Environ Monit Assess (2018) 190:427 https://doi.org/10.1007/s10661-018-6764-6

A multivariate examination of ‘artificial mussels’ in conjunction with spot water tests in freshwater ecosystems S. Dahms-Verster

&

N. J. Baker & R. Greenfield

Received: 4 October 2017 / Accepted: 31 May 2018 # Springer International Publishing AG, part of Springer Nature 2018

Abstract Metal pollution in aquatic systems is considered a serious environmental issue globally due to their ability to accumulate in aquatic environments. Wetlands are vulnerable to this pollution as they are known to trap toxins, removing them from the water. Artificial mussel technology, originally developed for marine environments, was applied to this freshwater system and spot water samples were collected. The Nyl River floodplain (Ramsar classified) is one of the largest and most ecologically significant wetlands in South Africa. The aims of this study were to determine metal contamination along the Nyl River system by means of artificial mussels (AM) and water ICP-MS analysis and to determine whether the use of AMs in conjunction with spot water testing could give more insight into the pollution in freshwater wetlands. The concentrations of Al, Cd, Cr, Co, Cu, Fe, Mn, Ni, Pb and Zn were determined at seven sites. It was determined that the levels accumulated by the AMs differed in spatial and temporal trends when compared to spot water samples. It was determined that there were high levels of some metals found in the spot water tests that were not corroborated by the AMs results, which could indicate isolated pollution events. The use of AMs in conjunction with spot water testing was determined to be beneficial in gaining deeper insight into water metal conditions in dynamic freshwater systems. S. Dahms-Verster (*) : N. J. Baker : R. Greenfield Department of Zoology, University of Johannesburg, PO Box 524, Auckland Park 2006, South Africa e-mail: [email protected]

Keywords ICP-MS . Wetland . Artificial mussels . Nylsvley . Water quality . Ramsar

Introduction The importance of wetlands in South Africa is highlighted by the increase of water pollution over the past decade, defying the efforts made globally to improve the quality of natural water resources. Metal contamination has been a popular topic of research in ecotoxicological studies because most metals have potentially harmful effects in concentrations above threshold levels (Claassens et al. 2016; Gerber et al. 2018; Greenfield et al. 2012; Pheiffer et al. 2014). There are many methods of determining metal contamination including water, sediment and bioindicator studies (Echols et al. 2009), as well as many analytical procedures to analyse samples including ICP-OES and ICP-MS (Boss and Fredeen 2004). Each method has its own positive and negative attributes and cannot be used as a stand-alone method of determining metal pollution. Essentially, it is necessary to incorporate various methods to acquire a full understanding of the metal pollution in a certain aquatic ecosystem. The levels of metals in a system can be influenced by many factors such as pH and total hardness of water as well as the synergistic interactions of different metals when in contact with one another (DWAF 1996). Metals occur in different speciation forms according to the water matrix and what it is constituted of. When considering environmental pollution holistically, it is

427

Page 2 of 17

evident that inland water systems such as rivers, lakes and wetlands receive pollution directly by industry and municipalities releasing wastewater directly into rivers, but that they are also subject to pollutants from the air and sediment through various natural or man-made processes (Echols et al. 2009). Amongst the pollutants commonly found in wetland systems, are metals which end up in water systems through these various processes (Sanders 1997). It is because of the toxicity of metals in high concentrations that certain metals were placed on the US EPA Priority Pollutant List (Echols et al. 2009). Amongst the metals on the list are Cd, Cr, Cu, Ni, Pb and Zn (Echols et al. 2009). When studying the metal content of water, it is important to note that the results acquired will only be an indication of the contaminants in the water at the precise moment of sampling (Prosi 1979). Conditions may differ hourly, daily and due to individual pollution events, which makes the use of spot water sampling for metal analysis problematic (Phillips 1977). The use of live organisms as biological indicators (bioindicators) in monitoring environmental pollution has become standard practice over the last few decades (Holt and Miller 2010). Their use in metal pollution studies is prominent due to the ability of some species to bioaccumulate metals from aquatic environments (Holt and Miller 2010). This practice gives an indication of pollutant concentrations on a temporal scale seeing as other methods, such as spot water sampling, give only a snapshot of the conditions in the water body (Hellawell 1986). The use of bioindicators can, however, be challenging due to variations in physiological regulation, organism availability, inter-species differences and ethical considerations (Markert et al. 2003). These factors have led to the development of a passive sampling device called the artificial mussel (AM) by Wu et al. in 2007. Artificial mussels contain Chelex-100, a chelating resin that is the ingredient which passively accumulates the bioavailable fraction of metals from the water over time. AM technology was originally developed and evaluated under marine conditions; however, recent studies have used AMs to validate metal contamination in freshwater systems (Kibria et al. 2010; Hossain et al. 2015; Claassens et al. 2016). Previous studies have incorporated the use of live indigenous mussels into AM studies for comparison between the accumulation of metals between AMs and live mussels (Leung et al. 2008; Degger et al. 2011; Gonzalez-Rey et al. 2011). Each of the studies found similarities in the accumulation

Environ Monit Assess (2018) 190:427

of certain metals although the uptake profiles differed. According to previous studies, AMs are less affected by salinity and temperature than bioindicator species and can as such be compared between systems (Wu et al. 2007). A recent study by Claassens et al. (2016) compared the accumulation of AMs and native freshwater snail species Melanoides terbiculata. It was determined that there were similarities in spatial trends considering metal accumulation between the AMs and the snail species. Extensive research on AMs has proven that they are able to accumulate environmentally relevant concentrations of Cd, Cr, Cu, Pb and Zn. Wu et al. (2007) determined that the resin contained in the AMs accumulates and releases metals as the levels in the water rise and fall. A recent study has also shown that AMs are effective at accumulating anionic arsenic from the water (Claassens et al. 2016). The present study aims to further evaluate the use of AMs in determining metal contamination in freshwater environments through comparisons between spot water samples and AM accumulation in an ecologically significant floodplain wetland in South Africa.

Materials and methods Study area and site selection The Nyl River originates near Modimolle, Limpopo, and flows through the towns of Modimolle and Mookgophong (Greenfield et al. 2007). It continues through the town of Mokopane from where it is renamed the Mogalakwena River. From Mokopane, the Mogalakwena River flows northeast towards the Botswana border where it meets the Limpopo River. The Nyl River is dammed upstream of Modimolle to form the Donkerpoort Dam and just upstream of Mokopane to form Moorddrift Dam. The Nyl River floodplain is one of the largest floodplain wetlands in South Africa (McCarthy et al. 2011). It acts as a habitat for multiple threatened and endangered species and acts as breeding grounds for many Red Data bird species. The abundance of avian species in the area was one of the leading factors leading to its Ramsar accreditation (Haskins and Kruger 1997). The importance of the Nyl River floodplain is highlighted by an increase in flooding events in the area which can indicate that the wetland is not performing optimally (Vlok et al. 2006). The Nyl River and Nyl River floodplain are contained in the Waterberg area

Environ Monit Assess (2018) 190:427

which is a registered Biosphere Reserve. This fact further increases the local and international importance of this river and the floodplain which it forms part of. The Nyl River has been studied in recent years comprehensively by the Water Research Commission (WRC) (Vlok et al. 2006) and by environmental consultants for the purposes of Environmental Impact Assessments (EIA). The interest in this system is due to its sheer size and high biodiversity. In the WRC study, the focus was placed on the contamination of water and sediments with metals (Vlok et al. 2006). The WRC study determined that the levels of metals in the river were of little concern at the time of sampling (2001/2002). Furthermore, the report determined that the metal levels generally remained below the Guidelines for Aquatic Ecosystems (DWAF 1996). At the time of the study, the water was reported to be soft to medium regarding hardness and the pH of the water was mostly neutral throughout the study and system (Vlok et al. 2006). The concentrations of metals measured were also found to be high throughout the system which indicated that metal concentrations in the water and sediment were naturally high (Greenfield 2004). A specialist study on the surface water of the Nyl River system was conducted by the consulting firm AED for an Environmental Impact Assessment (EIA) considering a proposed platinum mine in the surrounding area. It was determined that the Nyl floodplain was threatened by the increase in anthropogenic impacts which could alter the balance of the inflow and release of water. A metal assessment in water conducted by AED revealed that the quality of the water was exceptionally good for sites downstream of the Nyl floodplain. Unfortunately, the study by AED only included sites downstream of the Nyl floodplain and direct comparisons are therefore not possible. Sampling sites for the study were located in the upper reaches of the Nyl River system from Modimolle to Mokopane, Limpopo (Fig. 1). The sites selected include seven sites chosen for their positions relative to possible sources of pollution. These sites also correlate to those used in the WRC study conducted in 2006 for the purposes of comparison (Vlok et al. 2006). Site 1 or Klein Nyl Oog (KNO) (24° 42.967′S, 28° 14.542′E) was located approximately 2 km below the origin of the Klein Nyl River. The sampling was conducted just upstream of a weir on a cattle farm. At this point in the river, there are limited impacts and the site is expected to have natural levels of metals.

Page 3 of 17 427

Site 2 or Donkerpoort Dam (DPD) (24° 40.542′S, 28° 20.019′E), located just downstream of the Donkerpoort Dam on the Klein Nyl River. The DPD weir as well as guest houses, game lodges and minor agricultural activity are the possible impacts at this point. Site 3 or Golf Course (GC) (24° 41.697′S, 28° 25.074′E) was located downstream of the Koro Creek golf course. The river flows through the golf course and then into the town of Modimolle. The golf course could be altering nutrient and metal levels in the water through the addition of fertilisers to the golf course grass and vegetation. Site 4 or Sewage Treatment Works (STW) (24° 42.335′S, 28° 25.747′E) can be found directly next to the Modimolle sewage treatment facility effluent discharge point. During the sampling period, the STW was not functioning optimally with only certain parts of the system in working condition. The STW had a properly functioning solid waste removal system at the time of sampling; however, the flocculation procedures were not taking place and the partially treated sewage was released into the river. During high flow periods, the incoming sewage exceeded the capacity of the facility allowing raw sewage to spill directly into the Nyl River. Site 5 or Jasper (JAS) (24° 42.536′S, 28° 28.786′E) is around 2 km downstream of the STW. The Nyl River flows through a wetland before JAS, which could have a purifying effect on the water from the STW if the wetland is functioning properly. Site 6 or Nylsvley Nature Reserve (NYL) (24° 38.960′ S, 28° 41.445′E) is located in the Nylsvley Nature Reserve at the Jacana bird hide. This site is located in the floodplain section of the wetland. The substrate at this site consisted mainly of dense floating organic material. Site 7 or Moorddrift Dam (MOOR) (24° 15.175′S, 28° 58.521′E) is located approximately 50 km downstream of NYL. Samples were taken upstream of the weir. Water levels in the dam remained high throughout the wet and dry sampling periods. The dam is located on the Moorddrift Dairy Farm which no longer houses any cattle but acts as a game reserve and lodge. Water from the Moorddrift Dam is used to supplement water supply to neighbouring municipalities. Sampling Water sampling took place twice for the wet season and twice for the dry season. Dry season sampling was conducted from 27 to 28 July 2014 and 21–22 August

427

Page 4 of 17

Environ Monit Assess (2018) 190:427

Fig. 1 Map of the Nyl River, Limpopo Province, South Africa, with the sampling sites indicated. Sampling sites include the Klein Nyl Oog (KNO), Donkerpoort Dam (DPD), Golf Course (GC), Sewage Treatment Works (STW), Jasper (JAS), Nylsvley Nature

Reserve (NYL) and the Moorddrift Dam (MOOR) indicated by dots. Triangles indicate the towns of Modimolle and Mookgophong

2014, respectively. Wet season sampling took place on 27–29 February 2014 and 1–2 April, respectively. The artificial mussels were deployed at the first sampling dates for the wet and dry seasons and retrieved during the second sampling trips 1 month later. The water samples were taken a month apart for the high- and low-flow seasons as to coincide with the deployment

and retrieval of the AMs. One sample of 1 L each was taken for each sampling trip. In situ water quality parameters The physicochemical parameters of the water were determined during each sampling trip using a Eutech

Environ Monit Assess (2018) 190:427

multi-probe water quality meter. The parameters measured include pH, temperature (°C), oxygen saturation (% and mg/L) and electrical conductivity (μS/cm). The instrument consisted of three different probes that each measured a specific parameter. Each probe was calibrated beforehand according to the supplied manual (Greenfield 2004).

Water metal analysis For each of the sampling trips, water samples were collected in the field at each site using clean, acidwashed (HCl) polyethene bottles (Martin et al. 1994). Water samples were immediately frozen for transportation back to the University of Johannesburg Spectrum Analytical Facility for preparation and analysis. Water samples were filtered through 0.45-μmpore-size cellulose nitrate filter paper using a vacuum filtration system. Samples were acidified to 1% nitric acid using 65% Suprapur nitric acid (Martin et al. 1994). The prepared water samples were then analysed for Al, Cd, Cr, Co, Fe, Mn, Ni, Pb and Zn using inductively coupled plasma mass spectrometry (ICP-MS). The total hardness of the water was also determined by means of a Spectroquant Pharo 100 spectrophotometer using a standardised Merck test kit. The ICP-MS analysis was conducted using the PerkinElmer Nexion X-series ICP-MS (Fisher 2011). The ICP-MS analysis included calibration standards of 0.5, 1, 5, 10, 20, 40, 60 and 80 ppb. Furthermore, the ICP-MS analysis included an internal calibration verification standard (ICVS), continuing calibration verification standard (CCV), continuing calibration blank (CCB) and two internal standards, namely rhodium (Rh) and lutetium (Lu). Certified reference materials were added to the analysis for the purposes of quality assurance/quality control (QA/QC). Dogfish Liver Tissue (DOLT-4) (National Research Council Canada); Lake Sediment (LKSD-3) (CANMET) and Freshwater Sediment (FWSD) (ACLASS) were analysed with the samples (Table 1). An ICVS was also added to the analysis for QA/QC purposes. Metal concentrations were then determined by means of ICP-MS. The detection limits (LODs) of the analysis are indicated in Table 2. LODs were determined using the following equation:

Page 5 of 17 427

LOD ¼ ð3  Sy:xÞ=Slope

ð1Þ

The recoveries of the internal standards ranged from 97.5 to 114.4% for Rh and 98 to 113.6% for Lu. Artificial mussels AMs are semi-permeable passive sampling devices used for determining bioavailable metal levels in a water body over time. Figure 2 indicates the structure of a constructed AM as developed by Wu et al. (2007) for use in marine environments. The AM construction has been adapted for freshwater environments by using Milli-Q water as a replacement for artificial seawater. Artificial mussels consist of a 25-mm Perspex tubing containing 0.2 g Chelex100 resin beads and a 1-cm Perspex spacer suspended in Milli-Q water. These elements are enclosed between two semi-permeable polyacrylamide gel plugs. The plugs consist of a mix of acrylamide to N,N-methylene-bisacrylamide, with ammonium peroxidisulfate and N,N,N ′,N′-tetramethylethylenediamine (TMEDA). Once the AMs were constructed, they were placed in Milli-Q water until deployment into the aquatic environment (Hossain et al. 2015). Ten AMs were deployed at each site by fastening them to the sides of plastic baskets using cable ties. The baskets were tied together and then secured to a permanent structure to protect the AMs and keep the basket in place. The AMs were submerged in the water body and left for 4 weeks to allow the AMs to reach their saturation point (Hossain et al. 2015). Sampling was conducted for one wet season (February to April 2014) and one dry season (July to August 2014). The saturated AMs were collected from the water body and placed in buckets containing water from the site for transportation back to the laboratory. The AMs were removed from the buckets and the biofilm was removed as thoroughly as possible. The content of each AM was then removed, and the Chelex-100 beads were filtered from the rest of the contents using a sintered glass vacuum filtration system and 0.45-μm-pore-size cellulose nitrate filter paper. The Chelex-100 beads from each individual AM from each site were placed in glass conical flasks in 20 mL 6 M nitric acid. The conical flasks were sealed using parafilm to prevent evaporation and volatilisation of metals from the flasks. The flasks were placed on a ‘shaker’ for 24 h to aid the detachment of metals from the Chelex-100 beads (Hossain et al.

427

Environ Monit Assess (2018) 190:427

Page 6 of 17

Table 1 Recoveries of CRMs from ICP-MS analyses including Dogfish Liver Tissue (DOLT-4), Lake Sediment (LKSD-3) and Freshwater Sediment (FWSD) analysed for QA/QC purposes Metal DOLT-4

LKSD-3

FWSD

Obtained values

Certified values

Recovery (%)

Obtained values

Certified values

Recovery (%)

Cd

24.42

24.30

100.51

n/a

n/a

n/a

n/a

n/a

n/a

Cu

34.03

31.20

109.05

35.99

35.00

102.82

14.86

16.10

92.28

Fe

1697.94

1833.00

92.63

n/a

n/a

n/a

18,283.61

17,100

106.92

Mn

n/a

n/a

n/a

1507.28

1440.00

104.67

171.81

183.00

93.88

Ni

1.04

0.97

107.28

n/a

n/a

n/a

14.97

17.50

85.56

Zn

146.21

116.00

126.04

152.98

152.00

100.64

46.48

69.90

66.50

2015). After detachment, individual samples were once again filtered to separate the Chelex resin beads from the acid solution. The acid solution was diluted to 50 mL volumetrically and stored in 50-mL Falcon tubes in the refrigerator until analysis. The levels of metals were determined using ICP-MS following the same protocol as the water sample analysis. Statistical analysis Raw data was sorted and the respective method blanks were subtracted from each reading. The AM data was converted to μg/g units by using the following formula (Fisher 2011):

½Metal μg=g ¼ ½reading μg:L−1 −blank  ðDilution=dry massÞ=1000

ð2Þ

Obtained values

Certified values

Recovery (%)

Water data was expressed in μg/L after analysis, blanks were subtracted from the concentration given and no further changes were required. Once data was converted, descriptive statistics were determined using IBM SPSS v.22. Shapiro-Wilks tests for normality indicated that the data were not normal, and therefore, all data were normalised by log transformation and one-way ANOVA with Tukey’s post hoc tests was performed to determine significant differences (p < 0.05) between sites for AMs (van Emden 2008). A Student’s t test was performed to determine significant differences between water samples per site and season (van Emden 2008). A bivariate correlation with Spearman’s correlation test was performed to identify similar trends in metal concentrations between AMs and water samples (van Emden 2008). Principal component analyses (PCAs) were performed on the data to determine whether AMs and water from the same sites and seasons showed similar metal concentrations or similar trends (Anderson 2003).

Table 2 ICP-MS LODs for water and AM sample analysis Water LOD Water half LOD AM LOD AM half LOD Al

14.491

7.245

5.348

2.674

Cr

1.379

0.690

1.563

0.782

Fe

1.647

0.823

0.551

0.275

Mn

1.339

0.670

0.935

0.467

Co

1.136

0.568

1.166

0.583

Ni

0.993

0.497

1.347

0.674

Cu

1.246

0.623

1.824

0.912

Zn

2.082

1.041

2.472

1.236

Cd

0.782

0.391

0.939

0.470

Pb

2.930

1.465

2.622

1.311

LOD numbers are expressed in μg/L

Results In situ water quality parameters Water quality parameters were measured for all sampling trips except the first low-flow sampling trip (Table 3). The conductivity for KNO ranged from 27.2 to 51.9 μS/ cm, which was the lowest of all sites. STW had the highest conductivity ranging from 144.1 to 577 μS/cm. The conductivity of the water at sites near the origin of the river remained low throughout the study period with levels as low as 27.2 μS/cm. The pH levels recorded in

Environ Monit Assess (2018) 190:427

Page 7 of 17 427

Fig. 2 A constructed artificial mussel as designed by Wu et al. (2007) for marine metal pollution studies. The AMs used in this study have been adapted for freshwater use by replacing artificial seawater with Milli-Q water

the system ranged from 6.3 to 8.4 and were lower during the dry seasons compared to levels recorded in the wet season. The pH values were generally higher near the origin of the river decreasing at the downstream sites. When considering the percentage values of the oxygen saturation, the readings ranged from 21 to 150.4%. The target water quality range (TWQR) proposed by the Department of Water Affairs and Forestry (DWAF) is 80–120% saturated (DWAF 1996).

Metals The levels of metals detected in water samples and AMs are represented graphically in Fig. 3. Statistically

significant differences, identified by means of one-way ANOVA analysis, are indicated on the graphs where common superscripts denounce significant differences (p < 0.05). Spearman rank order indicated no significant correlations (p < 0.05) between AMs and water from the same sites and seasons. Water hardness tests revealed that the water from all sampling sites and seasons was soft (Table 4). The levels of Al in water samples (40.74– 115.59 μg/L) were highest at NYL for the dry season. Aluminium levels at GC were significantly higher than levels at DPD for the wet season of 2014. The levels were found to be significantly higher at NYL compared to MOOR in the dry season of 2014. Aluminium levels ranged from 51.29 to 359.93 μg/g in AMs and were

Table 3 In situ water quality parameters including conductivity, pH and oxygen saturation measured in the Nyl River system for 01 March and 02 April 2014 in the wet season and 27 July and 22 August 2014 in the dry season pH March 2014

Conductivity (μS/cm) July 2014

August 2014

March 2014

July 2014

Oxygen saturation (mg/L) August 2014

March 2014

July 2014

August 2014 11.8

KNO

8.4

7.5

8.3

51.9

27.2

31.6

8.0

12.7

DPD

n/a

8.3

7.3

n/a

56.6

59.8

7.9

11.9

9.5

GC

7.8

6.4

7.6

152.9

85.3

88.6

7.3

14.8

10.6

STW

7.9

6.4

7.0

144.1

551.0

577.0

5.8

8.6

9.0

JAS

8.2

6.3

6.8

141.2

175.0

224.0

8.3

15.8

7.5

NYL

8.3

6.3

7.6

108.8

140.4

169.0

7.3

13.6

9.2

MOOR

6.9

6.5

7.0

167.4

206.0

221.0

1.9

9.5

7.6

427

Page 8 of 17

lower for the wet season, remaining relatively constant as one moves downstream for the wet season but showing a clear increasing trend towards the STW, decreasing significantly from STW to JAS. Sewage Treatment Works had the highest levels of Al and GC had the lowest. The levels of Al at STW were significantly higher than those at KNO and JAS for the dry season. The AMs accumulated relatively similar levels of Al compared to those detected in the spot water tests for the wet season but showed a clear difference in the dry season between the water and AMs with the AMs accumulating higher levels of Al than those detected in spot water tests. The AMs showed significantly higher levels at the STW for the 2014 dry season, indicating that the inefficiency of the facility was affecting the water quality over the entire accumulation period. At sites where AM accumulation data was available for both seasons, levels of Al were higher for the dry seasons for each instance. This result is expected because of the dilution factor during the wet season caused by major flooding and heavy rainfall during the sampling period (Gouws and Du Toit 2014). Cadmium concentrations in water samples (0– 0.76 μg/L) were generally below detection in the dry seasons except for MOOR. Levels of Cd were also below detection for GC and STW for the 2014 wet season. No significant differences were found between sites and seasons for this metal. The levels of Cd accumulated by the AMs were all below the detection limits of the ICP-MS. The levels of Cr in water ranged from 1.16 to 9.52 μg/L and showed an increasing trend from the origin of the river moving downstream for the dry season of 2014. Significant differences were identified between the wet and dry seasons for STW. Sewage Treatment Works was also found to have significantly higher concentrations than DPD, JAS and MOOR for the dry season of 2014. The AMs accumulated Cr in the range of 0.24–4.61 μg/g. The levels of Cr for KNO stayed constant for the wet and dry seasons for AMs, which were also the lowest levels in the system. The highest AM accumulated levels were detected at STW in the dry season. Levels of Cr for DPD and JAS were higher for the dry season compared to the wet season. No significant differences were found for Cr levels in AM accumulation. Cobalt levels (0–0.71 μg/L) were below detection for many sites. The highest concentrations were detected at NYL for the wet season. The levels of Co were below

Environ Monit Assess (2018) 190:427 Fig. 3 Metal concentrations from water (W) and AM (AM)„ sample ICP-MS analysis for Al, Cd, Cr, Co, Cu, Fe, Mn, Ni, Pb and Zn, from seven sites along the Nyl River in Limpopo, South Africa, expressed in μg/L for water and μg/g for the AMs. Sampling periods were from 27 February 2014 to 02 April 2014 for the wet season (HF) and 27 July 2014 to 22 August 2014 for the dry season (LF). The wet season or high flow season is denoted by HF, and the low flow season or dry season is denoted by LF. Water samples are denoted by W and artificial mussels by AM; therefore, HFW means high flow water, LFAM means low flow artificial mussel, etc. Common superscripts denote significant differences (p < 0.05)

detection for all sites except STW and GC in the dry season of 2014. Considering Co concentrations, only KNO and DPD for the wet seasons had levels above detection. Levels ranged from 5.140 to 1.363 μg/g in AMs. No significant differences were detected. The accumulation pattern of Co by the AMs supports this theory as Co levels for GC, STW and JAS were below detection for the wet and dry seasons of 2014. The AMs did accumulate Co from KNO and DPD that are located in close proximity to the origin of the river. Copper levels in water samples had a range of 2.10– 11.01 μg/L with DPD from the dry season having the lowest concentration and GC from the 2014 wet season having the highest concentration. Student’s t tests determined that DPD was significantly higher than JAS of the 2014 wet season. It was also determined that STW was significantly higher than DPD and significantly lower than JAS from the 2014 dry season. Copper concentrations accumulated by the AMs ranged from 1.785 to 12.009 μg/g with KNO having the lowest levels in the wet season and the highest levels in the dry season. KNO and DPD had higher levels of Cu in the dry seasons. Jasper had similar levels during the wet and dry seasons. No significant differences were detected. For the wet season, all sites upstream of STW had much higher levels in the spot water tests than the levels accumulated in by the AMs, which is indicative of an isolated pollution event around the time of sampling. Iron concentrations in water (34.88–915.65 μg/L) were higher than most other metals in this study due to its high natural occurrence in aquatic environments. The lowest concentrations were found at KNO which is located close to the origin of the river. Sewage Treatment Works was discovered to contain significantly higher levels of Fe than KNO and DPD for the wet season. The levels of Fe in AMs ranged from 7.74 to 8573.46 μg/g with KNO having the highest concentrations during the wet season and JAS having the lowest

Environ Monit Assess (2018) 190:427

Page 9 of 17 427

427

Environ Monit Assess (2018) 190:427

Page 10 of 17

Table 4 Total hardness of water samples from the Nyl River system, Limpopo, South Africa Sites

Season

CaCO3 (mg/L)

Hardness of water

KNO

HF

14.01

Soft

LF

9.51

Soft

HF

12.01

Soft

LF

20.02

Soft

GC

HF

27.52

Soft

LF

23.52

Soft

STW

HF

42.54

Soft

LF

62.05

Soft

HF

24.52

Soft

LF

42.53

Soft

HF

33.53

Soft

LF

29.03

Soft

DPD

JAS NYL MOOR

HF

41.54

Soft

LF

51.04

Soft

Measurements are given in mg/L CaCO3. Water hardness is determined according to the DWAF guidelines for aquatic ecosystems (DWAF 1996): < 60 mg/L is soft, 60–119 mg/L is medium, 120–180 mg/L is hard and > 180 mg/L is very hard

concentrations during the dry season. Iron concentrations in the wet season showed a decreasing trend moving downstream, but this trend is not carried through in the dry season. Levels of Fe from KNO for the wet season were significantly higher than all other sites and seasons. The concentrations of Mn in water (0.10– 16.70 μg/L) were lowest in the 2014 dry season for JAS. The highest levels were detected at STW in the dry season. Manganese levels from KNO for the 2014 dry season were found to be significantly higher than MOOR, NYL, JAS and GC from that season. Klein Nyl Oog in wet season 2014 was significantly lower than KNO dry season 2014. Moorddrift Dam had significantly higher concentrations of Mn than KNO for the 2014 wet season. Golf Course differed significantly between the wet and dry seasons of 2014 and also had significantly lower levels of Mn than MOOR for the 2014 wet season. Though not statistically significant, the levels of Mn found at STW for the dry season was notably higher than other sites and seasons. Manganese concentrations in AMs ranged from 2.21 to 107.74 μg/g with KNO having the highest concentration during the wet season and JAS having the lowest levels in the dry season. The wet season shows a decreasing trend moving downstream. A significant difference was found between the wet and dry seasons for

KNO, DPD and JAS with all three indicating higher levels in the wet season. Considering the wet season, KNO was significantly higher than DPD, GC and JAS. Donkerpoort Dam was significantly higher than JAS in the wet season. Considering the dry season, KNO and DPD were significantly higher than JAS. The levels of Mn accumulated by the AMs were much higher than those detected in water samples. Nickel levels in water had a range of 0.48– 10.25 μg/L with DPD for the 2014 dry season having the lowest levels and NYL for the 2014 wet season having the highest levels. The levels of NYL for 2014 dry season were discovered to be significantly higher than KNO, DPD, GC, STW, JAS and MOOR for that season. Nickel levels for STW were significantly higher than DPD for the 2014 dry season. Though not statistically significant, the levels of Ni were increased for the 2014 wet season for NYL. The AMs accumulated Ni in a range of 0.27–1.64 μg/g. Klein Nyl Oog had the lowest levels in the dry season and the highest levels in the wet season. Significant differences were detected between the wet and dry seasons for KNO. It was also determined that STW had significantly higher levels of Ni than KNO for the dry season. The concentrations of Pb (0.50–11.27 μg/L) from NYL from the dry season contain the lowest levels and MOOR of that season contain the highest levels. The levels of Pb were generally lower than concentrations from previous sampling trips. Significant differences were detected between DPD and GC for the 2014 wet season. Donkerpoort Dam from this season was also significantly higher than NYL. Lead from NYL was however significantly higher than GC for the wet season of 2014. Lead concentrations in the AMs ranged from 1.06 to 2.94 μg/g with the lowest levels found at GC during the wet season and the highest found at STW in the dry season. The levels for the wet season stay relatively constant moving downstream, whereas the dry season shows an increasing trend to STW and a decrease at JAS. Aluminium, Cr, Ni and Zn also follow this spatial trend for AMs in the dry season. No significant differences were found for Pb. Zinc levels had a range of 14.62–89.67 μg/L in water samples. Donkerpoort Dam in the 2014 dry season had the lowest levels of Zn, whereas KNO in the dry season had the highest levels of Zn. It was determined that KNO and STW had significantly lower levels of Zn than DPD for the 2014 wet season. The levels of Zn for DPD also differed significantly between seasons.

Environ Monit Assess (2018) 190:427

Page 11 of 17 427

1.0

The golf course had significantly higher concentrations of Zn than DPD for the 2014 dry season. The AMs accumulated levels of Zn ranging from 24.65 to 55.23 μg/g. The lowest levels were from JAS in the dry season, and the highest were also from JAS for the wet season which was also significantly different from one another. Considering the dry season, STW was found to be significantly higher than JAS. All other levels are relatively constant throughout the system. All sites and seasons contained concentrations of Zn that exceeded the TWQR. Zinc levels were lower for KNO, JAS and NYL than levels found in 2001/2002 by Vlok et al. (2006). The levels of Zn were higher for DPD, STW and MOOR compared to levels detected by Vlok et al. (2006). The accumulation of Zn by the AMs was generally lower than levels found in the water. There was less variation in Zn on a spatial scale in the AM accumulation, indicating that the levels are naturally high in the system, but that there was an input of Zn from other sources at the time of sampling seeing that levels in spot water samples were higher and that it is not corroborated in the AM accumulation. A PCA conducted on the spot water data (Fig. 4) indicates a grouping of sites upstream of the STW Mn STW2

Pb KNO2

MDD2

(KNO1, KNO2, DPD1 and DPD2) that are positively correlated to Zn and Cd and negatively correlated to Cr and Ni. These sites also show little to no correlation to Pb, Mn, Co and Cu. NYL1 showed an opposite metal influence profile to the reference and upstream sites KNO and DPD and was therefore positively correlated to Cr and Ni, and negatively correlated to Cd and Zn. STW2 was closely correlated to Mn and Pb, with both metals acting as strong drivers. A PCA conducted on the metal data from spot water tests shows that there is almost no correlation between Mn and Fe, where there is clearly a pattern of similarity for these two metals accumulated in the AMs. The same is reflected in the correlation between Pb and Al, where there is a negative correlation in the water samples and a very strong positive correlation in the AMs. NYL2 clearly separates due to strong drivers of Al and Ni, indicating high levels of these metals at this site. It is also clear that NYL is influenced by a similar metal profile for both seasons though there are differences in correlation between to two seasons for the different metals. The statistics related to the PCAs can be found in Table 5. A PCA conducted on the AM accumulation data can be seen in Fig. 5. It shows a clear grouping of metals into two groups with Al, Pb, Cu and Cr grouping together and Ni, Fe, Co, Zn and Mn grouping together. These metal groupings have almost no correlation, to slightly negative correlations to each other. KNO2 and DPD2 have grouped together with both sites having close correlations to this grouping of metals. STW2 shares Table 5 Eigenvalues, origin scores and total variance for PCAs

DPD1

Co

Axis Water PCA

MDD1

Zn

Water and AM PCA

Cd

Eigenvalues

DPD2

JAS1

Cr

KNO1

Cu

Ni

GC2

Fe

NYL1

STW1 JAS2

Al

Explained variation (cumulative)

GC1

-0.8

NYL2

-0.6

AM PCA

1.0

Fig. 4 A principal component analysis (PCA) showing the grouping of water samples for seven sites along the Nyl River system, Limpopo. Sampling periods were from 27 February to 02 April 2014 for the wet season (1) and 27 July to 22 August 2014 for the dry season (2)

Origin scores

Total variation (%)

1

0.39

0.70

0.58

2

0.21

0.15

0.18

3

0.17

0.09

0.10

4

0.10

0.04

0.06

1

38.51

70.27

58.30

2

59.67

85.47

76.18

3

76.21

94.54

86.32

4

86.22

98.6

91.86

1

− 3.82

− 2.51 − 2.59

2

− 0.70

− 7.07 − 0.05

3

0.62

− 2.95 − 8.44

4

8.31 42.23

2.00 − 4.89 39.22

79.00

427

Environ Monit Assess (2018) 190:427

0.8

1.0

Page 12 of 17 Pb Al

JAS1W

STW2

Cu

GC1W

Fe

Cr DPD2W

Cr DPD2 KNO2

Zn

Cd

KNO1W

DPD1W

Cu

STW2A STW2W

Pb Al

GC1A KNO2A

Zn

-0.6

DPD1 JAS1

DPD1A

-0.6

GC1

-0.6

KNO2W

Co

DPD2A

JAS2

-0.6

JAS2A

JAS1A

KNO1

Co Mn

KNO1A

JAS2W

Ni Fe

Ni

1.0

Fig. 5 A principle component analysis (PCA) showing the grouping of AM samples for five sites along the Nyl River system, Limpopo. Sampling periods were from 27 February to 02 April 2014 for the wet season (grey) (1) and 27 July to 22 August 2014 for the dry season (black) (2)

Mn

1.0

Fig. 6 A principle component analysis (PCA) showing the grouping of water and AM samples for seven sites along the Nyl River system, Limpopo. Sampling periods were from 27 February to 02 April 2014 for the wet season (grey) and 27 July to 22 August 2014 for the dry season (black). Dry season AMs are labelled 1A, wet season AMs 2A, dry season water 1W and wet season water 2W. Water is indicated with stars and AMs with circles

Hierarchical cluster analysis this metal profile but has separated out based on strong Al and Pb drivers. DPD1 shows a strong negative correlation to DPD2 and a low correlation to the other metal grouping. KNO1 is correlated strongly and separates completely from all other sites based on a strong correlation with Fe and Co. KNO1 also has a slightly negative correlation to KNO2 and STW2. JAS1 and JAS2 grouped together and had very low correlation to the Al, Pb, Cu and Cr metal profile and a strong negative correlation with the Ni, Fe, Co, Mn and Zn profile. The PCA with combined water and AM metal data can be seen in Fig. 6. There is a strong negative correlation between the water and AMs for the wet season with KNO1A being closely correlated with Fe and Co as shown in Fig. 5 and KNO1W having a strong correlation to Cr and Cd. Grouped with KNO1W are the water and AMs for JAS2, as well as the water for DPD2. These sites share a very low correlation to Ni, Pb and Al, and strong negative correlations with Fe, Co and Mn. The water and AMs from STW2 grouped together with water being more strongly driven by Mn. There was a strong negative correlation between the water and AMs from JAS1 with the water being driven strongly by Ni and the AMs being closely correlated to Pb. All the AMs generally grouped together except for KNO1A and JAS2A.

The dendrogram indicated that in most cases, water and AM samples for the same sites and sampling seasons did not cluster together. The Euclidian distances are great between water and AMs from the same sites and seasons for most of the sampling sites. The greatest Euclidian distance was found between KNO water and AMs from the wet season.

Discussion The water quality data in Table 3 shows a clear ascending trend in the conductivity for each of the sites except for KNO which stayed relatively constant throughout the sampling period. Moving downstream, conductivity gets progressively higher and reaches a peak at the STW that was releasing partially treated sewage intermittently during the study period. This could be due to the influx of untreated sewage from the STW (Daniel et al. 2002). The pH of water is an indirect measure of the concentration of free hydrogen ions in the water. The pH of deionised water at room temperature is completely neutral with a pH of 7 (DWAF 1996). Though pH levels fluctuate from approximately 6 to approximately 8 in

Environ Monit Assess (2018) 190:427

this study, it still falls within the range of historical data for this system dating from the 1980s (Greenfield 2004). Oxygen saturation varied notably, possibly due to the presence of high volumes of aquatic macrophytes and algae. The variation could also be explained by the differences in the times of sampling between sites and seasons. Considering the wet season, STW and MOOR had levels below the TWQR (DWAF 1996). The fact that some sites had levels both above and below the TWQR could indicate that hyperoxic and hypoxic conditions were occurring at these sites, respectively; however, due to the variation in the sampling times and water temperature, this cannot be confirmed. Aluminium is commonly found in high concentrations in aquatic ecosystems as it forms part of the natural makeup of sediment. It can, however, be introduced by human activity such as fossil fuel combustion and multiple industries in the area (Kempster et al. 1980). Aluminium is highly reactive to changes in pH; a lowering of the pH of water in an aquatic ecosystem can lead to an increase in the concentration of dissolved Al (Munk and Faure 2004). A decrease in pH also leads to changes in the form of Al available in the water from harmless Al ions to hexahydrate species that are highly toxic to aquatic biota (DWAF 1996). Under alkaline conditions, Al normally occurs in biologically unavailable forms. The effect of the pH on the levels of Al is reflected in the results as Al levels were increased and pH levels reduced over time in the AMs. Cadmium is a trace element that is non-essential to life. It does occur naturally in aquatic ecosystems but can be introduced by various human activities such as mining, manufacturing processes; agriculture especially fertiliser and pesticides; and the breakdown of cadmium-plated containers (DWAF 1996). The results of this study showed that Cd was generally below the limit of detection for most of the sites and seasons, which is reflected in the water as well as the AMs. This is also reflected in historical data in a previous WRC study (Vlok et al. 2006). The levels of Cd detected at the reference site were notably lower than levels found by Vlok et al. (2006) and AED. Chromium is a metal that is not commonly found in high concentrations in aquatic ecosystems (Greenfield 2004). Some forms of Cr are highly toxic and soluble regardless of the pH of the water (DWAF 1996). The levels of Cr in water samples were below the TWQR of 7 μg/L for all sites and seasons except

Page 13 of 17 427

MOOR in the dry season, showing that there may have been isolated Cr pollution events. Sewage Treatment Works in the wet season showed significantly higher concentrations than most of the other sites for that season. This could potentially be due to the inefficient functioning of the STW during that time, causing untreated urban runoff and sewage to enter the Nyl River. The AMs accumulated Cr in a similar range to levels detected in the water, but there were no significant differences detected for the AMs where there were some significant differences in the water samples. With regard to Cr, it is also evident that the AMs accumulated higher levels in the dry season compared to the wet season. Copper is a metal that occurs naturally in aquatic ecosystems in three different oxidation states. When this element occurs in the environment under certain conditions such as water containing high levels of CaCO3, it becomes highly toxic (Greenfield 2004). The TWQR for Cu under soft water conditions is 0.3 μg/L; therefore, the levels of Cu were well above the TWQR for all sites and seasons in 2014. High levels of Cu were detected close to the source of the river. Sources of Cu pollution include STW effluents, algicides and pesticides, corrosion of copper pipes, mining, smelting and metal refineries as well as iron and steel industries (DWAF 1996). The town of Modimolle receives water from the Roodeplaat Dam which has had water quality issues in the past (DWA 2011). The AMs accumulated Cu in levels notably lower than levels detected in water for the wet season. This could indicate that the influx of Cu at the time of water sampling was an isolated event. The increased levels of Cu found in the water could be due to agricultural activities in the upper region of the river and the Cu contained in their pesticides (DWAF 1996). The AMs did, however, accumulate levels of Cu closer to those found in the water for the 2014 dry season. Iron is a naturally occurring metal that is found in varying quantities in aquatic ecosystems (Vlok et al. 2006). The level of Fe found in aquatic systems is subject to the geology of the area in which the system resides. Though Fe occurs naturally and is naturally leached from iron ores, it can also be introduced into the system by various industries and can be found in household chemical products, fungicides and fertilisers. When compared to results reported by Greenfield (2004), it is evident that these levels are still within the range found by Greenfield

427

Page 14 of 17

(2004). These levels also fall within the 500–50,000-μg/L range for Fe in aquatic ecosystems determined by Galvin (1996). The AMs accumulated significantly higher levels of Fe for KNO for the 2014 wet season. These levels seem to decrease moving downstream from the source for AM accumulation. The increased Fe concentrations could be due to the high concentrations of Fe normally found in some fertilisers (DWAF 1996). The spatial trend of Fe accumulation coincides with the spatial trend found in Mn accumulation by the AMs for the same sampling period. Manganese is an essential element to most living organisms; however, in high concentrations, it can have negative effects on the environment (DWAF 1996). Manganese is mostly found in high concentrations in sediment and is used in many industrial processes. Anthropogenic sources of Mn that can lead to elevated Mn concentrations in the water include the steel, chemical and fertiliser industries (DWAF 1996). When considering the results of this study, it is important to note that Mn levels are closely associated with Fe concentrations and follow the same spatial trend. All sites and seasons showed levels below the TWQR of 180 μg/L (DWAF 1996). The AMs accumulated Mn in the same spatial pattern as Fe in the 2014 wet season, with significantly higher levels at the reference site, KNO. This could be due to fertilisers containing Mn, ending up in the stream via agricultural runoff as the lands surrounding this site are predominantly agricultural (DWAF 1996). Levels decreased moving downstream as they did with the Fe levels. There are also major differences in the accumulation of Mn between the wet and dry seasons. The dry season samples showed notably lower concentrations. Nickel is used in many industrial and commercial processes. This element is essential in sustaining life and when absent can have detrimental effects on the organisms inhabiting the environment (Cempel and Nikel 2005). The levels of Ni determined in this current study were similar to levels found by AED for the same area. The accumulation of Ni by the AMs had levels 10 times lower than those detected in the water samples, indicating that the levels in the water at the time of sampling were high possibly due to an isolated event. For KNO and JAS, the AMs accumulated higher levels of Ni in the wet season compared to the dry season. Lead exists in the aquatic environment in four oxidation states with divalent lead (Pb2+) the most toxic form, being bioavailable to aquatic species

Environ Monit Assess (2018) 190:427

(Fisher 2011). Lead is introduced into the aquatic environment through fossil fuel combustion, industrial processes and urban runoff. The bioavailability is known to decrease as pH decreases. The toxicity of Pb also decreases as water hardness increases, which is of concern as water total hardness tests for all sites and seasons revealed that the water was soft. It can increase in areas with high oxygen saturation, which is important as the water was supersaturated for some of the sites and seasons (Table 3). Most of the results of this study showed levels that exceed the TWQR for South Africa but were below the acute effect value (DWAF 1996). The AMs accumulated Pb in a narrower range than the water samples. It also showed notably lower levels of Pb at KNO for the wet season. The AMs have proven to accumulate significantly similar concentrations of Pb when compared to multiple bioindicator species and could, therefore, be an indication of the Pb levels also being isolated pollution events instead of constant pollution on a temporal scale (Kibria et al. 2010; Claassens et al. 2016). Zinc is an essential metal found in aquatic ecosystems in two forms, the metal and the divalent cation Zn2+ (DWAF 1996). In an aquatic ecosystem, it is the divalent cation form which is potentially toxic to biota. Zinc naturally occurs in rocks and can enter the environment by natural weathering, erosion and leaching from substrates (Fisher 2011). The levels of Zn were elevated for all sites for all sampling seasons. The water data was not supportive of the AM accumulated data and contained much higher levels of some metals than those accumulated in the AMs, as rivers are dynamic systems and the concentrations of metals are constantly changing. During the wet season sampling period, there were intense flooding events which could have caused an input of fertiliser-related metals into the system which could have attributed to the higher levels of these metals detected in the water samples (Gouws and Du Toit 2014). This information would not be available if only AMs were used in this study and the health of the system could be misinterpreted as all acute contamination would be disregarded. The HCA (Fig. 7) supports the argument that spot water tests should be analysed along with AMs to give a greater indication of variability in metal levels in dynamic freshwater systems, just as spot water samples should be taken when sampling bioindicators for assessment. The AMs serve to replace bioindicators in water quality

Environ Monit Assess (2018) 190:427

Page 15 of 17 427

Fig. 7 A hierarchical cluster dendogram indicating clustering of metals in water samples (W) and artificial mussels (A) for the wet season (1) and the dry season (2)

assessments and give an indication of long-term pollution levels in aquatic systems.

Conclusion When considering the results as a whole, the STW is not the largest contributor of metal pollution in the system as many of the metal levels were relatively high at sites upstream of the STW. It was speculated that the high levels of metals found at these sites could be due to fertiliser use in those predominantly agricultural land use areas. The AM concentrations showed that the metal levels were higher for the dry season than the wet season for Al, Cr, Cu and Pb. This can be attributed to the lack of dilution factor in the dry season. However, a decreasing trend is evident with AM accumulated levels of Co, Fe, Ni and

Mn possibly due to fertilisers washed into the river by heavy rainfall and flooding for the wet season and either decrease by the settling of these metals from the water body or because of dilution of the effluent moving downstream. The water data showed that many of the metals were present in concentrations that exceeded the TWQR (Cr, Cu, Pb and Zn), but that the levels were in some cases lower than historical data (Cr), or consistently high throughout the entire system (Cu and Zn). The differences in the concentrations found in water versus concentrations in AMs give more insight into the conditions in a water system. Acute increases in metal levels can have a host of negative effects on aquatic ecosystems, and whilst AMs are a beneficial alternative to using live mussels or snails in a system, it is recommended that they be used in conjunction with spot water tests when determining the health of an ecosystem.

427

Page 16 of 17

Acknowledgements We acknowledge Mrs. E. Kroukamp at the University of Johannesburg Analytical Facility (Spectrum) for the assistance with sample analysis and Mr. R.S. Musa for the fieldwork assistance. Funding information This study was financially supported by the University of Johannesburg and National Research Foundation of South Africa.

References Anderson, T. W. (2003). An introduction to multivariate statistical analysis (3rd ed.). New Jersey: Wiley. Boss, C. B., & Fredeen, K. J. (2004). Concepts, instrumentation and techniques in inductively coupled plasma optical emission spectrometry (3rd ed.). USA: Perkin Elmer. Cempel, M., & Nikel, G. (2005). Nickel: A review of its sources and environmental toxicology. Polish Journal of Environmental Studies, 15, 375–382. Claassens, L., Dahms, S., van Vuren, J. H. J., & Greenfield, R. (2016). Artificial mussels as indicators of metal pollution in freshwater systems: A field evaluation in the Koekemoer Spruit, South Africa. Ecological Indicators, 60, 940–946. Daniel, M. H. B., Montebelo, A. A., Bernardes, M. C., Ometto, J. P. H. B., Camargo, P. B. ., Krusche, A. V., Ballester, M. V., Victoria, R. L., & Martinelli, L. A. (2002). Effects of urban sewage on dissolved oxygen, dissolved inorganic and organic carbon, and electrical conductivity of small streams along a gradient of urbanization in the Piracicaba River basin. Water, Air, and Soil Pollution, 136, 189–206. Degger, N., Wepener, V., Richardson, B. J., & Wu, R. S. S. (2011). Application of artificial mussels (AMs) under south African marine conditions: A validation study. Marine Pollution Bulletin, 63, 108–118. Department of Water Affairs (DWA). (2011). National Assembly: Question 3416 for written reply. Internal Question Paper No. 36. Department of Water Affairs and Forestry (DWAF). (1996). South African water quality guidelines. Volume 7: Aquatic ecosystems. Echols, K. R., Meadows, J. C., & Orazio, C. E. (2009). Pollution of aquatic ecosystems II: Hydrocarbons, synthetic organics, radionucleotides, heavy metals, acids and thermal pollution. In G. E. Likens (Ed.), Encyclopaedia of Inland Waters (1st ed., pp. 120–128). Boston: Elsevier. Fisher, E. M. (2011). Metal bioaccumulation and biomarker responses in tigerfish, Hydrocynus vittatus, from three South African populations. Dissertation. University of Johannesburg. Galvin, R. M. (1996). Occurrence of metals in water: An overview. Water SA, 1, 7–18. Gerber, R. J. L., Smit, N. J., Van Vuren, J. H. J., et al. (2018). Biomarkers in tigerfish (Hydrocynus vittatus) as indicators of metal and organic pollution in ecologically sensitive subtropical rivers. Ecotoxicology and Environmental Safety, 157, 307–317. Gonzalez-Rey, M., Lau, T. C., Gomez, T., Maria, V. L., Bebianno, M. J., & Wu, R. (2011). Comparison of metal accumulation

Environ Monit Assess (2018) 190:427 between ‘Artificial Mussel’ and natural mussels (Mytilus galloprovincialis) in marine environments. Marine Pollution Bulletin, 63, 149–153. Gouws, C., Du Toit, G. J. (2014). Water level monitoring report for the flooding event at the proposed Volspruit opencast mining project, Mokopane, Limpopo. Escience associates. Report no. EAV-13-447. Greenfield, R. (2004). An assessment protocol for water quality integrity and management of the Nyl River wetland system. Thesis. University of Johannesburg. Greenfield, R., van Vuren, J. H. J., & Wepener, V. (2007). Determination of sediment quality in the Nyl River system, Limpopo Province, South Africa. Water SA, 33(5), 693–700. Greenfield, R., Van Vuren, J. H. J., & Wepener, V. (2012). Heavy metal concentrations in the water of the Nyl River system, South Africa. African Journal of Aquatic Science, 37(2), 219–224. Haskins, C., & Kruger, J. (1997). Information sheet for the site designated to the List of Wetlands of International importance especially as waterfowl habitat. Pietersburg: Chief Directorate Environmental Affairs. Hellawell, J. M. (1986). Biological indicators of freshwater pollution and environmental management. London: Elsevier Applied Science Publishers. Holt, E. A., & Miller, S. W. (2010). Bioindicators: Using organisms to measure environmental impacts. Nature Education Knowledge, 3(10), 8. Hossain, M. M., Kibria, G., Nugegoda, D., Lau, T. C., & Wu, R. (2015). A training manual for assessing pollution (trace/ metals) in rivers, estuaries and coastal water using innovative artificial mussel (AM) technology- Bangladesh model. https://doi.org/10.13140/RG.2.1.4384.4644. Available from: https://www.researchgate.net/publication/273775181_A_ Training_Manual_for_Assessing_Pollution_%28 traceheavy_metals%29_in_Rivers_Estuaries_and_Coastal_ waters-Using_Innovative_Artificial_Mussel_%28AM%29_ Technology_-_Bangladesh_Model. Kempster, P. L., Hattingh, W. A. J., & Van Vliet, H. R. (1980). Summarized water quality criteria. Department of Water Affairs and Environmental Conservation, Hydrol Research Instit. Report No. TR108. Kibria, G., Rose, G., Lau, T. C., Lung, Y. K., Chan, A. K. Y., & Wu, R. (2010). Monitoring heavy metals in GoulburnMurray waterways using passive sampling with artificial mussels (AM) –pilot study. Report No. 2806033. ISBN: 978-1-876356-19-4. Leung, M. Y., Furness, R. W., Svavarsson, J., Lau, T. C., & Wu, R. S. S. (2008). Field validation, in Scotland and Iceland, of the artificial mussel for monitoring trace metals in temperate seas. Marine Pollution Bulletin, 57, 790–800. Markert, B. A., Breure, A. M., & Zechmeister, H. G. (2003). Bioindicators & Biomonitors – Principles, concepts and applications. Trace metals and other contaminants in the environment, 6, 3–39. Martin, T. D., Creed, J. T., & Brockhoff, C. A. (1994). USEPA Method 200.2. Sample preparation procedure for spectrochemical determination of total recoverable elements. Revision 2.8. McCarthy, T. S., Tooth, S., Jacobs, Z., et al. (2011). The origin and development of the Nyl River floodplain wetland, Limpopo

Environ Monit Assess (2018) 190:427 Province, South Africa: Trunk—Tributary river interactions in a dryland setting. South African Geographical Journal, 1, 1–19. Munk, L., & Faure, G. (2004). Effects of pH fluctuations on potentially toxic metals in the water and sediment of the Dillon reservoir, Summit County, Colorado. Applied Geochemistry, 19, 1065–1074. Pheiffer, W., Pieters, R., Van Dyk, J. C., & Smit, N. J. (2014). Metal contamination of sediments and fish from the Vaal River, South Africa. African Journal of Aquatic Science, 39(1), 117–121. Phillips, D. J. H. (1977). The use of biological indicator organisms to monitor trace metal pollution in marine and estuarine environments- a review. Environmental Pollution, 13, 281–317. Prosi, E. (1979). Heavy metals in aquatic organisms. Metal pollution in the aquatic environment (pp. 271–323). Berlin: Springer.

Page 17 of 17 427 Sanders, M. J. (1997). A field evaluation of the freshwater river crab, Potamonautes warren, as a bioaccumulation indicator of metal pollution. Dissertation. Rand Afrikaans University. Van Emden, H. F. (2008). Statistics for terrified biologists. Victoria: Blackwell Publishing. Vlok, W., Cook, C. L., Greenfield, R., et al. (2006). Biophysical framework for the sustainable management of wetlands in the Limpopo Province with Nylsvley as a reference model. WRC Report No: 1258/1/06. ISBN No: 1–77005–462-6. Wu, R. S. S., Lau, T. C., Fung, W. K. M., Ko, P. H., & Leung, K. M. Y. (2007). An artificial mussel for monitoring heavy metals in a marine environment. Environmental Pollution, 145, 104–110.