Eng. J.

717 718 719 720 721 722

232, 237-248. McCarthy, J.F., McKay, L.D., 2004. Colloid transport in the subsurface: Past, present, and future challenges. Vadose Zone J. 3, 326-337. McKay, L.D., Sanford, W.E., Strong, J.M., 2000. Field-scale migration of colloidal tracers in a fractured shale saprolite. Ground Water 38, 139-147. Michel, E., Majdalani, S., Di-Pietro, L., 2010. How differential capillary stresses promote particle

36

ACCEPTED MANUSCRIPT 723

mobilization in macroporous soils: A novel conceptual model. Vadose Zone J. 9, 307-316.

724

Mills, T.J., Anderson, S.P., Bern, C., Aguirre, A., Derry, L.A., 2017. Colloid mobilization and

725 726 727

seasonal variability in a semiarid headwater stream. J. Environ.Qual. 46, 88-95. Mishurov, M., Yakirevich, A., Weisbrod, N., 2008. Colloid transport in a heteogeneous partially saturated sand column. Environ. Sci. Technol. 42, 1066-1071. Missong, A., Holzmann, S., Bol, R., Nischwitz, V., Puhlmann, H., von Wilpert, K., Siemens, J.,

729

Klumpp, E., 2018. Leaching of natural colloids from forest topsoils and their relevance

730

for phosphorus mobility. Sci. Total Environ. 634, 305-315.

731 732 733

RI PT

728

Mohanty, S., Saiers, J.E., Ryan, J.N., 2014. Colloid-facilitated mobilization of metals by freezethaw cycles. Environ. Sci. Technol. 48, 977-984.

Mohanty, S.K., Bulicek, M.C., Metge, D.W., Harvey, R.W., Ryan, J.N., Boehm, A.B., 2015a.

Mobilization of microspheres from a fractured soil during intermittent infiltration events.

735

Vadose Zone J. 14, doi:10.2136/vzj2014.05.0058

SC

734

Mohanty, S.K., Saiers, J.E., Ryan, J.N., 2015b. Colloid mobilization in a fractured soil during dry-

737

wet cycles: Role of drying duration and flow path permeability. Environ. Sci. Technol. 49,

738

9100-9106.

739

M AN U

736

Mondal, P.K., Sleep, B.E., 2012. Colloid transport in dolomite rock fractures: Effects of fracture

740

characteristics, specific discharge, and ionic strength. Environ. Sci. Technol. 46, 9987-

741

9994.

Morales, V.L., Zhang, W., Gao, B., Lion, L.W., Bisogni, J.J., McDonough, B.A., Steenhuis, T.S.,

743

2011. Impact of dissolved organic matter on colloid transport in the vadose zone:

744

Deterministic approximation of transport deposition coefficients from polymeric coating

745

characteristics. Water Res. 45, 1691-1701.

746

TE D

742

Pronk, M., Goldscheider, N., Zopfi, J., 2007. Particle-size distribution as indicator for fecal bacteria contamination of drinking water from karst springs. Environ. Sci. Technol. 41,

748

8400-8405.

750 751 752

Pronk, M., Goldscheider, N., Zopfi, J., Zwahlen, F., 2009. Percolation and particle transport in the unsaturated zone of a karst aquifer. Ground Water 47, 361-369.

AC C

749

EP

747

Rani, R.D., Sasidhar, P., 2011. Stability assessment and characterization of colloids in coastal groundwater aquifer system at Kalpakkam. Environ. Earth Sci. 62, 233-243.

753

Rebecca, A.B., Robin, D.M., Joseph, N.R., Menachem E., 2002. Mobilization of natural colloids

754

from an iron oxide-coated sand aquifer: Effect of pH and ionic strength. Environ. Sci.

755

Technol. 36, 314-322.

756

Rosseel, Y., 2012. An R package for structural equation modeling. J. Stat. Softw. 48, 1-36.

757

Ryan, J.N., Gschwend, P.M., 1994. Effects of ionic strength and flow rate on colloid release:

758

Relating kinetics to intersurface potential energy. J. Colloid Interf. Sci. 164, 21-34.

759

Schiperski, F., Zirlewagen, J., Scheytt, T., 2016. Transport and attenuation of particles of different

760

density and surface charge: A karst aquifer field study. Environ. Sci. Technol. 50, 8028-

37

ACCEPTED MANUSCRIPT 761 762 763

8035. Schulten, H.R., Leinweber, P., 2000. New insights into organic-mineral particles: Composition, properties and models of molecular structure. Biol. Fert. Soils 30, 399-432.

764

Seaman, J.C., Bertsch, P.M., Kaplan, D.I., 2007. Spatial and temporal variability in colloid

765

dispersion as a function of groundwater injection rate within Atlantic Coastal Plain

766

sediments. Vadose Zone J. 6, 363-372.

768 769 770

Shevenell, L., McCarthy, J.F., 2002. Effects of precipitation events on colloids in a karst aquifer. J.

RI PT

767

Hydrol. 255, 50-68.

Shiyan, L.N., Tropina, E.А., Machekhina, K.I., Gryaznova, E.N., An, V.V., 2014. Colloid stability of iron compounds in groundwater of Western Siberia. Springer Plus 3, 1-7.

Toran, L., Tancredi, J.H., Herman, E.K., White, W.B., 2006. Conductivity and sediment variation

772

during storms as evidence of pathways to karst springs. Geol. Soc. Am. Spec. Pap. 404,

773

169-176.

SC

771

Torkzaban, S., Kim, H.N., Simunek, J., Bradford, S.A., 2010. Hysteresis of colloid retention and

775

release in saturated porous media during transients in solution chemistry. Environ. Sci.

776

Technol. 44, 1662-1669.

777

M AN U

774

Tran, E., Teutsch, N., Klein-BenDavid, O., Weisbrod, N., 2018. Uranium and Cesium sorption to

778

bentonite colloids under carbonate-rich environments: Implications for radionuclide

779

transport. Sci. Total Environ. 643, 260-269.

Walshe, G.E., Pang, L., Flury, M., Close, M.E., Flintoft, M., 2010. Effects of pH, ionic strength,

781

dissolved organic matter, and flow rate on the co-transport of MS2 bacteriophages with

782

kaolinite in gravel aquifer media. Water Res. 44, 1255-1269.

TE D

780

Wang, H.L., Tang, X.Y., Zhang, W., Song, S.B., McKenzie, B.M., 2015. Within-year changes in

784

hydraulic properties of a shallow entisol in farmland and forestland. Vadose Zone J. 14,

785

doi: 10.2136/v14.11.0163.

787 788 789 790 791

Weisbrod, N., Dahan, O., Adar, E.M., 2002. Particle transport in unsaturated fractured chalk under arid conditions. J. Contam. Hydrol. 56, 117-136. Xing, Y.N., Chen, X.J., Zhuang. J., Chen, X., 2015. What happens when parmaceuticals meet

AC C

786

EP

783

colloids. Ecotoxicology 24, 2100-2114.

Xu, H.C., Xu, M.W., Li, Y., Liu, X., Guo, L.D., Jiang, H., 2018. Chracterization, origin and aggregation behavior of colloids in eutrophic shallow lake. Water Res. 142, 176-186.

792

Yang, P.H., Yuan, D.X., Ye, X.C., Xie, S.Y., Chen, X.B., Liu, Z.Q., 2013. Sources and migration

793

path of chemical compositions in a karst groundwater system during rainfall events.

794

Chinese Sci. Bull. 58, 2488-2496.

795

Yang, X.Y., Zhang, Y.M., Chen, F.M., Yang, Y.S., 2015. Interplay of natural organic matter with

796

flow rate and particle size on colloid transport: Experimentation, visualization, and

797

modeling. Environ. Sci. Technol. 49, 13385-13393.

798

Zhang, W., Tang, X.Y., Weisbrod, N., Zhao, P., Reid, B.J., 2015. A coupled field study of

38

ACCEPTED MANUSCRIPT 799 800 801 802 803

subsurface fracture flow and colloid transport. J. Hydrol. 524, 476-488. Zhang, W., Tang, X.Y., Xian, Q.S., Weisbrod, N., Yang, J.E., Wang, H.L., 2016. A field study of colloid transport in surface and subsurface flows. J. Hydrol. 542, 101-114. Zhang, W., Tang, X., Weisbrod, N., Guan, Z., 2012. A review of colloid transport in fractured rocks. J. Mt. Sci. 9, 770-787. Zhang, M.Y., Bradford, S.A., Simunek, J., Vereecken, H., Klumpp, E., 2017. Role of cation

805

valance and change on the retention and colloid-facilitated transport of functionalized

806

multi-walled carbon nanotubes in a natural soil. Water Res. 109, 358-366.

RI PT

804

807

Zhao, J.B., He, X.B., Shao, T.J., 2012. Material composition and microstructure of purple soiland

808

purple mudstone in Chongqing area. Ac. Pedol. Sin. 49, 212-219 (in Chinese with an

809

English abstract).

Zhao, P., Tang, X., Zhao, P., Wang, C., Tang, J., 2013. Identifying the water source for subsurface

811

flow with deuterium and oxygen-18 isotopes of soil water collected from

812

tensionlysimeters and cores. J. Hydrol. 503, 1-10.

M AN U

SC

810

813

Zhao, P., Tang, X., Zhao, P., Zhang, W., Tang, J., 2016. Mixing of event and pre‐event water in a

814

shallow Entisol in sloping farmland based on isotopic and hydrometric measurements,

815

SW China. Hydrol. Process. 30, 3478-3493.

816

Zhou, D.,Wang, D., Cang, L., Hao, X., Chu, Y., 2011. Transport and re-entrainment of soil colloids

817

in saturated packed column: Effect of pH and ionic strength. J. Soil Sediment 11, 491-503.

818

Zvikelsky, O., Weisbrod, N., 2006. Impact of particle size on colloid transport in discrete fractures.

EP

TE D

Water Resour. Res. 42, doi:10.1029/2006WR004873.

AC C

819

39

ACCEPTED MANUSCRIPT Figure captions Fig. 1. Distributions of the farmland plot and three lowland shallow wells (well 1#: Zhaoxingqiang well; well 2#: Zhangfei well; well 3#: Xinjing well) investigated in the Jieliu catchment.

RI PT

Fig. 2. Dynamics of the discharge, colloid, cation and DOC concentrations, pH and EC of the fracture flow in the 0.15 hm2 sloping farmland plot in 2013. Fig. 3. Particle size distributions of well water samples from the three lowland wells following storm events from 30 June to 1 July. US treatment denotes the ultrasound treatment (2 min, 100 W) of the fracture flow and well water samples in a water bath sonicator.

SC

Fig. 4. pRDA results of groundwater colloid dynamics explained by discharge/water depth and chemical factors.

M AN U

Fig. 5. Structural equation model showing the direct and indirect effects of flow discharge and flow chemistries on colloid dynamics in fracture flow. The solid and dashed arrows indicate significant and insignificant relationships, respectively. R2 denotes the proportion of explained variance of a certain variable by the model. Numbers adjacent to the arrows are path coefficients with the significance levels of *P<0.05, **P<0.01, ***P<0.001. A satisfactory fit of our data by the model is reflected by χ2=0.321, df=1, P=0.571, GFI=0.99 and RMSEA=0.00.

AC C

EP

TE D

Fig. 6. Comparisons of δ13C abundances between groundwater colloids (recovered from the fracture flow or well water) and the nearby soil and rock samples of upper locations in the farmlands and forestlands.

ACCEPTED MANUSCRIPT Table 1 Descriptive statistics of the hydraulic and physiochemical parameters of the fracture flow and colloid concentration in the sloping farmland plot

pH

EC (µS/cm)

Ca2+ (mg/L)

Mg2+ (mg/L)

DOC (mg/L)

Colloid (mg/L)

67.22 0.089 3.32 6.21 1.87

14.64 0.004 2.06 2.69 1.31

8.81 6.75 7.96 0.41 0.05

769 441 654 60.84 0.09

107.03 57.04 83.36 9.61 0.12

24.04 3.11 15.70 3.23 0.21

3.82 0.28 1.41 0.64 0.45

31.90 0.54 1.64 2.85 1.74

AC C

EP

TE D

M AN U

SD: standard deviation, CV: coefficient of variation

RI PT

Flow depth (mm/day)

SC

Max Min Mean SD CV

Discharge (L/min)

ACCEPTED MANUSCRIPT Table 2 Descriptive statistics of colloid mass and number concentrations of the well water

Nmax Nmin Mmax Mmin Mean SD CV

8

7.00×10 8.82×107 7.62 0.96 2.72 1.77 0.65

Zhaoxingqiang well

Zhangfei well

9

1.07×10 4.74×107 11.66 0.52 2.21 2.24 1.01

5.06×108 1.62×107 5.51 0.24 1.42 1.23 0.87

RI PT

Xinjing well

AC C

EP

TE D

M AN U

maximum and minimum mass concentration (mg/L)

SC

Nmax and Nmin: maximum and minimum number concentration (colloids/L); Mmax and Mmin:

ACCEPTED MANUSCRIPT Table 3 Percentages of colloids (<2 µm) and coarse particles (2-10 µm) and median diameter (d50) of suspended particles in lowland wells

<2 µm

<10 µm

d50a (µm)

XJ ZXQ ZF

5.2 (20.9) b 7.0 (24.7) 2.4 (35.4)

97.2 (99.8) 89.2 (98.8) 50.1 (99.2)

4.2 (3.0) 5.3 (3.2) 10.0 (2.5)

RI PT

Lowland well

SC

XJ, ZXQ and ZF denotes Xinjing well, Zhaoxingqiang well and Zhangfei well, respectively.

a: Median particle diameters denotes the particle size at which the cumulative percentage of

M AN U

particles reaches 50%.

AC C

EP

TE D

b: Numbers in the bracket indicate the situations after ultrasound treatment.

ACCEPTED MANUSCRIPT Table 4 Results of Pearson correlation and linear regression (stepwise method) analyses of colloid concentration with the physicochemical parameters of the fracture flow EC (µS/cm)

Ca2+ (mg/L)

Mg2+ (mg/L)

DOC (mg/L)

Discharge
0.756** 0.798 53.27 0.885** 0.926 58.09 0.097 0.226 21.67

0.087 0.179 11.95 0.052 0.113 7.09 0.280* 0.338 32.41

-0.339** -0.204 13.62 -0.553** -0.101 6.34 -0.087 --d --

-0.135* -0.111 7.41 -0.167 -0.142 8.91 -0.075 ---

-0.319** 0.110 7.34 -0.368** 0.214 13.43 -0.376** -0.288 27.61

0.210** 0.096 6.41 0.247** 0.098 6.15 0.100 0.191 18.31

SC

Discharge>d50 (n=117)

Ra Rscb NRsc (%)c R Rsc NRsc (%) R Rsc NRsc (%)

M AN U

All cases (n=238)

RI PT

pH

Discharge (L/min)

d50 denotes the median (50% percentile) discharge.

a: R represents Pearson correlation coefficient of colloid concentration with each parameter. b and c: Rsc and NRsc represent the standardized coefficients (Beta) from the multi-linear

TE D

regression analysis and the normalized values of each Rsc, respectively. d: -- indicates the corresponding variable is excluded in the standardized linear regression

EP

equation.

*Correlation is significant at the 0.05 level (2-tailed).

AC C

**Correlation is significant at the 0.01 level (2-tailed).

ACCEPTED MANUSCRIPT Table 5 Results of Pearson correlation and linear regression (stepwise method) analyses of colloid concentration with the physicochemical parameters of the well water

Ca2+ (mg/L)

Mg2+ (mg/L)

DOC (mg/L)

0.229 --

-0.522** --

-0.489** -0.285

-0.381* --

0.812** 0.720

--

--

--

--

28.36

--

71.64

0.263 --

0.139 --

0.066 --

-0.668** -0.667

-0.290 --

-0.146 --

0.435* --

Zhaoxingqiang well (n=27)

--

--

--

100

--

--

--

-0.247 --

-0.555** -0.555

-0.308 --

-0.216 --

0.201 --

-0.232 --

-0.290 --

Zhangfei well (n=27)

--

100

--

--

--

--

--

SC

Xinjing well (n=27)

R Rscb NRsc (%)c R Rsc NRsc (%) R Rsc NRsc (%)

M AN U

a

Tem

RI PT

EC (µS/cm)

-0.337 --d

(°C) 0.201 --

pH

WL (cm)

TE D

a: R represents Pearson correlation coefficient of colloid concentration with each parameter. b and c: Rsc and NRsc represent the standardized coefficients (Beta) from the multi-linear regression analysis and the normalized values of each Rsc, respectively.

AC C

equation.

EP

d: -- indicates the corresponding variable is excluded in the standardized linear regression

*Correlation is significant at the 0.05 level (2-tailed). **Correlation is significant at the 0.01 level (2-tailed).

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

Fig. 1.

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

Fig. 2.

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

Fig. 3.

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

Fig. 4.

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

Fig. 5.

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

Fig. 6.

ACCEPTED MANUSCRIPT

Highlights Fracture flow and shallow groundwater were monitored

simultaneously.

Hydraulic perturbation generally dominates colloid dynamics in fracture flow.

RI PT

Water chemistries dominate colloid dynamics in shallow wells.

AC C

EP

TE D

M AN U

SC

Multi-parameters were jointly used for groundwater colloid source identification.

ACCEPTED MANUSCRIPT Dear Editor,

The authors declare that we have no commercial or associative interest that represents a conflict of interest in connection with this work submitted (i.e., Dynamics and sources of colloids in shallow groundwater in lowland wells and fracture flow in sloping farmland

Best regards, Jun-Fang Cui

Department of Soil and Environment

Chinese Academy of Sciences No. 9, Block 4, Renminnanlu Road Chengdu610041, China

M AN U

Institute of Mountain Hazards and Environment

Tel.: +86 28 85213556; fax: +86 28 85222258.

AC C

EP

TE D

E-mail address: [email protected]

SC

RI PT

by Wei Zhang, Jian-Hua Cheng, Qing-Song Xian, Jun-Fang Cui*, Xiang-Yu Tang, Gen-Xu Wang).