Episodic hydrological events, such as snowmelt during spring, have a marked effect on stream water chemistry. Here we investigated how spring snowmelt affected δ34S values of sulfate in six streams situated in northern Sweden. Four streams had high δ34SSO4 values during base flow with values ranging from +11.9 to +8.6‰. During snowmelt the δ34SSO4 decreased to around +6‰. In one of the streams and in the forested upper reaches of a second stream, δ34SSO4 values were close to +5‰ during base flow and decreased to about +3.8‰ during the spring snowmelt. One stream, which drained cultivated postglacial sediments dominated by acid sulfuric soils, was differentiated from the other streams by low δ34SSO4 values (−5.0‰ to −0.5‰). We could identify two stream water SO4 sources: sedimentary sulfides and anthropogenic S. Bacterial dissimilatory sulfate reduction was identified as an important process affecting stream water δ34SSO4 values and suggests that in this boreal landscape, peatlands and possibly riparian zones have a large influence on the biogeochemistry of SO42− during base flow conditions. Our results suggest that during the spring snowmelt, snow S and desorbing SO4 of mainly anthropogenic origin are the two major S sources in four of the investigated streams. Two streams in forested areas also indicate that reoxidation of reduced S may be released during the spring flood. The stream in the cultivated area was found to be strongly influenced by the acid sulfuric soils independent of stream flow conditions.
 Recovery from acidification is a long-lasting process [Mörth et al., 2005; Prechtel et al., 2001] and elevated stream water SO4 concentrations can be expected to remain for decades in areas that have received high levels of anthropogenic S deposition. The biogeochemistry of S is thus still an important issue in water management in many European and North American regions despite a markedly decreased S deposition during the last decades [Stoddard et al., 1999]. In northern Sweden a number of studies have investigated the origin of stream water acidity and whether it is related to natural processes or to anthropogenic deposition [Laudon et al., 2000, 2001; Laudon and Bishop, 2002a]. Since anthropogenic acidification is tightly linked to the SO42− deposition there has been an interest in monitoring stream water concentrations of SO42−. Sulfate concentrations in the stream water alone are not, however, conclusive in determining the origin of stream water S. Instead, stable sulfur isotope ratios (34S/32S) can provide a more decisive test on the fate and sources of natural and anthropogenic SO42−. The use of stable isotope ratios can also help resolve how differences in the landscape affect the biogeochemistry of SO42−.
 The boreal landscape in the Västerbotten County encompasses a large variability including wetlands, forested areas, and cultivated areas. These landscape variations are likely to have a large influence on the isotopic composition of stream water sulfate. There are, however, few δ34S isotope studies available from stream waters in northern Sweden. Andersson et al.  investigated a watershed south of the county of Västerbotten and found stream water δ34SSO4 variations commonly between +5 and +7‰ and reported individual δ34S values up to +30‰. The variation in the Kalix River catchment, one of the larger rivers in northern Sweden was less with the δ34SSO4 values ranging between +5.3 and +7.4‰ [Ingri et al.,1997]. In the coastal areas of Finland, east of the Västerbotten County, Åström and Spiro  found variations in stream water δ34SSO4 from +3.9 to −8.0‰ reflecting the large effect that acid sulfuric soils can have. Similar soils are found in many coastal catchments on the Swedish side of the Bothnian Bay. The acid sulfuric soils are sulfide-bearing sediments deposited during earlier stages of the Baltic sea from about 7000 to 4000 BP [Öborn, 1989, and references therein]. The sulfides were formed particularly in shallow bays where conditions for bacterial sulfate reduction were favorable and the concentration of marine sea salt was relatively high. In northeastern Sweden, the fine-grained sediments consist of sulfide-bearing, black clay and silt sediments, often with a thickness of 5 m.
 The spring time snowmelt is the largest episodic event of the year and accounts for up to 50% of the total yearly water discharge in the boreal forest of northern Sweden [Laudon et al., 2004a]. The hydrological flow paths are likely to change during these events and may cause shifts in the δ34SSO4 values. Few studies are, however, available from forested catchments on variations in the δ34SSO4 values during snowmelt conditions, despite the importance of these episodic events. Shanley et al.  showed that the stream water export of S changed from a dominance of geological S during base flow conditions to a larger atmospheric influence during the snowmelt episode. In contrast, Campbell et al.  found only a minor influence of anthropogenic S sources during the spring snowmelt. Other studies have also indicated shifts during episodic events: Fitzhugh et al.  reported a decrease in δ34SSO4 values during low-flow conditions that were related to a deeper hydrological flow path increasing the contact with S-bearing minerals. Åström and Spiro  also demonstrated changes in stream water δ34SSO4 values related to stream flow conditions. These previous studies clearly demonstrate that stream water δ34SSO4 values provide a powerful tool to elucidate the origin of SO42− during the spring runoff.
 We selected six streams in the Västerbotten region with different coverage of forest, peatland and agricultural land and with known differences in SO42− concentrations. The objective of this study was to (1) determine how stream water δ34SSO4 values vary during the spring episode; (2) relate these changes to catchment characteristics; and (3) determine sources of S using a simple two-component mixing model.
2. Material and Methods
2.1. Study Sites
 The study was conducted on six streams in the county of Västerbotten (Figure 2) between April and June 1997, as part of the episode project in northern Sweden [Laudon et al., 2000, 2001]. The stream catchments range in size from 0.55 to 3279 km2. Two of the streams, Lillån and Stridbäcken, are coastal and drain directly into the Gulf of Bothnia (Figure 2). Both these streams are located below the highest postglacial coastline. In this region, Quaternary deposits have been reworked by marine wave action resulting in the transport of fine fractions from exposed slopes to the valley floor. The Stridbäcken catchment consists of frequent bare rock outcrops, thin soils and large peatland complexes in the upper reaches of the catchment. A major portion of the Lillån stream basin is covered by thick marine-derived sulfide-rich postglacial sediments. The northernmost catchment, Flarkbäcken, is also predominantly located below the highest postglacial coastline. The upper reaches of the catchment (upstream of Pellboda), covering about 1876 km2, are covered by forests (70%) and peatland (15%). The catchment area downstream of Pellboda (1403 km2) consists of agricultural fields covering about 31% of the area. The remaining area is mainly covered by forest (67%) and peatland (2%).The soils are mainly postglacial sedimentary deposits; sandy sediments dominate the forested areas while finer, mainly marine-derived sulfide-rich, silt sediments are found in the lower cultivated areas.
 The three remaining catchments, Svartberget, Mellansjöbäcken and Sörbäcken, are located inland predominantly above the highest postglacial coastline. The quaternary deposits in the inland are often several meters thick. The mean annual precipitation in the region is close to 600 mm, of which approximately 35% falls as snow. Detailed descriptions of the physiographical, hydrological and hydrochemical characteristics are provided by Laudon and Bishop [2002b] and Bishop et al. .
2.2. Water and Snow Sampling
 Water sampling was carried out by collecting grab samples. The sampling program was based on weekly to biweekly sampling prior to snowmelt and daily sampling of the spring flood episode. All samples were collected in acid washed 250 mL HDPE bottles. Discharge was computed on an hourly basis from water level measurements using pressure transducers connected to Campbell Scientific data loggers in Svartberget and Stridbäcken. In Sörbäcken, Mellansjöbäcken and Lillån, stage heights were recorded manually concurrent with stream water sampling. In Flarkbäcken no discharge measurements were conducted. Rating curves were derived using bucket method measurements in Svartberget and using hand-held current meters or a semiautomatic salt dilution technique in the remaining streams. The water samples were taken at the outlet of the streams. The Flarkbäcken stream was also sampled at two sites upstream (Gammbyn and Pellboda, Figure 1).
 Snow samples were collected from all catchments as 5 cm diameter snow cores at one occasion. The core was divided into an upper and lower half and kept separate. The snow sampling occurred prior to any snowmelt and after all major snow fall events. The samples were melted in a cool room in closed containers in order to avoid evaporation.
 The SO42− concentrations were measured with an ion chromatography system (Dionex, DX-300 equipped with an AS14 column using electrical suppression, Dionex Corp., Sunnyvale, California).
 Sulfate from water was collected on an ion exchange column (strong basic anion resin, Sigma-Aldrich, number 21.740-9, Dowex 1 × 8–50, 20–50 mesh). The collected SO42− was eluted with 0.5M NaCl and precipitated with 0.5M BaCl2 after boiling for one minute and adjusting the pH to about 2 with HCl. The solution containing the precipitate of BaSO4 was left in a water bath at about 90°C. After two hours the solution was cooled at room temperature and filtered through a polycarbonate filter (Millipore, Isopore, HTTP 02500 or HTTP 04700, depending on the amount of BaSO4 precipitated). The BaSO4 collected on the filters was washed with ultra clean water (Elga Maxima Analytica) to remove Cl− from the eluent and dried in an oven at 60°C overnight. Filters were then scraped to remove the BaSO4 and the amount collected was weighed to calculate recovery. The recovery is normally over 95%, based on a large number of samples processed at the Department of Geology and Geochemistry, Stockholm University, over the last 12 years.
 Isotope measurements on extractable SO42− are reported in δ34S values according to
where R refers to the ratio of 34S and 32S in the sample and the standard (Cañon Diablo Troilite (CDT)), respectively. CDT, IAEA-S1 and in-house references were used as standards. The sample and standard mass ratios were measured using isotope ratio gas mass spectrometry (IRMS) (Finnigan Delta+, Thermofinnigan Corp., Bremen, Germany). All sulfur isotope measurements were done on SO2 (g) converted from BaSO4, with continuous flow technique on line using an elemental analyzer (Carlo Erba, NC2500). The BaSO4 was mixed with V2O5 (1:1 by weight, normally about 300 μg each) and put in tin capsules. The reference gas was measured before every sample and the precision was better than ±0.2‰.
 To test if mixing between two sources could explain variations in stream water δ34SSO4 values we plotted the stream water δ34SSO4 against the inverse of the sulfate concentration. The intercept b in equation (2) represents the δ34S value of one of the end-members,
We used a simple two-component mixing model to estimate the relative contribution of two different end-members;
where a and b are the fractions of a mass or mass flux (a + b = 1), the suffixes a and b indicate the isotope values for the mass or mass flux fractions, and the suffix T denotes the isotope value of the mixed end-members.
 Linear regression was used to test for significance between the δ34SSO4 and the inverse of the sulfate concentration. Uncertainties are given as 95% confidence interval unless otherwise stated. Differences were regarded as significant for probability levels <0.05 unless otherwise stated.
3.1. The δ34SSO4 Values in Snow and Stream Water
 The average δ34SSO4 value of the snow across all sites was +5.9 ± 0.4‰ (n = 6) and ranged between +6.6 and +5.6‰ (average values for total snow cover in each site). No difference between the upper and lower section of the snow cover was found (p = 0.48, paired sample t test).
 Three distinct patterns were found among the six streams studied. The two downstream sampling sites along the Flarkbäcken stream, Sigridsrönningen and Gammbyn, exhibited negative δ34SSO4 values that ranged between −4.5 and −0.5‰ and the δ34SSO4 values with the highest values during peak flow (Figure 3). The δ34SSO4 values in the upper sampling site at Flarkbäcken (Pellboda) and Svartberget, ranged between +5.4 and +3.8‰ and their values decreased over time during the snowmelt period (Figures 3 and 4) . The δ34SSO4 values at Flarkbäcken (Pellboda) and Svartberget were always lower than the δ34SSO4 values measured in the snow from the two sites and the difference increased with increasing stream discharge.
 The δ34SSO4 values in the remaining four streams were always higher than the snow δ34SSO4 values (Figure 4) at the start of the measurements. Under base flow conditions, δ34SSO4 values exceeded +10‰ in the three streams: Stridbäcken, Sörbäcken and Mellansjöbäcken. A maximum value of +8.6‰ was recorded in the remaining stream, Lillån, also during base flow. The δ34SSO4 values decreased with increasing stream water discharge, with the lowest δ34SSO4 values typically observed during peak flow. These values were also close to the δ34SSO4 values of snow. Large shifts in δ34SSO4 values during snowmelt were observed in the three streams Stridbäcken, Sörbäcken and Mellansjöbäcken; the δ34SSO4 value decreased almost 5‰ during the spring snowmelt episode. The smallest change was found in Pellboda (1‰), while the change in the other two streams was between 2 and 3‰.
 There was a clear decrease in the δ34SSO4 values with an increasing percentage of cultivated areas in the Flarkbäcken catchment represented by the two downstream sampling sites (Figure 5a). No significant relationship between percentage peatland and stream water δ34SSO4 values was found (Figure 5b).
3.2. Stream Water Sulfate Concentrations and Export
 There was a large variation in sulfate concentrations in the six streams sampled. The highest concentrations were found in the outlet of the Flarkbäcken catchment (Sigridsrönningen), with concentrations ranging between 440 and 3375 μmolc dm−3 (Figure 3). The associated sulfate concentrations in the upstream Flarkbäcken location at Pellboda, ranged between 50 to 150 μmolc dm−3 and increased downstream toward the outlet at Sigridsrönningen (Figure 3).
 At all sites sulfate concentrations were highest during base flow or the start of the spring snowmelt, and decreased during the snowmelt episode. The largest dilution effect was found at the two downstream sites in Flarkbäcken (Gammbyn and Sigridsrönningen) where the SO42− concentration was diluted by a factor of 5 and 8, respectively. The average dilution for the other streams was 2.3, ranging between 1.5 and 3.2. Stream water Cl concentrations were generally below 100 μmolc dm−3 (data not shown) except for Lillån, were the highest Cl concentration was 294 μmolc dm−3 at the start of the spring snowmelt. Lillån also had the highest Cl/SO4 molar ratio with a maximum value of 3.
 The total S export during the spring followed in general the pattern of SO42− concentrations with the highest export of 16.4 kg S ha−1 during the month of spring from Sigridsrönningen (Table 1). The lowest export of 0.74 kg kg S ha−1 during spring was from Sörbäcken. The estimated S contribution from direct snowmelt (and rain), following the method proposed by Laudon et al. [2004b], was always less than 30% (Table 1).
Table 1. Sulfate Export (1997) in the Investigated Streams Including Flarkbäcken Subcatchments
3.3. Relationship Between SO42− Concentrations and δ34SSO4
 Linear relationships were generally observed between 1/SO42− and δ34SSO4 in snow and stream water samples, however, the nature of the relationship, as indicated by the intercept, varied. In the snow, the intercept was +4.9‰ (Figure 6). The streams showed a considerable variation, for instance, the intercept for Gammbyn and Sigridsrönningen, considered jointly, was −5.1‰, whereas the intercept for the three streams Stridbäcken, Sörbäcken and Mellansjöbäcken (considered as group), was +11.7‰ (Figure 7). The intercepts for the two stream sites at Pellboda and Svartberget were, +5.3‰ and +5.9‰, respectively, while the intercept for Lillån was +8.4‰ (Figure 7).
 Our results clearly demonstrate that the origin of sulfate in stream water can vary considerably within the landscape and during hydrological episodes such as the spring snowmelt. We suggest that these differences reflect: (1) differences in current land use, (2) the influence of anthropogenic deposition; and (3) natural processes related to the relative influence of areas with reducing conditions such as wetlands and riparian zones. The relative contribution of especially the two latter is strongly affected by the hydrological flow path during the spring snowmelt. Below we address how shifts in sulfate sources relate to changes in hydrology.
 One of the more striking differences was the shift in stream water δ34SSO4 values in the Flarkbäcken catchment moving downstream, encompassing an almost 10‰ variation in δ34SSO4 values. The downstream decrease in δ34SSO4 values and increase in stream water SO42− concentrations was closely related to the increase in agricultural activities and cultivated areas that are dominated by fine-textured acid sulfuric soils [Carlsson, 2003]. Cultivation and artificial draining of these soils results in the oxidation of reduced S and will cause acidification and elevated SO42− concentrations in stream waters [Åström and Spiro, 2005]. The isotope value of sedimentary sulfides has been reported to range between −48 and −5‰ [Knöller et al., 2005] in sedimentary soils in Germany although Åström and Spiro  also show that higher values may occur in acid sulfuric soils. Isotope fractionation during oxidation of sulfides is either nonexistent or minor [Nakai and Jensen, 1964; Toran and Harris, 1989] and the isotope value of the sulfides and the dissolved SO42− produced during oxidation are likely to correspond closely.
 The linear relationship between the SO42− isotope values and the inverse of the sulfate concentrations (Figure 7a) suggests that there is one dominant sulfur source in the lower reaches of the Flarkbäcken system. The intercept of −5.1‰ is similar to the isotope value reported for HCl-extractable S in an acid sulfuric soil in northern Sweden [Torssander, 1996]. This source dominates the stream water S export, especially during base flow conditions. Similar low stream water δ34SSO4 values have also been reported by Åström and Spiro , who investigated a stream on the Finnish side of the Bothnian Bay where acid sulfuric soils also have a marked influence.
 The drastic decrease in SO42− concentrations in the two downstream Flarkbäcken sites and the simultaneous increase in δ34SSO4 values during snowmelt shows that there is a source with a higher δ34S value diluting the sulfide signal. We applied a two-source mixing model (equation (3)) with one end-member represented by the intercept of the regression (−5.1‰, Figure 7) and the other end-member as the input of the upper reaches represented by Pellboda (+3.9‰; lowest value during spring flood). The resulting mixing model suggested that 70% of the sulfur in the stream water measured at the outlet at Sigridsrönningen has its origin in the acid sulfuric soils during the spring snowmelt. A similar estimate for base flow conditions suggested that more than 90% of the S export comes from the acid sulfuric soils which are thus very important point sources for the S export.
 The sulfate contributions from snow and forested areas of the catchment are shown to be small and thus cannot explain the observed change in the isotope value during the snowmelt episode, despite the sevenfold decrease in sulfate concentrations at Sigridsrönningen. The sulfate concentrations are still about 9 times higher at Sigridsrönningen than at Pellboda and we suggest that a more superficial flow of water during the snowmelt results in increased drainage of the cultivated layer of the farmland. This layer has lost part of its sulfur because more permanent oxidized conditions and sulfate concentrations are generally lower then the underlying B horizon [Öborn, 1989; Åström, 1998].
 The Pellboda (upper reaches of Flarkbäcken) and Svartberget sites were similar in that the stream water δ34SSO4 values (+4.9 and +5.4‰, respectively) decrease during the spring episode to a stream water value of about +3.9‰ at high flow. The intercept of the regression in Figures 7b and 7c for Pellboda and Svartberget were also similar, +5.3 and +5.9‰, respectively, and the S isotope value during base flow was similar to that of the snow δ34SSO4 (+5.9‰). The anthropogenic deposition in this region of Sweden is lower compared to the southern parts of Sweden, about 1.5 versus 9 kg S ha−1 a−1 respectively. The marine influence is, however, also minor [Ingri et al., 1997; Novák et al., 2003] and the isotope values of rain and snow are similar to S isotope values reported from southwestern Sweden [Carlsson et al., 1999; Mörth and Torssander, 1995]. This was also apparent in the δ34SSO4 values of the snow in the current study: the intercept of +4.9‰ is close to the anthropogenic end-member reported from southwestern Sweden around +4.3‰ [Mörth and Torssander, 1995]. The snow δ34S values in this study suggest that the marine influence is less than 10% assuming an anthropogenic end-member of +4.9‰ and a marine of +20‰ (Figure 1).
 The base flow values may reflect anthropogenic inputs: most likely indirectly via desorbing soil SO42− with an anthropogenic signature [see, e.g., Novák et al., 2005]. Isotopic values reported from weathering of granitic bedrock in northern Sweden were between −0.5‰ and 2‰ and would exclude weathering release from granite as a major S source [Torssander et al., 2006]. Pyrite-containing sedimentary gneiss can also be found in till soils especially above the highest coastline [Torssander, 1996]. The δ34S values varied between −2.85‰ and 1.79‰ i.e., similar to values reported by Torssander et al. . Larger variations in bedrock δ34S values have been observed by Bailey et al. . We should not exclude the possibility that there may be an influence of bedrock weathering, however, this would require bedrock δ34S values higher than those reported from this region so far.
 The decrease in stream water δ34SSO4 values during the spring snowmelt suggests that a source with a lower δ34S value is influencing the S export in Pellboda and Svartberget catchment areas and excludes SO42− accumulated in the snow. Studies on the hydrological pathways in the Svartberget catchment suggest that the riparian zone has an increased influence during episodic events and that the flow is more superficial in the soil [Laudon et al., 2004c]. A soil profile from a nearby site had extractable sulfate δ34SSO4 values that decreased with depth from +5.4 in the surface layer to +4.1‰ at a depth of 0.2 m [Novák et al., 2003]. The streambed in the Svartberget catchment consists of an 0.5–0.8 m thick Sphagnum-covered humus/peat layer [Bishop et al., 2004] and reducing conditions are likely during part of the year. Bacterial dissimilatory sulfate reduction (BDSR) may thus occur in the streambed, as has been reported from other studies [Eimers et al., 2004; Mörth et al., 1999]. The BDSR is strongly fractionating [Bottrell and Novak, 1997; Groscheova et al., 2000] resulting in a product enrichment of 32S (sulfides) and possibly organic sulfur components [Novák et al., 2005; Mandernack et al., 2000]. Upon reoxidation, isotopically light SO42− can be released, especially after drought periods [Schiff et al., 2005; Eimers et al., 2004; Mörth et al., 1999]. Eimers et al.  reported values between 0 and +4.4‰ in the streambed S from an upland-draining stream with stream water δ34SSO4 values similar to ours. A similar δ34SSO4 value for our streambed could explain the decrease in stream water δ34SSO4 that we observe. The most likely scenario is that reoxidation occurs in autumn under dry and warm conditions and the sulfate subsequently is flushed into the stream water during the spring discharge.
 The base flow δ34SSO4 values of between +9 to +12‰ in the remaining four streams Sörbäcken, Mellansjöbäcken, Stridbäcken, and Lillån suggest that an isotopically heavier source affects the S export in these catchments compared to the other studied streams. All four catchments contain various amounts of peatlands ranging from 14 to 40% of the total catchment area (Table 2). The intercept in Figure 7d suggests that an end-member of +11.7‰ for the three streams Sörbäcken, Mellansjöbäcken and Stridbäcken and +8.4 for Lillån (Figure 7e). Sulfate in peatland pore water will be enriched in 34S since the use of SO42− in BDSR is strongly fractionating (see above). For instance, Mandernack et al.  reported δ34SSO4 values of +13.6‰ in peat pore water. Similarly, Ingri et al.  reported values ranging from +7.1 to +13.9‰ (median +11.2‰) in water from five peatlands in northern Sweden. The higher δ34SSO4 values are always associated with higher stream water SO42− concentrations and base flow conditions. During the BDSR in peatlands SO42− concentrations in the soil solution will decrease and the δ34SSO4 value increase. This is likely to occur during optimal conditions for BDSR, i.e., during conditions of high carbon supply and anoxic conditions. Under these conditions the peatlands are most likely sinks for deposition S. Our data do, however, indicate that there is loss of S during base flow conditions (Table 1). A BDSR in an open system of only 20% of the deposited S during base flow conditions would require a fractionation factor of 25‰ (assuming a δ34SSO4 of 5.5‰ for the deposition S) to get a δ34SSO4 value 12‰. Such a fractionation factor is not unlikely [Krouse and Grinenko, 1991] and could thus explain why we see the higher δ34SSO4 values in many of the streams reported here and in the other studies from northern Sweden. The decrease in SO42− concentration during the spring flood is just a dilution effect and is decoupled from the processes in the peatland. Another possibility is a marine influence, since this also would give a higher δ34SSO4 value. Two of the streams, Lillån and Stridbäcken, drain into the Gulf of Bothnia (Figure 2). A marine influence would be indicated by Cl/SO4 ratio of about 20, however, the highest ratio value found was 3 during base flow in the Lillån stream.
Table 2. Selected Information on the Six Streams Sampled
Catchment Size, km2
Annual Discharge, mm a−1
 The shift in δ34SSO4 values from around +12‰ to +6‰ is indicates a shift from peatland influenced S export to another S source. The isotope value of +6‰ is close to the snow values measured. The contribution of snow S is small, however (Table 1), and another source must be involved. We suggest that a more likely source is desorbing sulfate or mineralized S from the O horizon in the forest. Mineralization of organic S plays a crucial role in the reversing of acidification and has been shown to have a large impact on the δ34SSO4 values in runoff [Mörth et al., 2005; Giesler et al., 2005; Novák et al., 2005]. The shift in δ34SSO4 values may thus be indicative of a change in the hydrological pathway with a larger influence from the forested areas during the spring snowmelt. Previous studies from S deposition affected areas have shown that adsorbed SO42− in the mineral soil and SO42− released during mineralization in the O horizon has an isotope value reflecting the anthropogenic value [Giesler et al., 2005]. Applying the mixing model and assuming that the end-member for the forest soil S is similar to the anthropogenic value (around +5‰) suggests that Stridbäcken switches from peatland S (85%) during base flow conditions to forest S (89%) during spring flood conditions. The contribution from the snow S in stream water is estimated to 29% in Stridbäcken (Table 1) and the contribution from forest soil S is thus 60%. Estimates from the other three streams suggest similar drastic shifts in S source during the spring flood. Even if the forest S end-member is set to the lowest forest stream δ34SSO4 value measured (+4‰) and the peatland end-member to +13.6‰ [Mandernack et al., 2000] the change will still be large. For example, Stridbäcken would shift from 70% peatland S to 79% forest S, with the latter including the snow contribution.
 Notably, the differences in δ34SSO4 values are not revealed by either position in the landscape or peatland coverage in the catchments (Figure 5b). Three of the catchments, Svartberget, Sörbäcken and Mellansjöbäcken, are situated in the inland of the Västerbotten County, whereas Lillån and Stridbäcken are coastal streams. The former three are situated above the highest coastline on till soil, whereas the latter two are below the highest coastline and contain postglacial sediments. Despite these large differences, Mellansjöbäcken and Stridbäcken show a similar stream water δ34SSO4 pattern. It seems likely that our δ34SSO4 patterns are not linked to geomorphologic features of the catchments. The influence of the peatlands seems to be a common feature for four of the streams although percentage peatland coverage was not a good predictor for its influence on stream water SO42− in this study. Earlier studies of δ34S from small forested catchments have also shown similar patterns to our observations with an increase in δ34S values during low discharge and a decrease during episodic events [Andersson et al., 1992; Hesslein et al., 1988].
 This study shows large regional differences in stream water S sources followed by large shifts during the spring snowmelt. These differences are not easily interpreted without isotopic measurements. One exception is the influence of acid sulfuric soils which can generate extremely high stream water SO42− concentrations. We identified two stream water SO42− sources; sedimentary sulfides and anthropogenic S. Bacterial dissimilatory sulfate reduction was also identified as an important process affecting stream water δ34SSO4 values and suggests that peatlands and possibly riparian zones have a large influence on the biogeochemistry of SO42−. Landscape features, such as the percentage peatland in a catchment, do not seem to be conclusive when estimating the influence these ecosystems have on S export. Regional assessments on the impact of anthropogenic inputs of S, and the processes controlling stream water export of S, should include isotopic studies to better understand S dynamics.
 Funding for R.G. was provided by the Swedish Research Council, contract 2003-2713.