autonomous water sampling for long-term monitoring...

Autonomous Water Sampling for Long-Term Monitoring of TraceMetals in Remote EnvironmentsHyojin Kim,†,* James K. B. Bishop,†,‡ Todd J. Wood,‡ and Inez Y. Fung†

†Department of Earth and Planetary Science, The University of California at Berkeley, 307 McCone Hall Berkeley, CA94720‡Lawrence Berkeley National Laboratory, 1 Cyclotron Rd, Berkeley, California 94720, United States

*S Supporting Information

ABSTRACT: A remotely controlled autonomous method for long-term high-frequency sampling of environmental waters in remote locations is described.The method which preserves sample integrity of dissolved trace metals andmajor ions for month-long periods employs a gravitational filtration system(GFS) that separates dissolved and particulate phases as samples are collected.The key elements of GFS are (1) a modified “air-outlet” filter holder tomaximize filtration rate and thus minimize filtration artifacts; and (2) thedirect delivery of filtrate to dedicated bottle sets for specific analytes. Depthand screen filter types were evaluated with depth filters showing bestperformance. GFS performance is validated using ground, stream, and estuarywaters. Over 30 days of storage, samples with GFS treatment had averagerecoveries of 95 ± 19% and 105 ± 7% of Fe and Mn, respectively; withoutGFS treatment, average recoveries were only 16% and 18%. Dissolved majorcations K, Mg, and Na were stable independent of collection methodology, whereas Ca in some groundwater samples decreasedup to 42% without GFS due to CaCO3 precipitation. In-field performance of GFS equipped autosamplers is demonstrated usingground and streamwater samples collected at the Angelo Coast Range Reserve, California from October 3 to November 4 2011.

■ INTRODUCTION

With the introduction of autonomous water samplingtechniques, dynamics of nutrients and major elements havebeen successfully monitored in various environmental settings,for example, see refs 1−3. However, studies of the temporalvariability of trace metals have been very restricted. Most tracemetal studies to date have focused either on contaminants,often in urban settings, and were limited to short period time,for example, see refs 4 and 5, or they were focused on totalmetal concentration and thus did not attempt to separateparticulate and dissolved phases (for example, ref 6) Never-theless trace metals, particularly iron and manganese, play acritical role in transport of nutrients and pollutants, in redoxreactions as electron donors and acceptors, and in aquaticecosystem as essential nutrients or toxicants. Thus, for betterunderstanding of aquatic environments and ecosystems, thedynamics of these elements and underlying processes need tobe studied for an adequate period of time in variousenvironments.Obtaining representative time-series of trace metals in

environmental waters is challenging because they are highlyreactive. Trace metals are lost from solution rapidly viaadsorption either onto the sampling bottles or onto particlesin solution.7,8 The rate and magnitude of adsorption are closelyrelated to the conditions of samples including pH, concen-trations of dissolved organic carbon (DOC), and suspendedsediments which change significantly in time.9 Trace metals,particularly iron, from reduced environments such as ground-

water precipitate out quickly in the presence of oxygen.10 Wheniron precipitates, other trace elements can coprecipitate.10

To preserve samples for trace metal analysis, immediatefiltration using 0.45-μm pore size filters and acidification ofsamples (pH < 2) is a recommended standard protocol.10,11

However recent studies have reported various filtration artifactsthat can significantly influence the “dissolved phase” of tracemetals.12,13 Horowitz et al. (1996), for example, documentedthat the dissolved/particulate partitioning of trace metals(especially Fe and Al) in high turbidity samples can be affectedby colloid aggregation and/or filter pore clogging, and bysorption/desorption exchange with filter materials.12

Despite of these artifacts, filtration remains the best methodto maintain integrity of samples for trace metals at remote fieldsites over long periods of observation. Filtration is cost-effective, easy-to-use and energy-efficient. Filtration artifacts aremostly due to the overloading of filters so it is possible tominimize such artifacts by proper selection of sample volumeand the type and pore-size of filters.12

In this paper, a method for autonomous water samplingdesigned for month-long high-frequency (12 hours to 2 days)monitoring of trace metal dynamics is described. Thismethodology has been developed during a multiyear study of

Received: February 15, 2012Revised: July 25, 2012Accepted: September 28, 2012Published: September 28, 2012

Article

pubs.acs.org/est

© 2012 American Chemical Society 11220 dx.doi.org/10.1021/es3006404 | Environ. Sci. Technol. 2012, 46, 11220−11226

pubs.acs.org/est

the dynamics of ground and streamwater geochemistry at a siteon a small catchment of the Elder Creek at the Angelo CoastRange Reserve. This site is 280 km north of San Francisco inthe headwaters of the south fork of the Eel River in California.Groundwaters at the site are frequently suboxic (O2concentrations < 30 μM) and highly enriched with CO2 (upto 8%) and thus are far from chemical equilibration with theatmosphere. The objective of this study is to evaluate theperformance of the new method for trace metals. The studyalso considers dissolved organic carbon (DOC) which plays acritical role in trace metals behavior and is prone to bedegraded via biological activity. Finally, the integrity of majorcations (Ca, Na, Mg, and K), which are thought to be stable forlong-periods of time, is also evaluated.

■ MATERIALS AND METHODSThe new methodology has two dimensions: (1) a new samplecollection protocol to maintain the sample integrity and (2) anautonomous sampling control system that permits adjustmentof the sampling timing and volume over the Internet to capturerapidly changing environmental events. Validation of the newsampling protocol in the laboratory and its demonstrationunder field conditions are the main elements of this study.(1). Field Setup and Autonomous Sampling Control

System. Rivendell is a small catchment at the Angelo CoastRange Reserve, northern California, and is intensivelymonitored for the climate (air temperature, humidity, pressure,wind speed, wind direction, rainfall, and total solar radiance),vegetation, soil moisture and subsurface hydrology (SupportingInformation (SI) Figure S1). All installed instruments arepowered by solar panels and communicate through Internet-connected Campbell Scientific Inc. (Logan, UT) data loggers.Four 24 position ISCO 6712 autosamplers (Teledyne ISCO,

Lincoln, NE) are installed at the Rivendell site; three forgroundwater at Wells 1, 3, and 10 and one at Elder Creek (SIFigure S1). Each ISCO autosampler is controlled via its RS232serial port by sequenced commands from a custom low powernetworked microcontroller developed at Lawrence BerkeleyNational Laboratory (LBNL). Resident command sequencefiles on the microcontroller may be modified over the Internet,enabling changes to sampling time, sample tubing rinse volume,and sample collection volume; furthermore the controlleruploads status information from the field site. UNIX servers atLBNL automatically process the status files and provide onlineupdates of sampling activities. The sampling frequency variesfrom 12 hours to 3 days, depending on the flow regime.For groundwater sampling, when the water table depth from

the surface is greater than 8 m, the ISCO peristaltic pumpcannot raise the water by itself; thus our controller operates anadditional all plastic submersible pump (Whale WP6012,Global Water Inc., Gold River, CA) which can raise water anadditional 18 m. To rinse the tubing before the sampling, theautosampler pumps water for 12 s to a draining waste bottle(bottle position 12). In order to prevent disturbance of thewell, on completion of sampling, the water in the tubing isallowed to drain slowly back into the well regulated by capillaryair leak. Creek sampling follows the standard ISCO back-flushrinsing protocol.(2). Gravitational Filtration System (GFS). The gravita-

tional filtration system (GFS) is comprised of a standard ISCOProPak bag sampler frame which is modified to support a 140mL syringe (Monoject, Kendall Healthcare), a Luer connected25 mm filter holder, and an in-line series of two polypropylene

(PP) bottles and a borosilicate PTFE/silicon septa vial. Allinterconnections are via 1.6 mm inside-diameter (ID) Teflonand 3.2 mm ID Tygon tubing (Figure 1a). In the case of turbidsamples, to avoid the impact of filter clogging, sets of either onePP bottle/borosilicate bottles or two PP bottles were used.

Two different kinds of filter holders were tested: a 25 mmMillipore filter holder (Figure 1b) and 25 mm Whatman filterholder (Figure 1c). The top of each Millipore filter holder wasdrilled and fitted with a 15 cm long Teflon (1.6 mm bore)capillary tubing and sealed with silicone O rings to remove airtrapped in the headspace above the filter surface. Without thismodification, trapped air can slow or entirely block filtration.These “air-outlet” holders enable the sample to wet the entirearea of the filter and thus minimize sample filtration time.All parts used in this study were leached in 5% HCl for 48 h

and rinsed abundantly with 18.2 MΩ Milli-Q water. Theinterconnecting Teflon and Tygon tubing were acid-rinsed with5% HCl, aided using a vacuum pump and rinsed with Milli-Qwater, abundantly. The GFS parts were soaked in Milli-Q waterfor 24 h to saturate the activated sorption sites.

(3). Filter Selection. In this study, two different types ofmembrane filter were used: a screen type membrane(Polycarbonate membrane filters, Whatman) and depthmembrane (Supor polyethersulfone membrane disk, Pall LifeScience). Screen membranes are generally used for studies ofparticle morphology but can become quickly clogged and thus

Figure 1. Panel (a), Gravitational Filtration System (GFS): (1) 140mL syringe and 25 mm filter; (2) and (3) 60 mL and 30 mLpolypropylene bottles; (4) a 40 mL borosilicate bottle; all elements areinter connected via Teflon and Tygon tubing; panel (b), Millipore“Air-outlet” modified filter holder; panel (c), Whatman filter holderused without any modification.

Environmental Science & Technology Article

dx.doi.org/10.1021/es3006404 | Environ. Sci. Technol. 2012, 46, 11220−1122611221

can significantly influence trace metal concentrations.13 Depthmembrane filters have significantly higher porosity, thus permittypically 6-fold faster filtration rates and much higher particleloading compared to a screen filter of the same pore size.14 1.0μm pore-size polycarbonate filters and 0.45 and 0.8 μm pore-size Supor filters were tested to evaluate suitability for theautonomous sampling. For Polycarbonate and Supor GFS setsthe Whatman and the air-outlet Millipore filter holders wereused, respectively. All filters were leached with 5% HCl for 48 hand rinsed abundantly with Milli-Q water. The Supor filterswere also soaked in the Milli-Q water for 3 days prior to use tocontrol DOC-leaching.15

(4). GFS Evaluation. The effectiveness of the GFS wasevaluated using (1) Fe-spiked Rivendell samples, which weremanually spiked using an inductively coupled plasma (ICP) Festandard solution (1000 ppm in 2% HNO3) and (2)environmental water samples taken from diverse systems: (1)Rivendell, representing a pristine environment; (2) theStrawberry Creek at the UC Berkeley campus, a turbid urbanstream; and (3) San Francisco (SF) Bay, estuarine waters(Table 1). The Strawberry Creek and the SF Bay samples weretransported to the laboratory within 30 min of sampling forprocessing. The GFS tests using the Rivendell samples werecarried out in the field.At the beginning of evaluation tests, samples for trace metals

and DOC were collected following recommended manualsampling methodology.11,12 Briefly, samples were immediatelypressure-filtered using a 10 mL all-plastic syringe and the sametype and pore-size filter as used for the GFS evaluation, andfiltrate immediately dispensed into acid-leached scintillationvials and borosilicate bottles, respectively (referred to“Reference” sample). Reference samples from Rivendell werestored in an ice-cooled box until transported to the laboratory.The “Control” for each GFS evaluation set was a freshly

collected water sample transferred to an acid washed ISCOProPak LDPE 1000 mL bag; These bags are commonly usedfor sample collection and for this reason were adapted in theearly stages of our work. “Control” samples were left stored

loosely covered at room temperature in the laboratory withoutany treatment.A simple modification of the LDPE bag method, referred to

below as “GFS-BAG”, involved inserting an acid cleaned 140mL syringe and nonmodified Millipore filter holder (Figure 1b)into the mouth of the bag to permit gravity filtration. Themodified air-outlet filter holder and the Whatman filter holdercould not be inserted through the mouth of the ISCO frameinto the bag and thus were not tested with bags.The GFS tests included nonmodified Millipore filter holder

GFS sets (referred to “GFS-NM”) to validate the effectivenessof the air-outlet modification. The GFS, GFS-NM, GFS-BAG,and control samples were processed 1−4 weeks later to mimicthe age of the field deployment (Figure 2). Details of samplehandling procedure are discussed in section 5 below.The recovery rate was calculated as follows:

= ×C Crecovery rate(%) / 100E ref

where CE and Cref are concentration of element in evaluationand reference samples, respectively.

(5). Sample Handling. In the laboratory, all trace metalReference samples were acidified using HNO3 (Optima*,Fisher Chemical) within a few hours after sampling and DOCReference samples were stored in a refrigerator until analysis.The Control samples were filtered in the laboratory andprocessed identically. The GFS (including GFS-NM) samplesin the second PP bottle (no. 3 in Figure 1a) and the GFS-BAGsamples were acidified using Optima HNO3 and stored for atleast for 24 h before transfer to acid-leached scintillation vials.The GFS borosilicate vial samples (no. 4 in Figure 1a) wereanalyzed for dissolved organic carbon (DOC) withoutadditional treatment. In the case of the GFS samples with thetwo PP bottle set and the GFS-BAG set, an aliquot fromacidified samples was taken for the DOC analysis.

(6). Analytical Methods. Major and minor cations andtrace elements were analyzed using a Finnagan Element IImagnetic sector ICP-mass spectrometer (ICP-MS) at LBNL.Samples were run diluted and spiked with indium (In) as aninternal standard. To validate the accuracy and precision of the

Table 1. Basic Properties and Mean Concentrations of Selected Elements (Standard Deviation) 0.45 μm Supor Filtrates ofSamples Used for the GFS Evaluation

Rivendell

Fe-spiked RivendellGWb Strawberry Creek SF Bay Well 1 Well 3 Well 10 Elder Creek

na 3 3 1 10 6 6 6samplingperiod

2011 May andSeptember

2011 May−September

2011.June

2011 April− 2011October

2011 April− 2011October

2011 April −2011October

2011 May −2011October

pH 7.81 8.13 8.11 7.04 6.34 5.61 8.01turbidity turbid turbid variable variable variable variableFe 438 194 40 21 496 38 36(nM) (395) (154) (−) (11) (473) (11) (17)Mn 89 164 440 2433 6722 449 12(nM) (79) (84) (−) (2642) (7540) (403) (3)Al 121 207 62 37 71 142 542(nM) (84) (124) (−) (11) (66) (90) (37)Ca 0.24 1.00 9.98 1.53 1.42 0.28 0.32(mM) (<0.01) (0.3) (−) (0.07) (0.76) (0.23) (0.03)DOC 2522 1704

(−)766 1034 477 1512

(ppb) (126) (261) (605) (186) (768)aNumber of evaluation. bGroundwater.



ICP analysis, Certified River Water Reference Material forTrace Metals, SLRS-5 (National Research Council (NRC),Ottawa, ON, Canada) was used. The percent error of each ICPanalysis was calculated as shown:

= | − | ×− −C C C%error / 100SLRS 5 A SLRS 5

where CSLRS‑5 and CA are the concentration of each elementidentified by NRC and analyzed in this study, respectively.DOC was analyzed by the nonpurgeable organic carbon

(NPOC) method using a Shimadzu TOC Analyzer outfittedwith a Parker Balston TOC Gas Generator. Prior to DOCanalysis, 2 M HCl was added to samples adjust the pH below 2and then samples were purged 30 min. In case of the HNO3acidified samples, the acidification step was not required. Blank(Milli-Q water) and standard potassium hydrogen phthalate(KHP) solutions were analyzed every 10 samples.

■ RESULTS AND DISCUSSIONIn total, 36 samples from seven different locations (Table 1 anddetailed data available in SI Table S1) were used for the GFSevaluation: Fe-spiked Rivendell (number of samples, n = 3),Strawberry Creek (n = 3), San Francisco Bay (n = 1), andRivendell Wells 1 (n = 10), 3 (n = 6), and 10 (n = 6) and ElderCreek (n = 6). The Supor evaluation sets (pore-size 0.45- and0.8 μm) and Polycarbonate evaluation set (pore-size 1.0 μm)were run in triplicate and duplicate, respectively. The pH ofthese samples ranged from slightly acidic (pH 5.61, Well 10,Rivendell) to basic (pH 8.13, Strawberry Creek).The percent error of Fe, Mn, Al, major cations, Sr and Ba

calculated using SLRS-5 were 3%, 0.4%, 10%, 7%, 1%, and 6%,respectively.(1). Filter Performance. Series of 20 mL Reference

samples were taken from the same volume (40−60 mL) ofwater as the GFS samples and analyzed for Fe and Mn (Figure3; detailed data available in SI Figure S2). Fe and Mnconcentration ranges were from <2 nM to 1300 nM and from<2 nM to 1700 nM, respectively. Mn showed no change with

progressive volume filtered; while Fe decreased by up to 35%only when concentrations were extremely low (10−20 nM),(Figure S2). However, Fe concentrations vary 2 orders ofmagnitude at each wells throughout seasons (monthly referencesamples collected over 2 years); thus, reference sample volumesfrom 40 to 60 mL are not significantly biased by filter cloggingeffects.

(2). Trace Metals. The GFS improved the sample integrityfor trace metals in a major way (Figure 4). Over 30 days ofstorage, the GFS treatment recovered Fe and Mn, on average,95 ± 19% (Standard Deviation, SD) and 105 ± 7% whereas thecontrol samples recovered 16 ± 16% and 18 ± 14%,

Figure 2. Schematic overview of the GFS evaluation procedure andassigned sample names.

Figure 3.Mean of normalized Fe and Mn in 0.45 μm Supor filtrates ofReference samples as a function of cumulative sample volume.

Figure 4. Comparison of selected trace metals (Fe, Mn, and Al) andmajor cation (Ca) concentrations in GFS evaluation samples (y-axis)vs in the reference samples (x-axis). (note: † Gravitation FiltrationSystem treatment; ‡ nonmodified Swinnex Filter holders with GFS; $ISCO LDPE sample bag with gravitation filtration; * Control).



respectively. The recovery of Fe depended on the filter typeand required the use of air-outlet filter holders while Mn wassuccessfully recovered in all GFS sets.Fe is more rapidly oxidized than Mn17 and therefore is more

sensitive to sampling treatment. Once Fe precipitates, colloidalphase contributes significant portion of the element. Colloidsstart to be concentrated above the filter surface onset offiltration; therefore the probability of colloidal coagulationincreases and even can be higher when the filtration rate isslow.18 The impact of filtration artifacts also increases when thefilter surface area is smaller. The Polycarbonate filters havelower pore area than the Supor filters; therefore it is moreprone to filter clogging by high loading of colloids and particles.The results confirmed this discussion. First, the average

recovery of Supor GFS of Fe and Al, which showed similarbehavior with Fe, were 95 ± 19% and 116 ± 28%, respectively(Figure 4). In contrast, Polycarbonate GFS sets achieved only38 ± 36% and 42 ± 35% recovery of Fe and Al, respectively(Figure 4). The filtration rate of air-outlet GFS was higher (>5mL min−1) than that of GFS-NM (0.4−1.5 mL min−1); as aconsequence, the GFS showed no more than 35% and 27% lossof Fe and Al at low concentrations while the GFS-NM showedup to 85% and 52% loss, respectively (Figure 5 and Al in SIFigure S3).In contrast, Mn in Rivendell groundwater may be mostly in

dissolved form during the filtration, likely Mn (II) because theMn concentration was high (>1 μM) and dissolved oxygenconcentration was low (<30 μM).16 Thus Mn is less affected bythe filtration artifacts. The main process that causes Mn lossfrom the solution is oxidative precipitation and GFSsuccessfully separated the dissolved and precipitated phasesbefore this process has time to operate.Figure 4 further shows that GFS-BAG samples often yielded

Al concentrations a factor of 10 higher than reference levels.Tests with 1% HNO3 acidified Milli-Q water in acid washedISCO ProPak sample bags yielded Al blank levels as high

as1500 nM confirming that the bags were the origin of theelevated Al.

(3). Ca and Other Cations. The concentration of Ca insome Rivendell groundwaters (Well1 and Well3), unlike othermajor cations such as Na, K, Mg decreased by 18−42% withoutthe GFS due to CaCO3 precipitation (Figure 4). The pCO2 ofRivendell subsurface often is 20−100 times higher than theatmosphere; thus, these samples are prone to CO2 degassingand pH increase after the sampling. The concentration of Ca ofthese groundwater samples was high enough to be super-saturated with respect to CaCO3 after the equilibrium with theatmosphere. The GFS fully recovered this precipitated CaCO3,regardless of the air-outlet modification.When Ca precipitates in some groundwater samples, the

concentrations of other divalent elements of the Well 1samplesSr and Badecreased by 7% and 8%, respectively.This change may be due to the CaCO3 precipitation bysubstituting Ca.19

(4). Dissolved Organic Carbon. Previous studies haveaddressed the preservation of dissolved organic carbon (DOC)along with nutrient species, for example, see refs 1 and 20, andhave found that DOC preservation was determined by samplematrix, DOC concentration, and character of DOC.20,21 Giventhe importance of DOC to trace metal dynamics, theperformance of the GFS for this parameter was also evaluated.The preservation of DOC was closely related to the pore-size

of filters; sample types and the concentration of DOC (Table2). The 0.45 μm Supor GFS set of Rivendell groundwatersamples with low DOC concentrations (400−900 ppb C) andsurface water samples (Strawberry Creek and SF bay) with highDOC (1500−2500 ppb C) recovered 96% and 93% of theDOC, on average, respectively. In contrast, Strawberry Creekand SF bay samples filtered with the 1.0 μm polycarbonate GFSdecreased substantially, up to 30% over 4 weeks. This differencemay be due to microbial activity because the larger pore-sizefilters were not able to properly separate out microbes, allowingDOC utilization. In cases of Rivendell groundwater samples

Figure 5. Comparison between Fe concentration in 0.45 μm Supor filtrates of Reference samples versus the GFS and GFS-NM samples.

Table 2. Summary Results of DOC (Standard Deviation) Preservation of the GFS Evaluation

sample SB Creek SF Bay Elder Creek Well 1 Well 3 Well 10

Conca.(ppb) 2522 1704 1512 766 503 1632 477Aging period (days) 27 27 32 (8) 36 (5) 31 (8) 31 (8) 33 (8)Preservation (%)0.45-μm Supor 95 (−) 90 (−) 78 (19) 96 (24) 107 (9) 53(1) 122 (7)0.8-μm Supor 88 (−) 70 (−)1.0-μm PC 83 (−-) 73 (−)

aConcentration.



with high DOC (<1500 ppb C) values, even with the 0.45 μmSupor GFS set, the recoveries were by 53% on average (Table2). Fellman et al. (2008) reported that the chemical quality ofDOC (i.e., specific ultraviolet absorbance) significantly affectedon the preservation efficiency of DOC by freezing method.21

Our low recovery rate of DOC in some groundwater samplesmay be related to the quality of DOC; it requires furtherinvestigation.Controlling the DOC blank of the GFS has been a challenge

because all components of the GFS are plastic, particularlyfilters. Blanks are especially important to manage sinceRivendell groundwaters generally have subppm DOC levels.Kaplan (1994) reported that polysulfone, the substrate forSupor membrane filters, was a source if significant amount ofDOC;22 a more recent study demonstrated that soaking filtersin Milli-Q water for 3−4 days eliminates the source of leachableDOC.15 Our work shows that this treatment method reducesDOC blanks from plastic parts and Supor filters to undetectablelevels.In contrast, an evaluation of DOC blanks from ISCO ProPak

sample bags using Milli-Q showed that DOC was produced atan almost constant rate of 23 ppbC/day. ISCO ProPak samplebags cannot be used for DOC.(5). Field Deployment/Evaluation of GFS Autosam-

pling. The GFS was employed at the Rivendell Well3 fromOctober 3 to November 2, 2011 (Figure 6). In this test, fourreference samples were simultaneously collected with GFSsamples from October 2 to 7. The GFS recovered Fe and Mn,on average, 93 ± 7% and 103 ± 2%, respectively. The high Mnconcentrations at Rivendell Well3 are consistent with Mn(II)favorable, suboxic conditions. This result confirms that GFS issuitable for the trace metals sampling, even in the mostchallenging environments (i.e., anoxic, suboxic) which cancause rapid loss of trace metals via oxidative precipitation.

■ IMPLICATIONS

The GFS developed in this study has been proven to maintainthe dissolved/particulate partitioning of reactive trace metals,especially Fe and Mn, over month long periods both in thelaboratory and in the field. The GFS thus enables autonomoushigh-frequency sampling of trace metals in systems far fromatmospheric equilibrium (high CO2 and low O2) without theconstant attention of personnel in the field.Over eight months of field operation, few negatives to the

technique have been found, but all these can be addressed.First, approximately 4% of the GFS samples have been lostmainly due to the disconnection of the Tygon tubinginterconnecting the filter holder and the sample bottles. TheTygon tubing has been replaced by a female Luer-Lock-barbed

PP tube coupling (McMaster, tube ID 4.0 mm) which is moreresistant to acid than the Tygon tubing. Second, the GFS canbe prone to cross-contamination between adjacent bottles byoverpumping sample because the GFS can contain only 140mL. It is difficult to take an exact sample volume, particularlyduring the rainy season when the water table depth fluctuatessignificantly. To address this problem, a couple of holes havebeen be drilled near the top end of the syringe to drain theexcess from the samples. Third, since the method is new,reference samples should be collected in any new environmentwhere the method is applied in order to validate results.There are many positives. The GFS employs analyte specific

bottle sets instead of a single bag or bottle, thus enablingchemical analyses requiring different sample pretreatments(acidification vs nonacidification). The GFS method can bewidely used since all parts are commercially available, low cost,and easily assembled. The GFS further yields samples of filteredparticles. The filters can be leached or completely digested orsubjected to nondestructive analysis techniques, such asscanning-electron microscopy or X-ray diffraction to studythe morphology, element-phase association, and mineralogy ofthe particles. Thus, the GFS has significant potential to improvethe representativeness of water chemistry data by providingsamples on temporal/spatial scales heretofore unexplored.

■ ASSOCIATED CONTENT*S Supporting InformationFigures S1−S3 and Table S1. This material is available free ofcharge via the Internet at http://pubs.acs.org.

■ AUTHOR INFORMATIONCorresponding Author*Phone: 510 277 5616; fax: 510 643 9980; e-mail: [email protected].

NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThis study was funded by the W.M. Keck Foundation (KeckHydroWatch Center award) and the National ScienceFoundation award (ATM-0628678). We are also grateful tothe University of California Natural Reserve System forproviding the Angelo Reserve as a protected site for ourresearch. We thank the three reviewers for their contributionsto the final form of this paper. We acknowledge Michael Fongfor helping with DOC blank evaluation and for field support.We thank Nolan Wong, Tim Ault, and Ernesto Martinez forassistance with preparing the experiments and field campaign.

Figure 6. Time- series of Fe and Mn in 0.45 μm Supor GFS samples at Rivendell Well 3 from October 3 to November 2, 2011 at the beginning ofrainy season of water year 2012. Reference samples were manually collected from October 3 to 7.



http://pubs.acs.org

mailto:[email protected]

mailto:[email protected]

Support for all students was through the UC BerkeleyUndergraduate Research Apprentice Program.

■ REFERENCES(1) van Verseveld, W.; McDonnell, J. J.; Lajtha, K. A mechanisticassessment of nutrient flushing at the catchment scale. J. Hydrol. 2008,358, 268−287, DOI: 10.1016/j.jhydrol. 2008/06.009.(2) Ribolzi, O.; Karambiri, H.; Bariac, T.; Benedetti, M.; Caquineaux,S.; Descloitres, M.; Aventurier, A. Mechanisms affecting stormflowgeneration and solute behavior in a Sahelian headwater catchment. J.Hydrol. 2007, 337, 104−116, DOI: 10.1016/j.jhydrol.2007.01.019.(3) Inamdar, S.; Rupp, J.; Mitchell, M. Groundwater flushing ofsolutes at wetland and hillslope positions during storm events in asmall glaciated catchment in western New York, USA. Hydrol. Process2009, 23 (13), 1912−1926, DOI: 10.1002/hyp.7322.(4) Morrison, M. A.; Benoit, G. Temporal variability in physicalspeciation of metals during a winter rain-on-snow event. J. Environ.Qual. 2005, 34, 1610−1619, DOI: 10.2134/jeq2004.0324.(5) Gammons, C. H.; Nimick, D.; Parker, S. R.; Cleasby, T. E.;McClesky, R. B. Diel behavior of iron and other heavy metals in amountain stream with acidic to neutral pH: Fisher Creek Montana,USA. Geochim. Cosmochim. Acta 2005, 69 (10), 2505−2516; DOI10.1016/j.gca.2004.11.020(6) Isaac, R. A.; Gil, L.; Cooperman, A. N.; Hulme, K.; Eddy, B.;Ruiz, M.; Jacobson, K.; Larson, C.; Pancorbo, O. C. Corrosion indrinking water distribution systems: A major contributor of copperand lead to wastewaters and effluents. Environ. Sci. Technol. 1997, 31(11), 3198−3203.(7) Tramontano, J. M.; Scudlark., J. R.; Church, T. M. A method forthe collection handling and analysis of trace-metals in precipitation.Environ. Sci. Technol. 1987, 21 (8), 749−753.(8) Subramania, K. S.; Chakrabarti, C. L.; Sueiras, J. E.; Maines, I. S.Preservation of some trace-metals in samples of natural waters. Anal.Chem. 1978, 50 (3), 444−448.(9) Stumm, W.; Morgan, J. J. Aquatic Chemistry: Chemical Equilibriaand Rates in Natural Waters, 3rd ed.; Wiley: New York, 1996.(10) Nielsen, D.; Nielsen, G. L. The Essential Handbook of Ground-Water Sampling. CRC Press/Taylor & Francis: Boca Raton, FL, 2007.(11) Method 1669: Sampling Ambient Water for Trace Metals at EPAWater Quality Criteria Levels; United States Environmental ProtectionAgency, Office of Water, Engineering and Analysis Division:Washington, DC, 1996; http://purl.access.gpo.gov/GPO/LPS44979.(12) Horowitz, A. J.; Lum, K. R.; Garbarino, J. R.; Hall, G. E.;Lemieux, C.; Demas, C. R. Problems associated with using filtration todefine dissolved trace element concentrations in natural water samples.Environ. Sci. Technol. 1996, 30 (3), 3398−3400.(13) Morrison, M. A.; Benoit, G. Filtration artifacts caused byoverloading membrane filters. Environ. Sci. Technol. 2001, 35 (18),3774−3779.(14) Bishop, J. K.; Lam., P. J.; Wood, T. J. Getting good particles:Accurate sampling of particles by large volume in-situ filtration.Limnol. Oceanogr.: Methods 2012, (Accepted).(15) Khan, E.; Subramania-Pillai, S. Interferences contributed byleaching from filters on measurements of collective organicconstituents. Water Res. 2007, 41 (9), 1841−1850, DOI: 10.1016/j.waters.2006.12.028.(16) Sposito, G. The Chemistry of Soils, 2nd, ed.; Oxford UniversityPress: Oxford; New York, 2008.(17) Davison, W. Iron and manganese in lakes. Earth-Sci. Rev. 1993,34, 119−163.(18) Buffle, J.; Leppard, G. G. Characterization of aquatic colloidsand macromolecules: 2. Key role of physical structures on analyticalresults. Environ. Sci. Technol. 1995, 29 (9), 2176−2184.(19) Lorens, R. B. Sr, Cd, Mn and Co Distribution coefficient incalcite as a function of calcite precipitation rate. Geochim. Cosmochim.Acta 1981, 45 (4), 553−561.(20) Ensign, S. H.; Paerl, H. W. Development of an unattendedestuarine nutrient monitoring program using ferries as data-collectionplatforms. Limnol. Oceanogr.: Methods 2006, 4, 399−405.

(21) Fellman, J. B.; D’Amore, D. V.; Hood, E. An evaluation offreezing as a preservation technique for analyzing dissolved organic C,N and P in surface water samples. Sci. Total Environ. 2008, 392 (2−3),305−312, DOI: 10.1016/j.scitotenv.2007.11.027.(22) Kaplan, L. A. A field and laboratory procedure to collect,process, and preserve fresh-water samples for dissolved organic carbonanalysis. Limnol. Oceanogr. 1994, 39 (6), 1470−1476.



http://purl.access.gpo.gov/GPO/LPS44979

autonomous water sampling for long-term monitoring...

Documents