Particle fluxes in groundwater change subsurface

Earth and Planetary Science Letters 500 (2018) 180–191

Contents lists available at ScienceDirect

Earth and Planetary Science Letters www.elsevier.com/locate/epsl

Particle fluxes in groundwater change subsurface shale rock chemistry over geologic time Hyojin Kim a,b,∗ , Xin Gu c , Susan L. Brantley a,c a b c

Earth and Environmental Systems Institute, the Pennsylvania State University, University Park, PA 16802, USA Department of Groundwater and Quaternary Geology Mapping, the National Geological Survey of Denmark and Greenland, Denmark Department of Geosciences, the Pennsylvania State University, University Park, PA 16802, USA

a r t i c l e

i n f o

Article history: Received 18 March 2018 Received in revised form 12 July 2018 Accepted 23 July 2018 Available online xxxx Editor: L. Derry Keywords: subsurface particle loss chemical weathering physical erosion

a b s t r a c t Most models of landscape evolution posit that particles leave the land surface by physical weathering (i.e., erosion) and solutes leave the subsurface by chemical weathering. They assume that erosion does not affect the rock and soil chemistry. However, in this study of a shale catchment we discovered that particles are mobilized out of soil and weathering rock and are transported through the subsurface, resulting in changes in the rock and soil chemistry. We studied the solute and particle fluxes during six storms in the Susquehanna Shale Hills Critical Zone Observatory in Pennsylvania, USA and compared those to the record in regolith chemistry. The stream’s suspended particles primarily consisted of platyshaped, μm-sized illite, commonly coated with patchy, amorphous, submicron-sized Al-, Fe- and Si-rich oxides. The chemistry of the stream particles always differed from that of surface soils except during intense dry-season rainstorms. Stream particles were chemically similar to the laboratory-extracted soil colloids at high discharge but to groundwater particles at low discharge, implying a central role of flow path variations in controlling subsurface particle transport. Zr was effectively immobile in Shale Hills. Regolith chemistry revealed that the cumulative effects of particle loss to depths of 5–8 m in the fractured bedrock zone were estimated to account for 58% of K and 24% of Mg losses. In shale landscapes, we propose that subsurface particle transport must be considered in landscape evolution models as an important contributor to changes in rock and soil chemistry over geologic time periods. © 2018 Elsevier B.V. All rights reserved.

1. Introduction Losses of solutes and particles from the critical zone shape its architecture. Most conceptual and quantitative models of critical zone evolution describe loss of solutes throughout the weathering profile by chemical weathering and loss of particles from the surface by physical erosion (Anderson et al., 2007; Dixon et al., 2009; Riebe and Granger, 2013; Yoo and Mudd, 2008). However, many lines of evidence show that particles are also lost from the subsurface in some locations and that this loss depends on physical, chemical, and hydrological conditions (DeNovio et al., 2004; McCarthy and McKay, 2004; Ryan and Elimelech, 1996). Here, “particle transport through the subsurface” is used to refer to the suspension and movement of particles of any size through soil and bedrock: we do not refer here to episodic mass movement such as landslides or other such processes.

*

Corresponding author at: Earth and Environmental Systems Institute, the Pennsylvania State University, University Park, PA 16802, USA. E-mail address: [email protected] (H. Kim). https://doi.org/10.1016/j.epsl.2018.07.031 0012-821X/© 2018 Elsevier B.V. All rights reserved.

In most weathering models physical erosion has been assumed to not affect soil chemistry. Given that assumption, chemically immobile elements such as Zr, Ti, and Hf or relatively insoluble minerals have been used to quantify elemental losses by chemical weathering (Anderson et al., 2002; Brimhall and Dietrich, 1987). For example, the non-dimensional mass transfer coefficient (τ j ,i ) is now commonly used to quantify the loss or accumulation of a mobile element ( j) by comparing its concentrations in parent and weathered materials normalized by the immobile element (i) (Anderson et al., 2002; eqn. (1)):

τ j ,i =

C j, w C i, p C j, p C i, w

−1

(1)

here C is a concentration of an immobile element (i) or a mobile element ( j) of parent (p) or weathered material (w). The τ values vary from −1 (complete depletion) to 0 (no change) to positive values (accumulation). Riebe et al. (2003) further proposed that by combining the concentration of immobile elements and estimates of the total denudation rate ( D ) from cosmogenic nuclide meth-

H. Kim et al. / Earth and Planetary Science Letters 500 (2018) 180–191

ods (e.g., 10 Be), the relative importance of chemical and physical weathering fluxes can be quantified (eqn. (2)):

W = D 1−

[Zr] p [Zr] w

(2)

here W , D, [Zr] p , and [Zr] w refer to the chemical weathering rate, total denudation rate (chemical plus physical losses), and the Zr concentrations of parent and weathered materials, respectively. [Zr]

Furthermore, the authors define (1 − [Zr] p ) as the “chemical deplew tion fraction (CDF)” (Dixon et al., 2009; Riebe et al., 2004, 2003). Physical erosion is assumed to mobilize particles that have the same chemistry as the originating soil. However, particles that are enriched in Al, Fe, Si, and sometimes Ti are observed to be redistributed from one horizon to another within soil profiles and have also been documented to be lost from catchments (Aguirre et al., 2017; Bern et al., 2011; Jin et al., 2010; Kaup and Carter, 1987; Taboada et al., 2006; Trostle et al., 2016; Yesavage et al., 2012). Such particle losses, furthermore, may be significant enough to alter Al and Fe concentrations in soils (Jin et al., 2010; Yesavage et al., 2012). For instance, Jin et al. (2010) investigated the chemistry of soil, bedrock, and stream water at the Shale Hills catchment in the Susquehanna Shale Hills Critical Zone Observatory (CZO) in Pennsylvania. They reported that Al and Fe of the soil and fractured rock were depleted with respect to the parent material, but the solute concentrations of these elements in the stream (0.45 μm filtered) were negligible. They hypothesized that these elements were likely mobilized as particles. At the same site, Yesavage et al. (2012) also analyzed the chemistry of soil pore water, groundwater, macro-pore water (i.e., preferential flows through the soil), and stream waters. They found that Al and Fe concentrations of the soil-pore waters showed no differences between unfiltered and filtered samples (0.45 μm) while those of the other sampled waters were much higher in the unfiltered samples than in the filtered samples. The authors, therefore, postulated that the particles transporting these elements were larger than the pore-sizes of the lysimeter cups (1.3 μm). Recently, different approaches have been developed to quantify elemental losses via solutes and particles (Bern et al., 2015; Hasenmueller et al., 2017; Sullivan et al., 2016). For instance, Hasenmueller et al. (2017) proposed using Al as the immobile element to quantify the losses of the solute and particle fractions of a mobile element ( j ), τ j ,Al . The authors argued that Al was mobilized mainly as particles: therefore, the elemental loss as solute was revealed by using Al as the immobile element. They then assumed that Zr was immobile and argued that τ j ,Zr estimated the total loss of the mobile element. The difference between τ j ,Zr and τ j,Al becomes the fraction of elemental loss as particles ( P j ):

P j = τ j ,Zr − τ j ,Al

(3)

The authors employed this method to quantify the elemental loss as particles in the Missed Grouse catchment, a catchment that lies next to Shale Hills, in the Shale Hills CZO: more than half of the potassium (up to 75%) and magnesium (up to 63%) fluxes were attributed to particle loss. In another treatment, Bern et al. (2015) proposed a dual-phase mass balance model. They used a ratio of two elements such as Ti and Zr that primarily move via colloids (i.e., water dispersible colloids; WDCs) but that display different affinity to the colloids. The authors adopted a definition of colloid as particles smaller than 1 μm (Sposito, 2006). They operationally defined these colloids (material between 3 kDa and 1 μm) by extracting them from soil samples by shaking soil samples in deionized water for 10 min in the laboratory. They assumed that, as weathering progresses, Ti and Zr concentrations in parent bedrock, weathered material, and

181

WDCs develop characteristic Ti/Zr ratios. Then, the authors quantified the elemental loss and gain as colloids via the mass balance approach using the Ti/Zr ratios of these materials. Bern and Yesavage (2018) used this model to study colloidal loss in Shale Hills. They extracted and analyzed WDCs from the Shale Hills soils and estimated that more than 90% of the total mass loss was via WDCs in the catchment. They also concluded that Zr was mobile and that the loss of WDCs resulted in Zr depletion in the soil by 12–51% with respect to the parent rock. However, these previous treatments of subsurface particle losses and weathering have been indirect and only limited to the soil layer. The importance of particle transport in the evolution of the entire critical zone structure, particularly the underlying fractured/weathered rock zone, has not been previously assessed. In addition, in Shale Hills particles that are larger than a micron—and therefore do not qualify as colloids—may play a key role in Al and Fe losses (Jin et al., 2010). Here, we return to Shale Hills to explore the importance of losses of particles in the context of critical zone evolution. The primary objectives are to identify mechanisms of mobilization and transport of particles of all size, and to quantify the importance of subsurface particle loss at the catchment scale in both short- and long-term time scales. We accomplished these goals by direct observations of mobile particles in groundwater and in the stream at various hydrologic conditions and by comparing the mobile particle chemistry to the regolith chemistry. 2. Methods 2.1. Study site The Shale Hills catchment (drainage area: 0.08 km2 ) lies in central Pennsylvania (U.S.A.) and is underlain by Rose Hill Shale. Illite was the most dominant mineral of the parent bedrock (DC1) followed by quartz, chlorite, and trace amounts of feldspar and Fe-oxides (Jin et al., 2010). In the deepest layer (>37 m below the land surface under the northern ridge) in the parent bedrock, ankerite was found while near the stream outlet, the carbonaterich layer lies at 6–8 m deep (Brantley et al., 2013). The mean annual precipitation is 1090 mm and the erosion rate estimated from meteoric 10 Be measurements of stream sediment is 15 m Myr−1 (Jin et al., 2010). The stream is ephemeral and discharge is measured at a weir located near the outlet of the catchment at a 10-min interval. Stream and groundwater chemistry observations have demonstrated that the contribution of regional groundwater to the stream discharge may be small and may become important only at low flow (Sullivan et al., 2016). Instead, most stream discharge is fed by water that has quickly flowed through the upper layers of disaggregated soil (∼1.4 m thick) and through a highly fractured bedrock zone that extends to 5 or 8 m depth throughout the catchment (Fig. 1; Graham and Lin, 2011; Jin et al., 2011a; Lin et al., 2006; Sullivan et al., 2016). Jin et al. (2011b) proposed that the fractures of this layer may have developed primarily by freeze–thaw processes since the Last Glacial Maximum. Because of these porosity contrasts, the perched groundwater table (i.e., interflow) transiently develops at the interface between the highly fractured rock and underlying less fractured zone (Sullivan et al., 2016). At high flow, the dominant flow path is preferential flow via soil macro-pores and soil horizon interfaces in the unsaturated zone (Graham and Lin, 2011; Jin et al., 2011a; Lin et al., 2006) and interflow. During highly intense rainstorms or at the beginning of snow-melt, overland flow occurs (Lin et al., 2006). During the dry season, the stream dries out. 2.2. Sampling We collected water and suspended particle samples at 0.5–48 h intervals during 6 storm events using three ISCO samplers in the

182


Fig. 1. (a) Overview of the Shale Hills Critical Zone Observatory in Pennsylvania. Water and suspended particle samples of stream and interflow (W3, W5, and W11) were collected for six storm events using ISCO samplers. (b) Sullivan et al. (2016) measured the groundwater table depths at a monthly interval from 2013 to 2014. The highest (wet season) and lowest (dry season) groundwater tables along the central channel of the catchment were plotted (A–A transect). Note the vertical axis is exaggerated by three-fold. In Shale Hills, preferential flow paths through soil macropores and through interfaces of the soil horizons and at the perched groundwater table (i.e., interflow) at the interface of the highly fractured rock and less weathered rock are the primary flow paths. The contribution of regional groundwater to stream discharge at the catchment outlet is less than ∼10%. For interpretation of the colors in the figures, the reader is referred to the web version of this article.

Shale Hills catchment from August 2016 to May 2017 (Supplementary Table 1). One ISCO sampler was dedicated to stream sampling and was deployed near the outlet of the catchment. The other two were moved among interflow wells (i.e., W3, W5, and W11; Fig. 1). The sampling frequencies for the stream were 30 min for the summer storms and 10–15 h for the winter storms. Interflow samples were collected at 36–48 h intervals. For every ISCO sampling, we collected two bottles of samples. For solutes, we used a Gravitational Filtration System (GFS; Kim et al., 2012): the GFS filters water through 0.2 μm pore-sized Supor membrane filters by gravity immediately upon sampling. For suspended particles, approximately 1 L of bulk sample was collected using a Propak ISCO sample bag. Here, we operationally defined “dissolved species” as species that passed the 0.2 μm filter and a “particulate phase” as material that did not. These definitions are valid for Si and major solute cations; however, the 0.2 μm filtrates may overestimate the truly dissolved concentrations of Al and Fe because of nano-sized colloids (e.g., Horowitz et al., 1996, 1992). Bern and Yesavage (2018) also reported an abundance of such nano-sized colloids in the Shale Hills soils. However, other studies in Shale Hills postulated that Al and Fe leave soils mainly as relatively large particles rather than colloids (Jin et al., 2010; Yesavage et al., 2012). This hypothesis leads us to assume that the concentrations of Fe- and Al-containing particles in the 0.2 μm retentate are much greater than in the filtrate. Therefore, we concluded that our definitions are acceptable for the quantification of solute and particle concentrations. When the ISCO sample bottles were full, the samples were retrieved. The GFS water samples were acidified using nitric acid

(2% v/v) at least 24 h before transferring to storage bottles. In this way, any precipitated and adsorbed species were re-solubilized. The GFS retentate (i.e., samples on filters) was used for qualitative analysis of mineralogy and particle morphology. We collected particle samples by filtering the 1-liter bulk samples using the same filters as in GFS in the laboratory. The bulk samples were rigorously mixed by hand and sonicated for a minute before filtration. Then, the filters were air-dried and stored in plastic sample bags until analysis. All the filtration apparatus including filter membranes were acid leached overnight in 10% HCl and cleaned with Milli-Q water (18.2 M/cm) prior to use. 2.3. Particle characterization Selected filter samples from GFS for stream and groundwater wells W3, W5, and W11 were analyzed for mineralogy, morphology, and particle size distribution. Mineralogy was analyzed by Xray diffraction (XRD) using a PANalytical Empryean X-Ray Diffractometer (PANalytical Ltd., The Netherlands) at 45 kV and 40 mA with a Cu Kα radiation in the Material Characterization Laboratory (MCL) at the Pennsylvania State University. Stepwise scanning measurements were performed at a rate of 4◦ min− 1 in the range of 5–70◦ 2θ . A blank filter was analyzed as well to define background. We investigated particle morphology using a Talos F200X transmission electron microscopy (TEM) with a dispersive X-ray spectroscopy (EDS) in MCL (Thermo Fisher Scientific, USA). The operation voltage was 200 kV. TEM samples were prepared by resuspending the GFS retentate. The samples were sonicated in the ethanol solution for 10 min and were deposited on TEM copper


183

grids. We also analyzed particle size distribution using a Malvern Mastersizer (Malvern Instruments Ltd, UK). For these analyses, particles from selected GFS filters were re-suspended by sonication in Milli-Q water for 20–30 min and the solutions were injected into the Mastersizer.

Here the subscript d refers to the dissolved concentration of each element. Similarly, the total particle flux (F p ; μg/s) was calculated by multiplying the mass of particle per volume of filtered water (M; mg/L), the sum of the particle elemental compositions (% w/w), and discharge:

2.4. Water and particle chemistry analysis

F p = [Ca p ] + [Mg p ] + [Na p ] + [Si p ] + [K p ]

Water chemistry was analyzed using an inductively coupled plasma-atomic emission spectroscopy (ICP-AES) for major solute cations and Si and using an inductively coupled plasma-mass spectroscopy (ICP-MS) for Al, Ti, Mn, Fe, and Zr at the Laboratory for Isotopes and Metals at the Pennsylvania State University. Particles were digested via the standard lithium metaborate (LiBO2 ) fusion method. First, the filter membranes of the particle samples were removed by ashing them at 900 ◦ C overnight. This procedure completely removed the filter material, which is composed of primarily carbon, oxygen, and sulfur. Then, the ashed samples were mixed with LiBO2 and heated at 1000 ◦ C in ultra-clean graphite crucibles. The resultant beads were re-dissolved in nitric acid and then analyzed using ICP-AES for major elements (Al, Ca, Fe, Mg, Na, K, Ti) and using ICP-MS for Zr. We used four U. S. Geological Survey certified rock standards (BHVO-1, W-2, GSP-1b, SRB-1) and a LiBO2 blank to define calibration curves and verify the analysis results. The standard LiBO2 fusion usually fuses 100 mg of sample; however, the particle samples that we collected via ISCO yielded 1–10 mg of particle per L of water. We therefore evaluated the effects of the sample mass on the uncertainty of the chemical analysis using the Shale Hills’ soil samples. To yield reproducibility of 10% for reference samples, we found we had to use a minimum sample mass of 10 mg (Supplementary Fig. 1). Therefore, we combined 2–10 particle samples that were collected at similar discharge levels and also during similar hydrographic regimes (e.g., rising limb vs. peak discharge vs. falling limb) to obtain at least 10 mg of samples for analysis. 2.5. Variations in chemistry of suspended particles with discharge The temporal variability of the chemistry of suspended particles was investigated through inspection of variations of τ j ,Zr as a function of stream discharge. We calculated τ j ,Zr of stream and groundwater suspended particles using eqn. (1). Zr was used as the immobile element. The elemental concentrations of fresh bedrock of Shale Hills were reported in Sullivan et al. (2016), which averaged four fresh bedrock values from deep samples of boreholes CZMW 8 and DC1. The variations of τ j ,Zr with stream discharge were investigated and compared with other solid materials in Shale Hills (i.e., soils; Jin et al., 2010) as well as for WDCs reported by Bern and Yesavage (2018). These elemental compositions reported in the previous studies are based on the total mass, which include loss of ignition (LOI). However, the suspended particle chemistry of this study is based on ashed material, which excludes LOI. Therefore, we converted the elemental compositions of all the solid phases into ashed mass. 2.6. Quantifying solute and particle fluxes To calculate the total solute and particle fluxes, we first calculated the total solute flux (μg/s) of each sampling event (F d ) by multiplying the sum of all major cation concentrations (μg/L; ppb), and the measured discharge during sampling (D; L/s):

F d = [Cad ] + [Mgd ] + [Nad ] + [Sid ] + [Kd ] + [Fed ] + [Ald ] D (4)

+ [Fe p ] + [Al p ] M D

(5)

Here the subscript p refers to the elemental concentration of each particle sample. The solute and particle elemental concentrations for the ISCO sampling gaps were interpolated from the concentration–discharge relationships (Supplementary Fig. 2). The annual total fluxes of solutes and particles (t/km2 /year) were estimated by dividing the sum of F d and F p for the hydrological year by the catchment area. 3. Results 3.1. Particle chemistry and mineralogy The stream particles were enriched with Al, Ca, Fe, K, and P as compared to soils from the south planar hillslope (all topographic positions) and also as compared to the fresh bedrock (Table 1). On the other hand, these particles were largely depleted in Si, Ti, and Zr compared to those samples (Table 1). Stream particles showed similar elemental compositions as the WDCs (Table 1). Using X-ray diffraction, illite, kaolinite, and chlorite/vermiculite were identified in the stream particles, i.e., a similar mineralogy to that of soil and the WDCs. Quartz peaks in these diffractograms were weak particularly at low flow (Supplementary Fig. 3). The median size of stream particles varied between 0.4 and 2 μm. At high flow, most particles were larger than 1 μm while at low flow, most particles were smaller than 1 μm (Fig. 2). Results from the TEM/EDS analyses were consistent with the XRD and bulk chemistry results: large (0.5–1 μm), platy-shaped illite particles were observed to be commonly coated with amorphous, submicron-sized particulate Al-, Fe-, Si-rich oxides, and sometimes Ti-rich oxides (Fig. 3). These oxides were interpreted as secondary products of chemical weathering because of their amorphous appearance under TEM. Elemental compositions of interflow particles from wells W3, W5, and W11 differed significantly from the stream particles but were similar to those of soils and the DC1 core (Table 1). It is always possible that such a similarity is due to sampling artifacts (McCarthy and McKay, 2004; Ryan and Elimelech, 1996): ISCO samplers purge the tubing before sampling, which might disturb the wells resulting in mobilization of soil particles around the well casing. On the other hand, the samples could represent mobile particles in the fractured bedrock zone. This latter interpretation is consistent with observations from Hasenmueller et al. (2017) who studied the chemistry of particles that were observed in fractures at depth in soil pits in a catchment that neighbors Shale Hills. They found that the fracture-filling particles displayed similar chemistry to that of soil, which is consistent with our observations. They concluded that these particles may be generated in-situ by weathering. If such particles are occasionally mobilized by water through the fractures, this could explain our observations. Nonetheless, we cannot rule out the possibility of the effects of sampling artefacts. We, therefore, mainly focus on the stream particle chemistry. 3.2. Solute and particle concentration–discharge relationships The concentration ranges of solutes (Ca, Mg, Si, K, Fe) measured in this study were consistent with previously reported values (2006–2013; Supplementary Fig. 4; Herndon et al., 2015; Sullivan et al., 2016). Although these previous studies analyzed the

184


Table 1 Summary of elemental compositions of Shale Hills solid materials. # of obs.

Al

Ca

Fe

K

Mg

Mn

Na

P

Si

Ti

% (w/w)a

Zr ppm

Suspended particles Stream Groundwater W3 W5 W11

19 1 9 9

14.37 14.67 12.24 8.86

0.59 0.89 0.27 0.22

7.60 9.16 6.44 5.12

3.94 3.71 3.38 2.99

1.11 1.02 1.06 0.72

0.12 0.19 0.22 0.09

1.25 1.18 0.88 0.51

0.16 0.11 0.12 0.05

24.7 23.3 28.3 32.1

0.55 0.51 0.69 0.60

153.9 187.3 203.1 294.7

Soilb South Planer Ridge Top (SPRT) South Planer Mid Slope (SPMS) South Planer Valley Floor (SPVF)

3 5 11

8.34 8.08 9.34

0.13 0.11 0.13

4.56 4.05 4.57

2.45 2.46 3.08

0.63 0.62 0.78

0.36 0.13 0.08

0.39 0.47 0.40

0.11 0.08 0.05

33.21 33.94 32.11

0.73 0.72 0.70

264.7 301.0 246.6

13.34

0.14

6.40

3.20

0.80

0.22

–

0.11

27.8

0.34

128.7

11.27 (0.56)

0.19 (0.16)

5.98 (0.87)

4.07 (0.42)

1.33 (0.06)

0.04 (0.01)

0.32 (0.07)

0.04 (0.01)

28.66 (1.04)

0.62 (0.03)

155.0 (11.5)

Water dispersible colloids extracted from the soilc Soil colloids (