Sources and downstream variation of surface water chemistry in the dammed Waitaki catchment, South Island, New Zealand

ABSTRACT

This article presents the results of a pilot study exploring downstream changes in water chemistry in the Waitaki catchment, which drains the eastern slopes of the Southern Alps. The headwaters of this catchment are glaciated, and both natural glacial lakes and hydroelectric reservoirs occur in the catchment. The dominant lithology is calcite-poor quartzo-feldspathic metasedimentary rock. We sampled surface waters from streams and the inflow and outflow of lakes in the catchment on four occasions over the course of 1year, at c. 3-monthly intervals between summer 2013 and spring 2014. We also sampled ice released from the terminus of the Tasman Glacier. Small but measurable variations were observed in the major element (Ca, Mg, K, Na, Si), trace element (Al, Fe, Rb, Sr) and Sr isotopic composition of the surface waters in the catchment. The composition of the waters is interpreted to reflect the change from calcite-dominated weathering near the top of the catchment toward more silicate weathering down the catchment that has been observed in other studies. The composition of samples collected from the uppermost part of the catchment displays more temporal variability than those samples collected from dammed lakes, suggesting that seasonality is dampened by anthropogenic modification of the catchment.

KEYWORDS:

• Weathering
• Sr isotopes
• dissolved cations
• dissolved Si
• seasonality
• hydroelectric dams
• Southern Alps

Introduction

The denudation of mountain belts takes place via chemical and physical erosion, the relative intensities of which reflect both climatic (e.g. precipitation, temperature and runoff) and tectonic (e.g. exhumation, landscape surface age, mechanical weathering and sediment transport) factors. The chemical weathering of silicates is generally considered to be responsible for regulating the carbon cycle on geological time scales, which in turn has control on global average temperature (e.g. Berner et al., 1983). However, in settings where the residence time of fresh waters in a catchment is short, carbonate weathering becomes an important control on their chemical and isotopic composition. Furthermore, the relationship between physical and chemical weathering in a catchment depends on the rates of uplift and weathering. For example, previous work on the Southern Alps of New Zealand has shown that c. 90% of total weathering is physical, whereas the remainder is chemical (Lyons et al., 2005).

The record of ⁸⁷Sr/⁸⁶Sr in marine deposits has been used to make inferences about the relative drawdown of CO₂ from the atmosphere by silicate weathering, such as the hypothesis that weathering of the Himalaya has led to an increase in the ⁸⁷Sr/⁸⁶Sr of seawater in the late Cenozoic (e.g. Raymo and Ruddiman, 1992). However, due to the higher solubility and faster dissolution rates of carbonate minerals in comparison with the generally much more abundant silicate minerals, it has become apparent that river water chemistry in active orogenic belts is often dominated by carbonate weathering. Mountain belts often have disseminated or vein-associated calcite, which dissolves orders of magnitude more rapidly than silicates (Blum et al., 1998; Jacobson et al., 2003). Uplift continually exposes new minerals to dissolution, allowing Ca and associated Sr concentrations in river water to become elevated (Palmer and Edmond, 1992; Lyons et al., 2005; Chamberlain et al., 2005). Silicate weathering rates are generally slower in mountainous areas due to a combination of short residence times of waters and low temperatures (White and Blum, 1995; Dalai et al., 2002).

Study area

The Southern Alps of New Zealand are formed by the convergent component of the motion between the Australian and Pacific plates (Toy et al., 2012; Figure 1). Displacement has been along the Alpine Fault, which thrusts the Pacific Plate over the Australian Plate, forming steep topography on the west coast and relatively shallow topography in the east. The Alpine Fault has accommodated c. 480 km of dextral strike-slip motion and > 35 km of vertical uplift, much of it over the past 5 million years (Norris et al., 1990; Cox and Sutherland, 2007; Toy et al., 2012). The Pacific Plate at the Alpine Fault experiences average uplift rates in the order of 7–8 mm/yr (Norris and Cooper, 2001); in the Main Divide area, this is thought to be in dynamic equilibrium with the high rates of physical erosion (Figure 1C; Koons, 1990).

Figure 1. Maps showing the location and setting of the Waitaki catchment. A , Generalised geological map with water sampling locations indicated with Station number. B , Location of the catchment in the South Island of New Zealand and line of section for C, idealised cross section showing the topography relative to the Alpine Fault, with the highest elevation in the Southern Alps corresponding to the location of the main divide (after Koons, 1990).

To the east of the Alpine Fault, a splay of roughly NE-SW trending reverse faults accommodate the remaining 10–40% of plate motion not taken by the Alpine Fault, creating basin and range topography (Berryman and Beanland, 1991; Norris and Cooper, 2001). Uplift in this area is c. 1.7–0.5 mm/yr. These faults are cemented with calcite-rich fault gouge, and some travertine deposits form at discharge sites of both warm and cold springs (Jacobson et al., 2003). The dominant rock types on the eastern side of the Southern Alps, including this study area, are greywackes and mudstones (argillites) of the composite Torlesse Composite Terrane (Adams and Graham, 1997; Forsyth, 2001). Regions of this terrane were metamorphosed to the prehnite–pumpellyite facies during the Permian–Jurassic, and are primarily composed of quartz, albite, chlorite and muscovite derived from the erosion of a continental granitoid source (MacKinnon, 1983). Quartz–calcite veins that are centimetres to tens of centimetres thick have been deposited by hydrothermal groundwaters accompanying the recent uplift of the Southern Alps (Koons and Craw, 1991). Overall, hydrothermal calcite has been estimated to compose up to 3% of the bulk rock composition (Jacobson et al., 2003), but it is not evenly distributed. Bedload sediments in the nearby Clutha and Haast catchments have only trace amounts of calcite (Kautz and Martin, 2007).

The Southern Alps form a barrier across the prevailing westerly airstream from the Tasman Sea, striking almost perpendicular to it and creating an extremely asymmetric rainfall pattern in which the western side receives high rainfall (> 6 m/yr), while the eastern side receives relatively little precipitation (≥ 0.36 m/yr). Vegetation east of the Main Divide is predominately tussock grasslands, although in the Mackenzie Basin intensive farmland and exotic forests are rapidly expanding. Maximum mean air temperatures in the Mackenzie Basin are 8 °C in winter and 23 °C in summer (NIWA, 2015).

The Waitaki catchment, which drains from the highest point in the Southern Alps east to the Pacific Ocean (Figure 1) and has an area of 12 787 km², is the most regulated in New Zealand with respect to discharges for hydroelectricity and for irrigation. The headwaters of the catchment are located in pristine, temperate glaciated terrain, but its downstream hydrology has been greatly modified from its natural state with a number of dams and canals (Figure 2). Hydroelectric power stations in the Waitaki catchment collectively produce 35–40% of New Zealand’s electricity (Purdie and Bardsley, 2010). In addition to the importance of the Waitaki catchment in supplying hydroelectricity (Purdie and Bardsley, 2010), it supplies irrigation waters for agriculture, hosts an aquaculture salmon fishery, and is used for recreational activity.

Figure 2. Schematic depiction of the hydrology of the Waitaki catchment. Water sampling locations are indicated with station numbers (after Pettinger, 2016).

Because dams within the catchment allow water to accumulate in the hydro-lakes during the summer and autumn high-inflow months (Purdie and Bardsley, 2010), the dissolved load of each lake is not simply representative of each individual storm event but is rather a weighted average of the events taking place on a seasonal basis. Damming alters the residence time of waters in different sections of a catchment relative to its natural flow conditions (Table S1; K. Walter, NIWA, pers. comm. March 2016, and E. Stead, Meridian Energy Ltd., pers. comm., February 2016). Furthermore, as the water reaches the lakes, its velocity decreases and the suspended load falls as sediment to the bottom of the lake (McGowan and Sturman, 1997). In winter, as water is released from the dams to satisfy the increased requirement for energy, lake levels decrease, and the streams may pick up new sediment as the water travels further downstream. All these factors have the potential to affect water chemistry within the catchment (Gross et al., 2001).

Sample collection and analytical methods

Field sampling

Fresh water samples were collected quarterly (December 2013 to September 2014) from nine stations (Table 1) along the c. 215 km length of the catchment.

Table 1. Analytical results for water and ice samples in the Waitaki Catchment. Station numbers correspond to locations in Figure 1.

Sample season	Location (station no.) distance from mouth	T (°C)	pH	Ca (µmol/L)	Mg (µmol/L)	K (µmol/L)	Na (µmol/L)	Si (µmol/L)	Al (µmol/L)	Fe (µmol/L)	Rb (nmol/L)	Sr (nmol/L)	^87Sr/^86Sr
Summer	Tasman Glacier Ice 214 km			6.0	6.6	6.5	<4.3		19	8.6	15	22.7	0.71074
Autumn				63.9	<4.1	8.0	11.1	51.1	3.9	1.7	5.7	75.4	0.70933
Winter				7.9	<4.1	3.9	4.6	<2.7	0.2	0.1	1.9	3.40
Spring				88.0	<2.1	3.6	<2.2	2.9	1.2	0.8	3.1	19.5
Summer	Lake Hooker (1) 211 km	3.3	7.3	263	65.8	35.7	63.9		90	61	49	1224	0.70961
Spring		3.9	7.0	250	26.1	14.0	42.9	34.5	0.8	0.1	4.0	1184	0.70948
Summer	Stocking Stream (2) 210 km	11.0	7.2	131	8.8	7.1	28.9		2.2	1.2	4.5	535	0.70944
Autumn			7.6	185	11.0	7.5	20.0	15.7	0.4	0.1	3.7	768	0.70942
Winter			7.0	49	<4.1	5.3	9.3	7.3	1.3	0.5	2.7	201	0.70946
Summer	Lake Tasman (3) 208 km	3.8	8.1	288	38.2	40.1	92.1		65	22	5.7	1496	0.70950
Autumn		4.1	8.1	279	20.6	17.2	74.6	42.0	2.3	0.7	4.3	1380	0.70948
Winter		3.9	7.7	216	13.9	15.3	52.7	32.3	1.6	0.4	4.2	1119	0.70950
Spring		3.3	7.4	262	18.8	16.6	67.7	39.3	1.4	0.1	3.7	1450	0.70948
Summer	Lake Tekapo (4) 182 km	13.3	7.6	211	22.3	12.9	67.4		0.5	0.1	3.3	740	0.70998
Autumn		14.5	7.8	212	22.2	13.6	67.4	67.7	<0.1	<0.05	3.4	768	0.70999
Winter		10.2	7.5	193	18.1	13.1	55.0	68.6	<0.1	<0.05	3.9	706	0.70100
Spring		8.1	7.3	194	19.5	11.1	59.1	68.9	0.3	<0.05	3.2	763	0.71001
Summer	Lake Pukaki (5) 150 km	14.9	8.1	247	24.6	17.8	71.5		0.7	0.1	5.1	1034	0.70959
Autumn		15.0	7.9	245	23.7	16.6	69.1	58.6	0.8	0.1	4.0	1039	0.70961
Winter		10.6	7.8	219	20.2	15.9	58.7	57.4	1.1	0.3	4.7	903	0.70956
Spring		8.6	7.7	227	21.2	14.0	63.1	58.6	0.6	0.1	3.4	1023	0.70965
Summer	Lake Ruataniwha entrance (6a) 144 km	12.9	8.1	252	25.6	17.5	71.0		0.60	0.1	4.9	1016	0.70958
Autumn		13.0	7.9	252	25.1	17.8	71.2	60.1	1.00	0.2	5.3	1053	0.70956
Winter		10.1	7.7	231	25.0	13.5	57.1	71.9	0.3	0.1	3.5	836	0.70953
Spring		8.1	7.3	229	21.8	14.7	64.5	60.0	0.5	0.1	3.7	1040	0.70962
Summer	Lake Ruataniwha exit (6b) 138 km	13.1	8.1	249	27.2	17.9	71.2		4.9	2.3	7.0	1018	0.70959
Autumn		13.0	8.0	243	23.9	16.2	68.2	59.2	0.8	0.1	4.6	1015	0.70957
Winter		10.4	7.5	230	23.5	14.4	57.3	66.5	0.4	0.1	3.7	885	0.70955
Spring		8.1	7.1	228	22.0	14.5	64.2	59.5	0.5	0.1	3.6	1029	0.70960
Summer	Lake Aviemore (7) 86.9 km	15.3	7.9	226	35.8	16.2	90.6		1.4	0.4	4.1	802	0.70942
Autumn		14.0	7.7	261	41.9	16.4	104.3	66.2	0.3	0.1	3.7	913	0.70935
Winter		10.0	7.3	228	42.4	14.2	109.9	91.6	<0.1	0.1	3.1	798	0.70914
Spring		8.4	7.8	228	37.4	14.1	94.4	68.3	0.2	0.1	2.9	895	0.70932
Summer	Lower Waitaki (8) 5.17 km	16.0	7.7	228	37.3	17.5	95.7		2.1	1.2	4.7	865	0.70941
Autumn			7.6	246	34.0	17.7	87.4	56.5	0.3	0.2	4.1	910	0.70948
Winter		10.5	7.8	265	42.5	18.5	107.2	66.5	0.2	0.1	3.5	959	0.70929
Spring		9.5	7.8	215	31.6	14.1	81.0	59.2	0.3	0.1	3.1	878	0.70943

All elemental concentrations analysed by q-ICP-MS, except Si which was analysed by spectrophotometry (see text). Empty cells indicate that data were not collected. Uncertainty in ⁸⁷Sr/⁸⁶Sr ratios is ≤ 0.00002.

Sampling locations are indicated on Figures 1 and 2 and Table 1 (with further information on the sites and water fluxes and residence times given in Tables S1 and S2). Three sub-alpine locations include the glacial terminal lakes Tasman and Hooker and the alpine Stocking Stream. Lake Tasman was sampled at all four seasons, but Stocking Stream was not running in spring, while Lake Hooker could only be accessed in summer and spring. The other water sample locations were associated with the major hydro-lakes in the Mackenzie Basin and Waitaki Valley. In total, 33 water samples were collected from the nine locations. All water samples were collected from flowing water; lake samples were usually taken from the primary egress of the lake, which was assumed to provide a representative sample of the entire lake volume. Both the entrance and exit of Lake Ruataniwha (Stations 6a and 6b, respectively) were sampled to explore whether within-lake geochemical processes changed the composition of the water to a measurable extent.

At each water sampling location, the pH, conductivity and temperature were measured using a YSI Professional Eh-pH metre. Uncertainty for these measurements is approximately ± 0.02 pH units, ± 0.02 µS/cm and 0.1 °C. The metre was calibrated with standard solutions prior to sampling at each station. Samples were collected in acid-cleaned HDPE bottles after rinsing and discarding the water three times. Water samples were kept in the dark and chilled until return to the laboratory.

Glacial ice that had calved into terminal lake Tasman (Station 3) was collected from a boat on each sampling trip, for four samples in total. The ice surfaces were cleaned with quartz-distilled HCl immediately after collection, and the cleaned ice triple-bagged in clean resealable bags.

Rock samples were collected from the terminal moraine in the Tasman valley (Station 3). Boulders weighing up to about 10 kg were chosen to gather a representative set of the lithologies present within the catchment. Rocks from two prominent fault systems within the Mackenzie Basin (Irishman Creek and Ostler faults) were also sampled. All the rock samples were crushed inside HDPE bags using a steel mallet, and the < 2 mm internal fragments were powdered in an agate shatterbox. Residual rock material is archived in the Geology Department at University of Otago.

Laboratory and analytical methods

Upon returning to the laboratory, water samples were vacuum filtered using Millipore Sterifil units to 0.22 µm with mixed cellulose acetate filters in order to obtain the dissolved load, which probably includes some colloidal material. The ice samples were allowed to melt and then filtered in the same way as the water samples. Filtrates were subsequently acidified to 2% with concentrated quartz-distilled nitric acid (QD HNO₃). Splits of the waters were taken for analysis of major and trace elements, and a mixed isotope standard solution was added. Based on the Sr concentration obtained, further splits were taken to obtain sufficient Sr for isotope analysis. Sr isotope splits were dried down and then taken up in 3 N HNO₃, which was purified on Sr.Spec resin columns. A second pass through the Sr.Spec resin was undertaken to ensure the removal of interfering isotopes in the mass spectrometer.

Rock powders were subjected to a weak leaching process intended to obtain the carbonate and exchangeable ion fraction of the samples (e.g. Jacobson et al., 2003; Moore et al., 2013). Dilute acetic acid (0.5 N) was added to a weighed split of rock powder, and the acidified samples were agitated on a shaking table for 24 h at room temperature, and then centrifuged. The leaching procedure was repeated in order to ensure that the reactions had gone to completion. Splits were taken, dried and then redissolved in 5% QD nitric acid. The leachates were then processed for elemental and Sr isotopic analysis using the same methods as for the water and ice samples.

Concentrations of major cations (Ca, Mg, K, Na) and several trace elements (Fe, Al, Rb, Sr) of all samples were measured on an Agilent 7500 quadrupole ICP-MS using standard methods. Sr isotopic compositions were measured on a Nu Plasma MC-ICP-MS. Both ICP instruments are located at the University of Otago Centre for Trace Element Analysis. Both international and in-house standards (SRM987, Tridacna, HPS Sr, BHVO-2 and BCR-2) were run periodically to ensure the accuracy and precision of the results. This study did not include mass-dependent Sr isotopic measurements, so the data were normalised to ⁸⁶Sr/⁸⁸Sr = 0.1194 to correct for analytical mass fractionation using an exponential law (Weis et al., 2006). Procedural blanks (filtration and subsequent steps for water samples; dissolution blanks for rocks) were negligible, so no blank corrections were required.

Determination of the dissolved silica concentration in the water samples was conducted on a Shimadzu UV-1800 Spectrophotometer in the Geology Department at the University of Otago, using a molybdate-blue spectrophotometry method modified from Strickland and Parsons (1972). Three measurements of each sample were taken to ensure accuracy and from these an average concentration was calculated. All reagent blanks displayed absorbances less than 0.001. The silica concentrations were determined by linear regression through a 10-point standard curve using the OPAL (Vbasic macro) extension with Microsoft Excel. Silica concentrations in the summer samples were not obtained, as the entire sample had been previously acidified with 2% QD nitric acid prior to taking the Si split, compromising the sample preparation for spectrophotometry.

Although we did not measure (HCO₃)⁻ concentrations directly, we are able to infer a maximum possible value based the assumption of charge balance. Jacobson et al. (2003) reported that was the dominant anion in South Island rivers, at a relative molar abundance of 89%, with subordinate amounts of (SO₄)²⁻ (7%) and Cl⁻ (4%). Total dissolved solids for our water samples were estimated using these anion proportions. For the TDS calculation, differences in ionic proportions do not make a great deal of difference as long as alkalinity is the dominant anion.

The saturation state of the waters with regard to primary minerals and weathering products (Tables S3 and S4) was calculated using the measured field parameters and major ion concentrations as input to PHREEQC (Parkhurst and Appelo, 2013). Note that these are based on the same anion concentration estimates and limits discussed above. The water samples were unable to be stored appropriately and hence were discarded.

Results

Results for the water and ice samples are reported in Table 1, and results of the rock leachates are in Table 2.

Table 2. Elemental and Sr isotope geochemistry of the acetic acid leach of terminal moraine and fault rocks in the Waitaki catchment.

Sample location	Rock type	Ca (µmol/g)	Mg (µmol/g)	K (µmol/g)	Na (µmol/g)	Si (µmol/g)	Al (µmol/g)	Fe (µmol/g)	Rb (nmol/g)	Sr (nmol/g)	Ca/Sr (µmol/ nmol)	^87Sr/^86Sr
Tasman	Greywacke	39.9	0.29	0.37	0.62	0.53	0.91	0.85	0.17	118	0.34	0.70965
Tasman	Greywacke (Pump^a or Chl^b rich)	97.9	0.33	0.17	0.15	0.27	0.47	7.15	0.79	270	0.36	0.70822
Tasman	Greywacke/Argillite	10.7	0.75	1.19	0.35	0.61	1.45	1.46	0.67	57.8	0.19	0.70909
Tasman	Argillite	6.2	1.33	0.99	0.43	1.77	2.04	1.86	0.80	29.7	0.21	0.70994
Tasman	Argillite (slate cleavage)	1.5	1.25	1.78	0.34	0.54	2.55	2.83	1.25	6.9	0.22	0.71234
Tasman	Chert	4.1	2.97	0.16	0.14	0.53	3.00	7.57	2.51	1.1	3.88	0.71468
Irishman Creek Fault	Cataclasite	629.5	21.1	5.79	0.42	2.20	1.81	3.86	3.04	5355.3	0.12	0.71003
Ostler Fault	Fault gouge	195.1	1.86	0.21	0.38	0.57	0.36	0.67	0.26	102.3	1.91	0.70915

All elemental concentrations analysed by q-ICP-MS. Uncertainty in ⁸⁷Sr/⁸⁶Sr ratios is ≤ 0.00002.

^aPump = pumpellyite.

^bChl = chlorite.

All the waters in this study are relatively dilute compared with most world rivers, as is typical of New Zealand (e.g. Kim et al., 1999; Jacobson et al., 2003), and illustrated on a Gibbs diagram (Figure 3). Figure 4 illustrates the variation of pH in the water samples at each station, plotted as distance above the river mouth and keyed by season. The waters range in pH from 7.0 to 8.1 (Figure 4), with the two lowest pH values found in the upper catchment (Stations 1 and 2). The exit of Lake Ruataniwha sampled in spring has nearly as low a pH as the alpine Stocking Stream. There is a tendency for the pH to decrease slightly at each site going from summer through to the following spring, and the variation between sites is similar to the seasonal range at any given site. The lower Waitaki River (Station 8) exhibits the smallest seasonal variation in pH (7.6–7.8), while the exit to Lake Ruataniwha (Station 6b) has the largest variation (7.1–8.1). Only the results for the exit of Lake Ruataniwha (Station 6b) are plotted on this and subsequent figures, to make them more directly comparable with other lake sampling sites. There is essentially no change in water pH from the entrance of Lake Ruataniwha to its exit (Table 1).

Figure 3. Plot of total dissolved solids (mg/L) versus molar Na/(Na + Ca), showing a number of world rivers and other water bodies (after Gibbs, 1970). The field encircled in red indicates the Waitaki catchment waters from this study, with shading of circles becoming darker with decreasing distance from the mouth of the river. New Zealand rainwater values from Jacobson et al. (2003).

Figure 4. Plot of pH of water at the sample stations as a function of position in the catchment as distance from the Waitaki River mouth, keyed to the sampling season. Samples taken on the same date are connected by dashed lines.

Figure 5 illustrates the variation of the major dissolved species measured in the water samples, again plotted as distance upstream from the mouth of the Waitaki River. Stocking Stream has the lowest concentrations of all the major solution species in the water bodies for all seasons; indeed, Stocking Stream in winter has lower concentrations of dissolved species than some of the glacial ice samples, which could be expected to be similar to rainfall. The major solute geochemistry of waters in this study is similar to results found in previous studies (Kim et al., 1999; Jacobson et al., 2003; Moore et al., 2013).

Figure 5. Downstream variation in freshwater composition. A–D, Major cations and E, Si concentrations at the sample stations plotted as a function of position in the catchment, keyed to the sampling season.

Terminal glacial lakes Hooker (Station 1) and Tasman (Station 3) have high dissolved Ca concentrations compared to the alpine stream (Station 2) and tend to be similar to the down-catchment sites. Lakes Hooker and Tasman also have higher concentrations in summer for most elements than in the other seasons. Lake Tekapo has slightly lower Ca and slightly higher Si than would be expected from the downstream pattern for the other mid to upper catchment samples. The lowermost sites in the catchment, Lake Aviemore and the lower Waitaki River, have Ca and K concentrations that are similar to the other sites, but have higher Na and Mg concentrations. Calcium is the dominant cationic species measured in all the waters (average 68%, range 58–83%), followed by Na (average 20%, range 9–28%) > Mg (average 8%, range 5–15%) > K (average 5%, range 3–9%). When measured, dissolved Si is generally similar in concentration to Na (Table 1, Figure 5).

Despite the very low solute concentrations, almost all the waters are supersaturated with respect to the secondary minerals gibbsite, kaolinite and Na-beidellite (Table S4). The exception is for autumn samples of Stocking Stream and Lake Tekapo, which are undersaturated with respect to Na-beidellite. The waters are all significantly undersaturated with respect to calcite (Figure 6) and albitic feldspar, and are close to saturation with quartz (Table S4).

Figure 6. Plot of dissolved Ca concentrations (µmol/L) versus measured pH for stream waters in this study. Lines shown are calcite saturation curves for Ca concentration at indicated total dissolved carbonate species in the water. All the samples plot below the saturation curve even for the highest possible carbonate in the river water calculated for charge balance (as discussed in text).

Two ice samples from the Tasman Glacier have very low concentrations of all dissolved elements, as would be expected for pristine atmospheric precipitation. In contrast, the other two ice samples have markedly higher concentrations of Ca and Si than the alpine stream.

The concentrations of minor and trace elements in the waters and their Sr isotopic composition are shown in Table 1 and Figure 7. The summer samples have markedly higher concentrations of Al, Fe and Rb at most sites, which is similar to the results for Mg and K. Lakes Tasman and Hooker have the highest Sr concentration, and Lake Tekapo has the lowest Sr concentration below the alpine stream. The ⁸⁷Sr/⁸⁶Sr ratios of the stream waters range between 0.70914 and 0.71001. Water in Lake Tekapo is more radiogenic than water from the other locations. The winter water samples tend to co-vary and have slightly less radiogenic ⁸⁷Sr/⁸⁶Sr than samples collected at the same location in other seasons. The data in this study are generally in good agreement with previous studies that included data for the Waitaki catchment (Jacobson et al., 2003; Moore et al., 2013). An exception is Lake Tekapo, which is consistently more radiogenic in this study than that of Jacobson et al. (2003). The reason for this difference is unclear.

Figure 7. Downstream variation in freshwater Sr concentration and isotopic composition. A, Sr concentration and B, ⁸⁷Sr/⁸⁶Sr at the sample stations plotted as a function of position in the catchment, keyed to the sampling season.

The leachates of the moraine rocks are indicative of the more labile elements that may contribute to the dissolved load of the stream waters. Of the rocks collected high in the catchment on the terminal moraine of the Tasman Glacier, the leachates of the greywackes have the highest concentrations of Ca (Table 2), argillites have the highest K, and a chert leachate contained the highest concentration of Mg and Rb but extremely small amounts of Sr. The ⁸⁷Sr/⁸⁶Sr ratios of the greywacke leachates are 0.70822 and 0.70965, whereas the argillite leachates have lower Sr concentrations and are more radiogenic, with ⁸⁷Sr/⁸⁶Sr of 0.70994 and 0.71234. The mixed greywacke-argillite has an intermediate Sr concentration and ⁸⁷Sr/⁸⁶Sr (0.70909). The chert sample has the most radiogenic leachate (⁸⁷Sr/⁸⁶Sr = 0.71468), accompanied by a very low Sr concentration. These results are within the range reported by Adams and co-workers for whole rocks from the Torlesse Terrane (e.g. Adams and Graham, 1997; Adams and Maas, 2004). The measured ⁸⁷Sr/⁸⁶Sr in the Waitaki catchment water samples is within a much more restricted range than either the whole rocks or leachates.

The leachate of the Irishman Creek fault rock has relatively high concentrations of cations, especially Ca, Mg and Sr, that are each an order of magnitude higher than any other sample. The Ostler Fault rock leachate has concentrations of major elements that are similar to those of greywackes except for their notably low Fe and Al concentrations (Table 2). The Ostler Fault rock leachate has very high Ca and a somewhat elevated Sr concentration, in conjunction with a relatively unradiogenic ⁸⁷Sr/⁸⁶Sr of 0.70915. As for the moraine samples, the fault rock leachates have a much larger range of ⁸⁷Sr/⁸⁶Sr than the water samples in this study.

Discussion

Downstream trends in major element concentrations and contributions

The alpine stream, glacial ice and glacial terminal lakes Tasman and Hooker are the most pristine localities in the catchment in terms of human impact. Of all the waters analysed, the alpine stream has the lowest concentrations of all the major elements measured. All the ice samples have lower Ca concentration than all the water samples, with the exception of the winter sample from Stocking Stream. Two of the ice samples have low concentrations of all the major dissolved species analysed. These low concentrations are similar to the concentration in rainwater samples reported by Jacobson et al. (2003), and are interpreted to retain the signature of the precipitation from which the glacial ice formed. The other two ice samples (sampled in autumn and spring) have higher concentrations than expected for pure precipitation (rain or ice). We note that the ice samples were melted prior to filtering out any entrained dust or scree particles in the ice. It is possible that reaction between those particulates and the ice either prior to or during melting affected the concentrations of the melt water (Anderson et al., 1997). The observation that the alpine stream is more dilute than the terminal lakes may likewise be attributed to direct input of reactive, finely pulverised glacial sediment into the lakes. For the most part, the terminal lakes have higher absolute and proportional concentrations of Ca than the dissolved loads lower in the catchment. This is what would be expected if calcite dissolution dominates in the headwaters. Given their pH, none of the waters are saturated in calcite (Figure 6); this may indicate that there is not enough calcite in the rocks that are being supplied by uplift for the waters to reach saturation, or that the residence times of the waters in the catchment are too short for the calcite supply to be completely consumed (Maher, 2010). In contrast, dissolved Si is higher in concentration below the terminal lakes, a trend that is generally attributed to increased water residence times in rivers with lower uplift and runoff rates (Jacobson et al., 2003; Maher, 2010). In a study of the Haast and Clutha rivers, Kautz and Martin (2007) were unable to positively identify any residual calcite in bedload sediments. This would support the interpretation that the calcite supply in the lower Waitaki catchment has been consumed by previous chemical weathering while the silicate minerals remain available for continued weathering.

Although some previous studies have attributed an excess of dissolved Ca relative to Na to the chemical weathering of trace amounts of calcite (Blum et al., 1998; Jacobson et al., 2003, 2015; Moore et al., 2013), as discussed above, chemical weathering of other Ca-bearing minerals such as Ca-bearing plagioclase, apatite, epidote and fluorite, may also contribute dissolved Ca (Blum et al., 2002; Oliva et al., 2003; Andrews et al., 2016). The Ca/Na ratio has been used to estimate the sources of Ca in the dissolved load of waters across the world (Blum et al., 1998; Gaillardet et al., 1999; Anderson et al., 2000; Jacobson and Blum, 2000; Lyons et al., 2005) including the Waitaki catchment (Moore et al., 2013). Using the Ca/Na molar ratio (c. 0.35) of sandstones in the region (Moore et al., 2013) it is possible to estimate how much of the riverine Ca comes from silicate weathering for waters of the Waitaki catchment in this study. Results of this calculation indicate that, on average, 90% of the dissolved Ca is attributable to weathering of carbonate minerals (range 83%–96%), meaning that 10% (range 4%–17%) of the Ca is coming from weathering of silicates (+/− other Ca-bearing non-silicate minerals such as apatite). Earlier work across the Southern Alps by Jacobson et al., (2003) found that on average carbonate weathering provides 91% of the Ca in the dissolved load.

Removing the contribution of Ca from carbonate changes the average major cation molar proportions to 51% Na (up from 20%), 18% Ca (down from 68%), 19% Mg (up from 8%) and 12% K (up from 5%) for water samples. The terminal glacial lakes Tasman and Hooker have a notably higher proportion of K than the other localities, which might reflect preferential removal of K from glacially pulverised muscovite in the alpine areas (Drever and Hurcomb, 1986; Axtmann and Stallard, 1995; Blum and Erel, 1997; Hodson et al., 2000; Lyons et al., 2005). Non-stoichiometric hydrolysis of silicate minerals in regions with high rates of glaciation, has been attributed to high flushing rates through the ice-marginal and subglacial environments where K-bearing minerals experience high physical erosion rates (Anderson et al., 1997; Hodson et al., 2000).

Looking in more detail at down-catchment variations, it can be seen that the intermediate lakes of the arid Mackenzie Basin, Tekapo and Ruataniwha, have low and relatively constant dissolved Na concentrations in comparison with samples from the lowermost catchment Lake Aviemore and Lower Waitaki (Figure 5). The elevated Na results in a greater calculated proportion of silicate weathering in the lowest part of the catchment. Albite is the most significant Na-bearing mineral in the region, and it is suggested that feldspar is the dominant silicate weathering. Other silicates such as epidote, prehnite, pumpellyite and zeolite group minerals would not impact as strongly on the Na concentrations and would also tend to increase the dissolved Ca concentrations (Table S3). The lower catchment has more abundant, vegetated soils and relatively high temperature, factors that tend to enhance weathering of silicate minerals (Velbel, 1993; Calvaruso et al., 2006; Jin et al., 2008; Smits et al., 2012; Jacobson et al., 2015). Enhanced silicate weathering downstream was put forward by Andrews et al. (2016) to explain stable Sr isotope variations in Fiordland. In addition, the residence time of the water in the intermediate to lower catchment may have increased due to the formation and enhancements of lakes by the emplacement of dams, allowing for increased chemical weathering of silicates. An additional source of Na from sea spray, which would be greater nearer to the coast, cannot be ruled out, in particular because we do not have an independent measure for sea spray (Cl⁻ concentrations). However, based on the association of elevated Na with higher Si concentration in Lake Aviemore, we prefer the interpretation that the contribution of Na from sea spray is secondary to its source from silicate weathering in the lower catchment.

Sr isotopes as a monitor of mixing of waters

The Sr isotopic composition of the dissolved Sr in the catchment reflects the mixing of waters that have reacted with different source rocks as a proportion of their Sr concentration and water flux (Table 1, Figure 7; Table S1). Because we sampled all the water sources in the upper catchment (Figure 2), we can use the Sr isotope and mass balance in Lake Pukaki as a test of this method to see if it may be applied downstream. The uppermost catchment waters from the terminal glacial Lakes Tasman and Hooker and the alpine Stocking Stream have a relatively restricted ⁸⁷Sr/⁸⁶Sr ranging from 0.70942 to 0.70961 (Figure 8). Lake Tekapo has a more radiogenic ⁸⁷Sr/⁸⁶Sr of 0.71000 ± 0.00002. Because of the hydro-dams, 76 m³/s of water is currently being discharged by canal from Lake Tekapo into Lake Pukaki. Based on the average discharge of Lake Pukaki (211 m³/s), its waters are a mixture of 35% from Lake Tekapo and 65% from the glacial lakes and alpine stream. Using these values and the weighted average concentration of Sr in the inputs, the calculated ⁸⁷Sr/⁸⁶Sr of Lake Pukaki would be 0.70964. The present-day waters in Lake Pukaki have an average measured ⁸⁷Sr/⁸⁶Sr ratio of 0.70960 ± 0.00005, which agrees with the calculated value within the uncertainties. It should also be noted that some of the water from lakes Tasman and Hooker and Stocking Stream could be sourced from groundwater rather than solely from surface flow, as long as it has equilibrated with the same rock sources.

Figure 8. Plot showing the mass balance for Sr isotopic composition (⁸⁷Sr/⁸⁶Sr) in waters at each location in the Waitaki catchment. The range of ⁸⁷Sr/⁸⁶Sr measured at each sampling station is plotted as a function of its distance from the mouth of the Waitaki River (as in Figure 7B), except that the ⁸⁷Sr/⁸⁶Sr for Stations 1–3 are combined. Curves are labelled with the yearly average water fluxes (in m³/s) (from Table S1). Inferred values for locations where samples were not collected are in italics. The range of ⁸⁷Sr/⁸⁶Sr in leachates of rocks from the terminal moraine and the fault zones are indicated by the vertical line to the left of the samples, but note that the values are beyond the limits of the comparatively restricted range in the water samples.

Given the agreement between the measured and calculated values for the waters in Lake Pukaki, we should be able to calculate missing inputs in the Sr isotope balance for the rest of the catchment. The results of these calculations are also illustrated in Figure 8. At the next downstream location, Lake Ruataniwha (Station 6) is composed of a mixture of the entire outflow of Lake Pukaki, including the canal and river, along with an additional small inflow from the Tekapo River and an estimated additional input from Lake Ohau. The measured ⁸⁷Sr/⁸⁶Sr of Lake Ruataniwha is nearly identical to that of Lake Pukaki, as would be expected because it is dominated by the already mixed Pukaki by way of a canal, although an additional flux of water from Lake Ohau and its associated canal with an estimated ⁸⁷Sr/⁸⁶Sr similar to the uppermost catchment rivers could balance the small additional flux of more radiogenic water from Lake Tekapo. The next sampled location downstream, Lake Aviemore, with ⁸⁷Sr/⁸⁶Sr of 0.70931, is noticeably less radiogenic than Lake Ruataniwha. This could be accommodated by an additional flux from the Ahuriri River, which we did not collect, and other sources to the west with ⁸⁷Sr/⁸⁶Sr of approximately 0.70705. This is reasonable given the geology of the area, which may include more of the less-radiogenic greywackes and argillites such as those that were leached for this study (Table 2; Adams and Graham, 1997; Adams and Maas, 2004), or alternatively of mafic volcanic sources (Forsyth, 2001). The lowermost Waitaki sample is almost identical to Lake Aviemore, or perhaps slightly more radiogenic, which would be reasonable for a small flux of more radiogenic Sr similar to Lake Tekapo or from Cenozoic sediments, which occur in the north and eastern side of the catchment around the Hakataramea River (Forsyth, 2001). These calculations were done using average yearly fluxes, but it is noted that most of the lower catchment is typified by less radiogenic Sr in the winter season (Figure 7, Table 1). It is unclear whether this change is due to the increased fluxes of water through the hydro-dams to generate more electricity in winter or if the more radiogenic sources are less accessible to the streams in the winter season. The fact that the Si concentrations increase in winter in the lowermost catchment suggests that sources with greater silicate weathering become more dominant in the river in winter, and perhaps this is the signature of groundwater. In any case, the low residence time of waters in the upper- and lowermost parts of the catchment (Table S1) mean that the chemical and isotopic compositions respond quickly to variations in proportions of input sources. In the middle part of the catchment the large hydro-lakes Tekapo, Pukaki and Ohau have residence times on the order of 1–2 years, and it is noted that the Sr isotopic (Figure 7) and elemental compositions (Figure 5) of these lakes are much less seasonally variable than at either the top or bottom end of the catchment, implying that they are well mixed.

Confirmation of the Sr isotope and mass balance will entail analysing the major surface waters that were not undertaken in this study, in particular lakes Ohau and Benmore and the Ahuriri and Hakataramea rivers and subcatchments.

Effects of human activity and temporal variability of sources

The already modified hydrology of the Waitaki catchment may be further amended in the near future by changes in land use. The data presented here can be used as a baseline against which the impact of further changes in the catchment can be compared.

Although the data set presented here only covers one year, by sampling on four different occasions in different seasons over the course of a year, we have begun to investigate the possibility of temporal (seasonal) changes in the composition of Waitaki River water. It is worth noting that there is no single straightforward pattern of composition that can be related to simple conservative mixing of waters with well-characterized sources. As noted above, the parts of the hydrologic system with the shortest residence time should show the greatest geochemical variability, reflecting short-term changes in inputs, whereas the larger lakes with long residence times effectively ‘see through’ any specific events to give an integrated average of the chemical weathering taking place in the catchment.

Conclusion

In this work, downstream variations in the elemental and Sr isotopic composition of surface waters in the Waitaki River catchment were observed as a function of location at quarterly intervals over a year. In common with previous studies, we have found that the chemical weathering of calcite dominates the dissolved Ca concentrations, particularly high in the catchment, whereas both the absolute and proportional contribution of silicate weathering increases lower in the catchment. The Sr isotopic composition and flux are well explained by mixing of waters that have interacted with the various lithologies within the catchment, particularly in the upper parts of the catchment. Additional data from the parts of the catchment that were not sampled for this study, both surface and ground waters, will test the robustness and consistency of these results.

Acknowledgments

We thank Claudine Stirling, Malcolm Reid and Dave Barr for their generous help with the analyses undertaken in the Centre for Trace Element Analysis. Christine McLachlan, Briar Taylor-Silva and Rebecca Parker guided VP through measuring silica by spectrophotometry. Katrin Wellnitz helped with the analytical work and PHREEQC. Luke Easterbrook-Clarke was a great help with making figures. We thank Michael Bau, Sylvia Sander, Nathalie van Havre and Birthe Kortner for their support during the first water collecting field trip. Thanks to hydrologists Kathy Walter from NIWA and Eddie Stead from Meridian Energy Ltd. for providing unpublished river and canal flow data. Nick Mortimer and two anonymous reviewers provided suggestions that significantly improved this manuscript.

Disclosure statement

No potential conflict of interest was reported by the authors.