The strontium isotope distribution in water and sh within major South African catchments

Strontium has 4 naturally-occurring isotopes (84Sr, 86Sr, 87Sr, 88Sr) all of which are stable (Faure, 1986). e correlation between the 87Sr/86Sr isotope ratio of lake water and sh n spine tissue was investigated in 23 lakes within 4 major South African catchments. Data showed that sh within a speci c lake all have the same Sr isotope ratio in their n spine tissue regardless of species, age, sex and condition. e origin of the dissolved Sr fraction in lake water can be either from the natural weathering of upstream geological units or from an upstream anthropogenic source. e Sr isotopic ratios of the water samples were, however, constant over a multi-year period suggesting that the main source may be the more consistent geological environment. e Sr isotope ratio of river and lake water generally increases along the course of the rivers within the tertiary catchments of the areas investigated. In large rivers like the Vaal, where pollution also plays a role, the pattern is much more complicated. In the Olifants River catchment, Lake Middelburg, Lake Witbank and Lake Doornpoort have a similar Sr isotope ratio, which is distinct from Lake Bronkhorstspruit. Lake Loskop which is downstream from these lakes has a Sr isotope ratio between these two extremes, indicating mixing of water from upstream sources. Similarly Lake Arabie (Flag Boshielo), which is even further downstream, shows a Sr isotope composition between the composition of Lake Loskop and the lakes in the Elands River.


INTRODUCTION
Strontium has 4 naturally-occurring isotopes ( 84 Sr, 86 Sr, 87 Sr, 88 Sr), all of which are stable (Faure, 1986).Strontium isotope ratios have been used in various applications to link biota to their physical environment.Strontium is an ideal element for this purpose as it is present in high concentrations suitable for analytical applications, while Sr isotope ratios remain independent of biological processes (Capo et al., 1998).e 87 Sr/ 86 Sr ratios of natural materials re ect the source of Sr available during their formation (Capo et al., 1998).
Strontium isotope ratios have successfully been used to link elephant bone and ivory to the environment in which these animals lived (Van der Merwe et al., 1990;Vogel et al., 1990).A similar study linked Sr ratio distribution in modern rodents to geology (Hoppe et al., 1999), while Beard and Clark (2000) showed that the Sr isotope composition of skeletal material can indicate the birthplace and geographic mobility of humans and animals.
Lakes are very special habitats as sh migration and movements are limited by dam walls and usually very shallow waters at the inlets.Adult sh are therefore expected to spend their entire lives within a relatively limited area.Penne and Pierce (2006), in a telemetry study of carp in Clear Lake, Iowa, USA, found that carp congregated in a relatively small area in winter, then moved to speci c spawning areas in spring and spread out somewhat during the summer and autumn months.Otis and Weber (1982) showed similar results in a carp telemetry study of the Lake Winnebago system, USA, and indicated that carp in a river system occupy restricted home ranges.
It can therefore be hypothesised that bone tissue from sh in South African lakes may have a similar Sr isotope ratio to the aquatic environment they live in. is has been demonstrated in Atlantic salmon from the Connecticut River, Massachusetts, by Kennedy et al. (2002) and in Dolly Varden char from the Yukon Territory by Outridge et al. (2002), using sh otoliths.Strontium is incorporated into bone mass as substituting for calcium in various microcrystalline sites.Palmer and Edmond (1992) state that otolith 87 Sr/ 86 Sr ratios directly re ect dissolved ambient ratios, which in freshwater habitats depend on the geological composition of the catchment.Walther and orrold (2006) found in an experimental study of juvenile marine mummichogs (Fundulus heteroclitus) that the water composition and not the food composition determine the isotope composition of Sr deposited in sh otoliths.Using available South African data (De Villiers et al., 2000;De Villiers and De Wit, 2007), stable geographical distribution patterns in Sr isotope composition of river water can be shown.Douglas et al. (1995) used Sr isotope ratios to link suspended particulate matter in the Murray-Darling River system, Australia, to weathering lithologies in the catchment areas.Preliminary work (Jordaan et al., 2006;Jordaan et al., 2009;Jordaan and Rademeyer, 2009) already indicates a correlation between the Sr isotope ratio of sh n spine tissue and lake water in South African lakes.
If only the elemental composition of lake or river water in a large catchment is considered, then the mixing of water from two sub-catchments, which can be described by existing mixing models using the water Ca/Sr ratio (Land et al., 2000), may also be re ected in the strontium isotope ratio of sh n spines.Using the water Cl − /SO 4 2− ratio in a mixing model may similarly add information about possible anthropogenic in uences.
e aim of this paper is to investigate the relationship in strontium isotope ratios between di erent sh species and water within selected South African lakes.e study was undertaken to develop a scienti c method (forensic tool) to minimize illegal entries at major South African freshwater shing tournaments (Jordaan, 2015).

Description of the project area
e project area consisted of selected lakes within the Vaal, Mgeni, Crocodile (West) and Olifants River catchments (Fig. 1).e selection criteria were di erent sizes, di erent sources of pollution and di erent underlying geological composition.

Collection and preparation of samples
Samples included: water taken on the surface of lakes and sh samples taken mainly by bank and boat angling as well as gill netting (Table 1).Water samples for isotope analyses were collected in 2-L high-density polyethylene (HDPE) containers, acidi ed with ultrapure HNO 3 , cooled and sent to the laboratory for analyses within 24 h.Water samples for anion and metal concentration analyses were collected in 2-L high-density polyethylene (HDPE) containers, not acidi ed, cooled and sent to the laboratory for analyses within 24 h.
Four sh species were targeted, i.e., common carp (Cyprinus carpio), sharptooth cat sh, (Clarias gariepinus), largemouth bass, (Micropterus salmoides) and Mozambique tilapia (Oreochromis mossambicus), although minor sh species were also included.Fish samples were collected in plastic containers, packed in ice and brought to the laboratory, where they were frozen to −5°C.
Fish spines from the dorsal or pectoral ns were removed and dried in an oven at approximately 80°C for 14 days.All so tissues were removed followed by pulverizing in a swing mill.

Chemical analyses
e analytical method for Sr isotope analysis of water samples consisted of ltering 2 L samples through 0.45 µm cellulose nitrate lters and drying in a drying box to concentrate the Sr.
e samples were re-dissolved in 10 mL concentrated nitric acid and evaporated to dryness in 50 mL disposable polypropylene beakers on a hotplate − 2 mL 6 M HCl was added, followed by drying on a hotplate.is step was repeated twice to ensure the samples were converted to chlorides.e samples were then centrifuged for 5 min at 2 500 r/min.e samples were puri ed using 6 mL Bio-Rad AG50Wx12, 200-400# cation resin packed in a 10 mm ID quartz glass column.e resin was cleaned using 30 mL 2.5 M HCl. e samples were carefully transferred onto a resin bed, taking care not to disrupt the resin.Hydrochloric acid was passed through the column, collecting Rb, Sr and the rare earth element fraction.e puri ed samples were then analysed for Sr isotope ratios using a Finnigan MAT 261 thermal ionization mass spectrometer.
e instrument is controlled by Run It 26X so ware created by Spectromat (Germany).
e analytical method for Sr isotope analyses of sh n spine samples consisted of weighing 1 g of pulverized sample into clean, static-free Savillex beakers; 2 mL concentrated nitric acid was added such that the samples were thoroughly wetted. 2 mL 6 M HCl was added and the beakers closed with screw caps.
e sample was le to dissolve overnight on a hotplate set at  (Jordaan and Maritz, 2010).Samples were analysed either undiluted or at a dilution factor of 7 times on a Dionex QIC analyser ion chromatograph.
95°C. e samples were removed from the hotplate and allowed to dry to half volume a er which another 1 mL of concentrated nitric acid was added and allowed to dry.A further 2 mL of 6 M HCl was added and allowed to dry. is step was repeated twice to ensure the sample was converted to chlorides.e sample was then centrifuged for 5 min at 2 500 r/min.e samples were puri ed using Bio-Rad AG50Wx12, 200-400# cation resin, followed by analysis on a Finnigan MAT 261 thermal ionization mass spectrometer.
e analytical method for concentration analyses of Sr and Ca in water samples consisted of rst ltering samples through 0.45 µm cellulose nitrate lters (Jordaan and Maritz, 2010).Water samples were then diluted 5 times to add the internal Quality assurance e standard reference material NIST987, used for calibration of the Sr isotope ratio method, produced the following average 87 Sr/ 86 Sr ratio: 0.710213, range: 0.709915-0.710414,standard deviation: 0.000067, n = 96, which is within the published range of 0.71034±0.00026.A set of samples were analysed at the University of Cape Town on a NU Instruments MC-ICP-MS for veri cation.An 87 Sr/ 86 Sr ratio of 0.710300 (n = 8) was obtained for the reference material.A detection limit of 0.2 µg/L Sr, 0.03 mg/L Cl − and 0.10 mg/L SO 4 2− was achieved for concentration analyses of water samples.
To evaluate the method used to determine concentrations in water, liquid samples from the SABS Water-Check (Group 1 and Group 3) inter-laboratory pro ciency test (South African Bureau of Standards, 2010a, 2010b, 2010c, 2010d, 2010e, 2010f) were regularly analysed.e elements analysed included Al, Ba, Be, B, Cd, Cr, Co, Cu, Fe, Pb, Mn, Hg, Mo, Ni, Si, Sr, V, Zn, As and Se for Group 1, and chloride and sulphate for Group 3. e average Group 1 z-scores obtained for this evaluation are as follows: April 2010, 0.73; July 2010, 0.73 and October 2010, 0.69.e average Group 3 z-scores obtained for this evaluation are as follows: June 2010, 0.77; September 2010, 0.70 and December 2010, 0.60.All z-scores between −2 and 2 are considered satisfactory.

Statistical analyses
e sample codes used are explained in Table 1.A summary of analytical data for water and sh samples (all species) is listed in Table 2.A one-way ANOVA test was used to assess the correlation of Sr isotope ratios between water and sh from lakes in the project area (Table 2).Similarly, the largest dataset from Lake Loskop sh, as well as the results of a one-way ANOVA test used to assess the correlation in Sr isotope ratios between carp and other sh species within the lake, are listed in Table 3. e Single Factor ANOVA function of the so ware package Microso O ce Excel was used for the calculation.

Sample identi cation
In Table 1 the number of species sampled and spine tissues collected for isotope ratio analyses from each lake are represented.e codes provided for the di erent lakes and sh species are used in all gures and tables.e Olifants River catchment was sampled during both the wet and dry season, while most of the lakes in the other catchments were sampled only during each of the seasons for Ca, Sr and anion concentration analyses.For the purpose of this project the period from August to November is considered the dry season and from December to July the wet season.

The Crocodile River catchment
Figure 2 shows the Sr isotope ratio variation along the Crocodile River system in Gauteng and the North West Province.is system is complicated due to complex geology,

Figure 2
Sr isotope ratios of lake water in the Crocodile River catchment superimposed on the 1:1 000 000 scale geology map of South Africa (Council for Geoscience, 2011), draped over the topography.Square symbols show the 87 Sr/ 86 Sr ratio of groundwater from McCa rey and Willis (2001).The colour of the dots/squares represents the 87 Sr/ 86 Sr isotope ratio as indicated in the legend.Rivers (2 nd order and higher) and catchments (ternary)   but in general shows a slight increase in Sr isotope ratios as water moves downstream within tertiary catchments.Lake Roodekopjes is inconsistent with this pattern, as it has a slightly lower value than Lake Hartbeespoort, which is approximately 50 km upstream.The Lake Vaalkop tertiary catchment contains the Pilanesberg alkali intrusive complex which is different in both composition and age to the geology of the Lake Hartbeespoort/Roodekopjes system.This may explain the very distinct Sr ratio of the Lake Vaalkop water.

The Olifants River catchment
The Sr isotope ratio of water samples from different localities in the Olifants River catchment shows a very distinct pattern (Fig. 3).Water within tertiary catchments of the Olifants River system generally shows a slight increase in the 87 Sr/ 86 Sr isotope ratio downstream.When water from two different tertiary catchments is combined the Sr isotope ratio shows an intermediate value between the values of the two sources.In the Olifants River catchment, water from the Witbank and Middelburg sub-catchments is similar.When combined with water from the Bronkhorstspruit sub-catchment the intermediate ratios of Lake Loskop result.Lakes in the Elands River catchment have much higher Sr ratios than that of the Olifants River catchment and result in the intermediate Sr isotope ratio of water in Lake Arabie, located downstream of the confluence of the two systems.Figure 3 shows that catchments with similar geology produce similar isotope ratios in the rivers that drain them, while different geology can produce quite distinct Sr ratios in others.

The Orange/Vaal River catchment
Figure 4 shows the Sr isotope ratios in the massive Orange/Vaal River system (some data from De Villiers et al., 2000;De Villiers and De Wit, 2007).When looking at the Upper Orange and Caledon River catchment a similar pattern emerges, with a slight increase in the Sr isotope ratio within tertiary catchments. is part drains the relatively uncomplicated Karoo stratigraphy and is very similar to the Mgeni River which drains eastwards from the same Karoo stratigraphy.e Lower Orange River shows a mixed ratio of water from the Upper Orange and the Vaal Rivers.Further downstream, variation in Sr isotope ratios is limited to con uences of major tributaries.e Vaal River (Fig. 4) shows the most complicated and varied Sr isotope ratio pattern of all the systems investigated.In this case smaller sub-catchments need to be evaluated to follow the variation in Sr isotope ratios.Figure 5 shows the Vaal River from Lake Vaal in the east to Lake Bloemhof in the west.e map indicates 2 nd and higher order rivers as well as tertiary catchments.Lake Vaal is very large and has two inlets.ere is a clear di erence in Sr isotope ratios of water from the Vaal River side versus the Wilge River side.Water is well mixed in the western part of the lake closest to the dam wall.e Vaal River between Lake Vaal and Lake Bloemhof shows a large variation in Sr isotope ratios.is is mostly related to the large number of tributaries joining the Vaal River in this area.Tributaries have very distinct Sr isotope ratios, possibly related not only to the underlying geology but also to industrial pollution from the Vereeniging and Vanderbijlpark areas.Lake Bloemhof (Fig. 5) also has two inlets but seems to be dominated by the Vaal River as it has a higher ow rate than the Vet River.e Vet River does have a distinct Sr isotope ratio 10 km upstream from Lake Bloemhof.Sr isotope ratios of lake and river water in the Orange/Vaal, Crocodile, Olifants and Mgeni River catchments, superimposed on the 1:1 000 000 scale geology map of South Africa (Council for Geoscience, 2011), draped over the topography.Some data from the Orange/Vaal catchment from De Villiers et al. (2000).Swaziland data from De Villiers and De Wit (2007).The colour of the dots represents the 87 Sr/ 86 Sr isotope ratio as indicated in the legend.Rivers and catchments are indicated as blue and black lines, respectively.

Figure 5
Sr isotope ratios of water in Lake Vaal and Lake Bloemhof superimposed on the 1:1 000 000 scale geology map of South Africa (Council for Geoscience, 2011), draped over the topography.Some data from the Orange/Vaal catchment from De Villiers et al. (2000).The colour of the dots represents the 87 Sr/ 86 Sr isotope ratio as indicated in the legend.Rivers (2 nd order and higher) and catchments (ternary) are indicated as blue and black lines, respectively.

The Mgeni River catchment
Figure 6 shows the Sr isotope ratios of water from 4 lakes within the Mgeni River in KwaZulu-Natal. is system appears much simpler as it is contained within one tertiary catchment which shows a slight increase in Sr isotope ratios moving downstream from Lake Midmar towards Lake Inanda.e river is approximately 130 km long and passes over several distinct geological units which may contribute to the Sr isotope ratio of the river water.

Figure 6
Sr isotope ratios of lake water in the Mgeni River catchment superimposed on the 1:1 000 000 scale geology map of South Africa (Council for Geoscience, 2011), draped over the topography.The colour of the dots represents the 87 Sr/ 86 Sr isotope ratio as indicated in the legend.Rivers (2 nd order and higher) and catchments (ternary) are indicated as blue and black lines, respectively.

Fish
Data obtained during the project indicated a correlation (although not in all cases statistically signi cant), between the 87 Sr/ 86 Sr isotope ratio of sh spine tissue and the lake water in which the sh lived while developing these tissues (Table 2).In the Mgeni and Orange/Vaal River catchments the Sr ratio of sh and water were statistically the same.In the Olifants and Crocodile River catchments population sizes were too small for comparison, except at Lake Loskop where the Sr ratios of sh and water were statistically di erent.In Lake Loskop the Sr ratios of di erent sh species were not signi cantly di erent (Table 3).
When plotting the average 87 Sr/ 86 Sr ratios of sh spine tissue against the average 87 Sr/ 86 Sr ratios of water (Fig. 7), a signicant linear regression (r = 0.98, p < 0.005) was obtained.e error bars indicate the total range of values for each locality.
e range of values for water samples from any speci c lake is generally smaller than the range of values for sh samples from that same lake.It is, however, not in uenced by the speci c sh species (Table 3), the sex of the sh, the age of the sh and the season in which the sh were caught, as sh were collected from both sexes, both seasons and di erent ages, yet still produced a limited range of Sr isotope ratios for each lake.e Sr ratios were in some cases determined over a 4-year period in which time there were no signi cant changes in these ratios (Table 1).
Fish spine tissue from Lake Rust de Winter showed a relatively large variation in 87 Sr/ 86 Sr ratio, while the water had a relatively small range.in uence on the natural water chemistry.Correlating sh to lake water should therefore be possible despite the fact that the water may contain signi cant anthropogenic components.

Geological signi cance and Sr isotope ratios
If Rb and Sr are incorporated into rock at its formation and the system remains closed, the amount of 87 Sr increases over time as radioactive 87 Rb decays.Older rocks will therefore, in general, have higher 87 Sr/ 86 Sr ratios than younger rocks with the same initial Rb/Sr ratio (Capo et al., 1998).e utility of the Rb-Sr isotope system results from the fact that di erent minerals in a given geological setting can have distinctly di erent 87 Sr/ 86 Sr ratios as a consequence of di erent ages, original Rb/Sr values and the initial 87 Sr/ 86 Sr ratios.Strontium derived from minerals through weathering reactions will have the same 87 Sr/ 86 Sr as the initial mineral.Di erences in 87 Sr/ 86 Sr in catchment waters will therefore either depend on di erences in mineralogy along contrasting owpaths or on the relative amounts of Sr weathered from the same suite of minerals (Kendal et al., 1995).According to Faure (1986) Rb-bearing minerals are generally more resistant to weathering than Sr-bearing minerals.Strontium is therefore more readily lost from rocks exposed to weathering than Rb and the 87 Sr/ 86 Sr ratio of Sr that goes into solution is generally lower than the 87 Sr/ 86 Sr ratio of the unweathered rock.Faure (1986) does however note exceptions to this generalization.
Figure 4 shows 4 th and higher order rivers superimposed on the 1:1 000 000 geology map of South Africa.When evaluating the regional systems, it is noted that the origins of the Caledon, Orange, Mgeni and Wilge Rivers are all situated in the middle and upper regions of the Drakensberg.In all of these rivers a very similar Sr isotope pattern emerges, regardless of the direction in which the river drains (some data from De Villiers et al., 2000).ese rivers originate in or close to the Drakensburg basalts, then ow into older successions of the Karoo sedimentary sequence until they nally reach signi cantly older basement lithologies.e 87 Sr/ 86 Sr ratios of water samples increase from approximately 0.709 at the top to 0.715 at the base of the Karoo sequence.Where the Vaal River drains lithologies older than the Karoo sediments it has a Sr ratio of approximately 0.715 to 0.720, which is maintained in the Lower Orange River.ere are however many local variations to this pattern.Fisher and Strueber (1976) found a similar situation in the Susquehanna River and its tributaries that drain large areas of Pennsylvania and Maryland in the USA.ey observed that the 87 Sr/ 86 Sr ratio varies irregularly along the river as a result of mixing with water from tributaries.Pollution from anthropogenic sources may play a role in portions of the Vaal River, but this is likely a minor contributor to the Sr ratio of a large river.Several inter-catchment water transfers to and from the Vaal River may also contribute to a more irregular Sr isotope pattern.
e Swaziland rivers show a very complex pattern as they drain a very complicated geology (data from De Villiers and De Wit, 2007).e patterns are very di erent from the rivers draining the Drakensberg lavas and sediments.Primary river water in an area subject to active erosion will show a much closer relationship with the geology than massive rivers that ow for hundreds of kilometers and show only average Sr isotope ratios.e Crocodile River system (Fig. 2) originates mostly in the Transvaal sediments and then continues through the Bushveld complex. is gives all the tributaries a similar Sr isotope ratio.Lake Vaalkop has two inlets and is very di erent to the rest of the system.e Hex River (southern inlet) is similar to the rest For codes see Table 1.
the regression line in Fig. 7. Samples from Lake Loskop, Lake Roodeplaat, Lake Bon Accord and Lake Doornpoort all plot slightly above the regression line in Fig. 7.Even though only a limited number of spine samples were analysed from each individual lake, the data range for each lake indicates that this shi in the data may be real and not related to an analytical artefact.e relatively large spread in data from Lake Vaal (Fig. 7) is due to a di erence in Sr isotope ratios between the Vaal and the Wilge Rivers that feed the lake.Fish were mainly collected during shing competitions at the Jim Fouché resort on the Wilge River inlet side, although sampling was not restricted to this area.Lake Bloemhof shows a similar situation although the lake is dominated by the Vaal River.

In uence of inter-catchment water transfers on Sr isotope ratios
Middleton and Bailey (2005) describe 28 major inter-catchment water transfer schemes for South Africa.ese systems include domestic and industrial water supply schemes, hydro-electric power generation schemes and irrigation schemes.ese systems either transport water directly from one catchment to another or transport water for domestic or industrial use that reaches destination rivers as waste water (DWAF, 2004).
e most signi cant of these schemes is water from the Lesotho Highlands Project which is transferred into the Vaal River catchment and then pumped from Lake Vaal to Johannesburg and Pretoria to end up in the Crocodile River catchment.Water from the Lesotho mountains has a distinctly low 87 Sr/ 86 Sr isotope ratio (0.709) compared to other more local Lake Vaal tributaries (0.715) (Fig. 4).erefore the Wilge River side of Lake Vaal, where the water from Lesotho is received, also has a lower 87 Sr/ 86 Sr isotope ratio than the Vaal River side.
Water is also transferred to the Olifants River catchment to supply coal-red power stations in the Upper Olifants River region. is water is, however, not returned to the Olifants River system (Van der Merwe, 2011).ese pumping schemes should therefore contribute in some degree to the chemical characteristics of the water in rivers and lakes in South Africa.
e amount of water displaced through these schemes is, however, relatively constant and should therefore have a constant  All water data from project area.Orange and Vaal River data from De Villiers et al. (2000).For codes see Table 1.

Figure 9
All water data from project area.For codes see Table 1.
of the catchment but the Elands River (western inlet) originates near the Pilanesberg alkali complex. is may contribute to the much lower Sr isotope ratio of Lake Vaalkop.e Crocodile River system is also the destination of inter-catchment water transfers from Lake Vaal. is may contribute to the di erence between water that originates locally and water that has a larger component of water from Lake Vaal.
In the Olifants River catchment (Fig. 3), tributaries originate in, geologically, very distinctive sequences.e Middelburg and Witbank sub-catchments originate in the older portions of the Karoo sequence and are very di erent from the Bronkhorstspruit sub-catchment, which originates in mostly Transvaal sequence lithologies.is can clearly be seen in the Sr isotope ratios, which combine to give Lake Loskop an intermediate Sr isotope ratio.e Elands River originates in the Bushveld igneous complex and is very di erent from the Olifants River.e Sr isotope ratio of the Elands River is therefore higher (0.747) than the water in Lake Loskop (0.729).A mixing of these sources results in the Sr ratio of 0.738 at Lake Arabie further downstream (Table 2).e Sr isotope ratio of water in a river system is therefore mostly determined by the geology at the origin of the river were most weathering occurs (De Villiers et al., 2000).Further downstream it is determined by the contributions from tributaries which gain their Sr isotope ratio in a similar fashion.Sr isotope data from rocks in the upper catchment of the Olifants River are limited.Barton et al. (2004) investigated Sr isotope ratios in glauconite from quartzrich sediments of the Witbank coal eld above the Number 4 and Number 5 coal seams.e 87 Sr/ 86 Sr ratios ranged from 0.751 to 0.922 for glauconite from the Vryheid formation, from 1.165 to 1.241 for coarse-grained detrital muscovite, and was 1.201 for coarse-grained detrital biotite from strata closely associated with the glauconite-bearing beds.
Published Sr isotope data from global lakes and rivers is very limited.Comparing the Sr isotope data from South African rivers on a mixing diagram (Fig. 8) to the limited amount of data from international examples listed by Faure (1986), Edmond (1992), Harris et al. (1998), Galy et al. (1999), Dalai et al. (2003), Négrel et al. (2004) andDe Villiers andDe Wit (2007), it can be seen that the Olifants and Crocodile River systems correspond to rivers draining the Himalayas, while the Orange/Vaal and Mgeni River systems fall within the eld of global rivers.De Villiers and De Wit (2007) noted that rivers draining Swaziland also correspond to Himalayan examples.Rivers draining catchments dominated by K-granites have higher 87 Sr/ 86 Sr ratios (average 0.767) while Achaean Na-rich gneisses (average 0.731) and volcano-sedimentary sequences (average 0.735) have lower values.Edmond (1992) concluded that the average uvial 87 Sr/ 86 Sr ratio (excluding Himalayas) is 0.710.
A linear correlation on a mixing diagram is typical of many river systems where a high Sr, low 87 Sr/ 86 Sr end member, derived from carbonate, is mixed with a low Sr, high 87 Sr/ 86 Sr end member, derived from silicate (Palmer and Edmond, 1992).e Mgeni River system shows a gradual increase in the Sr isotope ratio as well as in the Sr content moving downstream from Lake Midmar, to Lake Albert Falls, then to Lake Nagle and Lake Inanda (Fig. 8).e Vaal River system does not show such clear trends as the system is massive and has numerous tributaries with very di erent Sr characteristics.e present data however correspond very well with the Vaal River data of De Villiers et al. (2000).e data from De Villiers et al. (2000) for the Lower Orange River plots centrally within all the available data for the upstream parts of this river system, possibly indicating that the Lower Orange River composition represents mixing of all the upstream components (Fig. 8).e Crocodile River system shows much clearer mixing properties.Lake Loskop falls on a mixing line between Lake Bronkhorstspruit and Lake Witbank/Middelburg.Lake Arabie falls on a mixing line between Lake Loskop and Lake Rhenosterkop.is is a precise representation of the eld observations.Lake Rhenosterkop, the Elands River and Lake Rust de Winter show the same linear downstream trend as the Mgeni River system.Variation in Sr concentration data in the Crocodile River system is limited.Clear linear mixing trends are di cult to identify.Land et al. (2000) used element ratios rather than element concentrations for mixing models when investigating a small catchment situated within the Kalix River watershed in northern Sweden.Figure 9 (a er Land et al., 2000) shows the 87 Sr/ 86 Sr vs. Ca/Sr for the project area.Mixing trends in data from all four large catchments are essentially similar to trends described on the 87 Sr/ 86 Sr vs. 1/Sr plot (Fig. 8). Figure 10 shows the 87 Sr/ 86 Sr vs. Cl − /SO 4 2− data on a mixing diagram.It includes the possibility of chloride and sulphur from possible anthropogenic pollution sources.Again, mixing trends in data from all four large catchments are essentially similar to trends described on the 87 Sr/ 86 Sr vs. 1/Sr plot (Fig. 8).Data are however less spread out and mixing relationships are clearer.

CONCLUSIONS
Sr isotope ratios of water samples were in all cases determined on the dissolved Sr fraction.e origin of this fraction can therefore be either from the natural weathering of upstream geological units or from upstream anthropogenic sources.An investigation of the Vaal and Orange Rivers by De Villiers et al. (2000) showed that much of the Sr isotope ratio of the river is determined by the isotope ratio of the predominant geological strata in the upper part of the catchment (where active weathering is taking place).In the upper Olifants River system there is certainly ample proof of additions to the river water from mine, municipal or industrial sources.e Sr isotopic ratios of the water samples were however constant over a 3-year period (2007)(2008)(2009), suggesting that the main source may be the more consistent geological environment.Anthropogenic Sr sources vary along the course of a river, but individual anthropogenic sources may be more constant over time than expected.Most of the larger sh that were analysed (carp above 4 kg) were also older than 5 years (Balik et al., 2006;Sedaghat, 2013) indicating that a relatively constant Sr source may have been available to them for several years.In larger lakes with only one major inlet the Sr isotope ratio is very constant, even close to the inlet.In large lakes with two major inlets like Lake Vaal or Lake Bloemhof, slight di erences may exist between the inlets.
e Sr isotope ratio of sh spine tissue shows a remarkable correlation with the Sr isotope ratio of the lake water in which these sh lived.It correlates much better than any single element and may therefore be used to establish a link between the sh and lake water.It also shows that the Sr isotope ratio systematics is constant over the multi-year span of the project and that the season in which samples were taken did not contribute any variability.e establishment of a chemical link between sh and lake water provides a forensic tool to eliminate illegal entries at major South African shing tournaments.
Inter-catchment water transfers are relatively constant and should have a constant in uence on the natural water chemistry.Correlating sh to lake water should therefore be possible despite the fact that the water may contain anthropogenic components.
In the Olifants River catchment, Lake Middelburg, Lake Witbank and Lake Doornpoort have a similar Sr isotope ratio, which is distinct from Lake Bronkhorstspruit, primarily due to di erences in catchment geology.Lake Loskop, which is downstream from these lakes, has a Sr isotope ratio between these two extremes, clearly indicating mixing of water from upstream sources.Similarly, Lake Arabie, which is even further downstream, shows a Sr isotope composition between the composition of Lake Loskop and the lakes in the Elands River, again indicating mixing.e Sr isotope composition of a single lake may therefore be the result of several factors that may give a lake and the sh living within it a distinctive character.

Figure 3
Figure 3Sr isotope ratios of lake and river water in the Olifants River catchment superimposed on the 1:1 000 000 scale geology map of South Africa (Council for Geoscience, 2011), draped over the topography.The colour of the dots represents the 87 Sr/ 86 Sr isotope ratio as indicated in the legend.Rivers (2 nd order and higher) and catchments (ternary) are indicated as blue and black lines respectively.
It is not clear what causes this large range in sh spine tissue data or why most of the data plot below http://dx.doi.org/10.4314/wsa.v42i2.05Available on website http://www.wrc.org.zaISSN 1816-7950 (On-line) = Water SA Vol.42 No. 2 April 2016 Published under a Creative Commons Attribution Licence

Figure 7
Figure 7Average Sr isotope ratios of water and sh from the project area.Solid line indicates 1:1 reference line.A linear regression of all data produced a R 2 value of 0.98, whether forced through zero (y = 1.00x) or not (y = 0.93x + 0.05).Error bars indicate total range of data for each locality.For codes see Table1.

Figure 8
Figure 8All water data from project area.Orange and Vaal River datafrom De  Villiers et al. (2000).For codes see Table1.

TABLE 2 Average (± standard deviation) Sr isotope ratio of lake water and sh n spines (all species) from lakes in the project area. Average (± standard deviation) Ca, Sr and anion content of lake water from the project area. Codes for the lakes and number of samples (n) are explained in Table 1. Catchment Lake Water 87 Sr/ 86 Sr Spine 87 Sr/ 86 Sr p-value Cl (mg
http://dx.doi.org/10.4314/wsa.v42i2.05Available on website http://www.wrc.org.zaISSN 1816-7950 (On-line) = Water SA Vol.42 No. 2 April 2016 Published under a Creative Commons Attribution Licence