Effects of forestry operations on dissolved organic carbon concentrations and export in boreal first-order streams



[1] The purpose of this study was to quantify the effects of clear-cutting and site preparation on dissolved organic carbon (DOC) concentrations and export in four boreal headwater streams in northern Sweden. The data set included intensive stream water monitoring from 2 years of pretreatment conditions (2004–2005), a 2 year post-clear-cut period (2006–2007), and a 2 year period after site preparation (2008–2009). To investigate differences in [DOC], an analysis of variance on ranks was performed on the data sets. Clear-cutting increased the median DOC concentrations significantly from 15.9 to 20.4 mg L−1, which represents a net increase (treatment versus control) of 3.0 mg L−1 in the 2006–2007 period. Site preparation had an even more profound effect on DOC levels; an increase from 20.4 to 27.6 mg L−1was found in the site-prepared catchments, whereas the control sites increased slightly from 17.4 to 21.4 mg L−1during the wetter years of 2008–2009. Riverine C fluxes increased significantly by 100% after clear-cutting and by 79% after site preparation (92% and 195%, respectively, if compared to pretreatment conditions). When comparing these yearly C fluxes (183 kg C ha−1 yr−1after clear-cutting; 280 kg C ha−1 yr−1after site preparation) to the net ecosystem exchange (NEE) of a forest in the region, the DOC flux represented 10% of NEE before harvest, increased to 18% after the clear-cut, and increased to 28% after site preparation. These results underline the large impact of forestry operations on stream water quality as well as DOC exports leaving managed boreal forests.

1. Introduction

[2] Increased leaching of terrestrial carbon into surface waters has been observed in many regions of the temperate and boreal climate [Erlandsson et al., 2008; Evans et al., 2005; Monteith et al., 2007]. The major sources of this transported carbon (C) are the catchment soils. The mobilization of C from the terrestrial into the aquatic ecosystem is in particular critical in boreal watersheds where, on a global scale, large amounts of C are stored in soil C pools [Batjes, 1996; Jobbágy and Jackson, 2000].

[3] A large interest in stream water dissolved organic carbon (DOC) concentrations arises from the fact that DOC plays an integral role in the biogeochemistry and ecology of surface waters across forested regions of the world. DOC directly affects the food web structure in lakes [Jansson et al., 2007], exerts control on the acid-base chemistry of surface waters [Buffam et al., 2008], and influences metal export and speciation in streams and rivers, such as the DOC-related mobilization of aluminum [Cory et al., 2006] as well as the strong coupling of DOC and total mercury concentrations in stream water [Dittman et al., 2010; Grigal, 2003]. Furthermore, [DOC] also plays an important role in catchment carbon budgets [Dinsmore et al., 2010; Nilsson et al., 2008; Roulet et al., 2007].

[4] For the mobilization of DOC from terrestrial to aquatic ecosystems the role of headwaters is especially relevant [Bishop et al., 2008; Temnerud et al., 2007]. Effects taking place on this comparably small scale can alter larger downstream ecosystems [Laudon et al., 2011; Ågren et al., 2007]. First-order streams are also often neglected in larger-scale carbon models, even if they represent a major fraction of the total stream length [Bishop et al., 2008].

[5] In first-order streams DOC mobilization is assumed to be primarily controlled by runoff generation; the hypothesized mechanism of “transmissivity feedback” [Bishop et al., 2004; Kendall et al., 1999] in which the lateral flow of saturated soil water to the stream rises to the more superficial soil layers during high-flow conditions. This process is directly associated with DOC mobilization from riparian soils, which often have decreasing DOC concentrations with soil depth [Bishop et al., 2004; Seibert et al., 2009] so that hydrological activation of more superficial soil layers enhances DOC transport [Sebestyen et al., 2008]. Additional variation in stream DOC concentrations which cannot be explained by the variation in groundwater (GW) tables in riparian soils and wetlands are often related to climatic controls such as air temperature [Boyer et al., 1996; Köhler et al., 2009; Sebestyen et al., 2009], soil temperature as well as variation of soil frost in riparian wetlands [Haei et al., 2010; Ågren et al., 2010].

[6] Forestry is of strong economic interest in Sweden. Forestry accounted for 11.1% of the industrial production in 2007, which equals 2.7% of the Swedish “Gross National Product” in the same year [Swedish Forest Agency, 2010]. However, at the same time as forestry is important for the economy, forestry causes a large disturbance of the forest ecosystems which alters the biogeochemical processes and nutrient cycling of these ecosystems [Kreutzweiser et al., 2008]. Forest harvesting is known to influence the microclimate of forests with higher snow accumulation [Murray and Buttle, 2003; Sørensen et al., 2009b] and lower evapotranspiration. This often results in higher GW levels [Pothier et al., 2003] and more direct runoff components during the spring flood [Monteith et al., 2006] which may facilitate transport of DOC to surface waters [Kreutzweiser et al., 2008].

[7] Final harvesting is typically assumed to be followed by increased nutrient availability in soil water [McLaughlin et al., 1996], which can be attributed to increased litter inputs from logging residues [Hyvonen et al., 2000] as well as increased decomposition of soil C (as, e.g., hypothesized by Lamontagne et al. [2000]). In turn, increased concentrations in soil solutions will increase the mobility of nutrients into surface waters [Kreutzweiser et al., 2008; Lamontagne et al., 2000]. As a result, DOC concentrations are observed to increase after harvesting in numerous types of streams in northern catchments [Lamontagne et al., 2000; Laudon et al., 2009; McHale et al., 2007; Zoltai and Martikainen, 1996] as well as after site preparation [Mannerkoski et al., 2005; Piirainen et al., 2007], which is commonly performed in boreal regions before replanting. Increased DOC concentrations attributed to forestry are particularly prominent in landscapes with peat rich soils, such as drained peatlands [Mannerkoski et al., 2005; Nieminen, 2004].

[8] However, stream water DOC concentrations at a larger landscape scale are typically found to be strongly controlled by different landscape units, which are (1) uplands, (2) wetlands, and (3) open water bodies [Schelker et al., 2011; Ågren et al., 2007], and forestry effects at these larger scales are difficult to assess [Mallik and Teichert, 2009] if the small-scale impacts are not well quantified. It is therefore crucial to quantify the effects of forestry operations on a small, but at the same time representative scale, which will allow upscaling to the landscape level, where differential gain and loss processes of DOC may take place [Cole et al., 2007].

[9] Overall the DOC export from terrestrial to aquatic ecosystems after logging is dependent on two major mechanisms [Kreutzweiser et al., 2008]. First, changes of factors controlling soil solution concentrations of DOC which, if increased, will also increase the export from these soils. Second increased water fluxes may also increase the nutrient export simply by increasing the transport after clear-cutting. A third factor will be that much of the increase in water export will occur along more superficial flow paths where the concentrations of DOC are higher. These mechanisms can act concurrently to increase the overall impact on downstream water quality.

[10] Despite the importance of boreal ecosystems for carbon storage, the economical role of timber production and the high sensitivity of aquatic ecosystems to factors regulating DOC concentrations, little has been done to quantify the direct effects of the physical anthropogenic disturbance with the highest areal coverage; the effects of commonly used forestry operations on stream water DOC concentrations. We define forestry operations here as two operations with a major impact on catchment soils: (1) clear-cutting and (2) site preparation. The objective of this study was therefore to study how these forestry operations influence the dynamic of DOC concentrations as well as the total DOC export from a boreal catchment in northern Sweden.

[11] We hypothesize that (1) clear-cutting of a boreal forest increases DOC concentrations in stream water significantly compared to reference sites, (2) even higher stream DOC concentrations will occur after applying common site-preparation techniques (disk trenching), owing to the additional soil disturbance, and (3) increased stream DOC concentrations as well as increased water fluxes along more superficial flow paths from treated areas will increase the total riverine organic C export as both concentrations and water fluxes act simultaneously.

2. Site Description

[12] The Balsjö paired catchment study (277 Balsjö) in northern Sweden (N 64° 1′ 37″ E 18° 55′ 43″) was set up to investigate the effects of forestry operations on water quality [Löfgren et al., 2009]. Within the first stage of the experiment it was shown that clear-cutting has a direct effect on hydrology [Sørensen et al., 2009b], major ion concentrations in stream water [Löfgren et al., 2009] as well as leaching of DOC during the summer season [Laudon et al., 2009]. In contrast, mercury mobilization showed small changes after clear-cutting [Sørensen et al., 2009a].

[13] The experimental set up is located approximately 70 km west of the Baltic Sea coast. The landscape consists of low-relief river valleys and hilltops. It is also marked by features from the last glaciation. In some places bedrock, which typically consists of pegmatite with aplitic granite and aplite, can be found at the soil surface. Soils are typically till dominated, and are well drained. In valley bottoms, riparian wetlands with organic rich peat soils can be found along the streams. In addition large amounts of peat are located in mires, often in nearly flat areas. Such mire peatlands account typically for approximately 25% of the area in this region [Ågren et al., 2008].

[14] The study area consists of four headwater catchments (Figure 1) that drain into the River Balån. This study is focusing on these four headwater catchments in which the stream monitoring began in April 2004 in all streams. One of the two treatment catchments was a clear-cut with a small (>5 m width on each stream side) discontinuous buffer strip of trees left along the stream (BF-5), and the other was a clear-cut with no riparian buffer (CC-4). The two reference sites (Table 1) are of different size: whereas the southern one is larger (RF-3), the northern one (NR-7) is small with more wetland influence and drains into the BF-5 catchment [Laudon et al., 2009]. Clear-cutting took place in spring 2006. To avoid driving damage, this was performed on frozen soils and driving within a 10 m zone along streams was avoided. All clear-cut sites were site prepared by disk trenching in May 2008 (seeLöfgren et al. [2009] for more information).

Figure 1.

Map of the 277 Balsjö paired catchment experiment located in northern Sweden. The two treated catchments are named CC-4 (the site which was harvested entirely to the edge of the stream) and BF-5 (the one harvested leaving a discontinuous riparian buffer strip). The two control sites are named RS-3 (the southern, larger control catchment) and NR-7 (the northern reference area). Thin black lines represent the stream network, and thick black lines represent the catchment boundaries; black dots indicate the location of sample drawing as well as the location of stream gauges for the CC-4, BF-5, and NR-7 sites, respectively.

Table 1. Catchment Areas of the Four Headwater Streams Used in This Study in the 277 Balsjö Experiment
NameSite NameArea (ha)Area HarvestedRiparian Buffer
Southern clear-cutCC-440.563.7%No, harvested to stream
Northern referenceNR-724.2<5.0%
Northern clear-cut (buffer)BF-515.688.2%Discontinuous, ∼10 m wide at each side
Southern controlRF-3156.2<8.0%

[15] Vegetation in all catchments is typical for the boreal regions in Scandinavia. Norway spruce (Picea abies) dominates in the lower and middle elevations, while Scots pine (Pinus sylvestris) can be found on higher locations with drier soils. In wetter areas, as, for example, along stream channels, birch (Betula sp.) is also present. The ground vegetation mainly consists of herbs and shrubs such as wood sorrel (Oxalis acetosella), common cow wheat (Melampyrum pratense), bilberry (Vaccinium myrtillus), and cowberry (Vaccinium vitis-idaea). Closer to the stream, cloudberry (Rubus chamaemorus) can also be found. In the harvested catchments where the forests have been clear-cut, grasses such as wavy hair grass (Deschampsia flexuosa) have become abundant but cowberry as well as heather (Calluna vulgaris) are also widespread. Naturally regenerated seedlings of spruce, pine, birch, and some sporadic aspen (Populus tremula) were found throughout the harvested catchments prior to ground scarification. After the site preparation the ground vegetation was strongly disturbed and full ground cover was not prevalent at the time of this study, even if pine tree seeds (Pinus sylvestris) were planted during the scarification process to promote forest regeneration.

3. Methods

3.1. Stream Sampling

[16] During the pretreatment period (2004–2006) as well as for the first year after the harvest, stream samples were taken as grab samples at a 2 week interval (regular sampling) from early spring to late fall. During winter (typically November–March), when cold winter conditions lowered variations in stream flow, the sampling interval was lowered to once a month. In 2007 the sampling frequency was increased by adding four ISCO automated samplers to the regular sampling scheme. The samplers were typically set up in early spring and operated over the entire growing season. ISCO sampling frequencies range from a minimum of 12 h in 2007 and 2008 to an 8 h interval in 2009 during high-flow conditions. Low-flow sampling frequencies were set to a minimum of one sample every 2 days during the 2007–2009 growing seasons.

[17] In addition a sampling of riparian soil water was performed in August 2009. Samples were collected along three different transects located within the BF-5 catchment by preemptying the PVC piezometer tubes two days before sample collection. Samples were then taken using a peristaltic pump.

[18] All samples were stored cool (<7°C) and frozen within 2–3 days after sampling. Analyses were performed using a Shimadzu TOC 5050 for all samples from 2004 to 2006 and a Shimadzu TOC-VCPH after 2006. Even if the sample analysis was performed without a filtering step, the measured concentrations were assume to represent DOC concentrations, because particulate fractions represent typically less than 3% of the total organic carbon (TOC) content in stream water in this region [Laudon et al., 2011]. Even if the sample uncertainty was quantified as <5%, we assume an overall sample uncertainty of 5% for all DOC samples to cover other eventual uncertainties.

[19] Similarly to a previous study [Laudon et al., 2009], we assume that the thin, discontinuous buffer strip along the BF-5 stream is negligible from a DOC perspective. Therefore we assume the two reference sites (RF-3 and NR-7) as well as both treated sites (CC-4 and BF-5) to act as replicates. Differences in concentrations between treated and control sites are then expressed as the difference in average concentrations according to

equation image

Differences in specific discharge (ΔQspec) as well as differences in soil temperatures (Δ Soil Temp) are defined similarly to ΔDOC, as the difference in averages for each time step.

3.2. Hydrometric Measurements

[20] Hourly discharge time series were calculated from water levels measured at each of the catchments except the RS-3 catchment, where no gauging station is present. Rating curves were based on earlier versions given bySørensen et al. [2009b], which were then updated with numerous water level and discharge measurements performed during 2008–2009. A more detailed description of hydrological measurements and the quantified hydrological changes in the pre-clear-cut and post-clear-cut periods, respectively, is given bySørensen et al. [2009b].

[21] Soil temperatures were measured at an hourly resolution by using temperature probes connected to a Campbell Scientific data logger system (TOJO Skogsteknik, Sweden). Probes were placed at different soil depth (15 cm, 30 cm, and 50 cm) at two different sites on a western slope, one within the clear-cut area (BF-5 catchment) and the other one in the control forest (NR-7 catchment). Both sites have a distance of approximately 40–50 m from the stream.

[22] Catchment areas used in our specific discharge and load calculations, respectively (Table 1), vary slightly from those used in earlier studies at the 277 Balsjö sites [Laudon et al., 2009; Löfgren et al., 2009; Sørensen et al., 2009b]. These areas were recently updated on the basis of a combination of additional field observations, updated digital elevation data from airborne laser scanning (LIDAR) as well as water-balance comparisons.

3.3. Flux Calculations

[23] Fluxes were based on daily average concentrations and hourly discharge values which were derived from the hourly water level data. Concentrations between the sampling intervals were interpolated linearly, if the sampling frequency was below one sample per day; average values were used for high-flow samplings. For parts of the pretreatment period (2004–2005) where discharge data is missing for the control sites, “gap filling” was performed on the flux time series of the control sites, by using the specific export of the entire northern catchment, which is the combined outlet of the NR-7 and the BF-5 catchments together.

[24] Uncertainties in flux calculations were based on a Gaussian error propagation. By assuming all uncertainties to be normally distributed and noncorrelated, the general form of this error model is given as

equation image

with sf2 being the variance of the output of the model f and xi being the variable i (i = 1…n) with the variance of si2. For two correlated variables this model extends to equation (3) given by Beven [2008]:

equation image

with ρij being the correlation coefficient and sxixj2 being the covariance of the two linearly dependent variables xixj, respectively.

[25] Uncertainties in discharge are not often found to be normally distributed. Therefore we based our error model on the general form of the flux equation in the logarithmic scale (equation (4)), in which the discharge log(Q) is given by log transformation of Q, which is calculated from a two-parameter rating curve of the formQ = a hb with h being the water level and a and b being the parameters. This type of rating curve has been applied earlier in the 277 Balsjö catchment, and residuals of log(Q) calculated with an uncertainty model based on the given rating curves did not show any indication of nonnormality or heteroscedasticity for any of the gauged sites [Sørensen et al., 2009b]:

equation image

with Ef being the specific export of a substance with the concentration C and A being the catchment area. By assuming all terms of equation (4) to be relevant in equation (3), a given linear relationship between log(Q) and log(C), and log(A) being an independent variable, the resulting error model is given as

equation image

with slogEf being the resulting standard deviation of log(Ef). Uncertainties were calculated for each time step of the flux time series to predict confidence intervals. Uncertainty intervals of total annual exports were calculated by averaging the relative uncertainties, retransformation into nonlog space and then multiplying with the total summed export. Uncertainties of input variables were set to slogC = 5% for DOC concentrations (see also section 3.1), which is a common estimate [Ågren et al., 2007]. Catchment areas were estimated to have an uncertainty (SlogA) of 2.5%; uncertainties of log(Q) (slogQ) were set to 5%, which is also a common estimate for the Balsjö catchments [Sørensen et al., 2009b].

3.4. Statistical Analysis

[26] For all four sites DOC concentrations were determined from analysis results. If several samples were taken within one day (e.g., during summer events 2009, when ISCO units were set to an 8 h sampling interval) daily DOC concentrations were calculated as the average of all samples taken during this day. Concentrations for the buffer catchment (BF5) were calculated on the basis of a mass balance approach similar to the calculation used by Laudon et al. [2009]. Average DOC concentrations for the two treated sites (CC-4 and BF-5) as well as the two reference sites (RF-3 and NR-7) were then calculated as the average for the treated and control sites for this day, respectively, and were only used in the statistical analysis if a pair of values (one value for controls and one for the references, respectively) was available.

[27] Within each of the three time periods evaluated in this study (pretreatment period, after clear-cutting, and after site preparation), each of the data sets was tested for normality (Shapiro-Wilk) as well as for equal variance. Whereas none of the data sets were found to be normally distributed, all data sets passed the test for homoscedasticity. To further compare the three different time periods, we used a combination of a nonparametric Kruskal-Wallis one way analysis of variance on ranks (ANOVA-R), which was combined with Dunn's test to investigate significant differences in median values. This combination allows analyzing nonnormally distributed data sets with different sample sizes. All statistical analyses and calculations were performed by Sigmaplot 11.0 (Systat Software, Inc.) or Matlab 7.8 (Mathworks Inc.). To investigate the effects of each treatment, one ANOVA-R was performed for all possible combinations (clear-cut versus pretreatment, site preparation versus pretreatment, site preparation versus clear-cut).

4. Results

4.1. Changes in DOC Concentrations

[28] DOC concentrations showed a strong annual variation in all catchments (Table 2), primarily controlled by hydrological conditions. The median values of the pretreatment period were quantified as 15.9 mg L−1 for both the treated and the control catchments (Figure 2). Median concentrations increased after the clear felling operation performed in spring 2006 to a concentration of 20.4 mg L−1 in average for the treated catchments, whereas the median of the control for the same time period was 17.4 mg L−1. Site preparation resulted in an even larger effect on stream DOC concentrations. An increase in the control concentrations to a value of 21.4 mg L−1was observed, whereas the median for the site-prepared catchments was as high as 27.6 mg L−1.

Table 2. DOC Concentrations of the Four Headwater Streams of the 277 Balsjö Experiment During the Three Different Treatment Periodsa
TreatmentTime PeriodRF-3CC-4Northern CatchmentNR-7
  • a

    The “Northern Catchment” represents the first-order stream draining sites BF-5 and NR-7 together (see alsoFigure 1). Concentrations of the BF-5 site used for statistical analysis are calculated on the basis of a mass balance approach. Values are given as average concentrations (mg L−1) followed by their standard deviations in parentheses (mg L−1) and the number of samples taken (n).

PreharvestApril 2004 to March 200616.1 (5.9); n = 10218.3 (9.5); n = 11418.2 (7); n = 11218.6 (8); n = 114
Clear-cutApril 2006 to May 200816.7 (7.1); n = 22621.6 (10.5); n = 21521.4 (9.3); n = 20921.5 (10.5); n = 210
Site preparationMay 2008 to October 200921.0 (7.1); n = 21931.5 (13); n = 26527.0 (9.2); n = 25925.8 (11.1); n = 247
OverallApril 2004 to October 200918.3 (7.2); n = 54725.4 (12.8); n = 59423.2 (9.5); n = 58022.8 (10.7); n = 571
Figure 2.

Box plot of mean DOC concentrations in stream water for the two treated sites and the two control sites for the different time periods during which forestry operations were performed at the Balsjö paired catchment. Gray boxes represent the 25th to 75th percentiles, and error bars indicate the 10th to 90th percentiles. Black dots indicate the 5th to 95th percentiles. The median is shown as a horizontal line. Sample sizes are given as npre-treatment = 87, nclear-cut = 342, and nsite-prep = 388, with nbeing the sample size of control plus the treated catchments for the pretreatment, clear-cutting, and site-preparation periods, respectively.

[29] The result of the statistical analysis confirmed the pattern visible in Figure 2. The ANOVA-R for comparing the period after clear-cutting to the pretreatment period showed that the change in concentration for the treated sites was significantly higher compared to all sites before clear-cutting as well as to the reference sites after cutting (Table 3). When comparing the site-preparation period with the pretreatment period the same pattern was generally valid. Significant increases of the DOC concentrations at the treated sites compared to the reference was observed, as well as compared to both sites in the pretreatment period. There was, however, also a significant increase in concentrations at the control sites after site preparation compared to both areas in the pretreatment period. This change may be addressed to additional drivers such as the year-to-year variation in environmental conditions, as observed variations in precipitation and therefore also in stream runoff.

Table 3. Statistical Analysis of All Multiple Pairwise Comparisons for Treated and Untreated Catchments for All Three Different Time Periods
ComparisonDifference of RanksQP < 0.05
Clear-Cut Period Versus Pretreatment Period
BF-5 + CC-4, clear-cut versus RS-3 + NR-7, before76.4563.642Yes
BF-5 + CC-4, clear-cut versus BF-5 + CC-4, before64.2422.979Yes
BF-5 + CC-4, clear-cut versus RS-3 + NR-7, clear-cut45.4543.375Yes
RS-3 + NR-7, clear-cut versus RS-3 + NR-7, before31.0021.506No
RS-3 + NR-7, clear-cut versus BF-5 + CC-4, before18.7870.887No
BF-5 + CC-4, before versus RS-3 + NR-7, before12.2140.459No
Site-Preparation Period Versus Pretreatment Period
BF-5 + CC-4, site-preparation versus RS-3 + NR-7, before164.1437.22Yes
BF-5 + CC-4, site-preparation versus BF-5 + CC-4, before147.3616.302Yes
BF-5 + CC-4, site-preparation versus RS-3 + NR-7, site-prep.63.0564.524Yes
RS-3 + NR-7, site-preparation versus RS-3 + NR-7, before101.0884.455Yes
RS-3 + NR-7, site-preparation versus BF-5 + CC-4, before84.3053.612Yes
BF-5 + CC-4, before versus RS-3 + NR-7, before16.7830.57No
Site-Preparation Period Versus Clear-Cut Period
BF-5 + CC-4, site preparation versus RS-3 + NR-7, clear-cut225.43510.405Yes
BF-5 + CC-4, site preparation versus BF-5 + CC-4, clear-cut156.3796.868Yes
BF-5 + CC-4, site preparation versus RS-3 + NR-7, site-prep.99.4114.643Yes
RS-3 + NR-7, site preparation versus RS-3 + NR-7, clear-cut126.0245.846Yes
RS-3 + NR-7, site preparation versus BF-5 + CC-4, clear-cut56.9682.513No
BF-5 + CC-4, clear-cut versus RS-3 + NR-7, clear-cut69.0563.015Yes

[30] ANOVA-R results for the site preparation versus the clear-cutting period showed that the site preparation also governed significant increases in DOC concentrations of the treated sites compared to all sites in the post-clear-cut period. At the same time controls also showed significant increases of [DOC] during the post-site-preparation period compared to pretreatment period, which indicates, again the importance of additional factors controlling DOC levels (Table 3).

4.2. Seasonal Variation in DOC Concentrations

[31] DOC concentrations as well as hydrological conditions vary strongly at all 277 Balsjö sites during seasons (Figure 3). Whereas DOC concentrations are normally low during winter low-flow conditions, there is generally an increase to higher values in spring during snowmelt. Highest concentrations are observed during late summer high-flow events, which are often of short duration.

Figure 3.

(top) Time series of DOC concentrations for the two treated sites and the two nontreated sites. (top middle) Differences (ΔDOC) between treated and untreated for each sampling as well as cumulatively daily difference in concentrations. (bottom middle) Specific discharge for the three gauged catchments within the 277 Balsjö paired catchment experiment. (bottom) The variation in specific discharge. Missing discharge data for the sites during the early period is primarily caused by the different timing of the installation of the weirs, whereas later periods of lacking data are mainly caused by malfunction of pressure transducers and data loggers.

[32] Differences in DOC concentrations between the treated areas and the reference sites (ΔDOC) indicate a small variation between the sites during the pretreatment period (Figure 3, top middle). After the clear-cut, the differences are strongly positive during the first and second summer (2006 and 2007, respectively) but decrease during the fall. This pattern was also observed byLaudon et al. [2009], who used a subset of the data set presented in this study to investigate the changes during the summer 2007. During post-clear-cutting winter seasons, the difference is smaller and some negative values can be found. In the period after site preparation (2008 and 2009), however, the spring and summer ΔDOC levels are high with no clear signs of a decrease in the fall, whereas winter differences are close to zero with some lower values in early spring.

[33] Differences in specific discharge (ΔQspec) between the sites were small in the pretreatment period (Figure 3, bottom). Strong differences can be found after clear-cutting: here the response of the first spring flood (2006) is delayed, possibly owing to the effects of snow compaction from harvesting machines (compare also toSørensen et al. [2009b]). However, the period after the first spring flood shows strongly positive differences; a net increase of flows, especially during low-flow conditions, is prevalent [see alsoSørensen et al., 2009b]. The period after the disk trenching is also primarily characterized by positive differences during the snowmelt and summer seasons, whereas there is little difference during winter. Cumulative differences in ΔQ (cum. ΔQspec) indicate the same pattern: whereas the first snowmelt after clear-cutting is negative, there is a steady increase in the following years, with low gradients during winter seasons and higher ones during spring, summer and fall.

4.3. Changes in Riverine Organic C Export

[34] The riverine DOC export in the pretreatment period was quantified as 95 (+39; −26) kg C ha−1 yr−1 for the treated catchments, whereas the flux of 73 (+19; −13) kg C ha−1 yr−1 of the controls was slightly lower during the same period of time (Figure 4); these differences were, however, not significant. After the clear-cutting the fluxes of the treated sites increased to 183 (+75; −50) kg C ha−1 yr−1 for the treated sites, whereas the export of the controls was slightly higher than the pretreatment conditions (92 (+24; −17) kg C ha−1 yr−1), which represented a net increase (treatment versus control during the period after clear-cutting) of 100%. Site preparation gave an even higher increase in fluxes; here the flux was calculated as 280 (+115; −76) kg C ha−1 yr−1, which corresponds to an increase of 79% compared to the export of 157 (+41; −28) kg C ha−1 yr−1 from the control streams during the same period of time as well as an increase of 195% if compared to the pretreatment fluxes of the treated sites. However, the specific DOC transport of the reference sites also varied over time. When comparing the period after site preparation to pretreatment conditions a net increase of 116% was observed, indicating that the interannual variation controlled by factors other than the treatments performed are important.

Figure 4.

Yearly specific riverine organic carbon fluxes of the three different time periods investigated in this study: the pretreatment period (2004–2005), the time after the clear-cut (2006–2007), and the time after site preparation (2008–2009). The average flux of the two treated catchments is shown as a gray bar, whereas the two nontreated catchments are shown as a white bar for each time period, respectively.

5. Discussion

[35] The results of this study show strong increases in DOC concentrations after forestry operations in boreal first-order streams. Changes were quantified to an average increase of 3.0 mg L−1after clear-cutting, and average ΔDOC of 6.2 mg L−1 after site preparation. This corresponds well with results found in other regions of the world as, for example, 2–5 fold increases reported for many North American catchments (see Kreutzweiser et al. [2008] for a summary) or studies of Fennoscandia, where, for example, Nieminen [2004] quantified differences of 8.4 to 22.8 mg L−1 in catchments with different harvest intensities during the growing season. That was the period where the differences are also largest in this study (compare to section 3.4). However, if taking the hydrological variation (Figure 3, bottom) into account, the years with especially wet conditions, such as 2009, seem to promote equally strong variations of [DOC] at the control sites, relative to the extent of increases attributed to clear-cutting effects during the 2006–2007 period. So there is a chance that the effects of site preparation may be superimposed with the effects of clear-cutting if those first became fully apparent during the first wet year in 2009, 3 years after the harvest, but one year after site preparation. Overall the key question which remains from these results is which drivers are causing these increases in [DOC] after clear-cutting and after ground preparation, respectively. We hypothesize that a number of different mechanisms may increase DOC mobilization from soils to streams.

5.1. Drivers of DOC Mobilization

[36] In general, hydrology is assumed to be the primary driver of DOC mobilization in northern Swedish streams [Bishop et al., 2004; Laudon et al., 2011]. The mechanism governing a higher transport of DOC is that increased GW tables increase the lateral flow at higher soil layers, which favors surface near runoff generation. This concept was formalized as the “transmissivity feedback mechanism” [Bishop et al., 2004]. Here strong increases of DOC concentrations in streams during high-flow conditions are caused by the lateral flow through riparian zones, in which exponentially increasing DOC concentrations of riparian soil water toward the soil surface are prevalent. This mechanism is, for example, observed in the nearby Krycklan Catchment [Bishop et al., 1990; Lyon et al., 2011].

[37] In Balsjö, the results from the snapshot sampling for TOC concentrations of riparian soil water performed in August 2009 (Figure 5) suggest that similar exponentially shaped concentration profiles in riparian soils are prevalent (sampling was performed in three transects within the BF-5 catchment).

Figure 5.

Soil water total organic carbon (TOC) concentrations sampled at different depths along three transects within the BF-5 catchment during a snapshot campaign performed on 31 August 2011. Transects represent up to 10 sampling locations within the riparian zone, which are located at different distances from the stream (0.1–20 m). Diamonds represent the average concentration, and whiskers indicate the standard deviation. The number of samples at each depth is given byn.

[38] The second component of the hydrological driver, increased GW levels, is commonly observed after clear-cutting [Mannerkoski et al., 2005; Monteith et al., 2006]. Data presented by Sørensen et al. [2009b]suggest that increases in GW levels also took place in the Balsjö catchments after harvesting. These results showed relative increases in discharge at low- and medium-flow conditions, respectively, which clearly can be associated with increasing GW contributions to the streams.

[39] In summary, the combined effects of increased GW levels and DOC mobilization from higher soil layers would give several times higher DOC concentrations in the stream for each incremental increase in lateral flow entering the stream. In turn this highlights the importance of a changing water balance for water quality after forestry operations.

[40] A secondary control of increased DOC concentrations after forestry operations could be attributed to higher decomposition of soil organic matter due to increased soil temperatures [Liski et al., 1998]. Temperature-dependent buildup of DOC in soils [Boyer et al., 1996; Hornberger et al., 1994] has been hypothesized for boreal catchments after clear-cutting [Lamontagne et al., 2000] and is also seen as a suggested driver of increased DOC levels in undisturbed streams in northern Sweden [Köhler et al., 2009]. Further, recent work from northern Sweden [Winterdahl et al., 2011] clearly indicates how the secondary control of soil temperature acts concurrently with the superimposed hydrological drivers. Soil temperatures have been observed to increase in the Balsjö catchments after clear-cutting by up to 5.5°C at 30 cm soil depth (Figure 6), which strongly suggests this driver to be another important component of the increased DOC mobilization after forestry operations. However, a clear and causal linking of the sensitivity of stream water DOC levels to changing soil temperatures after clear-cutting is, according to our knowledge, still absent.

Figure 6.

(top) Variation of soil temperatures measured at different depths during 2009. (bottom) Differences in soil temperatures measured at different depths during 2009. Both measurement sites are located on a western slope with an approximate distance of 40–50 m from the stream; the clear-cut site is situated within the BF-5 catchment, whereas the measurements at the forest site were performed within the NR-7 catchment.

[41] Third, and possible more hypothetical, a similar mechanism to that of increased soil temperatures could also be prevalent during the winter seasons in which lower canopy coverage [Devito et al., 2005] may be linked to deeper soil frost: a variable which has recently been discovered to be positively correlated to [DOC] in riparian soils [Haei et al., 2010] as well as DOC export in northern Swedish streams during the spring freshet [Ågren et al., 2010]. However, typically observed higher snow accumulation in clear-cuts [Buttle et al., 2005] with its concomitant insulation effect, which is also found in Balsjö [Sørensen et al., 2009b] (Figure 6), acts counter to this hypothesis.

[42] Deciphering the relative importance of the different drivers seems therefore to be the logical next step when aiming to transfer the results of this study into larger watersheds of the landscape scale, where the full impact of forestry on water quality may be strongly dependent on the relative contributions and timing of DOC flushing from different headwater streams [Öhman et al., 2009].

5.2. Lateral Dissolved C Flux

[43] This study further accentuates the role of [DOC] for the lateral C export in boreal streams (Figure 4). These increases in fluxes are primarily controlled by increases in water discharge; 90% of the increased C fluxes after clear-cutting are caused by the increases in runoff. This is in agreement with other studies of the region in which strong year-to-year variations of dissolved C fluxes were found to be primarily controlled by water fluxes rather than changes in concentrations [Laudon et al., 2004; Nilsson et al., 2008; Ågren et al., 2007]. After site preparation, however, the percentage of increases in DOC flux explained by increasing discharge is lower (78.6%) which in turn indicates that the strong soil disturbance during site preparation may cause increased DOC concentrations in soils, which then play a more crucial role for the total lateral C export from soils to the streams and to downstream ecosystems. Alternatively the strong increases in [DOC] after site preparation could be caused by the fact that 2009 was the first very wet summer after all forestry operations performed and these summer events could have mobilized large surface near C pools, which were not flushed into the stream during the drier years prior to the site preparation, but after clear-cutting.

[44] Recent studies have further emphasized the importance of DIC as a component of the C budgets of boreal catchments. Boreal first-order streams are permanently supersaturated in partial pressure of CO2 (PCO2) in northern Sweden [Öquist et al., 2009], and dissolved inorganic carbon (DIC) can be an important fraction of the lateral C flux [Wallin et al., 2010], which may be an additional component of the riverine C flux in boreal forests which are subject to forestry operations. However, even if DIC concentrations are not included in this study, recent research also shows that the DOC-based lateral C export on its own can account for 20%–24% of the net ecosystem exchange (NEE) in northern, peat-rich catchments [Dinsmore et al., 2010; Nilsson et al., 2008; Roulet et al., 2007]. Even though no direct measurements of NEE were performed at the 277 Balsjö sites, data for the nearby Flakaliden research forest (distance = 36.0 km) are available. The values of this Norway spruce (Picea abies) dominated forest at an age of 43 years are given as −1008 (±128) kg C ha−1 yr−1 for the years 2001–2002 [Lindroth et al., 2008]. Assuming a similar productivity of the older stands in Balsjö, the role of DOC export in the 277 Balsjö catchments would account for a fraction of approximately 10% (average of all controls and the treated sites during pretreatment period) of NEE. This fraction would then increase to 18% after clear-cutting and would reach 28% of NEE, respectively, in the period after site preparation. It should be noted here, that NEE also varies over time. This variation can be caused by variations of climate variables [Lindroth et al., 2008; Nilsson et al., 2008] and by stand age [Schwalm et al., 2007]. However, the carbon balance is in either case strongly affected by forestry operations and forests can even acts as net carbon sources to the atmosphere. Annual NEE values in the range of 500–1300 kg C ha−2 yr−1are reported for boreal forest sites located in central Saskatchewan, Canada after clear-cutting and fire disturbance [Amiro et al., 2006]. This underlines in turn the relative importance of the riverine C fluxes as a second pathway leaving the ecosystem for at least the first decade after forestry operations.

6. Conclusion

[45] This study indicates a strong impact of forestry operations on DOC concentrations and riverine C fluxes in boreal first-order catchments. Both forestry operations, clear-cutting and site preparation, are commonly used in the boreal regions. Nevertheless, there is little attention in larger-scale studies on DOC mobilization originating from these disturbances [see, e.g.,Erlandsson et al., 2008]. This work hence emphasizes that forestry operations affecting larger spatial areas may have a profound effect on the biogeochemistry of streams at the landscape scale. Further we suggest that the results presented in this study should favor the development of management techniques that minimize the effect of forestry on water quality in larger river ecosystems.


[46] Financial support for this work has been provided by CMF, Future Forests, and the Formas (ForWater). We also thank Lenka Kuglerova, Peder Blomkvist, Ida Taberman, and Viktor Sjöblom for help in the field and laboratory, as well as Karin Öhman and Thomas Grabs for useful comments on an earlier version of this paper.