Interactive comment on “ Spatial patterns of some trace elements in four Swedish stream networks ”

Four river basins in southern Sweden, with catchment sizes from 0.3 to 127 km2 (median 1.9), were sampled in October∼2007. The 243 samples were analysed for 26 trace elements (Ag, As, Au, Ba, Be, B ...


Introduction
The importance of headwaters as a resource for biodiversity and human welfare is increasingly recognised (Lowe and Likens, 2005;Bishop et al., 2008).One reason is that headwaters make up most of the watercourse length and hence provide a large proportion of water and solutes to downstream sites (Alexander et al., 2007).For example, streams with catchment size < 2 km 2 make up approximately 80 % of the total length of all perennial watercourses in Sweden (Nisell et al., 2007).Headwaters also provide the most direct indication of what the terrestrial landscape is contributing to surface waters before in-stream and in-lake processes begin to alter the chemistry of runoff water.It is widely known that the variability in water quality changes with catchment size, typically small watercourses show the highest variability in both space (Wolock et al., 1997;Temnerud and Bishop, 2005) and time (Nagorski et al., 2003;Buffam et al., 2007).Significant efforts have been made to quantify the variability of headwaters (e.g., Hutchins et al., 1999;Smart et al., 2001;Likens and Buso, 2006).The distribution and concentration of a large number of trace elements in the landscape is very instructive.This information may be valuable for terrestrial catchment output heterogeneity (Lofts and Tipping, 2000), lake processes (Johansson et al., 2001;Taillefert and Gaillard, 2002) and downstream transformations in general that can be of value for land-use and water quality management (Reimann et al., 2009).Quantifying landscape variation and potential drivers of water quality is relevant for management of forestry practices and nutrient leaching.Furthermore, as trace metals are increasingly used and spread it is important to try to define current values against which future changes can be judged (Grahn et al., 2006;Karlsson et al., 2006;Reimann et al., 2009).
Some studies estimating the spatial variability of trace elements in Scandinavian, in the range of 10-100 km 2 , have been published (Tarvainen et al., 1997;Ed én et al., 1999;Andersson et al., 2006;Dahlqvist et al., 2007;Reimann et al., 2009).Focus in this study has been headwaters, < 2 km 2 , in combination with larger streams in the network.The boreal Northern Hemisphere is also rich in lakes and they need to be accounted for when spatial patterns of Swedish stream networks are evaluated.Lakes can significantly change the concentration of total organic carbon (TOC), Fe, Al and trace elements in streams, as the water velocity decreases and water depth increases which allows biological, chemical and physical processes, to proceed further than they can in running waters (Drever, 1997).Examples of such processes are: flocculation (von Wachenfeldt and Tranvik, 2008) and sedimentation of particles, changes in pH and Introduction

Conclusions References
Tables Figures

Back Close
Full The aim of this study is to characterize the spatial variability of a range of trace elements (Ag, As, Au, Ba, Be, Bi, Cd, Co, Cr, Cu, Ga, Ge, In, La, Li, Mo, Ni, Pb, Sb, Se, Sn, Tl, Ti, U, V and Zn) entering the drainage network in headwaters and downstream changes in concentrations from four river networks varying in land-use, landscape type, geographical location and impact by acid rain.In each of these basins exists Swedish long-term monitoring sites.This makes it possible to locate the sampling occasions in the long-term variation of other water chemistry parameters of importance for trace elements including pH, Fe, Al and TOC.The main hypothesis in this study is that trace elements and TOC (as well as Fe and Al) show the same general pattern of catchment outputs and downstream changes.Concentration changes downstream are then, mainly due TOC sources and sinks, with the position and volume of lakes being a major factor, since TOC (in combination with Fe and Al) are the main carriers of trace elements (Lyv én et al., 2003).Furthermore, we expect that a number of elements show significant co-variation due to their geochemical resemblance.In addition we want to test if the acid deposition history or seasalt may impact metal mobilisation at the landscape level.

Sampling approach
The synoptic surveys used in this study were designed to provide a snapshot of the water chemistry in the stream networks (Fig. 1 and Table 1).All sampling was carried out in October 2007 over a one to three day period per catchment when the weather was stable.In in total 243 grab samples were taken (Fig. 2 and Table 1).The majority of the samples were taken at headwaters (Table 1), which in this study are defined as sites with a catchment size of < 2 km 2 .Figures

Back Close
Full
The dominant vegetation type is coniferous forest (> 80 %), mainly Picea abies, Norway spruce, on till soils (Table 1).Mires and small humic lakes made up most of the remaining parts of the catchments, while the proportions of agricultural and developed areas were minimal (< 1 %).The mean annual air temperature  in the catchments ranged from 7 -median Q for the whole period 1990-2010)/interquartile range of Q for the whole period 1990-2010 (Fig. 2).Runoff at the time of sampling in October 2007 was compared to percentiles based on October 1990-2010 (Table 1).Marine impacts, such as chloride-deposition, were largest at R. A. and in decreasing order R. L., R. D. and R. G. Medians for bulk deposition of chloride (m Eq l −1 ) during 2007 were 0.053 (G årdsj ön), 0.020 (Aneboda) and 0.012 (Kindla) (L öfgren et al., 2011).(In the cited work G årdsj ön is a site inside R. A., Aneboda is approximately 50 km south-west of R. L. and Kindla is inside R. D.) Liming activities have taken place in all rivers, to the largest extent in R. A. and then in decreasing order R. L., R. D. and R. G.When estimating the degree of acidification in 2007, using metaMAGIC (Erlandsson et al., 2008), no sites in R. G. Figures

Back Close
Full deposition over the country that is highest in the south-west and decreasing in both north and east (Moldan et al., 2004).The R. A. catchment has larger median coverage of lake surface and wetlands than the other 3 river catchments (Table 1), and a more heterogeneous soil mostly consisting of bare rock, while the other catchments were dominated by till.The arable land areas were small, with the largest portion in R. L., which also had the largest volume of forest.The R. D. catchment has the smallest percentage of lake surface coverage and the forest has equal volumes of pine and spruce.The smallest catchment, R. G., has the highest percentage of wetlands and the forest is dominated by pine.For other details see Temnerud et al. (2009).

Chemical analyses
After collection, all water samples were kept dark and cool until they were analysed.Total organic carbon (TOC) was measured as CO 2 after combustion using a Shimadzu TOC-VPCH analyser, after acidification and sparging to remove inorganic carbon.It has previously been shown that dissolved organic carbon (DOC) and TOC differ on average by less than 5 % (Ivarsson and Jansson, 1994;K öhler, 1999), so TOC is essentially identical to DOC in these waters (see also Gadmar et al., 2002).Total nitrogen (N) is the sum of Kjeldahl nitrogen and of nitrite and nitrate (analysed according to the standard SS-EN ISO 13395).Total phosphorous (P) was analysed according to SS-EN ISO 6878.

Back Close
Full ICP-MS (Agilent 4500) equipped with an ultrasonic nebulizer (U-6000AT+, CETAC).Samples for anion analysis (chloride (Cl), fluoride (F) and sulphate (SO 4 )) were stored frozen until analysed on a Dionex ion chromatograph system.The typical precision in anion and cation analyses based on measurements of certified standards was better than 2 %.For limit of detection (LOD), 3 times the standard deviation of the blank was used for the trace elements, see Table 2. Acid neutralizing capacity (ANC) was calculated as the difference between strong base cations and strong inorganic acid anions (K öhler et al., 2001).ICP-OES and anions analysis where performed at the Department of Geological Sciences, Stockholm University, ICP-MS and TOC at Örebro University and N, P and pH at Vattenlaboratoriet, Uppsala Municipality.

Map material
To relate trace elements to catchment character, we used land-use maps on the scale 1 : 100 000 (SNLS, 2002) and soil maps, along with the kNN-database of vegetation that contains forestry variables estimated from Landsat 2000 satellite photos (Reese et al., 2003) (Table 1).The variable "organic soil type" is referred to as "peat" in this study.A land-use data set called CORINE, provided by the European Environment Agency, was also used (EEA, 2002, http://www.eea.europa.eu/publications/COR0-landcover).Digital soil data were available for R. A. (SGU, 2001).For the other catchments, photographs of these soil maps, with GIS coordinates, were digitalised manually; R. L. year 1876 and scale 1 : 200 000, R. D. year 1879 and scale 1 : 50 000, R. G. year 1930 and scale 1 : 50 000 and for a small part (∼ 10 %) map from year 1872 and scale 1 : 50 000.The percentages of each variable on each catchment from the land-use and soil maps were used in the modelling.In addition to the data read directly from each map, three variables were calculated from the map information: stream length (m), drainage density (km −1 ), and altitude (m).Drainage density is the ratio of the total length of streams within a catchment to the total area of the catchment.

Statistical analysis
In order to estimate the variability of water chemistry, between catchments, robust coefficient of variation (CV) was used.Robust CV = 100× normalised IQR/median, where IQR is interquartile range (75-25 percentile) and normalised IQR is IQR multiplied by 0.7413 (PTA, 2006).The factor 0.7413 makes IQR comparable to standard deviation; if the distribution is normal, normalised IQR values would equal standard deviation.Since the data are not normally distributed (tested using one-sample Kolmogorov-Smirnov test) and there are generally few samples, normalised IQR was used.A Piper diagram was used to evaluate the differences in major cat-and anions (Piper, 1944), the diagram was made using the software GW Chart version 1.23.3.0 for MS Windows (Winston, 2000).
Common factor analysis (FA), also called principal axis factoring, was performed to check for latent structures in trace elements, major chemistry and physical parameters using the oblique (non-orthogonal) rotation method Oblimin (δ = 0) and Kaiser normalisation (SPSS for Windows v19.0).The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy tests whether the partial correlations among variables are small.KMO should be greater than 0.6 in order to make FA valid to perform.Another indicator of the strength of the relationship among variables is Bartlett's test of sphericity, which is used to test the null hypothesis that the variables in the correlation matrix are uncorrelated.When the significance level is small enough to reject the hypothesis (p < 0.0001), the relationship among variables are strong, and FA can be conducted.In FA the variance of a single variable is decomposed into common variance that is shared by other variables included in the model and unique variance that is specific to a particular variable (Gauch, 1982).Principal component analysis (PCA), a similar multivariate procedure, considers the total variance and makes no distinction between common and unique variance (Gauch, 1982).Oblique rotation allows the factors to correlate, if the factors are truly uncorrelated, orthogonal and oblique rotation produces similar results (Costello and Osborne, 2005) Chloride was omitted from all FA-analyses since it made the matrix indefinite, i.e. the eigenvalues were not positive.Of all variables in Table 1 the following were used: altitude, forest volume, lake surface, drainage density, forest, wetland, open fields (arable, pasture and open field), because the other variables made the matrix indefinite or gave KMO < 0.6.Likewise, F, N, Na and P were omitted for R. D. and Ca, F, K, Mg, Mn, N, Na, P, SO 4 , for R. G.
Statistical power analysis was performed using SPSS for Windows (v19.0).The post hoc analysis of the difference in means between headwaters and downstream using the two-tailed Student t-test was run in accuracy mode calculation.H0 was that there is no difference between headwater and downstream sites.H1 was that there is one.The statistical power (1 − β) gives the probability that the hypothesis is correctly rejected, a value above 0.80 indicates that H0 was correctly rejected.
Median and robust coefficient of variation (CV) of the difference in stream water chemistry between 40 stream reaches (all four rivers put together), 53 stream junctions with 107 tributaries vs. downstream comparisons and 36 lakes with 77 inlets vs. outlet comparisons (Table 1).To test if the difference was significant both Wilcoxon and Paired Student t-test were used.

Results
In Sweden groundwater tables are usually low in September and October.During 2007 the groundwater tables were lower than usual for this time of the year for the 4 rivers (SGU, 2012).The weather in September 2007 was normal (compared to 30 yr averages) while it was 1 • C colder and precipitation was 50 % lower than usual in October

Variability of chemical variables
The Piper diagram shows that the majority of the samples are dominated by the cations Na and K and by the anions Cl, and to a lesser degree SO 4 (Fig. S1 in the Supplement).R. A. and R. L. that are located closer to the sea and that are more heavily impacted by acid rain had on average twice as high average concentrations of the marine derived ions Na, Cl, Mg and SO 4 .This was also the case for the elements Li, K, Cu, Ni and Sr. R. D. and R. G. had higher concentrations of U, Bi, V and Be.The following elements were under the LOD: Ag and Au < 0.1, In < 0.5 and Se < 500 (all in ng l −1 ).
The rivers (Table 2) are discussed in the following order: R. A., R. L., R. D, and R. G.This is in accordance with natural gradient from the Swedish south-west coast (R. A.), up north-east to R. G. (Fig. 1).The median value decreased monotonically for Cl, Li, Na and pH (Table 2), from R. A. to R. G., while the opposite trend was observed for As, Bi, Ga, Pb, Sn, Tl, U and V.The elements with the highest robust CV (> 100), using all 243 sites, were in decreasing order: U, Cl, V, Fe, Bi, and Ga (the column denoted "All" in Table 2).At the other end of the scale, the chemical variables with the lowest robust CV (< 41) were in increasing order: pH, Sb, Li, As and SO 4 (Table 2).Robust CV for Al and TOC were 65 and 69, respectively, meaning they are essentially "median" robust CVs of all chemical variables.Five chemical variables were chosen to exemplify trace elements variation with catchment size and pH; Li, V, Ga, Sb and La.Higher variability with lower catchment size was generally seen for those chemical variables (Fig. 3).Introduction

Conclusions References
Tables Figures

Back Close
Full shown FA-analyses shown were valid to perform, although the values were borderline for R. G.
In the factor analysis three groupings were identified (Fig. 4a): -Near the far end of the first factor, opposite to pH, was a cluster of Al, As, Bi, Cr, Fe, Ga, Ge, Si, Tl, TOC, Pb and V; mainly substances associated with low pH conditions.This cluster is hereafter termed FA1 Chem 1+.
-A third cluster, FA2 Chem 1+, includes Mg, K, Sr, La, Na, ANC, SO 4 and Ca located at the positive end of the second factor.
Along the first factor lake surface was the map information with the highest negative eigenvalue (Fig. 4a), opposite of FA1 Chem 1+.Positive values along the first axis were forest and altitude (Fig. 4a).Cluster FA2 Chem 1+ co-varies positively with open fields (such as arable, pasture and open fields), while forest coverage and altitude were opposite to FA2 Chem 1+ (Fig. 4a).The patterns observed in Fig. 4a could, to some degree, be observed in Fig. 4c-f, although the pattern for R. A. (Fig. 4e) was roughly a mirror image of the FA-analysis for the other 3 rivers.Perhaps the most notable FA result was that Cu did not co-vary with TOC and Zn (Fig. 4a, c-f).Based on the score plot some few sites in R. A. and R. D. showed differences between the four rivers, but most sites overlap based on scores, indicating that they have similar spatial pattern (Fig. 4b).
The Piper plot confirms the similarity between the rivers, except one site at R. L. that behaved differently (Fig. S1).

Headwaters compared to downstream
The variability was generally higher in the headwaters (< 2 km 2 ) compared to sites further downstream.The median concentration values decreased with increasing Introduction

Conclusions References
Tables Figures

Back Close
Full catchment size, with some exceptions.The catchment with fewest significant chemical differences between headwaters and downstream sites was R. L. The biggest differences were observed for R. G., with both increasing (ANC, Ca, K, pH and SO 4 ) and decreasing (Al, As, Ba, Be, Cd, Co, Cr, Fe, Ga, Ge, Li, N, Ni, P, Pb, Si, Sn, Ti, Tl, TOC, V and Zn) concentrations with increasing catchment size (Table 2).All significant changes in R. D. showed lower downstream concentrations (Al, Ba, Be, Cd, Co, Cr, Ga, Ge, Si, Ti and Tl).In R. A., both significant increases (Li, Mo, pH and Ti) and decreases (Cd, Co and Zn) were detected (Table 2).

Comparison of stream reaches, stream junctions and lakes
When comparing concentrations within the same stream channel (n = 40) the following showed significant differences: Cl, K, Mg, Mo, Na, Ni, Sb, Sr, Ti and V (Table 3) which all increased in concentration except Sb, which decreased.The 53 stream junctions had in total 107 tributaries and the following elements changed significantly when comparing inlets with the site downstream the junction: Cd, Co, Ge, pH and Tl increased while the others decreased in concentration (Table 3).For 36 lakes, comparisons of 77 pairs of inlets vs. outlets concentration gave significant differences for: Al, As, Ba, Bi, Ca, Cd, Cl, Co, Cr, Fe, Ga, Ge, K, N, Na, Ni, P, Pb, pH, Si, Sn, SO 4 , Ti, Tl, TOC, U, V and Zn (Table 3); decreasing concentrations except for Ca, K, pH and SO 4 which showed an increase.

Comparison with long-term time series and analysis of sampling technique
We used data, covering the period October 1996-2009, from 132 reference sites at the Swedish national environmental monitoring programme for watercourses in Southern Sweden for comparison with observed concentrations in the present study (same sites as in Huser et al., 2011).Concentrations of TOC, Fe, Co from this study were usually higher than median of the national programme, while this study values were lower for Mo, Cu and Zn.In general the observed concentrations of Al, Fe, Pb and TOC were Introduction

Conclusions References
Tables Figures

Back Close
Full higher in our study compared to the median of water samples in the Geochemical Atlas of Europe (the column denoted "Europe median" in Table 2) (Salminen et al., 2005).While Ca, K, Mo, pH and SO 4 were lower in our study compared to the median in the Geochemical Atlas of Europe.This study was based on unfiltered samples because the filtering process in these humic and iron rich systems may create artefacts (Sholkovitz, 1992).However, in the lower parts of R. A. there were silty soils and livestock, potentially influencing the amount of inorganic particles suspended in the samples.Acidifying (using HNO 3 ) the unfiltered samples at R. A. could be the reason why some rather similar trace elements had the highest or the lowest robust CV.Thus the lack of filtering could be the cause of the apparent increases in Li and Mo in downstream sites.This was the only river where we found significant differences between filtered (at 0.45 µm) and unfiltered absorbance at 420 nm (absDIFF) (Fig. S2 in Supplement).At all other rivers absDIFF were smaller than 7 % on average.

Discussion
Trace elements may originate from anthropogenic point or diffuse sources or from the bedrock and soil.They are mobilized under favourable geochemical conditions (typically acidic pH) in the presence of colloidal carrier phases such as organic carbon or sesquioxide particles (Al, Fe, Mn oxy-hydroxides) and in the form of other inorganic complex such as chlorides, fluorides, hydroxides, carbonates or sulphates.Organic carbon is a major controlling factor on both metal mobility and pH in most of the running waters and lakes in this study.The main hypothesis in this study was that trace elements follow the same spatial pattern as TOC, and that concentration decreases from upstream to downstream sites, mainly due to the position and volume of lakes.We can not reject the hypothesis for: As, Ba, Be, Bi, Cd, Co, Cr, Ga, Ge, Ni, Pb, Sn, Ti, Tl, U, V and Zn (Fig. 4a), as they follow similar spatial pattern as TOC (as well as Al and Fe) when comparing lake inlets and outlets (Table 3).The largest absolute changes in Figures sites downstream lakes were observed for sites in R. D. and R. G. that have a lower lake surface area and also lower specific average runoff.The trace elements that did not show a significant decrease in concentration when comparing inlets with outlets at lakes were Cu, La, Li, Mo, Sb and Sr (but also Mg and Mn) (Table 3) most of which are discussed below.Significant increases in lakes were detected for Ca, K, pH and SO 4 .These results suggest that there could be a connection between the levels of Cu, La, Li, Mo, Sb, Sr and those of Ca, K, pH and/or SO 4 .In R. G. there was a significant decrease in TOC between headwater and downstream sites, along with Al, Ba, Be, Bi, Cd, Co, Cr, Fe, Ga, Ge, Li, Ni, Pb, Si, Sn, Ti, Tl, U, V and Zn while ANC, Ca, K, pH, SO 4 increased.In the other rivers there were significant downstream decreases in Cd and Co.At R. D. the concentrations of Al, Ba, Be, Cr, Ga, Ge, Si, Ti and Tl decreased downstream and for R. A. there was an increase in concentration of Li, Mo, pH and Ti (Table 2).
Elements where significant downstream losses were observed at all rivers include the elements Al, Be, Fe, Ga, Ge, Si, Ti and Tl.All these elements have in common that they bind comparingly strong to hydroxide ions (Stumm and Morgan, 1996).They may form neutral species in water when pH increases such as may be observed during the passage of a lake with internal alkalinity production.Elements that mainly form cationic species such as La, Sr or predominantly exist in the anionic form such as As, Sb, Mo, U or V are not removed during a lake passage or are removed only when iron is removed simultaneously.
For R. A., R. L. and R. D. no significant changes in TOC were detected between headwater and downstream sites and as well in pH for R. L. and R. D. In as much differences/similarities in the behaviour in the spatial patterns of chemistry between the four rivers could be due to differences/similarities in runoff situation during sampling (Fig. 2 could indicate that the latter elements were mobilized depending on geological variations of trace minerals while Pb and Zn were present but immobilised under higher pH. The overall FA analysis indicates that there are roughly three clusters of elements, major elements to the upper left, TOC and Fe related trace elements to the right and a number of elements in between those two groups.This indicates various drivers and element behaviour also varies in between rivers.TOC, Fe and Al have a strong negative correlation to pH in many Swedish surface waters indicating TOC as a major driver for pH, Fe and Al mobility (K öhler et al., 2000;Buffam et al., 2007;Sj östedt et al., 2010), which is why these two factors are confounded.A detailed analysis of this behaviour is outside the scope of this article and will thus not be presented here.Instead we want to focus here on some specific observations only.The metals described here may be divided in many different ways according to their physical and chemical properties.

Major soluble cations (Ca, K, Mg and Na)
The major cation concentrations in the streams are controlled by deposition, weather- two latter are heavily influenced by agricultural practice and plant uptake while the concentration variations in F probably reflect variations in F bearing minerals such as apatite or muscovite etc. Si was mainly in soluble form.Comparisons of molybdate reactive Si and Si determined with ICP-OES at R. A. and R. L. (slope = 1.00 ± 0.05, n = 44, R 2 = 0.92) do not indicate any differences (Fig. S2).We thus assume that the possible contribution from Si by clay particles (< 0.2 ppm) is negligible.Initially Si is derived from weathering and does not show any geographical pattern.Recently the importance of Si recycling from biologically derived material was documented in boreal areas (Engstr öm et al., 2010).Our finding of Si clustering together with TOC-Fe may be a hint that significant amounts of Si have been recycled and are mobilized together with TOC from organic rich riparian soils.

4.3
Trace cations in stable oxidation states (Al, Ba, Be, Bi, Cd, Co, Cs, Cu, Ga, Ge, La, Li, Ni, Sn, Sr, Pb, Ti and Zn) Our overall and site specific FA analysis confirm many of the known drivers for metal mobility such as the effect of pH for trace metal concentration in general, the strong binding of Co to MnO 2 , Cr to Al, Pb to Fe(OH) 3 .Many of those latter elements relate in a very linear fashion at all sites (for Co and Mn respective Al and Cr see Figs.S3 and  S4).The liming activity (both pH and Ca) could have an impact on the spatial pattern at other sites (Egeberg and H åkedal, 1998;Lydersen et al., 2002) but are considered to be low in our case, since none of the trace elements were related to dose of liming in R. A. and R. L. (data not shown).However, Fe concentrations decrease with increasing lime dose at both rivers.In R. A., the river that is most affected by marine sea salt spray, we observed strong correlations of Cl to a large number of trace elements.At this river a significant correlation between SO 4 and Cu was also observed.This may indicate an effect of soil solution ionic strength from marine sea salt sources on the mobilisation of elements at that river and/or mobilisation of Cu due to the impact of acid rain.According to our FA-analysis (Fig. 4), Cu does not co-vary strongly with Al, Fe, Zn or TOC, despite that organic ligands are known to strongly bind Cu (Bresnahan et al.,5734 Introduction

Conclusions References
Tables Figures

Back Close
Full 1978; Hart, 1981).Zn and Cu behaved differently during the passage from the inlet to the outlet of lakes, possibly due to differences in sedimentation (see Table 11.5 in Stumm, 1992).Similar findings for Cu were observed in a Swedish national survey of 1200 lakes (see PCA-plots on page 196 and 202 in SEPA, 1999).Cu behaved differently compared to Zn, and had no significant correlation with TOC, when evaluating Swedish time series of trace elements in 139 streams (Huser et al., 2011).Other articles had dealt with this difference in behaviour between Cu and Zn (e.g., Ross and Sherrell, 1999;Lawlor and Tipping, 2003;Tipping et al., 2010).
The non hydrolysed cations La and Sr are not removed during river or lake passage.Elements potentially binding strongly to colloidal phases of Al, Fe, Mn and/or organic matter such as Co, Pb, Cr and maybe even Ga and Ba are removed when pH, TOC or alkalinity changes occur.Surprisingly this is not the case for Cu and U that both are known to bind strongly to these colloidal phases.Potentially both La, Cu and U bind very strong to fractions of organic matter that are not easily removed during lake passage.Alternativly in the case of La, the strong relationship of La to ANC may be due the co-variation of La with alkalinity (R 2 = 0.76) at R. A. potentially indicating La mobilisation from carbonates in solution (Johannesson et al., 2004).

Trace cationic elements in varying (I-IV) oxidation states (Mn, Cr, Fe and Tl)
We have no analytical data on the oxidation state of these elements despite this being a strong driver for their mobility and binding.Fe and Mn are known to partly exist as Mn(II) and Fe(II) in presence of organic matter preventing hydrolysis.We assume though that the majority of both of those elements are mostly present in their oxidized form Fe(III) and Mn(IV+) in these well oxygenated surface waters.The strong correlation of Co with Mn may be an indicator that Mn is present as MnO 2 colloid to which Co is known to bind strongly (Pokrovsky et al., 2011).In a similar way we may argue for Cr being in the anionic form as Cr(V) bound to aluminum or organic aluminum colloids (cf.strong correlation between Al and Cr in the Fig. S3 in Supplement).Introduction

Conclusions References
Tables Figures

Back Close
Full The anionic elements As, Mo, Sb, U and V do not seem to behave in a similar manner.While As and V are removed at river R. G. potentially together with Fe in accordance with the predicted speciation calculations by W ällstedt et al. ( 2010), both Mo and U do not show any removal.Very little information is available for Mo in surface waters while U is known to strongly bind both to carbonates, organic matter and Fe colloids (Tipping et al., 2011).

Land-use
In our analysis we did not find any of our available landscape variables adding significant additional explained variance to the FA analysis (Fig. 4).The information contained in our landscape data is probably too unspecific for exploring geochemical variations quantitatively and instead represent only chemical, physical and biological processes.
The strong removal of a large number of metals when passing lakes is though evident in the FA analysis where lake surface plots opposite to many of those metals (Fig. 4).Forest volume does not respond in a similar systematic fashion and, surprisingly, the amount of wetland does not relate strongly to either TOC or Fe at any of the rivers.More detailed spatial analysis including extent and thickness of the riparian strips might help to understand this better.Our analysis is a first step in better understanding processes that drive metal concentrations and metal retention in the Swedish landscape.Many of the trace metals studied here are co-transported with organic matter and a better understanding of quantitative organic carbon, Al and Fe removal in stream reaches and lakes will aid in the understanding of the fate of many trace elements e.g. the systematic study of the transformation conditions of dissolved to particulate forms of Al and Fe such as described in Sj östedt et al. (2010).Both systematic studies of temporal and spatial variation of filtered and unfiltered metal concentrations (Huser et al., 2012) and speciation modelling of trace elements are under way to better elucidate metal removal Introduction

Conclusions References
Tables Figures

Back Close
Full processes.Reasonable data for natural background concentrations of a large number of elements is required by the new framewater directive.Defining reliable background concentrations based on geographical information only seems a very challenging task.Our analysis implies that lake position, morphology and volume are the most important drivers for metal concentration as compared to many other natural landscape features in the investigated catchments.

Conclusions References
Tables Figures

Back Close
Full Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

(
SMHI, 2012).During the sampling the HYPE modelled runoff was low and stable in R. A. and in R. G., while for R. L. and R. D. the sampling took place during the descending limb of a minor runoff episode (Fig.2).FiguresBack CloseFull Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | ) are unclear.Classic heavy metals such as Cd, Co, Cu, Pb or Zn do not show any strong geographical or sea salt controlled pattern contrary to what we expected.Only Pb and Zn seem to be mobilized preferentially under acidic conditions while Co, Cu and Ni display varying concentrations irrespective of pH.This behaviour Discussion Paper | Discussion Paper | Discussion Paper | ing, biomass uptake and ion-exchange.Detailed descriptions of these processes and fitted models to cation fluxes of three sites inside the catchments R. A., R. L. and R. D. may be found in K öhler et al. (2011).Higher deposition fluxes of both antropogenically derived acids and marine derived sea salt both lead to increasing salt loads in R. A. and R. L. as compared to R. D. and R. G. sites.In R. A. elevated Ca concentrations are due to liming activity.After excluding limed sites almost constant Ca/Mg ratios were observed at almost all other sites indicating somewhat homogeneous net cation mobilization at most sites.4.2 Major soluble components in anionic form (C, Cl, F, N, P, S and Si) At the rivers R. A. and R. L. that are close to the sea, marine derived S and Cl are much higher than at the other rivers while this is not the case for F, N and P. The Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Tipping, E., Rothwell, J. J., Shotbolt, L., and Lawlor, A. J.: Dynamic modelling of atmospherically-deposited Ni, Cu, Zn, Cd and Pb in Pennine catchments (Northern England), Environ.Pollut., 158, 1521-1529, doi:10.1016/j.envpol.2009.12.026, 2010.Tipping, E., Lofts, S., and Sonke, J. E.: Humic Ion-Binding Model VII: a revised parameterisation of cation-binding by humic substances, Environ.Chem., 8, 225-235, doi:10.1071/DiscussionPaper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Table

Figures
Figures

Fig. 1 .
Fig. 1.Map of Sweden with the four rivers included.The black dots are sampled sites.For stream junctions one dot could indicate that samples were taken upstream and downstream of the junction.O indicates the outlets of the rivers.
Discussion Paper | Discussion Paper | Discussion Paper |

Table 1 .
Map material for each catchment, median with 5th and 95th-percentiles in parentheses.Outlet runoff is the daily HYPE modelled runoff during sampling compared to percentiles (perc) for October 1990-2010.For more details see the Method section.