Does Process Matter?

An Evaluation of the Effect of Sample Processing Treatments on Alkalinity Measurements

Amy Williams

Presented in partial fulfillment for the degree of Bachelor of Science in Earth and Environmental Sciences

Furman University

2007

Acknowledgments

Thanks to Connie Gawne and Lee Mitchell, South Carolina Department of Natural Resources, for their assistance in locating and sampling Gowensville wells, to Wes Dripps for assistance in field sampling, and to Suresh Muthukrishnan for GIS assistance. Special thanks to Tracy Jones, Department of Geology, University of Tennessee-Chattanooga, for her assistance in the field and other logistical support. Thanks to Chris Priedemann for baking cookies for the night spent in the lab. Very special thanks to Ivan Irizarry for assistance with field sampling, alkalinity analysis, GIS mapping and being there. Also very special thanks to Selena “Spang” Pang to assistance in field sampling and maintaining sanity. Thanks to Lori Nelsen for her guidance and assistance in laboratory analyses and listening to the griping. Finally, the most thanks goes to my advisor, Brannon Andersen, for suggesting the project and pushing me to work harder than I ever thought possible.

Funding was provided by a NSF-Research Experiences for Undergraduates grant (NSF-REU Grant #EAR-0453205), NSF-MRI Grant #EAR-0116487, Saluda Reedy Watershed Consortium Grant C2-04, and by Furman University.

Abstract

The concentration of bicarbonate, the dominant anion in river water, is calculated using measured alkalinity concentrations. Although the Gran Titration is the standard research method for measuring alkalinity, there is variation in how water samples are processed. The purpose of this study was to determine whether variation in processing methods leads to statistically significant differences in alkalinity concentrations in 1) stream and ground waters influenced by high-grade silicate metamorphic rocks and in 2) stream waters influenced by carbonate rocks.

Samples were collected from the piedmont region of northwestern South Carolina and the ridge and valley region of eastern Tennessee. Aliquots of each sample were collected in triplicate using four processing treatments: filtered and refrigerated, filtered and unrefrigerated, unfiltered and refrigerated, and unfiltered and unrefrigerated. All aliquots were analyzed within 24 hours using the Gran Titration method. Two-way ANOVA results indicate that there were predominantly no statistically significant differences among the various treatments any of the water samples. Few samples demonstrated some statistically significant differences, but the difference between the means was less than 3%. The alkalinity concentrations of all samples were determined again after 17 to 51 days of storage to determine if storage time affects alkalinity measurements.

Our results suggest that alkalinity concentrations measured on surface water and ground water samples collected in the piedmont region and surface waters collected in the valley and ridge region are not dependent on processing treatment. Our recommended treatment of fresh surface water samples is filtered and refrigerated to maintain the chemical integrity of the samples.

Introduction

Rivers serve as a major carbon transportation system between terrestrial sources and the ocean (Ramesh et al. 1995; Edmond et al. 1995). They are also a major source of carbon flux to the atmosphere during transit (Richey et al. 2002; Oechel et al. 2000). Thus it is important to have an understanding of the biogeochemical cycling of carbon in rivers and the geologic sources and sinks that feed this cycle. Forms of riverine carbon can include dissolved organic carbon (DOC), particulate organic carbon (POC), particulate inorganic carbon (PIC), and dissolved inorganic carbon (DIC) (Meybeck et al. 2003). In the waters we studied, DIC was typically in the form of bicarbonate (Dawson et al. 2002; Meybeck 1993).

Bicarbonate is generally the most important anion in 98% of world rivers flowing into the ocean (Meybeck 1993) and is the dominant carbonate species in 50% of river basins and 40% of U.S. ground waters (Meybeck et al. 2003; Eby et al. 2004). In the typical pH range (pH 6-8.2) for surface and ground waters, bicarbonate DIC can contribute a significant amount to the carbon flux between terrestrial waters and the atmosphere. The two sources of DIC are i) silicate weathering by carbonic acid (including oxidation of sedimentary organic matter found in some rocks and uptake of atmospheric and soil CO₂ during weathering (Meybeck 1993)) and ii) dissolution of carbonates, such as calcite and dolomite (Meybeck 2004). In rivers influenced by carbonates, DIC is controlled by both Pco₂ in soils and the ambient temperature (White, 1984).

Of the forms of dissolved carbon, it is most problematic to quantify the flux of DIC because it consists of 3 carbonate species and easily degases. Thus the method of sample processing before analysis is important to preserving the integrity of the alkalinity. Total dissolved inorganic carbon (DIC) can be measured directly (Raymond et al. 1997), although this method is less common. More commonly, DIC and inorganic carbon species are calculated using a thermodynamic model utilizing stream or ground water temperature, pH, and alkalinity as model variables. Of these three variables, pH and alkalinity have the greatest uncertainty, with temperature measurements being generally reliable. The uncertainty in pH measurements in natural waters (Piñol et al. 1992; Raymond et al. 1997) can be attributed to high sample Pco₂ (partial pressure of carbon dioxide) values (Cai et al. 1998), variable temperatures, errors due to using a large buffer range (Howland et al. 2000), and analytical error, i.e. electrode aging and variation in liquid junction (Pytkowicz et al. 1966). Despite these uncertainties, the focus of this study will be on the determination of alkalinity.

Alkalinity often is the key to the calculation of Pco₂, charge balance, and carbonate speciation. The Pco₂ in rivers is important to the global carbon cycle because it can be used to describe carbon fluxes to the atmosphere and the ocean (Neal et al. 1998b). Charge balances provide us with an error range to check the accuracy of our results. The carbonate speciation is calculated from measurements of pH, temperature, and alkalinity (McDowell et al. 1994; Raymond et al. 1997), provided alkalinity is primarily in the form of bicarbonate and carbonate. The variation in carbonate speciation is a function of the natural weathering of the underlying geology, the contributions from headwater streams, and the introduction of anthropogenic sources (Richey et al. 1990).

Gran Titration is the standard method for measuring alkalinity, but there is some variation in how surface water samples are processed. In the Standard Methods (1995), it is recommended that alkalinities are measured within 24 hours of collection on unfiltered samples, with a maximum of 14 days of refrigerated storage. A variety of methods have been used to demonstrate a range of treatment methods with conditions affecting titration, filtration, refrigeration, depth of grab samples, and method of alkalinity measurement (Table 1). There is no consensus on which is the most reliable and there is no consistent method that is used in all of the following papers, although the Gran Titration method of alkalinity measurement was used in 26 of the 38 papers surveyed.

Samples can be titrated in the field (Pinol et al. 1992; Barth et al. 2003; Richey et al. 1990; Cameron et al. 1995; Grimaldi et al. 1999; Hoffer-French et al. 1989; Katz et al. 1985), or in the lab (Cai et al. 1998; Zhang et al. 1995; Standard Methods 1995; Rice et al. 1995; Grosbois et al. 2001; Guasch et al. 1998; Helie et al. 2002; Hill et al. 2002; Hoffer-French et al. 1989; Howland et al. 2000; Huh et al. 1998; Jones Jr. et al. 1998; Kim et al. 1996; Lewis Jr. et al. 1979; Lewiss Jr. et al. 1987; McDowell et al. 1994; Pamde et al. 1994; Ometo et al. 2000; Edmond et al. 1995). Samples can be filtered (Neal et al. 1994; Grosbois et al. 2001; Huh et al. 1998; Rice et al. 1995; Zhang et al. 1995; Jarvie et al. 2002; Katz et al. 1985; Pande et al. 1994; Williams et al. 2001; Wu et al. 2005; Cameron et al. 1995; Finley et al. 1997; Howland et al. 2000; Kim et al. 1996; Edmond et al. 1995) or not (Standard Methods 1995; Dawson et al. 1995; Hill et al. 2002; Hoffer-French et al. 1989; Jones Jr. et al. 1998; Lesack et al. 1991; Lewis Jr. et al. 1979; Neal et al. 2000, Edmond et al. 1995), and refrigerated before anaylsis (Rice et al. 1995; Standard Methods 1995; Billett et al. 1996; Dawson et al. 2001; Guasch et al. 1998; Helie et al. 2002; Hill et al. 2002; Hoffer-French et al. 1989; Jarvie et al 2002; Jones Jr. et al. 1998; Katz et al. 1985; Kim et al. 1996; Lesack et al 1991; Raymond et al 1997; Neal et al. 2002) or not (Barth et al. 2003; Howland et al. 2000; Lewis Jr. et al. 1979; Edmond et al. 1995). With all of the water treatment methods available, a study of reliable treatment methods is necessary to establish realistic expectations for the treatment of waters intended for alkalinity measurement.

Table 1. Summary of treatment methods from previous water chemistry studies. ?= not reported; In lab=no time limit specified
Author	Titration occurred:	Filtration occurred:	Refrigeration conditions:	Method of Alkalinity Measurement
Piñol 1992	On site	Nr	Nr	End Point
Neal 1994	Nr	On site	Nr	Gran Acidity
Neal 1988	2 hours	Nr	Nr	Modified Gran Titration
Cai 1998	In lab	Nr	Nr	Gran Titration
Barth 2003	On site	Nr	Field Temperature	End Point
Zhang 1995	In lab	On site	Nr	Gran Titration
Rice 1995	In lab	Yes	At 4°C	Modified Gran Titration
Standard Methods 1995	24 hours	No filtering	At 4°C	End Point
Richey 1990	On site	Nr	Nr	Gran Titration
Billett 1996	24 hours	Nr	At 4°C	Nr
Cameron 1995	On site/ 24 hour	12 hours	Nr	Nr
Dawson 2001	Nr	No filtering	At 4°C	End point (pH 4.5 and 4.0)
Finley 1997	4 hours	Same day	Nr	Gran Titration
Grimaldi 1999	On site	Nr	Nr	Gran Titration
Grosbois 2001	In lab	On site	Nr	Gran Titration
Guasch 1998	In lab	Nr	At 4°C	Gran Titration
Helie 2002	In lab	On site	At 4°C	Gran Titration
Hill 2002	In lab	No filtering	At 4°C	Modified Gran Titration
Hoffer-French 1989	On site/ In lab (within 48 hours)	No filtering	On ice when in lab	Gran Titration
Howland 2000	In lab	Same day	Room temperature/ in the dark	Gran Titration
Huh 1998	In lab	On site	Nr	Gran Titration
Jarvie 2002	24 hours	On site	At 4°C	Modified Gran Titration
Jones Jr. 1998	In lab	No filtering	At 4°C	End Point
Katz 1985	On site	On site	At 4°C	Gran Titration
Kim 1996	In lab	12 hours	At 4°C	Gran Titration
Lesack 1991	24 hours	No filtering	At 4°C	Gran Titration
Lewis Jr. 1979	Just after collection in lab	No filtering	Room temperature	End Point
Lewis Jr. 1987	In lab	Just after collection	Nr	Gran Titration & End Point
McDowell 1994	Just after collection in lab	Nr	Nr	IR spectroscopy (DIC)
Pande 1994	In lab	On site	Nr	Titrimetry
Ometo 2000	In lab	Yes	Nr	Gran Titration
Raymond 1997	24 hours	nr	At 4°C	Syringe gas stripping method (Stainton 1973) (DIC)
Risacher 2002	nr	Yes	Nr	Electrometric titration
Williams 2001	In lab	On site	Nr	Gran Titration
Wu 2005	In lab	On site	Nr	Gran Titration
Neal 2000	3 days	No	At 4°C	Modified Gran Titration
Neal 2002	In lab	Nr	Nr	Modified Gran Titration
Edmond 1995	In lab (weeks later)	5 hours- several days	No	Gran Titration

The purpose of our study was to determine whether variation in processing treatments (filtration, refrigeration, and storage time) lead to statistically significant differences in alkalinity concentrations. This research was conducted on surface and ground water samples from the piedmont region of northwestern South Carolina and the valley and ridge region of eastern Tennessee. If alkalinity measurements do not show a meaningful statistically significant difference between treatment methods, then we assume that samples can undergo any treatment method without reducing the accuracy of charge balance, carbonate speciation and PCO₂ calculations. Not having a meaningful difference is defined as having a change in alkalinity measurements so small that the change in the final calculations based on these measurements does not affect the outcome of the study (from sample locality NC02, a change in alkalinity of 1 mg CaCO₃/L does not affect the results of the study). Our previous studies demonstrated that alkalinity values measured in the lab provided charge balances well within 10%. Based on this data, we did not compare the results of field versus lab measurements.

This study found that there were no meaningful statistically significant differences between filtration, refrigeration, or storage time for alkalinity measurements from piedmont and carbonate surface and ground waters with low turbidity and undersaturated with respect to calcite. Because the range of alkalinities sampled in this study represent over 85% of world surface waters (Table 2), and 40% of U.S. ground waters (Table 3), we concluded that, while this research is widely applicable, it is not to be used without considering other geochemical effects.

Table 2. Bicarbonate concentrations from rivers around the world. (from Berner et al., 1996)
Continent	River	HCO3- (mg/L)	Reference
N. America	Colorado 1960s	135	A*
	Columbia	76	A
	Mackenzie	111	A
	St. Lawrence	75	A
	Yukon	104	A
	Mississippi 1905	116	A
	Mississippi 1965	118	A
	Frazer	60	A
	Nelson	144	A
	Rio Grande:Laredo	183	B*
	Ohio	63	B
Europe	Danube	190	A
	U. Rhine:unpolluted	114	C*
	U. Rhine:polluted	153	C
	Norwegian rivers	12	A
	Black Sea rivers	136	A
	Icelandic rivers	35.5	A
S. America	U. Amazon:Peru	68	D*
	U. Amazon:Brazil	20	D
	L. Negro	0.7	D
	Madeira	28	D
	Parana	31	A
	Madgdalene	49	A
	Guyana rivers	12	A
	Orinoco	11	A
Africa	Zambeze	25	A
	Congo (Zaire)	13.4	E*
	Ubangui	19	E
	Niger	36	A
	Nile	134	A
	Orange	107	A
Asia	Brahmaputra	58	F*
	Ganges	127	F
	Indus	90	A
	Mekong	58	A
	Japanese rivers	31	A
	Indonesian rivers	26	A
	New Zealand rivers	50	A
	Yangtze (Changjiang)	120	G*
	Yellow (Hwanghe)	182	G
	Ob	79	H*
	Yenisei	74	A
*A= Meybeck 1979, B=Livingstone 1963, C=Zobrist & Stumm 1980, D=Stallard 1980, E=Probst 1992, F=Sarin 1989, G=Zhang 1990, H=Gordeev & Siderov 1993.

Table 3. Bicarbonate concentrations from US ground waters. (from Eby et al., 2004)
State	Dominant Rock	HCO3^- (mg/L)	Reference
Central Florida	Carbonate	124	A
Central Pennsylvania	Carbonate	279	B
Montana	Sandstone	2080	C
New Mexico	Gypsum	143	C
California	Serpentine	1300	C
Rhode Island	Granite	38	C
Maryland	Gabbro	37	C
Hawaii	Basalt	84	C
New Mexico	Rhyolite	77	D
North Carolina	Mica schist	69	D
West Virginia	Sand and gravel	101	D
Alabama	Limestone	146	D
*A=Back & Hanshaw 1970, B=Langmuir 1971, C=Matthess 1982, D=Hem 1970.

Study Area

Surface and ground water samples were collected from seven localities in the piedmont region of South Carolina (Fig.1) and four localities in the valley and ridge region of eastern Tennessee (Fig.2). Sample localities were chosen to represent the greatest range of alkalinities possible, dependent on the underlying geology, i.e. non-soluble compared to highly soluble minerals.

Piedmont Region Localities

The piedmont region is characterized by metamorphic and igneous silicate rocks (granites and gneisses) with low solubility minerals (Overstreet et al. 1965). Ultisols composed of sandy clay loam saprolites dominate the watershed (Byrd, 1972). Thus surface and ground waters have low conductivity and low alkalinity (Lewis et al. 2007). The three surface water sample localities were chosen from the Saluda and Reedy River Basins to represent a characteristic range of alkalinities (high at CL01 to low at US64).

Ground water samples were collected from four wells ranging from 5.2 to 101.8 meters deep located in the Reedy, Enoree, and Pacolet River Basins. The localities include shallow and deep wells chosen to represent a characteristic range of alkalinities (high at GV W1 to low at BA W1). Locality GV W1 was chosen for depth and being screened in bedrock. Locality SA W1 was chosen as a comparison for GV W1 because it is within 1 meter of GV W1 and is shallowly screened in saprolite.

Valley and Ridge Region Localities of East Tennessee

The carbonate localities in the valley and ridge area of the Lower Tennessee River Basin are characterized by Paleozoic age carbonate rocks (Pavlicek et al. 1996). The three carbonate surface water samples collected from sections of the North Chickamauga Creek were chosen to represent a characteristic range of alkalinities (high at NC02 to low at NC01). The first locality, NC01, was chosen because it drains the carbonates of the Newman Limestone. It is also about one mile downstream of three mines which have contributed acid mine drainage to the system. The second locality, NC02, drains the carbonates of the Knox Dolomite. The third locality, NC03, drains the carbonates of the Copper Ridge Dolomite.

One sample was chosen from Douglas Lake (DL01) in the Lower French Broad River Basin Watershed because it drains the carbonate-cemented Tellico Sandstone and is chemically influenced by the Lenior Limestone. We chose to sample DL01 because our first limestone sample, NC01, was influenced by acid mine drainage and did not yield the expected alkalinity measurement.

Figure 1. Map of the surface and ground water localities sampled in the South Carolina piedmont. Surface water samples include US64, FU-C 03, and CL01. Ground water samples include BA W1, BY32 W1, GV W1 and SA W1. Watersheds included are the Saluda, Reedy, Enoree, and Pacolet.

Figure 2. Map of the surface localities sampled in the east Tennessee carbonates. Surface water samples from the North Chickamauga Creek in the Lower Tennessee River Basin Watershed include NC01, NC02, and NC03. The surface water sample DL01 was collected from Douglas Lake in the Lower French Broad River Bain Watershed.

Methods

Samples were collected from the following locations: 3 from piedmont surface waters, 4 from carbonate surface waters, and 5 from piedmont ground waters. Piedmont surface waters were chosen because they demonstrated low, medium, or high alkalinity, known from previous unpublished data. Carbonate surface water and piedmont ground water samples were chosen based on field crew assumptions of the local geology’s effects on expected alkalinity.

In the field, surface water samples were collected in a 4L pre-cleaned, acid-rinsed high density polyethylene bottle. Prior to collection, each bottle was pre-contaminated three times with the sample water and emptied downstream from the locality. A separate sample was collected for turbidity analysis. Groundwater samples were collected in four 500mL bottles (to be filtered) and twelve 125mL bottles (left as unfiltered). To minimize degassing, all samples were collected with zero headspace, either by submerging the bottle and capping under water or filling the bottle to overflowing and then capping. pH, conductivity, dissolved oxygen, and temperature were measured in the field with a Fisher Scientific AP62 Accumet pH meter, YSI 30 salinity/conductivity meter and YSI 55 dissolved oxygen meter.

In the laboratory, samples were processed using four different methods with three replicates per method. Each method consists of a different combination of filtering and refrigeration. Samples were filtered through a 0.45µm membrane filter using an N₂ gas positive pressure filtration system. Refrigerated samples were stored at 4°C in the dark for both 24-hour and timed analyses. The four method combinations were filtered-refrigerated (FR), filtered-unrefrigerated (FU), unfiltered-refrigerated (UR), and unfiltered-unrefrigerated (UU). The FR method functions as our control, as this is the standard method in our research program.

Alkalinity was measured with an Accumet AR-25 dual channel pH/ ion meter using an Accumet electrode and a Kimax 5mL buret. Sample alkalinities were measured within 24 hours and again after a storage period ranging from 17-51 days to determine the effect of sample storage. Alkalinity was measured using the Gran Titration method. All piedmont surface water and ground water samples and carbonate sample NC01, were titrated with 0.02N H₂SO₄. All other carbonate samples were titrated using 0.2N H₂SO₄.

Between 17 (Conestee Lake) and 51 (Furman Lake) days after their respective collection dates, the unrefrigerated piedmont surface water samples were measured again to test for time-based alkalinity concentration changes. Between 18 (Brushy Creek 2.13.07) and 50 (Bunched Arrowhead) days after their respective collection dates, all carbonate surface water samples and sealed piedmont ground water samples were measured again to test for time-based alkalinity concentration changes.

The chemical analyses were run within one week of sample collection. From the original sample, an aliquot was preserved with trace metal-grade nitric acid for cation analysis. Cation concentrations (Na⁺, Ca²⁺, K⁺, Mg²⁺, Si⁴⁺, Fe²⁺) were measured with an ICP-AES. Another aliquot was preserved with trace metal-grade sulfuric acid for TDN analysis with an Alpkem Flow Solution IV. One unpreserved FR sample from each locality was used to measure anions, NH₄⁺, and DOC. Anion concentrations (F^-, Cl^-, Br^-, NO₃^-, NO₂^-, H₂PO₄^-, SO₄^2-) were measured with a Dionex DX-120 Ion Chromatograph, ammonium (NH₄⁺) with a 10-AU Fluorometer, and dissolved organic carbon (DOC) with a Tekmar Dohrmann Phoenix 8000. Turbidity was measured the same day of collection with a LaMotte 2020 turbidity meter. Charge balance was calculated using the method of Freeze and Cherry (1979).

Using SigmaStat software, two-way ANOVA tests were used to determine statistically significant differences in the treatments of 24-hour and timed samples independently. T-tests, Wilcoxon Rank Sum tests, and Mann-Whitney Rank Sum tests were used to determine any statistical differences in treatments between 24-hour and timed samples.

Equilibrium constants were calculated using the equations of Plummer and Busenburg (1982), and activity coefficients were calculated using the extended Debye-Huckel equation. The ratio of PCO₂ (aq) in river water to PCO₂ in the atmosphere indicates the degree of CO₂ supersaturation in the river relative to the atmosphere (Piñol et al. 1992; Huh et al. 1998). We refer hereafter to this ratio as PCO₂ saturation.

The equation for the calculation of the partial pressure of carbon dioxide (PCO₂), which uses measurements of water temperature, pH, and HCO₃^- concentration, was not modified to account for the dissolution of calcite in our carbonate research (e.g. Neal et al. 1998b), because the waters were undersaturated with respect to calcite (Fig.13).

Results

Piedmont Surface Waters

Within these sites, field pH ranged from 5.26 to 7.08 and conductivity ranged from 16.2 µS/cm (US64) to 98.9 µS/cm (CL01). The lowest dissolved oxygen content measured was 2.7 mg/L (CL01) and the average temperature ranged from 9.14°C (CL01) to 17.03°C (US64)(Table 4). Ternary plots of major solutes show that US64 was more Na+K rich and silicon-rich, while CL01 was more Ca-rich and almost completely silicon-depleted, and FU-C 03 has a higher concentration of bicarbonate and was more Ca-rich (Fig.3).

The PCO₂ saturation graph demonstrates that, while both lake samples FU-C 03 and CL01 had much lower PCO₂ saturation levels than US64 (7 compared with 73), PCO₂ saturation still demonstrated a 1:1 relationship between 24-hour and timed samples (Fig.4).

The 24-hour alkalinity ranged from about 3.5 to 32 mg CaCO₃/L, as did the timed alkalinity, also ranging from about 3.5 to 32 mg CaCO₃/L (Fig.5a,b). One timed US64 FR sample (measured 21 days after collection) had a value of 6 mg CaCO₃/ L, which when included in two-way ANOVA and t-tests, made the results significant. The sample alkalinity was measured again (27 days after collection) and when included in two-way ANOVA and t-tests, it did not make the results significant. The error was attributed to a faulty electrode. The bicarbonate concentration ranged from about 4.5 HCO₃^-/L to 38.5 mg HCO₃^-/L (Table 8).

When measuring alkalinity, we asked whether 1) alkalinity varies with method, 2) alkalinity varies with time, and 3) alkalinity varies between samples run within 24 hour and those stored. The results of two-way ANOVA for filtration and refrigeration between only 24-hour samples indicated that there were no statistically significant differences. The results of two-way ANOVA for filtration and refrigeration between only timed samples indicated that there were no statistically significant differences. Finally, t-tests between 24-hour and timed samples indicated that there were no statistically significant differences between any of the treatment methods for Piedmont surface waters (Table 6).

A comparison of 24-hour to timed alkalinity measurements demonstrates that, while both lake samples FU-C 03 and CL01 had much lower alkalinity measurements than US64, a 1:1 relationship between 24-hour and timed measurements still exists, suggesting no change in alkalinity over time (Fig.6).

A charge balance graph shows the range to be ±10% for 24 hour samples and ±12% for stored samples (Fig.7). These poor charge balances can be attributed to the low alkalinity and conductivity of these headwaters. T-tests between 24-hour and timed samples indicated that there were no statistically significant differences between any treatment methods for US64 and FU-C 03. A Mann-Whitney Rank Sum test indicated that there were no statistically significant differences between any treatment methods for CL01 (Table 7).

Figure 3. Ternary plot illustrating the variation in hydrochemical facies in the waters sampled. Notice that FU-C 03 had a higher concentration of bicarbonate and was more Ca-rich, US64 was more Na+K rich and silicon-rich, and CL01 was more Ca-rich and almost completely silicon-depleted.

Figure 4. A comparison of the PCO₂ saturation levels of 24-hour samples versus timed samples demonstrates a 1:1 relationship, suggesting no change in PCO₂ saturation over time.

Figure 5a,b. Both graphs demonstrate the range of alkalinity (mg CaCO₃/L) of 24-hour and timed samples in Piedmont surface waters collected between October 2005 and March 2006. None of the 24-hour or timed samples demonstrated any statistically significant differences.

Figure 6. A comparison of the alkalinities of 24 hour samples versus timed samples demonstrated a 1:1 relationship between the measurements, suggesting no change in alkalinity over time.

Figure 7. Range of charge balance for 24 hour and timed samples, illustrating our confidence in the alkalinity measurements. 24-hour and timed samples were both within ±12%. There is no explanation for individual localities’ charge balances being consistently negative or positive.

Carbonate Surface Waters

Within these sites, field pH ranged from 5.85 to 9.15 and conductivity ranged from 56 µS/cm (NC01) to 217.3 µS/cm (NC02). The lowest dissolved oxygen content measured was 5.91 mg/L (NC03) and the average temperature ranged from 21°C (NC03) to 30.27°C (DL01)(Table 4). In contrast to piedmont surface water, the ternary plots of major solutes in carbonate surface waters showed that all four samples were Ca-rich and Na+K, Mg, and silicon depleted and all but NC01 had a high bicarbonate concentration. NC01 had a higher concentration of Cl+SO₄ due to the influence of acid mine drainage. NC02 and NC03 were both underlain by dolomite and had almost the same concentration of bicarbonate (Fig.8)

The PCO₂ saturation graph demonstrates that, while both samples underlain by dolomite had much higher PCO₂ saturation levels than DL01 or the acid mine drainage influenced NC01 (0.11 compared with 17), PCO₂ saturation still demonstrated a 1:1 relationship between 24-hour and timed samples (Fig.9).

The 24-hour alkalinity ranged from about 0.80 to 130 mg CaCO₃/L, while the timed alkalinity ranged from about 0.20 to 110 mg CaCO₃/L (Fig.10a,b). One 24-hour NC03 FR sample had a value of 156.16 mg CaCO₂/L, which when included in two-way ANOVA and t-tests, made the results significant. When the sample was not included in these statistical tests, the results were not statistically significant. The error was attributed to bottle contamination. The bicarbonate concentration ranged from about 0.7 HCO₃^-/L to 133 HCO₃^-/L (Table 8).

The results of two-way ANOVA for filtration and refrigeration between only 24-hour samples indicated that there were some statistically significant differences. The results of two-way ANOVA for filtration and refrigeration between only timed samples indicated that there were some statistically significant differences. Finally, t-tests and a Wilcoxon Rank Sum test between 24-hour and timed samples indicated that there were some statistically significant differences between filtration and refrigeration in carbonate surface waters (Table 6).

A comparison of 24-hour to timed alkalinity measurements demonstrates that, while both samples underlain by dolomite had higher alkalinity than DL01 or the acid mine drainage influenced NC01, a 1:1 relationship between 24 hour and timed measurements still exists, suggesting no change in alkalinity over time (Fig.11).

A charge balance graph shows the range to be ±10% for 24-hour samples and ±8% for stored samples with the one 24-hour NC03 FR sample at about -20% (Fig.12). T-tests between 24-hour and timed samples indicated that there were no statistically significant differences between any treatment methods for DL01. Mann-Whitney Rank Sum tests indicated that there were no statistically significant differences between any treatment methods for NC01, NC02, and NC03 (Table 7).

The calcite saturation graph indicates that none of the localities sampled were supersaturated with respect to calcite, although several carbonate surface water samples were close to saturation (Fig.13).

Figure 8. Ternary plot illustrating the variation in hydrochemical facies in the waters sampled. Notice that all four carbonate samples were Ca-rich, Na+K, Mg, and silicon depleted and all but NC01 had a high bicarbonate concentration. NC01 had a higher concentration of Cl+SO4 due to the influence of acid mine drainage. NC02 and NC03 were both underlain by dolomite and had almost the same concentration of bicarbonate.

Figure 9. A comparison of the PCO₂ saturation levels of 24 hour samples versus timed samples demonstrates a 1:1 relationship, suggesting that there is no change over time. Notice that DL01 had almost no PCO₂ saturation, suggesting that the water was nearly at equilibrium with the atmosphere.

Figure 10a,b. Both graphs demonstrate the range of alkalinity (mg CaCO₃/L) of 24-hour and timed samples in carbonate surface waters collected between July and August 2006. Depending on the treatment method, some samples demonstrated statistically significant differences, indicating changes in alkalinity dependent on treatment method.

Figure 11. A comparison of the alkalinities of 24 hour samples versus timed samples demonstrated a 1:1 relationship between the measurements, suggesting no meaningful change in alkalinity over time.

Figure 12. Range of charge balance for 24-hour and timed samples, illustrating our confidence in the alkalinity measurements. 24-hour and timed samples were both within ±10%, except for the one 24-hour NC03 FR sample at about ---20%.

Figure 13. This calcite saturation graph with error ranges demonstrates that none of the samples in this study were oversaturated with respect to calcite. For this reason, we did not need to account for the effects of calcite precipitation.

Piedmont Ground Waters

Within these sites, field pH ranged from 4.9 to 7.74 and conductivity ranged from 32.47 µS/cm (SA W1) to 118.2 µS/cm (GV W1). The lowest dissolved oxygen content measured was 2.27 mg/L (GV W1) and the average temperature ranged from 12.1°C (BY32 W1 2.13.07) to 19.47°C (GV W1)(Table 4). Ternary plots of major solutes show that BY32 W1 11.30.06, BY32 W1 2.13.07, and SA W1 were more Na+K rich. All of the samples but BY32 W1 2.13.07 were bicarbonate rich with GV W1 being almost completely composed of bicarbonate. All of the samples were silicon depleted (Fig.14). During storage, the BY32 W1 samples began to precipitate Fe²⁺. Significant amounts were detected when these samples and SA W1 were re-tested for Fe²⁺.

The PCO₂ saturation graph demonstrates that, while all of the groundwater samples had much lower PCO₂ saturation levels than BA W1 (3.1 compared with 306), PCO₂ saturation still demonstrated a 1:1 relationship between 24-hour and stored samples (Fig.15).

The 24-hour alkalinity ranged from about 6 to 60 mg CaCO₃/L, while the timed alkalinity similarly ranged from about 8 to 59 mg CaCO₃/L (Fig.16). The electrode broke in one 24-hour SA W1 FU sample, and the sample was re-measured about one hour after the bottle was initially opened. The bicarbonate concentration ranged from about 8.5 HCO₃^-/L to 73 HCO₃^-/L (Table 8).

The results of two-way ANOVA for filtration and refrigeration between only 24-hour samples indicated that there were no statistically significant differences. The results of two-way ANOVA for filtration and refrigeration between only timed samples indicated that there was one statistically significant difference in the SA W1 sample. Finally, t-tests and a Mann-Whitney Rank Sum test between 24-hour and timed samples indicated that there were some statistically significant differences between filtration and refrigeration in piedmont ground waters (Table 6).

A comparison of 24-hour to timed alkalinity measurements suggests that these are very complicated systems. Three groundwater localities demonstrated a 1:1 relationship between 24 hour and timed measurements, suggesting no change in alkalinity over time. The locality BY32 W1, sampled twice, does not fit this 1:1 line, suggesting that there may be some change in alkalinity over time (Fig.17).

A charge balance graph shows the range to be ±10% for 24-hour and timed samples with the outlying 24-hour BY32 W1 samples between -10% and -20% and the SA W1 samples between -5% and -15% (Fig.18). T-tests between 24-hour and timed samples indicated that there were some statistically significant differences between treatment methods. A Mann-Whitney Rank Sum test indicated that there were some statistically significant differences between treatment methods (Table 7). Charge balances for both of the BY32 W1 samples and SA W1 were found to improve between 3.92 μmeq/L (BY32 W1 11.30.06) and 0.13 μmeq/L (SA W1) when Fe²⁺ was accounted for in the charge balance.

Figure 14. Ternary plot illustrating the variation in hydrochemical facies in the waters. Notice that three of the samples are Na+K rich and that all but BA W1 have a high bicarbonate concentration.

Figure 15. A comparison of the PCO2 saturation levels of 24-hour samples versus timed samples demonstrates a 1:1 relationship, indicating that there is no change in pCO₂ saturation over time.

Figure 16a,b. Both graphs demonstrate the range of alkalinity (mg CaCO₃/L) of 24-hour and timed samples in Piedmont ground waters collected between November 2006 and March 2007. Depending on the treatment method, some samples demonstrated statistically significant differences, indicating that there is no change in alkalinity dependant on treatment method.

Figure 17. A comparison of the alkalinities of 24 hour samples versus timed samples demonstrated a 1:1 relationship between the measurements of three ground water localities, suggesting no change in alkalinity over time. The locality BY32 W1, sampled twice, does not fit this line, suggesting that there may be some change in alkalinity over time.

Figure 18. Range of charge balance for 24 hour and timed samples, illustrating our confidence in the alkalinity measurements. 24-hour samples were within ±20% and timed samples were within ±15%.

Table 4. Field data from sample localities.
Locality	pH	Average Temperature (°C)	Dissolved Oxygen (mg/L)	Conductivity (µS/cm)
Piedmont Surface Water
US64 10.20.05	5.26	17.03	9.37	16.20
FU-C 03 12.01.05	6.93	10.73	9.04	47.00
CL01 01.29.06	7.08	9.14	2.70	98.90
Karstic Surface Water
NC01 07.05.06	5.85	25.60	7.43	56.00
NC02 07.06.06	7.25	21.10	6.09	217.30
NC03 07.06.06	7.22	21.00	5.91	193.00
DL01 8.30.06	9.15	30.27	7.76	155.00
Piedmont Ground Water
BY32 W1 11.30.06	6.14	15.00	3.76	90.30
BA W1 01.15.07	4.90	16.63	6.50	75.50
BY32 W1 02.13.07	6.07	12.10	2.55	80.80
GV W1 03.10.07	7.74	19.47	2.27	118.20
SA W1 03.10.07	5.96	19.27	6.67	32.47

Table 5. Chemical analysis of samples. All samples reported in mg/L. NA= not analyzed. BD= below detectible levels.
Locality	DOC	Na^+	K^+	Ca^2+	Mg^2+	Si^4+	Fe^2+	F^-	Cl^-	NO2^-	Br^-	H2PO4^-	NO3^-	SO4^2-	NH4^+
Piedmont Surface Water
US64 10.20.05	NA	1.62	0.60	0.53	0.31	6.14	0.11	0.02	0.94	BD	BD	BD	0.04	0.65	NA
FU-C 03 12.01.05	NA	3.05	2.52	5.42	1.67	2.45	0.12	0.07	3.06	BD	BD	BD	0.13	3.80	NA
CL01 01.29.06	NA	10.51	3.59	9.98	1.50	0.41	NA	0.15	15.21	0.31	0.07	BD	2.61	8.91	NA
Karstic Surface Water
NC01 07.05.06	0.92	0.59	0.83	3.96	1.85	2.04	NA	0.04	1.54	0.00	0.21	0.00	0.75	17.15	0.00
NC02 07.06.06	1.57	1.34	0.88	32.49	7.60	2.60	NA	0.05	2.82	0.79	0.00	0.00	2.03	17.26	0.03
NC03 07.06.06	1.32	1.01	0.87	31.25	5.01	2.53	NA	0.05	2.58	0.00	0.00	0.62	1.20	17.37	0.03
DL01 08.30.06	2.99	0.95	2.49	16.07	7.24	3.32	NA	0.13	8.42	0.00	0.00	0.81	0.00	11.00	0.01
Piedmont Ground Water
BY32 W1 11.30.06	2.69	4.79	2.52	4.85	1.42	5.98	0.72	0.10	4.61	0.00	0.07	0.00	0.14	0.11	0.65
BA W1 01.15.07	3.93	2.42	4.69	6.47	1.08	4.85	NA	0.18	7.69	0.10	BD	0.31	17.24	0.52	0.07
BY32 W1 02.13.07	1.66	4.55	2.64	4.69	1.42	5.57	0.63	0.10	4.78	NA	0.05	NA	0.12	0.13	0.47
GV W1 03.10.07	0.45	7.48	4.65	9.86	3.80	1.54	NA	0.06	1.23	NA	0.26	NA	0.06	0.96	0.02
SA W1 03.10.07	0.42	3.22	1.6	0.94	0.37	8.30	0.01	0.04	1.02	NA	NA	NA	1.39	0.76	0.02

Table 6. Statistical tests used on sample alkalinities and results on statistical significance. *Y= yes, N= no, P= passed, F= failed
Location	T-test	Rank Sum Test	Normality Test	Equal Variance	Statistically Significant?
Piedmont	Surface	Water
US64 10.20.05	Y*	-	P*	P	N*
FU-C 03 12.01.05	Y	-	P	P	N
CL01 01.29.06	Y	-	F*	P	N
Carbonate	Surface	Water
NC01 07.05.06	Y	-	P	-	Y
NC02 07.06.06	Y	-	P	-	N
NC03 07.06.06	Y	-	P	-	N
DL01 08.30.06	N	Wilcoxon	F	-	Y
Piedmont	Ground	Water
BY32 W1 11.30.06	N	Mann-Whitney	P	F	Y
BA W1 01.15.07	Y	-	P	P	N
BY32 W1 2.13.07	N	Mann-Whitney	F	-	Y
GV W1 03.10.07	Y	-	P	P	Y
SA W1 03.10.07	Y	-	P	P	N

Table 7. Statistical tests used on sample charge balance and results on statistical significance. *Y= yes, N= no, P= passed, F= failed
Location	T-test	Mann-Whitney Rank Sum Test	Normality Test	Equal Variance	Statistically Significant?
Piedmont	Surface	Water
US64 10.20.05	Y*	N*	P*	P	N
FU-C 03 12.01.05	Y	N	P	P	N
CL01 01.29.06	N	Y	F*	-	N
Carbonate	Surface	Water
NC01 07.05.06	N	Y	P	F	N
NC02 07.06.06	N	Y	F	-	N
NC03 07.06.06	N	Y	F	-	N
DL01 08.30.06	Y	N	P	P	N
Piedmont	Ground	Water
BY32 W1 11.30.07	N	Y	F	-	Y
BA W1 01.15.07	Y	N	P	P	N
BY32 W1 2.13.07	N	Y	P	F	Y
GV W1 03.10.07	Y	N	P	P	Y
SA W1 03.10.07	Y	N	P	P	N

Table 8. Average alkalinity and bicarbonate concentrations in surface and ground waters from this study. Alkalinity measured in mg CaC03/L, bicarbonate measured in mg HCO3^-/L.
Location	Alkalinity (mg CaCO3/L)	Bicarbonate (mg HCO3^-/L)
Piedmont Surface Water
US64	4.06 ± 0.37	4.96 ± 0.45
FU C 03	17.93 ± 0.41	21.94 ± 0.49
CL01	30.65 ± 0.94	37.4 ± 1.14
Karstic Surface Water
NC01*	0.80 ± 0.24	0.98 ± 0.29
DL01	49.61 ± 2.82	115.47 ± 9.03
NC03	94.65 ± 7.40	115.47 ± 9.03
NC02	106.07 ± 2.94	129.41 ± 3.59
Piedmont Ground Water
BA W1	7.89 ± 0.83	9.63 ± 1.01
SA W1	11.79 ± 0.64	14.39 ± 0.78
BY32 W1 11.30.07	29.63 ± 3.20	36.31 ± 3.80
BY32 W1 02.13.07	29.68 ± 5.85	36.20 ± 7.14
GV W1	59.18 ± 0.71	72.20 ± 0.87
*Water chemistry influenced by acid mine drainage.

Discussion

When measuring alkalinity, we asked whether 1) alkalinity varies with method, 2) alkalinity varies with time, and 3) alkalinity varies between samples run within 24 hour and those stored. Based on our study, we can conclude that there is no meaningful statistically significant change in alkalinity due to methods or storage time. Although a few samples showed statistically significant changes in alkalinity, these were not meaningful differences and the benefits of taking the sample for calculation of charge balance, PCO₂, or carbonate speciation far outweighs the minor alkalinity changes that may occur during unfiltered/unrefrigerated storage. The final calculations will not change appreciably, so the samples should be taken and measured for alkalinity no matter the sampling conditions.

Previous studies on storage of filtered vs. unfiltered samples have confirmed that waters sampled from the Guayana Shield area of the Orinoco basin in South America, and in the Amazon basin in Brazil, did not undergo statistically significant changes in regards to water chemistry in over a month of storage. (Stallard, 1980; Edmond et al. 1995). This serves as one more indicator that, for many world waters, sample processing does not affect alkalinity measurements.

By comparing our treatment methods to those surveyed in Table 1, we can see that there is a general trend in methodology similar to our protocol that most researchers follow. Neal et al. (1994) stored water samples in sealed glass bottles with zero headspace so that CO₂ degassing could be minimized. He performed alkalinities on unfiltered samples using electrometric techniques and maintained the alkalinity-measuring materials at field temperature to reduce temperature variation interference. He used the Gran Acidity method to measure his samples, (with sodium hydroxide as the titrant) and kept the entire operation under an N₂ atmosphere to prevent atmospheric CO₂ contamination. In a separate study, Neal et al. (1988), used the Gran Titration method and filtered samples only when suspended sediment was visible to minimize degassing. Cai et al. (1998) collected samples from 1 meter below the surface and sealed them so there was no contact with the air to preserve the high Pco₂ levels. Dawson et al. (2001) collected samples for alkalinity measurements in separate plastic bottles and avoided CO₂ degassing by collecting pH samples in plastic luer-lock syringes with a stopper fitted underwater. Barth et al. (2003) refrigerated samples at 4°C and then titrating with 1.6 N H₂SO₄. Finally, Raymond et al. (1997) collected samples in gas-tight BOD bottles and either transported them in a cooler to the lab or measured alkalinity directly in the field. Most of the methods surveyed followed the same treatment trends of refrigeration and titration in the lab, with optional filtration.

Additionally, the practice of not sampling for alkalinity due to preservation concerns must be addressed. In a study of the rivers of the Taklimakan Desert in northwest China, alkalinity samples were not taken because of the remote location of the sample site. The researchers believed that the alkalinity measurements would change with sample storage, so the alkalinities were estimated based on the remaining charge balance after ion analysis (Zhang et al. 1995). The same practice was observed in a study of eleven watersheds in Japan (Nakagawa et al. 2000). We would recommend that in situations such as these, when the waters are undersaturated with respect to calcite, they should be sampled regardless of the ability to refrigerate or filter and be measured whenever the samples can be returned to lab.

The range of our bicarbonate concentrations for surface waters encompass over 85% of world rivers (Table 2), and our range for ground waters encompass about 40% of U.S. ground waters (Table 3). We use this data to suggest that while our research is widely applicable to many surface and ground water situations, it is not to be used without considering other geochemical effects. Change in alkalinity due to calcite precipitation is the main consideration when measuring alkalinity on stored samples. Streams are typically supersaturated with respect to the atmosphere, being up to 20 times the atmospheric partial pressure (Kling et al. 1991; Jarvie et al. 1997; Cole and Caraco 2001, Aston 1984). Conversely, all of our samples were undersaturated with respect to calcite, so this was one source of variation that we did not have to consider. We also did not titrate samples with high turbidity, as this would only titrate the minerals in the water, providing a false alkalinity value.

We still recommend the method of filtering and refrigerating samples intended for alkalinity measurement. This is the method frequently utilized in studies of aqueous geochemistry (Neal et al. 1994; Grosbois et al. 2001; Huh et al. 1998; Rice et al. 1995; Zhang et al. 1995; Jarvie et al. 2002; Katz et al. 1985; Pande et al. 1994; Williams et al. 2001; Wu et al. 2005; Cameron et al. 1995; Finley et al. 1997; Howland et al. 2000; Kim et al. 1996; Edmond et al. 1995; Andersen 2002; Standard Methods 1995; Billett et al. 1996; Dawson et al. 2001; Guasch et al. 1998; Helie et al. 2002; Hill et al. 2002; Hoffer-French et al. 1989; Jones Jr. et al. 1998; Kim et al. 1996; Lesack et al 1991; Raymond et al 1997; Neal et al. 2002 ). Our results provide reasonable confidence that filtration, refrigeration, and storage time do not affect charge balance, carbonate speciation, or Pco₂ calculations. Our method of filtration and refrigeration is the best for several reasons, the main one being that it is convenient to be able to use water filtered for other analytical procedures (e.g., ion chromatography or ICP spectroscopy) for alkalinity measurements as well.

Continuing research in this topic should include a study of undersaturated carbonate ground waters and surface waters that are supersaturated with respect to calcite. We would also like to conduct the same study of treatment methods on samples intended for chemical analyses that do not require filtration, such as ions, NH₄⁺, and DOC.

References

American Public Health Association, American Water Works Association, and Water Environment Federation., 1995. "Collection and Preservation of Samples." Standard Methods for the Examination of Water and Wastewater. Ed. Andrew D. Eaton, Lenore S. Clesceri, and Arnold E. Greenberg. 19th ed. Washington, D.C.: American Public Health Association, 1-22.

Andersen, C. B., 2002. "Understanding carbonate equilibria by measuring alkalinity in experimental and natural systems." Journal of Geoscience Education 50.4: 389-403.

Aston, A.R. "The effect of doubling atmospheric CO2 on streamflow: A simulation." Journal of Hydrology 67 (1984): 273-280. 

Back, W., Hanhaw, B.B., 1970. "Comparison of chemical hydrology of Florida and Yucatan. Journal of Hydrology. 10: 360-368.

Barth, J.A.C., Cronin, A.A., Dunlop, J., Kalin, R.M., 2003. "Influence of carbonates on the riverine carbon cycle in an anthropogenically dominated catchment basin: evidence from major elements and stable carbon isotopes in the Lagan river (N. Ireland)." Chemical Geology. 200: 203-216.

Berner, E.K., Berner R.A., 1996. Global Environment: Water, Air, and Geochemical Cycles. Robert A. McConnin. New Jersey: Prentice-Hall Inc.

Billett, M.F., et al. "The influence of parent material on small-scale spatial changes in streamwater chemistry in Scottish upland catchments." Journal of Hydrology 187 (1996): 311-331. 

Byrd, H., 1972. Soil survey of Pickens County, South Carolina. Washington, DC. U.S. Department of Agriculture Soil Conservation Service.

Cai, Wei-Jun, Wang, Y., 1998. "The chemistry, fluxes, and sources of carbon dioxide in the estuarine waters of the Satilla and Altamaha Rivers, Georgia." Limnology and Oceanography. 43.4: 657-668.

Cole, J.J., and N.F. Caraco. "Carbon in catchments: connecting terrestrial carbon losses with aquatic metabolism." Marine Freshwater Research 52 (2001): 101-110.

Dawson, J.J.C., Billett, M.F., Neal, C., Hill, S., 2002. "A comparison of particulate, dissolved and gaseous carbon in two contrasting upland streams in the UK." Journal of Hydrology 257: 226-246.

Dawson, J.J.C., M.F. Billett, and D. Hope. "Diurnal variations in the carbon chemistry of two acidic peatland streams in north-east Scotland." Freshwater Biology 46 (2001): 1309-1322.

Eby, G.N., 2004. Principles of Environmental Geochemistry. Pacific Grove: Brooks/Cole-Thomson Learning.

Edmond, J.M., et al. "The fluvial geochemistry and denudation rate of the Guayana Shield in Venezuela, Colombia, and Brazil." Geochimica et Cosmochimica Acta 59.16 (1995): 3301-3325. 

Finley, J.B., and J.I. Drever. "Chemical mass balance and rates of mineral weathering in a high-elevation catchment, West Glacier Lake, Wyoming." Hydrological Processes 11 (1997): 745-764. 

Freeze, R.A., Cherry, J.A., 1979. Groundwater. Englewood Cliffs, NJ: Prentice Hall.

Gordeev, V.V., Siderov, I.S., 1993. "Concentrations od major elements and their outflow into the Laptec Sea by the Lena River.” Marine Chemistry. 43: 33-45.

Grimaldi, Catherine, and V. Chaplot. "Nitrate depletion during within-stream transport: effects of exchange processes between streamwater, the hyporheic, and riparian zones." Water, Air, and Soil Pollution 124 (2000): 95-112.
 
Grosbois, C., et al. "An overview of dissolved and suspended matter fluxes in the Loire River basin: Natural and anthropogenic inputs." Aquatic Geoochemistry 7 (2001): 81-105. 
 
Guasch, H., et al. "Diurnal variation in dissolved oxygen and carbon dioxide in two low-order streams." Water Resources 32.4 (1998): 1067-1074. 

Helie, J.-F., C. Hillaire-Marcel, and B. Rondeau. "Seasonal changes in the sources and fluxes of dissolved inorganic carbon through the St. Lawrence River- isotopic ad chemical constraint." Chemical Geology 186 (2002): 117-138.

Hem, J.D., 1970. Study and Interpretation of the Chemical Characteristics of Natural Water, 2nd Ed. Geological Survey Water-Supply Paper 1473. Washington, D.C.: United States Government Printing Office, 363 pp.
 
Hill, T.J., R.A. Skeffington, and P.G. Whitehead. "Recovery from acidification in the Tillingbourne catchment, southern England: catchment description and preliminary results." The Science of the Total Environment 282-283 (2002): 81-97. 
 
Hoffer-French, K.J., and J.S. Herman. "Evaluation of hydrological and biological influences on CO2 fluxes from a karst stream." Journal of Hydrology 108 (1989): 189-212.
 
Howland, R.J.M., et al. "Distributions and seasonal variability of pH and alkalinty in the Tweed Estuary, UK." The Science of the Total Environment 251-252 (2000): 125-138.

Huh, Y., et al. "The fluvial geochemistry of the rivers of Eastern Siberia: I. Tributaries of the Lena river draining the sedimentary platform of the Siberian craton." Geochimica et Cosmochimica Acta 62.10 (1998): 1657-1676. 

Huh, Y., et al. "The fluvial geochemistry of the rivers of Eastern Siberia: II. Tributaries of the Lena, Omoloy, Yana, Indigirka, Kolyma, and Anadyr draining the collisional/accretionary zone of the Verkhoyansk and Cherskiy ranges." Geochimica et Cosmochimica Acta 62.12 (1998): 2053-2075.
 
Jarvie, H.P., et al. "Patterns in nutrient concentrations and biological quality indices across the upper Thames river basin, UK." The Science of the Total Environment 282-283 (2002): 263-294.
 
Jarvie, H.P., et al.. "Major ion concentrations and the inorganic carbon chemistry of the Humber rivers." The Science of the Total Environment 194-195 (1997): 285-302. 
 
Jones, J.B., Jr, and P.J. Mulholland. "Influence of drainage basin topography and elevation on carbon dioxide and methane supersaturation of stream water." Biogeochemistry 40 (1998): 57-72.

Katz, B.G., O.P. Bricker, and M.M. Kennedy. "Geochemical mass-balance relationships for selected ions in precipitation and stream water, Catoctin Mountains, Maryland." American Journal of Science 285 (1985): 931-962. 

Kim, J.P., et al. "Aqueous chemistry of major ions and trace metals in the Clutha River, New Zealand." Marine Freshwater Research 47 (1996): 918-929. 

Kim, K., 2005. "Evaluation of geochemical processes affecting groundwater chemistry based on mass balance approach: A case study in Namwon, Korea." Geochemical Journal 39: 357-369.

Kling, G.W., G.W. Kipphut, and M.C. Miller. "Artic lakes and streams as conduits to the atmosphere: implications for tundra carbon budgets." Science 251 (1991): 298-301.

Langmuir, D., 1971. "The geochemistry of some carbonate groundwaters in central Pennsylvania.” Geochimica et Cosmochimica Acta. 42: 1023-1045.

Lesack, L.F.W., and J.M. Melack. "The deposition, composition, and potential sources of major ionic solutes in rain of the central Amazon basin." Water Resources Research 27.11 (Nov. 1991): 2953-2977. 

Lewis, G.P., Mitchell, J.P., Andersen, C.B., Haney, D.C., Liao, M.-K., Sargent, K.A., 2007. "Urban Influences on Stream Chemistry and Biology in the Big Brushy Creek Watershed, South Carolina." Water, Air, and Soil Pollution 182: 303-323.

Lewis, W.M., Jr., and M.C. Grant. "Relationships between stream discharge and field of dissolved substances from a Colorado mountain watershed." Soil Science 128.6 (Dec. 1979): 353-363.
 
Lewis, W.M., Jr, et al. "Major element chemistry, weathering and element yields for the Caura River drainage, Venzuela." BIogeochemistry 4 (1987): 159-181. 

Livingstone, D.A., 1963. "Chemical composition of rivers and lakes.” USGS Professional Paper 440G, 64.

Matthess, G., 1982. The Properties of Groundwater. New York: John Wiley & Sons, 406 pp.

McDowell, W.H., and C.E. Asbury. "Export of carbon, nitrogen, and major ions from three tropical montane watersheds." Limnology and Oceanography 39.1 (1994): 111-125.

Meybeck, M., 1979. "Concentrations des eaux fluiales en éléments majeurs et apports en solution aux océans." Reviews de Géologique Dynamique et Géographique Physique. 21.3: 215-246

Meybeck, M. "5.08 Global Occurrence of Major Elements in Rivers." Treatise on Geochemistry. Ed. H. D. Holland and K. K. Turekian. Vol. 5. Paris: Elsevier Ltd., 2004. 207-223.

Meybeck, M. "C, N, P and S in Rivers: From Sources to Global Inputs." Global Environmental Change: Interactions of C, N, P and S biogeochemical cycles and global change. Ed. R. Wollast, F.T. Mackenzie, and L. Chou. Vol. 4. Global Environmental Change 1. Berlin Heidelberg: Springer-Verlag, 1993. 163-193.

Musgrove, M., Banner, J.L., 2004. "Groundwater Evolution in Two Karst Aquifers: Vadose Zone Processes, Soil Influences, and Climate Connections." Epikarst 9: 62-70.

Nakagawa, Y., and G. Iwatsubo. "Water chemistry in a number of mountainous streams of east Asia." Journal of Hydrology 240 (2000): 118-130. 
 
Neal, C., et al. "The water quality of the river Thames at a rural site downstream of Oxford." The Science of the Total Environment 251-252 (2000): 441-457.
 

Neal, Colin, House, W.A., Down, K.. ”An assessment of excess carbon dioxide partial pressures in natural waters based on pH and alkalinity measurements.” The Science of the Total Environment. 210/211 (1998): 173-185.

Neal, Colin, and Susan Hill. "Dissolved inorganic and organic carbon in moorland and forest streams: Plynlimon, Mid-Wales." Journal of Hydrology 153 (1994): 231-243. 

Neal, C. "Determination of dissolved CO2 in upland streamwater." Journal of Hydrology 99 (1988): 127-142. 

Oechel, W.C., et al. "A scaling approach for quantifying the net CO2 flux of the Kuparuk river basin, Alaska ." Global Change Biology 6.Suppl.1 (2000): 160-173. 

Ometo, J.P.H.B., et al. "Effects of land use on water chemistry and macroinvertebrates in two streams of the Piracicaba river basin, south-eat Brazil." Freshwater Biology 44 (2000): 327-337. 

Overstreet, W.C., Bell III, H., 1965. Geologic map of the crystalline rocks of South Carolina. U.S. Geological Survey Miscellaneous Geologic Investigations map I-413, I sheet, scale 1:250,000

Pande, K., et al. "The Indus river system (India-Pakistan): Major ion chemistry, uranium and strontium isotopes." Chemical Geology 116 (1994): 245-259. 

Pavlicek, 2006. Comprehensive Stormwater Pilot Master Plan for North Chickamauga Creek Watershed. 30 Mar. 2005. City of Chattanooga, Stormwater Management, Department of Public Works-Engineering. 2 May 2007 <http://www.chattanooga.gov/Files/NPDES-PilotMasterPlanNChick.pdf>.

Piñol, J., Avila, A., 1992. "Streamwater pH, alkalinity, pCO2 and discharge relationships in some forested Mediterranean catchments.” Journal of Hydrology 131: 205-225

Plummer, L.N., Busenburg, E., 1982. “The solubilities of calcite, aragonite, and vaterite in CO2-H2O solutions between 0 and 90 degree systems and an evaluation of the aqueous model for the system CaCO3-CO2-H2O.” Geochimica et Cosmochimica Acta, 46: 1011-1040.

Probst, J.-L., Tardy, Y., 1989. “Global runoff fluctuations during the last 80 years in relation to world temperature change.” American Journal of Science. 289: 267-285.

Pytkowicz, R.M., D.R. Kester, and B.C. Burgener. "Reproducibility of pH Measurements in Seawater." Limnology and Oceanography 11.3 (July 1966): 417-419.

Ramesh, R., G.R. Purvaja, and V. Subramanian. "Carbon and phosphorus transport by the major Indian rivers." Journal of Biogeography 22 (1995): 409-415.

Raymond, P.A., Caraco, N.F., Cole, J.J., 1997. "Carbon dioxide concentration and atmospheric flux in the Hudson river." Estuaries 20.2: 381-390.

Rice, K.C., and Bricker, O.P., 1995. "Seasonal cycles of dissolved constituents in streamwater in two forested catchments in the mid-Atlantic region of the eastern USA." Journal of Hydrology 170: 137-158.

Richey, J.E., et al. "Outgassing from Amazonian rivers and wetlands as a large tropical source of atmospheric CO2." Nature 416 (2002): 617-620. 

Richey, Jeffery E., Hedges, J.I., Devol, A.H., Quay, P.D., Victoria, R., Martinelli, L.., Forsberg, B.R., 1990. "Biogeochemistry of carbon in the Amazon River." Limnology & Oceanography 35.2: 352-371.

Risacher, F., H. Alonso, and C. Salazar. "Hydrochemistry of two adjacent acid saline lakes in the Andes of northern Chile." Chemical Geology 187 (2002): 39-57.

Sarin, M.M., Krishnaswami, S., Dilli, K., Somayajulu, B.L.K., Moore, W.S., 1989. "Major ion chemistry of the Ganga-Bramaputra river systems, India." Geochimica et Cosmochimica Acta. 53: 997-1009.

Shugg, A., Knight, M.J., 1994. "Hydrogeology of the Mineral Springs of Central Victoria." Water Down Under. 94 2.part A: 215-221.

Schlesinger, William H. "The Global Carbon Cycle." Biogeochemistry: An Analysis of Global Change. 1991. 2nd ed. San Diego: Academic Press, 1997. 345-357.

Stallard, R.F., 1980. "Major Element geochemistry of the Amazon River system". Ph.D. dissertation, MIT/Woods Hole Oceanographic Institute, WHOI-80-29, 366.

White, W.B. "Rate processes: chemical kinetics and karst landform development." Groundwater as a Geomorphologic Agent. Ed. R.G. Lafleur. Boston: Allen and Unwin, 1984. 227-248.

Williams, A., Andersen, C.B., and Lewis, G.P., 2006. “Effect of sample process methodology on alkalinity concentrations and carbonate speciation calculations.” Geological Society of America Abstracts with Programs, Southeastern Section Meeting, 38: 61.

Williams, M.R., et al. "Precipitation and river water chemistry of the Piracicaba river basin, southeast Brazil." Journal of Environmental Quality 30 (2001): 967-981.

Worrall, F., and A. Lancaster. "The release of CO2 from riverwaters – the contribution of excess CO2 from groundwater." Biogeochemistry 76 (2005): 299-317.

Wu, L., et al. "Chemical weathering in the Upper Huang He (Yellow river) draining the eastern Quinhai-Tibet Plateau." Geochimica et Cosmochimica Acta 69.22 (2005): 5279-5294.

Zelewski, L.M., 1999. "Spatial and temporal patterns of trace metal concentrations and partitioning in a small headwater stream." Diss. University of Wisconsin- Madison.

Zhang, J.W., Huang, W., Lin, M.G., Zhon, A., 1990. “Drainage basin weathering and major element transport of two large Chinese rivers (Huanghe and Changjiang).” Journal of Geophysical Research. 95:13277-13288.

Zhang, J., et al. "Water chemistry of the rivers around the Taklimakan Desert (NW China): Crustal weathering and evaporation processes in arid land." Chemical Geology 119 (1995): 225-237.

Zhang, J., Huang, W.W.,, Létolle, R., Jusserand, C. "Major element chemistry of the Huanghe (Yellow River), China - weathering processes and chemical fluxes." Journal of Hydrology 168 (1995): 173-203.

Zobrist, J., Stumm, W., 1980. "Chemical dynamics of the Rhine catchment area in Switzerland: extrapolation to the "pristine" Rhine rive input into the ocean". In Martin, J.M., Burton, J.D., and Eisma, D., (eds.). River Inputs to Oceans Systems, SCOR/UNEP, UNESCO Review and Workshop. Rome, FAO, 52-63.