Dissolved organic carbon (DOC) is of environmental and biological significance to Precambrian shield streams. Our objective was to determine the extent of transformations to the concentration and molecular weight distribution (MWD) of DOC along the length of Harp 4A stream, from the headwater swamp to the outflow into Harp Lake. The average MWD shifted to a smaller weight by 0.48 kDa. A rapid decline in the concentration and MWD of DOC was observed within 120 m from the main DOC source, a headwater swamp. The most significant decline in concentration and MWD was observed within the first of a series of three beaver ponds. DOC within waters downstream of the beaver ponds was resistant to further changes in concentration and MWD. The MWD within each of four stream sections with Hp 4a remained stable over 6 months, from May to October 2001.
 Dissolved organic carbon (DOC) in natural waters consists of a heterogeneous mixture of organic compounds, with ill-defined molecular structures and a wide range of chemical and physical characteristics. The concentration and quality of DOC in natural waters has numerous abiotic and biotic consequences. The largely hydrophobic character of humic substances, comprising approximately 50% of DOC, has been shown to reduce the bioavailability of PCBs [Knulst, 1992], and the toxicity of polycyclic aromatic hydrocarbons [Chin et al., 1997] by hydrophobic adsorption to DOC. DOC aids in increasing the attenuation of UV and visible light in natural waters [Scully and Lean, 1994], providing protection to aquatic organisms from potentially damaging radiation even at low DOC levels [Williamson et al., 1996]. Smaller, generally nonhumic molecules comprising DOC are an important source of energy for aquatic organisms [Wetzel, 1995; Hessen, 1992].
 The DOC concentration in lakes has been positively correlated to percent wetland cover within the surrounding catchment [Curtis, 1998; Dillon and Molot, 1997]. Less attention has been devoted to in situ changes in DOC concentration and alterations to DOC chemistry as water passes through the stream, from headwaters to the outlet, usually a lake [Mulholland and Hill, 1997; McKnight and Bencala, 1990; McKnight et al., 2002]. Numerous potential sources, sinks and transformations can modify the concentration and quality of DOC as it moves downstream. In particular, beaver ponds, a common landscape feature of the Canadian Shield, are known to alter stream hydrology, chemistry, organic carbon retention [Naiman et al., 1986], and nutrient dynamics [Devito and Dillon 1989]. DOC in natural waters can be mineralized by photooxidation [Köhler et al., 2002], degraded by microbial colonies [Kieber et al., 1990] or lost to the sediment by coagulation processes. Alternatively, DOC can be produced autochthonously by microbial respiration [Hessen, 1992; Moran and Hodson, 1990]. DOC can also enter a stream through seepage and overland flow. However, the strong relationship between DOC concentration in Precambrian shield lakes and percent weltand cover (R2 = 0.80 [Dillon and Molot, 1997]), suggests that wetlands are the primary source of DOC to streams and lakes. Thus identifying the extent of spatial in-stream processing of DOC from wetlands to a lake is of particular importance for predictive, management, and modeling purposes.
 The molecular weight distribution (MWD) of DOC is a vital bulk property that dictates both its environmental and biological influence within aquatic systems (see review by Cabaniss et al. ). High molecular weight DOC is generally more aromatic and hydrophobic, allowing for better affinity for organic contaminants such as atrazine, diazinon and lindane [Saint-Fort and Visser, 1988]. Small DOC molecules are more soluble in water, and are hypothesized to be more readily metabolized by biological organisms [Lee and Wakeham, 1992]. The size of aquatic humic substances also will affect the complexation of trace metals [Bartschat et al., 1992]. DOC with a molecular size too large to pass a biological membrane will be restricted from biological uptake, along with any affiliated organic and inorganic contaminants.
 In this study, we examine the variation in DOC concentration and MWD along two streams draining catchments in the Canadian Precambrian Shield. One stream is surrounded by a wide riparian zone and contains no beaver ponds. The other drains a headwater swamp, possesses a thin riparian zone and passes through several beaver ponds. We tie changes in DOC concentration and MWD to processes affecting DOC within the stream system. The study asks if changes to the average molecular weight, and hence structure of DOC, are detectable with movement along the length of a stream.
2. Study Sites
 Harp Lake (45° 23′N 79° 08′W) is a small (71 ha) oligotrophic headwater lake, located on the Precambrian shield in south central Ontario, with a total catchment area of 542 ha. Two of the six main catchment areas comprising the Harp Lake basin, Harp 4 (Hp 4) and Harp 3 (Hp 3), were selected for this study (Figure 1a). In Harp 4, we studied only the Hp 4A tributary (Figure 1b), a first-order stream draining an area of 73 ha with a length of approximately 1500 m. Hp 3 is a first-order stream draining a catchment of 26 ha with a length of approximately 200 m.
 The geomorphology, vegetation and hydrology of Hp 3 and Hp 4 catchments are typical of Precambrian shield catchments and have been described in detail previously [Devito and Dillon, 1993; Dillon and Molot, 1997; Dillon et al., 1991]. Briefly, the underlying silicate bedrock is covered by glacial till with depths ranging from 0 to 15 m, mainly composed of Brunisolic and Podzolic soils. The vegetation is primarily mixed deciduous-coniferous forests including sugar maple (Acer saccharum), beech (Fagus grandifolia), yellow birch (Betula alleghaniensis), aspen (Populus tremuloides), white pine (Pinus strobus), and balsam fir (Abies balsamea). Over 25 years, Hp 4A has had 0 dry days, whereas Hp 3 has had an average of 25 ± 29 dry days per year.
 Hp 4A stream was separated into four stream sections: headwaters, upstream, beaver ponds, and downstream (Figure 1b). A total of 27 sites were originally sampled, and subsequently reduced to fifteen sites, thus the site numbering is not always continuous. Headwater sites included Hp 4A-1, a large swamp draining a mixed deciduous and coniferous forest, and Hp 4A-2, an ephemeral stream that drained it. Upstream sites included Hp 4A-4, a wetland that was previously a beaver pond and sites Hp 4A-5, Hp 4A-7 and Hp 4A-9 located downstream from the wetland. Hp 4A-9 is the inflow to beaver pond 1. Beaver pond sites included beaver ponds 1, 2, 3 and the outflow from pond 3, denoted as study sites Hp 4A-10 to 13 respectively. Hp 4A-10, the largest of the three beaver ponds (3.8 ha), has an average depth of 1.2 m, and drains the upper 62 ha (84%) of Hp 4A. Beaver pond 1 has been described in detail by Devito and Dillon  and Devito et al. . Downstream sites Hp 4A-15, Hp 4A-19, Hp 4A-23 and Hp 4A-27, were located 40, 130, 250 and 370 m downstream from the outflow of beaver pond 3 (Hp 4A-13). About 30 m downstream from site Hp 4A-27, the Hp 4A stream converged with another tributary of the Hp 4 watershed, which subsequently flows though the Hp 4 weir, and drains into Harp Lake (Hp Lk), where samples were collected from the littoral zone.
 Hp 3 drains a headwater wetland (Hp 3-1) with a shallow, wide riparian zone spanning the full length of the stream. Sampling sites Hp 3-2 to Hp 3-7 are located at 30 m intervals downstream from each other, and Hp 3-8 is located at a weir near the inflow to Hp Lk.
 Water samples (50 mL) were transported to the lab within eight hours of collection, filtered through 0.45 μm nylon syringe filters (Life Sciences, Peterborough, ON), and stored in the dark at 4°C. A new filter was used for each sample, and three Milli-Q water samples were filtered with each batch of 15 samples to correct for carbon leaching from the nylon filters. Total organic carbon analysis was performed with a Shimadzu 5000 TOC analyzer.
 High performance size exclusion chromatography (HPSEC) was performed using a Waters 600 controller (Waters Associates, Milford, MA) with a flow rate of 0.5 mL min−1, a Waters photodiode array 996 detector set at 254 nm, and a Rheodyne rotary injector valve equipped with a 200 μL sample loop. The void volume (Vo) was determined using a polystyrene sulfonate (PSS) standard with a MW of 2.7 million Da (2.7 × 103 kDa), and the total permeation volume (Vt) was determined with acetone (MW = 56 Da). MW calibration was based on PSS standards 2.7 × 103, 2.6 × 102, 34.0, 7.9, 4.0, 1.8, 1.4 kDa (American Polymer Standards, Mentor, OH) and 210 Da (Fluka, Canada). The mobile phase consisted of 0.1 M NaCl (Fisher Scientific, Nepean, ON) + 0.002 M KH2PO4 (Fisher Scientific, Nepean, ON) + 0.002 M K2HPO4 (BDH, Mississauga, ON), as previously used by Chin and Gschwend .
 Chromatograms were produced with a Waters YMC 60 silica diol column, with a narrow and specific calibration range from 7.9 kDa to 210 Da (Waters Associates, Milford, MA). For samples where a fraction of stream DOC material eluted near the void volume, a supplementary YMC 300 (Waters Associates, Milford, MA) column, with a larger gel bead pore size (300Å) and wider calibration range (2.62 × 103 kDa to 210 Da),was utilized to establish the maximum molecular weight of DOC. Maximum peak height, total peak area, peak start, peak end, and retention volume (Rv) for the sample peak(s) were determined by the Millenium software (Waters Associates, Millford, MA). Results were comparable to setting the high MW cutoff at 1%, and a low MW cutoff at 2% of the peak height or 50 Da, whichever was the higher value, as suggested by Zhou et al. .
 The MWD of DOC for samples was expressed in terms of; number average molecular weight (Mn), weight average molecular weight (Mw), and polydispersity (p) as described by Yau et al. . Mn and Mw are two methods of calculating an average value of a distribution. When Mn and Mw values are equal, the polydispersity, p, is equal to 1.0, and represents a pure substance. Exclusive examination of Mn, Mw and p for the comparison of distributions including unimodal and bimodal distributions, as found in this study, can result in an erroneous bias. Thus Mn and Mw were calculated for comparison of temporal patterns of peaks with reproducible shape. A commonly used reference material, Suwannee River Fulvic Acid (SRFA), was used to ensure standardization with other studies. MW estimates from studies with similar HPSEC conditions have reported SRFA values for Mn of 0.64 to 1.40 kDa, and Mw of 1.00 to 2.30 kDa [Zhou et al., 2000; Chin et al., 1998]. This study reports SRFA values within this range, with Mn and Mw values of 0.98 kDa and 1.32 kDa for the YMC 60 column, respectively. HPSEC chromatograms were also divided into four MW fractions (Figure 2), large (>5.0 kDa), medium-large (2.0 to 5.0 kDa), medium (1.5 to 2.0 kDa) and small (<1.5 kDa). A strong relationship exists between the total peak area (determined from UV absorbance at 254 nm) and total DOC concentration (R2 = 0.96, n = 87). Thus each of the four MW fractions were also assessed spatially based on relative percent abundance, and DOC concentration.
 Hp 3 and Hp 4 stream discharge was recorded using A71 Leupold-Stevens recorders at V notched weirs. Yearly DOC export for Hp 3 and Hp 4 was calculated using mean daily discharge by linear integration of the chart hydrographs and weekly DOC concentrations determined as outlined by the Ontario Ministry of the Environment .
4. Results and Discussion
4.1. Spatial and Temporal Variations in DOC Concentrations
 DOC concentrations, based on the average of the 15 sampling dates from May to October, 2001, within the Hp 4A stream rapidly declined from 35.9 ± 8.5 mg L−1 at the headwater swamp, to 25.0 ± 6.1 mg L−1 at the beaver pond inflow, located 120 m downstream of the swamp (Figure 3). DOC concentrations declined to an average of 9.0 ± 2.1 mg L−1 at the stream outflow into Harp Lake. Spatial DOC losses observed within the upstream section (4.7 mg L−1) over a 100 m section of Hp 4A were far greater than the section downstream of the beaver ponds (1.2 mg L−1), over 100 m. Yet, groundwater inputs into Harp 4A stream are more sizable in the downstream section, where dilution might be expected to lower DOC concentrations [Devito, 1995].
 The first in a series of three beaver ponds (Hp 4A-10) was responsible for the single greatest loss of DOC along the length of the Hp 4A stream. DOC concentrations dropped between sites Hp 4A-9 and beaver pond 1, site Hp 4A-10, by an average of 10.9 ± 6.5 mg L−1 (Figure 3) over the 6 month study. Beaver ponds 2 and 3, (Hp 4A-11 and 12) located downstream from beaver pond 1 failed to experience any subsequent decreases in DOC concentration (13.5 ± 1.3 and 13.5 ± 0.9 mg L−1) (Figure 3). Naiman et al.  reported no significant difference in DOC or POM concentrations entering and leaving a beaver pond located downstream of 10 other beaver ponds during a 2 year study. Minor variations in DOC concentration of Hp 3 stream (18.1 ± 2.8 mg L−1), sampled on two dates 1 month apart, may reflect the influence of riparian zone and the absence of beaver ponds. As riparian soils have been identified as rich sources of DOC to streams [Bishop et al., 1994; Fiebig et al., 1990], they may have contributed to the spatially consistent DOC concentrations along Hp 3, while Hp 4A stream obtained most of the DOC from a point source, the headwater swamp (Hp 4A-1).
 Distinct sections of the Hp 4A stream displayed different temporal patterns over 6 months (Figure 4). DOC accumulated within the dry headwater swamp over a 6 week summer drought, and was mobilized upon re-saturation of wetland and mineral soils in September, with concentrations peaking to an average of 38.4 ± 2.8 mg L−1. About one third (32.3%) of the total annual DOC yield of Hp 4 stream (5.3 g m−2 yr−1) was flushed from the catchment over 2 months, between 12 September 2001 and 12 November 2001. By late October, DOC stores within the headwater swamp were lowered and resembled spring-like concentrations 20.0 ± 1.6 mg L−1. Three beaver ponds (sites Hp 4A-10 to 12) maintained temporally stable average DOC concentrations of 13.6 ± 1.2 mg L−1, despite fluctuations in total stream discharge (Figure 4). All downstream sites of Hp 4A maintained continuous stream flow, even during the drought (Qmin = 1.7 ± 0.8 L s−1), unlike the headwater and upstream sites, which had zero discharge for a period of six weeks. Average DOC concentrations at downstream sites lowered from 10.4 ± 2.1 mg L−1 to 7.4 ± 0.8 mg L−1 (Figure 4) during the drought. Following the drought, DOC concentrations returned to predrought levels (11.0 ± 1.4 mg L−1), however the surge in DOC concentrations observed at headwater sites was not observed here at downstream sites. Consequently, ephemeral headwater stream sections can respond with a large postdrought peak in DOC concentrations, however, the peak may be dampened in downstream sections where flow is continuous. Hydrological flow paths and groundwater contributions were not monitored at each stream section, however discharge was monitored for the Hp 4 watershed, at the V notched weir just downstream of Hp 4A-27 (Figure 4).
4.2. Variations in the Molecular Weight Distribution of DOC
 The MWD of DOC was variable along the length of Hp 4A stream, but remained relatively similar at individual stream sections from mid-May to late October. Headwaters and upstream sites displayed a unimodal MWD (Figure 2a), with a polydispersity of 1.49 ± 0.15 (Table 1). Beaver pond sites, and stream waters downstream of the ponds showed bimodal size distribution (Figure 2b), with additional peak shoulders, and greater average polydispersity 1.81 ± 0.25 (Table 1). The shape of MWDs were reproduced over the 6 month study with exception of the drought.
Table 1. Number Average Molecular Weight (Mn), Weight Average Molecular Weight (Mw), and Polydispersity (p) for Four Sections in Stream Harp 4A, Stream Hp 3, and Harp Lakea
Standard deviation over the 6 month study period (n = 15).
Hp 4A headwaters
1343 ± 267
1972 ± 289
1.49 ± 0.15
Hp 4A upstream
1270 ± 196
1975 ± 350
1.57 ± 0.17
Hp 4A beaver ponds
1005 ± 210
1811 ± 351
1.80 ± 0.24
Hp 4A downstream
1084 ± 206
1962 ± 388
1.81 ± 0.26
1419 ± 16
1868 ± 52
1.32 ± 0.03
879 ± 178
1131 ± 170
1.29 ± 0.08
 Material categorized as large MW (>5.0 kDa) increased with movement downstream, ranging from <1% at the headwaters to 7% at downstream sites (Figures 5a, 5b, and 5c). Kaplan et al.  also detected a higher percentage of large molecules (>10 kDa) with movement downstream from the source, using ultrafiltration. DOC within the large MW fraction of Hp 4A included molecules up to 2.62 × 103 kDa, the upper calibration limit of a supplementary column (YMC 300). The concentration (maximum: 1.5 mg L−1, Figure 6a) and relative percent abundance of this large MW fraction (maximum: 7.1%, Figure 5b) is small compared to other MW fractions. However, the highly hydrophobic nature of these large molecules could lead to the adsorption of other organic constituents including organic contaminants, making them less available for biological uptake [Kukkonen, 1999]. Trace levels of this large MW material detected within the Lake (Hp Lk, <0.1 mg L−1) suggest removal by sedimentation.
 The proportion of medium-large and medium (1.5–5.0 kDa) DOC decreased with movement downstream, while the proportion of small MW material generally increased with movement downstream (Figures 5a, 5b, and 5c). Processing within beaver pond 1 (Hp 4A-10) led to a decrease in the proportion of medium-large MW DOC during the summer (from 24.8 to 11.1%), and to a lesser degree in the fall (from 12.1 to 6.6%). Medium MW DOC also decreased at the beaver pond site during the summer (from 26.9 to 19.7%) and the fall (from 19.6 to 10.1%). However, there was an increase in small MW DOC at beaver pond 1, Hp 4A-10 in the summer (from 44.5 to 66.3%), and the fall (from 67.4 to 81.0%). We were unable to find a shift in the MWD during the drought (Figure 5b) since headwater and upstream sites were dry, or merely stagnant pools of water. The concentration of medium-large, medium and small MW material decreased substantially at site 4A-10, beaver pond 1, with the exception of small MW material during the summer months (Figures 6b, 6c, and 6d). The pattern of concentration within each size fraction is similar between the two summer and fall dates (Figures 6b, 6c, and 6d), suggesting good reproducibility of the method.
 The shift in the MWD of DOC within beaver pond 1 of Hp 4A from larger to smaller MW indicates that mechanisms aside from simple hydrological dilution effects were responsible for the loss in total DOC concentration. Beaver pond 1 was determined to have an average water residence time of 47 days, ranging from 9 days in April to 242 days in the summer and fall of the 1987–1988 hydrological years [Devito and Dillon, 1993]. Consequently, beaver pond waters are ideal for sedimentation, microbial processing and photooxidation, particularly in summer months. The average molecular weight (Mn and Mw) of DOC in wetland pore waters has been found to be greater than the overlying water column [Chin et al.,1998; Wang et al., 1997]. Consequently, DOC molecules ranging from 1.5 to 5.0 kDa may have precipitated into the sediment layer. In addition, the high surface area to volume ratio of beaver ponds are ideal for UV light penetration and photochemical transformations, which have been shown to increase the proportion of low molecular weight DOC compounds [Bertilsson and Tranvik, 2000; Wetzel, 1995]. Microbial consumption and production of DOC within beaver pond 1 may have also contributed to the to the decrease in proportion of molecules 1.5 to 5.0 kDa and increase of <1.5 kDa. The exact MW distinction for bioavailable organic compounds is unknown since the chemical composition of both DOC and microbial cultures is highly variable. A review paper reports DOC bioavailability ranging from <1 to >75% of total DOC, for different molecular weight fractions, and different environments [Sun et al., 1997]. While beaver pond 1 is an ideal site for multiple processes, affecting DOC, we are unable to identify which is most important here. Further studies are required to determine the relative contribution of sedimentation, photodegradation and microbial processing of DOC within beaver ponds.
 This study supports the work of others who have considered the flux of DOC, or used carbon isotopes to study the origin, transport and fate of DOC [Devito and Dillon, 1989; Devito et al., 1999; Schiff et al., 1998, 1990] in suggesting that the beaver pond of Hp 4 catchment is not a source of DOC. Thus beaver ponds should be classified distinctly from wetlands such as conifer swamps, especially when assessing the effects of general landscape features on DOC export dynamics.
 Beaver ponds mature through several stages, young, adolescent, mature, senescent, marsh, and dry [Welsh, 1963], until eventually a forest is reestablished. Site Hp 4A-4, is an abandoned beaver pond at a more mature successional stage than Hp 4A-10, and possesses swamp-like characteristics. While there was a decline in the average DOC concentration (3.8 ± 5.4 mg L−1) and a shift in the MWD toward a smaller average size in Hp 4A-4, the alteration was not as clearly defined as the less mature, and larger pond, Hp 4A-10. Consequently, the maturity or successional stage, along with the size of a beaver pond may dictate the type, and extent, of DOC processing. The retention of organic material within beaver ponds suggests that a large pool of organic matter is sequestered within the sediment layer of the beaver pond. Disturbances such as lowering of water tables, or breaking a beaver dam could potentially remobilize this large pool of sequestered carbon.
 The smaller Hp 3 stream with a well-developed organic riparian zone did not display any significant changes to total DOC concentration (Figure 3) spatially. Each of four size fractions also remained spatially consistent in terms of proportion (large: 3.4 ± 0.9%, medium-large: 27.4 ± 1.1%, medium: 26.5 ± 0.6%, and small: 42.6 ± 1.1%; Figure 5d). The MWD of Hp 3 stream waters remained spatially unimodal, with an average Mn = 1.41 ± 0.02 kDa, (Table 1). The lack of structural modifications to DOC within Hp 3 suggests an insignificant level of in-stream processing.
 Over the 6 month study of Hp 4A stream, the shape of chromatographic distributions were reproducible within stream sections. Headwaters sites (Hp 4A-1 and 2) produced unimodal distributions over the full study period. Beaver pond sites were consistently bimodal with an average shift of the main sample peak of ∼0.48 kDa, compared to upstream sites. Downstream sites were also consistently bimodal with a further shift to lower MW in the main sample peak by ∼0.10 kDa. Examination of Mn and Mw revealed relatively minor temporal fluctuations within stream sections, with an increase observed on August 13 during the drought (ΔMn = 340 Da and ΔMw = 590 Da) (Figure 7). Polydispersity was greatest in summer months reaching values of 2.2 at sites Hp 4A-10 and Hp 4A-23 during June and July (Figure 7). Microbial and photodegradation processes are greatest during these summer months, resulting in a broad composite of organic by-products.
 We observed a sharp decline in the total DOC concentration and a shift toward smaller MW with movement along the length of a stream (Hp 4A) from the headwater swamp to the lake. The concentration and MWD of DOC altered between the headwater swamp and 120 m downstream from the swamp, suggesting that DOC originating from headwater swamps is highly decomposable. A large alteration to the MWD and DOC concentration occurred within the first of a series of three beaver ponds. Total DOC and MWD waters downstream of beaver ponds remained stable compared to upstream sites, suggesting that organic compounds were structurally more refractory after being processed within beaver pond waters. Beaver pond processing caused a significant shift in MW from larger to smaller molecules by ∼0.480 kDa. A nearby stream, Hp 3, with a wide riparian zone showed small spatial variations in DOC concentration and MWD downstream. These structural modifications to the MWD of DOC could have large implications in terms of providing energy to microbial communities, number of chromophoric sites available for light attenuation, contaminant and nutrient mobility, and buffering capacity.
 Thanks to Hayla Evans, Tim Moore, and two anonymous reviewers for reviewing the manuscript. Thanks to Jian Jun Yang, Heather Broadbent, Mark Dzurko, Joe Findeis, and Lem Scott for technical and field assistance. Special thanks to the staff, past and present, of the Dorset Environmental Science Center. Thanks for financial support provided by INCO and the National Science and Engineering Research Council (NSERC) of Canada.