Water quality in two Icelandic rivers: the influence of impoundment, agriculture, glaciation and permafrost

The physical and chemical quality of water within river systems is important as it directly influences their ability to be a resource for both ecological and anthropological activities (Neal et al. 2006). Recognition that water quality affects both public health and aquatic life has led to the establishment of legislation to improve and protect water quality. For example, under the European Union (EU) Water Framework Directive (WFD) all surface waters in Europe must achieve a ‘good’ ecological status by 2015, with a key contributing factor being water quality (Donohue et al. 2006).

River water quality is composed of a complex combination of macro- and micro-scale chemical, physical and biological factors interacting over space and time (Brainwood et al. 2004). Icelandic rivers are no exception and are controlled by a number of natural and anthropogenic factors interacting spatially and temporally. Amongst the many factors that are known to influence water quality, the following occur within Icelandic river catchments: the impoundment of rivers for hydro-electric power generation (Thórhallsdóttir 2007); agricultural land use (Gudleifsson 2002); glaciation (Louvat et al. 2010) and permafrost (Etzelmüller et al. 2007).

Over 40% of global rivers are now impounded by dams for human activities such as power generation, flood prevention and irrigation for agricultural land (Vörösmarty et al. 2003). Hydropower exploitation in Iceland is currently at around 10% of what is technically feasible, with approximately 50% of the technically feasible being economically viable (Sæþórsdóttir & Ólafsson 2010). Reservoirs are thought to influence several aspects of downstream water quality including: temperature, heavy metal concentration, dissolved oxygen concentration and suspended sediment (Ahearn et al. 2005). Suspended sediment levels are expected to decrease downstream of reservoirs because the flow velocity within reservoirs is extremely low, causing sediment that was carried in the flow to be deposited. This process is often referred to as sediment trapping (Lu & Siew 2006).

Agriculture is typically considered to be an important factor determining water quality in rural river catchments (Bengraïne & Marhaba 2003). This is because nitrate and phosphate fertilisers are used in agriculture to increase vegetation growth and when they enter river systems in high quantities they can cause eutrophication (Arheimer & Lidén 2000). Eutrophication is the process by which elevated nitrate and phosphate levels initiate the growth of algae whose respiration reduces the dissolved oxygen available to other aquatic flora and fauna (Prat & Munné 2000). Agriculture in Iceland is primarily permanent grasses that are fertilised annually and cut for hay or sometimes grazed (Gudleifsson 2002).

Glacial rivers are associated with dark and turbid waters carrying high quantities of suspended sediment. The influence that glaciers have over river suspended sediment loads varies with glacier size, glacial activity and the nature of the glacial drainage system (Hasnain 1996). For example, it has been shown that river catchments with higher glacial coverage experience greater levels of mechanical erosion rates, resulting in increased sediment available for transport (Eiriksdottir et al. 2008). The degree of glacial melting is also a key control on the levels of dissolved constituents within river flows. Dissolved solid concentrations are typically found to be at their highest concentrations during the winter and at their lowest during the summer. This is because the increased flows from melt-water during summer period leads to dilution of the dissolved solid concentrations (Eiriksdottir et al. 2008). Icelandic river catchments are often influenced by glaciers as they occupy 12% of the total surface area of Iceland, the biggest of which is Vatnajökull (Louvat et al. 2010).

In periglacial areas, permafrost soils can also influence water quality as they contain high insoluble element concentrations and organo-mineral colloids. These organo-mineral colloids aid in the mobilisation of insoluble elements such as aluminium and iron during the spring and summer floods. As a result, higher concentrations of aluminium and iron are typically found in river catchments with significant contributions from land covered by permafrost (Bagard et al. 2011). In the Icelandic highlands, several areas of sporadic permafrost (palsa) have been recorded (Etzelmüller et al. 2007).

The influence of these factors on the spatial and temporal variability of water quality within Icelandic river systems has, to-date, been reported on by a limited amount of academic literature (Eiriksdottir et al. 2008; Louvat et al. 2010; Oskarsdottir et al. 2011). This research article partially addresses this issue by presenting the results from an investigation exploring the spatial variations in water quality of two northern Icelandic Rivers; the Blanda and the Skjálfandafljót.

Iceland is a volcanic island located below the Arctic Circle on the North Atlantic mid-oceanic ridge. Predominately the coastal area is characterised by deep fjords, extending into large inland valleys. The climate is described as oceanic boreal, whereby it has cool summers and mild winters considering its location (Louvat et al. 2010). The mean temperature varies from −5 to 0 °C in the winter and from 5 to 10 °C in the summer. Precipitation levels vary from 400 to 4,000 mm/year, with the highest levels occurring at the highest altitudes and in the south (Louvat et al. 2010). The Blanda and the Skjálfandafljót (Figure 1) were the two rivers selected as the focus of this study as they are influenced by the four factors of interest to this investigation (impoundment, agricultural land cover, glacial land cover and permafrost). A summary of the key characteristics of each river is shown in Table 1.

The River Blanda flows from its source, the Hofsjökull Glacier, north into Húnaflói Bay at Blönduós. Along its course there are several reservoirs that were built in 1989 to provide water to the local hydroelectric power plant. These reservoirs lie between 400–600 m above sea level (asl) on the edge of the central highland plateau (Figure 2). For the majority of the year the flow passes through these reservoirs before being released back into the river. During the summer months (May–October) the reservoirs fill up and, once full, an overflow starts releasing water downstream. The overflow is located at 478 m asl and once the Blöndulón reservoir is full it covers 57 km² and holds 400 million m³ of water (Vilmundardóttir et al. 2010). Downstream of the reservoirs, an agricultural valley has developed which widens towards Húnaflói Bay. In the highlands sporadic permafrost is present with high ground ice content and thin overburden and exposed bedrock (Brown et al. 2002; Etzelmüller et al. 2007).

The River Skjálfandafljót flows north from the Vatnajökull glacier through the gravel flats and lava fields of Sprengisandur on the highland plateau, and cuts down with the Kvíahraun lava flow into the gorges of Fljótsdalur, Krokdalur and agricultural valley of Bárðardalur towards Skjálfandi Bay. The Skjálfandafljót is one of only three major glacial rivers in Iceland yet to be harnessed for hydropower production. No permafrost has been recorded in the catchment of the Skjálfandafljót (Etzelmüller et al. 2007).

The five main research questions that this project has attempted to answer are:

1. How does the water quality of the Blanda and the Skjálfandafljót vary longitudinally?

2. How does river impoundment impact upon their water quality?

3. How does agricultural land use impact upon their water quality?

4. How does the presence of glaciated areas impact upon their water quality?

5. How does the presence of permafrost impact upon their water quality?

Water quality sampling was undertaken over 4 weeks during the summer glacial flood period from 5th August to 2nd September 2011. Sampling on the Blanda took 8 days (5–13th August), whereas on the Skjálfandafljót sampling required 14 days to complete (18th August to 2nd September). The sample sites were spaced at approximately 10 km intervals, with additional samples taken at points that would provide information on the influence of the Blöndulón reservoir overflow and a spring-fed tributary. In total, 32 sample sites were established along the lengths of the Blanda (Figure 3) and the Skjálfandafljót (Figure 4). At each site three replicate measurements were taken and an average was calculated. Samples were collected at a distance of 2 m from the bank and at a depth of 30 cm in the water column (following the method of Bordalo et al. 2001). In Table 2 the water quality parameters recorded as a part of this study are listed along with their international guidelines. Due to the remote location and absence of a laboratory the methods used in this study are entirely field based.

Sample sites on the Blanda were split into four categories based upon how they were affected by the presence of the reservoir: those downstream of the impoundment that were sampled before the overflow was running (sample sites 15–12), those downstream of the impoundment that were sampled whilst the overflow was running (sample sites 11–9), those within the Blöndulón reservoir (sample sites 8–5), and those upstream (sample sites 4–1).

Agricultural land use was recorded as present or absent at each site using an adapted version of the UK Environment Agency's River Habitat Survey (EA 2003). Agricultural land was present downstream from sample site 11 on the Blanda and from sample site 11 on the Skjálfandafljót.

Global Positioning System (GPS) co-ordinates of each sample location, glacial extents from the CAMPGIC (Circum-Arctic Map of Permafrost and Ground-Ice Conditions) dataset (Brown et al. 2002), a 30 m resolution ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) DEM (Digital Elevation Model), and the hydrological toolset in ESRI's ArcGIS 10 were used to calculate the proportion of each sample site's catchment that was glaciated. These calculations may be slight overestimates of glacial extent due to recent glacial retreat in Iceland.

Spatial analysis using ArcGIS 10 and the permafrost extent described by the CAMPGIC dataset allowed for the presence or absence of permafrost within each sample site's catchment to be determined.

Once the data were collected, they were analysed using both visual and statistical methods. To establish if any longitudinal relationships existed in the water quality parameters measured in each river catchment, scatter plots of all the data were created and a Pearson's correlation analysis was performed. If any statistically significant relationships were observed then a regression analysis was performed in order to establish the nature of that relationship. This type of analysis was also performed to look for relationships between water quality parameters and the proportion of the catchment area that was glaciated. In order to analyse the influence of the presence of impoundment, agricultural land use and permafrost, comparative analysis was carried out using box plots and two-sample unequal variance t-tests.

Figure 5(a) shows that on the Blanda turbidity is at its highest (306.3 FTU) close to the source, it then decreases until levelling out within the Blöndulón reservoir. Pearson's correlation coefficient returned a value of –0.876 and a p-value of 0.001 indicating a statistically significant negative correlation between turbidity and distance downstream on the Blanda. A non-linear exponential regression equation (y = 581.43e^–4.863x) is the best fit (R² = 0.8791) for a negative correlation between turbidity and distance downstream. A sharp increase (53.3–206.0 FTU) is observed after the reservoir, while the reservoir overflow is operational, before decreasing and levelling out once more towards the mouth of the river, reaching its minimum (4.88 FTU). This results in a range of 301.4 FTU for turbidity along the Blanda.

Figure 5(b) shows a general decline in turbidity with distance downstream on the Skjálfandafljót. Pearson's correlation coefficient returned a value of −0.659 and a p-value of 0.006 indicating a statistically significant (99.4% confidence) negative correlation between turbidity and distance downstream. A non-linear polynomial regression equation (y = 25.466 × ^2–42.828x + 25.511) is the best fit (R² = 0.5213) for a negative correlation between turbidity and distance downstream. The range between the highest and lowest recorded turbidity is 18.8 FTU, with the highest value (21.1 FTU) recorded 14% downstream from the source and the lowest (2.3 FTU) at 64%. A sharp decline in turbidity between 42% and 64% is responsible for this range.

On the Blanda in Figure 5(c), despite some fluctuations, pH increases with distance downstream. Pearson's correlation coefficient returned a value of 0.572 and a p-value of 0.026 indicating a statistically significant positive correlation between pH and distance downstream on the Blanda. A linear regression equation (y = 0.7645x + 6.4584) is the best fit (R² = 0.3273) for a weak positive correlation between pH and distance downstream on the Blanda.

Figure 5(d) shows fluctuations in pH with distance downstream on the Skjálfandafljót, however, no clear relationship is visible. The Pearson correlation coefficient returned a value of −0.070 and a p-value of 0.796 indicating no statistically significant correlation between pH and distance downstream. The most notable fluctuation is a drop in pH from its highest value (8.19) at 72% to its lowest (5.98) at 77%.

No clear relationships were observed for Zn on the Blanda or the Skjálfandafljót (Figure 5(a) and (f)). Pearson's correlation coefficient returned a value of 0.044 and a p-value of 0.875 indicating no correlation between Zn and distance downstream on the Blanda. On the Skjálfandafljót, the Pearson correlation coefficient returned a value of −0.357 and a p-value of 0.175 indicating no statistically significant correlation between Zn and distance downstream.

Figure 5(g) indicates that on the Blanda the overall trend is for a decrease in Al concentrations with distance downstream. Pearson's correlation coefficient returned a value of −0.849 and a p-value of 0.001 indicating a statistically significant, negative correlation between Al and distance downstream on the Blanda. A linear regression equation (y = −0.5514x + 0.5656) is the best fit (R² = 0.721) for a strong negative correlation between Al concentration and distance downstream on the Blanda. Al concentrations remain consistently high (0.5 mg/L) until 35% downstream. The concentration then declines to 0.14 mg/L by 49% downstream, a sharp increase (0.14–0.45 mg/L) is observed for the points sampled after the reservoir while the reservoir overflow is operational. The points downstream of the reservoir that were sampled before the reservoir overflow was operating have extremely low Al concentrations, reaching a minimum value of 0.01 mg/L at 94% downstream.

On the Skjálfandafljót (Figure 5(h)), there are no apparent downstream trends and Pearson's correlation coefficient returned a value of −0.385 and a p-value of 0.141 indicating a statistically insignificant weak negative correlation between Al and distance downstream.

On the Blanda (Figure 5(i)), the TDS increases until reaching the Blöndulón reservoir (33%), at which point TDS decreases. Downstream of the reservoir (both during and before the reservoir overflow was operating) TDS increased with distance downstream. Pearson's correlation coefficient returned a value of 0.561 and a p-value of 0.030 indicating a statistically significant positive correlation between TDS and distance downstream on the Blanda. For the Blanda, a power regression equation (y = 0.0385×^0.122) is the best fit (R² = 0.3655) for a weak positive correlation between TDS and distance downstream.

No clear downstream trend exists between TDS and distance downstream on the Skjálfandafljót (Figure 5(j)), overall a slight decrease in TDS from 0.0501 to 0.0421 g/L was recorded. Pearson's correlation coefficient returned a value of 0.070 and a p-value of 0.797 indicating no statistically significant correlation between TDS and distance downstream. An increase in TDS is observed between 61% and 64% downstream, where a major direct runoff and spring-fed tributary enters the main stem.

No clear relationships were observed for either nitrates on the Blanda or the Skjálfandafljót (Figure 5(k) and (l)). Pearson's correlation coefficient returned a value of −0.003 and a p-value of 0.990 indicating no correlation between nitrates and distance downstream on the Blanda. Similarly, on the Skjálfandafljót, Pearson's correlation coefficient returned a value of −0.184 and a p-value of 0.495 indicating no statistically significant correlation between nitrates and distance downstream. A sharp increase can be seen for nitrates (0.27 mg/L) for two of the points in the Blöndulón reservoir.

For phosphates, no clear relationships were observed on the Blanda or the Skjálfandafljót (Figure 5(m) and (n)). Pearson's correlation coefficient returned a value of −0.301 and a p-value of 0.276 on the Blanda and a correlation coefficient of −0.126 and a p-value of 0.643 on the Skjálfandafljót.

Figure 5(o) and (p) illustrates how water temperature increases with distance downstream for both rivers. The temperature range for the Blanda (9.6 °C) is greater than the Skjálfandafljót (7.2 °C). On the Blanda, the lowest recorded temperature (5.5°C) was at 3% distance downstream and the highest (15.1 °C) at 86%, whereas on the Skjálfandafljót the lowest temperature (5.2 °C) was at 30% distance downstream and the highest (12.4 °C) at 93%. Both rivers experienced a drop in temperature near the mouth, where both rivers meet the Arctic Ocean. Pearson's correlation coefficient returned a value of 0.738 and a p-value of 0.002 indicating a statistically significant positive correlation between temperature and distance downstream on the Blanda. On the Skjálfandafljót, Pearson's correlation coefficient returned a value of 0.853 and a p-value of 0.001 indicating a statistically significant positive correlation between temperature and distance downstream. For the Blanda, a power regression equation (y = 13.571×^0.2589) is the best fit (R² = 0.6678) for a positive correlation between temperature and distance downstream. An exponential regression equation (y = 4.63e^0.9372x) is the best fit (R² = 0.7736) for a positive correlation between temperature and distance downstream on the Skjálfandafljót.

Figure 6 shows that the median turbidity is higher upstream (225.3 mg/L) than downstream (5.01 mg/L) of the Blöndulón reservoir. This difference is statistically significant at the 98% significance level according to the t-test (t = 5.51; p = 0.012). This is considered to be a result of the reservoir acting as a sediment trap as suggested by Lu & Siew (2006).

This trapping of sediment within the reservoir has several important management implications. Firstly, sedimentation within the reservoir reduces its capacity to contain water, making it less viable for hydro-power energy production. Secondly, the interruption to downstream suspended sediment transfer may have positive implications for downstream ecological health. This is because reduced suspended sediment loads can improve the feeding and respiration conditions for downstream aquatic fauna such as salmon (Bilotta et al. 2012). Finally, the perceived trapping by the reservoir has a positive impact on drinking water quality: the upstream levels are significantly higher than the recommended international guideline before water is considered safe for human consumption (5 FTU), whereas the downstream levels fall around the recommended level (Rickwood & Carr 2009).

Figure 7 shows that the median Al upstream (0.5 mg/L) is higher than downstream (0.03 mg/L) of the Blöndulón reservoir. This difference is statistically significant (t = 6.63; p = 0.029). This is considered to be the result of the reservoir acting as a trap for chemicals as well as sediment. This finding supports the studies of Higgins (1978), Kelly (2001) and Ahearn et al. (2005) who all observed this phenomenon occurring downstream of reservoirs. Again, this impact has implications for drinking water quality: the median upstream concentration is five times greater than the recommended safe drinking level of 0.1 mg/L suggested under the international guidelines (Rickwood & Carr 2009), whilst the median downstream concentration is lower than the guideline. In addition, this decrease in downstream Al concentrations is beneficial for ecological water quality as high concentrations are detrimental due to bioaccumulation within the food chain (Davie 2008).

Along with affecting the spatial distribution of water quality measurements in the Blanda, the Blöndulón reservoir was also noted to impact water quality temporally. This is as a result of the reservoir's overflow system being activated during the summer, when flows are increased due to the increased contribution from glacial melt-water. The overflow was operational in 2011 from 7th July to 8th August (Stefánsson 2012). This overflow creates a ‘flushing’ effect downstream of the reservoir, whereby TDS, nitrate and Zn concentrations were reduced and turbidity, Al and phosphate levels were increased. This is illustrated by Figure 8, which not only shows how water quality parameters were affected by the overflow, but also shows visually that when the overflow is off the water is clear and when the overflow is in operation the water is more turbid.

The increase in turbidity and Al levels when the overflow is activated is supported by the findings of Chung et al. (2008) who suggest this is the result of remobilisation of organic matter and suspended sediment. An additional factor that explains the increase in turbidity downstream of the reservoir when the overflow is active is that the water flows through the reservoir faster, reducing the amount of suspended sediment that is trapped (Eiriksdottir et al. 2008).

The decrease in TDS, nitrates and Zn downstream of the reservoir when the overflow is active is explained by Chellappa et al.'s (2009) theory that the release of additional water leads to a dilution of downstream constituents.

The wider management implications of the ‘flushing’ effect of reservoir overflows is somewhat contested. Some argue that it has a negative impact by disturbing the natural equilibrium that has established itself during the time it is not operational (De Vincenzo et al. 2011). In the case of the Blanda, it could be argued that the increase in suspended sediment could result in a detrimental effect on the salmon and other aquatic fauna through altering feeding success and respiration (Bilotta et al. 2012). However, Chellappa et al. (2009) found that ‘flushing’ caused by reservoir activity improved water quality and increased biodiversity as the dilution of nutrients reduced eutrophication and created a flux in water quality that was beneficial to biodiversity. An additional benefit is that the overflow has a positive effect of extending the lifespan of the reservoir by reducing the amount of sedimentation during the sediment laden summer floods (and potentially remobilising previously deposited sediment).

According to the results of the t-tests, the presence of agricultural land use within the catchment has no statistically significant influence on any of the water quality parameters in either river (all p-values >0.06). In addition, both nitrate and phosphate levels were generally found to be low throughout both the Blanda and the Skjálfandafljót. The highest median concentration of nitrates (0.035 mg/L) is well below the level at which eutrophication starts to occur (0.1 mg/L) (Hanrahan et al. 2003). This is a positive result for the farming communities operating within each of these river catchments as it suggests that their operation is having minimal impact on river water quality, probably due to the sparse nature and the limited use of fertilisers in Northern Icelandic agricultural practices (Gudleifsson 2002).

However, in making these conclusions, it should be noted that the apparent lack of influence that agricultural land use has on water quality may also be related to the fact that the data for this study were collected over the summer period. The temporal influence of climate may have affected water quality in three ways. Firstly, in the summer the greatest dilution occurs as a result of the increased discharge from snow and glacial melt water (Eiriksdottir et al. 2008). Secondly, due to minimal precipitation there is limited transportation of pollutants via land washing (Kang et al. 2010). Finally, summer is also the period at which the highest levels of pollutants such as nitrates and phosphates are immobilised by microbes and processes within vegetation (Foster et al. 1995).

It is apparent, both visually in Figure 9, and statistically as a result of the Pearson correlation tests (Blanda r = 0.927, p = 0.023; Skjálfandafljót r = 0.705, p = 0.002), that there is a positive correlation between the proportion of the catchment that is glaciated and the water turbidity within each of the two river systems. These relationships are described by the regression lines and equations shown in Figure 9. Turbidity is associated with the level of suspended sediment carried in the river's flow. Therefore, the results displayed suggest that as the proportion of the catchment that is glaciated increases the suspended sediment load carried by the river increases. As mentioned in the previous section glacial extents may be a slight overestimate given recent glacial retreat in Iceland.

This supports the findings of other studies (Hasnain 1996; Brown 2002; Eiriksdottir et al. 2008) which demonstrated that glaciers increase the amount of suspended sediment within fluvial systems. The suggested reason for this is that glaciers cause increased erosion via abrasion and the hydraulic action of turbulent melt water (Brown 2002). The increase in suspended sediment loads due to increased glacial runoff is important as it impacts upon the feeding and respiration conditions for downstream salmonids (Bilotta et al. 2012) as well as drinking water quality (Rickwood & Carr 2009).

No relationship between TDS and the proportion of catchment that is glacial-fed was apparent, either visually or statistically (Pearson's r = −0.047, p = 0.863), on the Skjálfandafljót. However, on the Blanda, a negative correlation was apparent both visually (Figure 10) and statistically (Pearson's r = −0.642, p = 0.010). A power equation (y = 0.0673x^−0.246, R² = 0.5735) provides the best explanation for this trend. This negative correlation is considered to be due to the increasing contribution of glacial melt water, which dilutes the concentration of dissolved solids. This agrees with the findings of Hindshaw et al. (2011) who found decreasing levels of TDS due to the associated increase in discharge from glacial melt water. This theory also explains why this trend is significant in the Blanda (total proportion of glacial area = 14%) but not the Skjálfandafljót (total proportion of glacial area = 3%).

To explore the impact that glacial runoff has on water quality further, water quality measurements were taken around a confluence between the glacial main branch of the Skjálfandafljót, and a spring/direct runoff fed tributary. Figure 11 illustrates the contrasting nature of the predominantly glacial runoff flow from the main branch and from the direct runoff and spring-fed flow from the tributary. The non-glacial tributary has a higher TDS concentration and lower turbidity level than the predominantly glacial main branch. As a result, the TDS concentration of the Skjálfandafljót downstream of the confluence is increased and the turbidity level decreased. These findings support the findings of Louvat et al. (2010) who observed that the highest TDS and lowest suspended sediment concentrations occur in spring-fed rivers.

As well as finding that glacial areas have an influence on the water quality of Icelandic rivers, this study's results also suggest that the presence of permafrost soils within the Blanda's catchment (Etzelmüller et al. 2007) also influences a river's water quality. In Figure 12, the median Al concentration of sample points on the Blanda upstream of the dam with permafrost in their catchment (0.50 mg/L) is 84% higher than those without on the Skjálfandafljót (0.08 mg/L). Figure 12 reiterates the potential influence of the dam has on downstream Al concentrations, even comparing all the points on both rivers, the median Al concentrations are higher on the Blanda (0.26 mg/L) than on the Skjálfandafljót (0.08 mg/L).

Al concentrations for sample points upstream of the dam whose catchments contained permafrost were significantly higher than those that did not (Figure 12, and t-test t = 7.63; p = 0.002). This is potentially important as these high levels of Al could have a detrimental effect on the gill bearing aquatic fauna as well as creating the potential for bioaccumulation within the food chain (Jonnalagadda & Mhere 2001).

This article has investigated some of the spatial and temporal changes in water quality along two Icelandic Rivers, the Blanda and the Skjálfandafljót, and has made several key findings.

The presence of the Blöndulón reservoir on the Blanda has reduced the downstream levels of turbidity and Al concentrations, decreasing the importance of the upper watershed to downstream water quality. This is the result of the reservoir acting as a sediment and chemical sink.

The Blöndulón reservoir was found to temporarily influence water quality during the summer months when the overflow switched on creating a ‘flushing’ effect. This flushing effect reduced downstream TDS, nitrates and Zn concentrations as a result of dilution by the increased flow. The flushing effect also increased downstream turbidity levels and Al concentrations because of a combination of the remobilisation of sediment previously deposited in the reservoir and reduced residence times in the reservoir.

A greater proportion of glacier-fed flow was found to increase water turbidity but decrease TDS. From this, it was inferred that the suspended sediment concentrations are positively correlated to the glacial area within a catchment as a result of increased erosion. The fall in TDS is considered to be the effect of dilution from the glacial melt water. These general findings were supported by detailed measurements around a confluence on the Skjálfandafljót between a predominantly glacial fed channel and a channel fed by direct runoff and springs.

It is important to recognise that this study has focused on just a few of the many factors that influence water quality within river channels. There are many other factors, including geology, climate, topography, soil, time of year and urban and industrial land use that were not investigated. As a result it has not been possible to fully represent the complex controls over water quality in Icelandic rivers.

Thanks go to Owen King and Edward Ramstedt who were part of the Ice cap to Ocean, Iceland 2011 expedition team and assisted in the field with data collection. In addition, the authors would like to thank the Royal Geographical Society (RGS), the Icelandic Centre for Research (RANNIS) and the University of the West of England (UWE) for their support and funding. As a part of the Ice cap to Ocean project, thanks go to Dr Chad Staddon, Guðmundur R. Stefánsson, Viðar Hreinsson and The Reykjavik Academy, Guðrun Tryggvadóttir and the Svartarkot project team and Dr Edward Huijbens who assisted with the logistics and planning of the expedition.