DOM samples were analyzed by means of high-performance liquid chromatography coupled to high-resolution mass spectrometry (HPLC-HRMS). Liquid chromatography was performed using an Agilent 1100 Series system with a polar C18 column (Kinetex^®, 2.1 × 150 mm, 2.6 µm bead size, 100 Å pore size) with mobile phase A (0.1 % formic acid in LCMS-grade water) and mobile phase B (0.1 % formic acid in 80 : 20 acetonitrile : LCMS-grade water (v/v)). Samples were diluted in 5 % v/v acetonitrile solution (LiChrosolv, Merck) and 20 µL was injected at an initial flow rate of 150 µL min⁻¹ with mobile phase A at 95 % and mobile phase B at 5 %. After 10 min, the acetonitrile mobile phase B was increased to 95 % for 2 min and then was decreased to 5 %, where it was held to be isocratic until 15 min. Mass spectrometric analysis was completed via an LTQ Velos Pro Orbitrap MS (Thermo Scientific, Germany) using an electrospray ionization source (ESI) operating in negative mode (spray voltage: -3.1 kV, capillary temperature: 275 °C). Blanks consisting of mobile phase A were injected periodically between samples. Each spectrum was internally calibrated in lock mass mode using three expected compounds, namely capsaicin, fusidic acid sodium, and glycyrrhizic acid ammonium salt, with 304.1921, 515.3378, and 821.3965 negative m/z, respectively, providing suitable accuracy and precision (<1 ppm) in the mass range of m/z 150–800. Data were collected at a resolution mode of 100 000. More detailed instrumentation parameters are reported elsewhere (Fonvielle et al., 2023).

Mass spectrometry data were exported from the mass spectrometer and converted to mzXML files with ReAdW and then were processed further using MATLAB (version 2019b). A MATLAB routine was developed in-house and is available with raw mzXML files. Molecular formulas were assigned between 150–800 Da masses. Formulas were limited with the following criteria: carbon of 4–50, hydrogen of 4–100, oxygen of 2–40, nitrogen of 0–2, sulfur of 0–1, H/C = 0.3 to 2.2, O/C = 0 to 1, and double-bond-equivalent oxygen (DBE-O) = -10 to 10. In addition, the valence electron needed to be equal to an even number, and formulas which contained both nitrogen and sulfur, ¹³C and nitrogen, or sulfur and ¹³C were removed. A mass error of 0.7 ppm was allowed for formula assignment. Peak intensities with formula assignments were normalized to sum up to 1×106 for each sample. Further descriptions for the different DOM parameters are given in Table 1. Compounds known as terrestrial peaks (referred to as “t-Peaks”), commonly found in rivers (Medeiros et al., 2016), are identified in our experimental mass spectrometry data by matching their molecular formulas and molecular masses to those of reported t-Peaks.

The change in POC concentrations from the start (t0) to the end of incubation (t1) of filtered (F) and unfiltered (UF) treatments followed a generally similar pattern (relative increase or decrease) throughout the year (Fig. 4). In winter (December and February), after 36 h of incubation, we measured lower POC concentrations in F water at t1 compared to at t0 (a decrease in the mean POC concentration of -2.55 µM ± 0.8 (around -50 % relative to t0)). There was an increase in POC during the incubation period in the filtered water (F) of April, June, and September, with a mean of +1.6 µM ± 0.5 (Fig. 4; April: 76 % increase relative to t0; June: 88 %; September: 70 %). The decrease in POC in winter and the relative increase in April followed in a similar manner in UF water (t0 is not available for UF in September). However, in June, the relative increase in POC concentrations was 4 times higher in UF compared to in F. For the F treatment, a t test revealed significant differences in the changes in POC concentrations between the winter and productive periods (p = 0.04, t value = 4.6, degrees of freedom (df) = 3), whereas this difference was not significant for the UF treatment (p>0.05, t value = 1.6, df = 1).

Bacterial biomass in F water did not show large differences after incubation relative to the start (t0) from September until April; however, it increased in June and August (Fig. 4b). The patterns were different in UF water, where, in winter, a decrease in bacterial biomass was observed, while, in spring and summer, there was an increase (Fig. S4). Despite performing the incubation in darkness, there was an increase in pico-sized phytoplankton cells in the unfiltered (UF) fraction in August. At this time the system had the highest concentrations of small phytoplankton (<2 µm). Picoeukaryotes and Synechococcus increased in August from 4000 cells at t0 to 24 000 cells mL⁻¹ after 36 h (Fig. S4), equalling a biomass increase of approximately 2 µM carbon (Fig. 4c).

In F water, bacterial activity (HNA / LNA) increased relatively constantly throughout the year; however, in UF water, little change was observed in winter and August, while April was characterized by a sharp increase in activity (Fig. 4d). Similarly to bacterial activity, EPSs always increased in both treatments and throughout the whole year (sampling apart from February; Fig. 4e). Conspicuously, while, for UF water, the highest increase in POC was observed in June, the highest increase in EPSs was observed in April. In F water, the highest increase in EPSs was present in June.

Initial (t0) DOC concentrations followed a similar pattern to the concentrations in the field (Figs. S2, S4). In September, concentrations increased in both F and UF water (about +60 µM in F and +180 µM in UF; Fig. 4f). In October and December, DOC concentrations still increased in F water (+73.6 and +27.2 µM, respectively) but increased very little or even decreased in UF water. In April and June, DOC concentrations decreased in F but increased in UF water (Fig. 4f).

The molecular composition of DOM was analyzed at the beginning and end of incubation experiments conducted during contrasting periods of the year. September incubations represent a productive period, whereas experiments in December and February indicate a winter period. Additionally, October incubations were also analyzed for DOM composition to assess the transitional period. Metrics for DOM composition were reported as signal intensity weighted average (wa) values of hundreds of DOM formulas per treatment (Table 2). A table presenting the individual mean values per sample for each DOM metric is provided in the Supplement (Table S2).

During the productive period, studied in the September incubations, average hydrogen-to-carbon atomic ratios (H/Cwa) decreased for both UF and F treatments (Fig. 5). Simultaneously, the average aromaticity (AI_mod wa) and oxygen-to-carbon atomic ratios (O/Cwa) increased for both F and UF treatments (Fig. 5). Average molecular weight (MW_wa) increased at the end of incubation in the F treatment and decreased at the end of incubation in the UF treatment. Notably, a removal of more saturated compounds (H/C > 1.5) is observed at the end of incubation (t1) for the F (Fig. 6a) and UF (Fig. S5) treatments in September. Additionally, low SPE–DOC recoveries were observed in September and October (Fig. S9). This indicates a higher proportion of hydrophilic material (Kirchman et al., 2001; Goldberg et al., 2009), which is made up of compounds that are not well retained by our SPE sorbent (Grasset et al., 2023).

Molecular weight patterns are shown in mass spectra (Fig. 7b) where a decrease in the relative intensities of low-molecular-weight compounds is observed from the start (t0) to the end (t1) of the September incubations for the F treatment. These low-molecular-weight compounds (m/z < 250) are highlighted in Fig. S6 and show formulas with higher relative intensities at H/Cwa ratios greater than 1.3.

Winter month (December and February) incubations showed a contrasting pattern, with an increase in hydrogen-saturated formulas (H/Cwa) in both F and UF treatments (Fig. 5a). During the same period, a decrease in average aromaticity (AImodwa) and in oxygen-rich formulas (O/Cwa) was observed in both F and UF treatments (Fig. 5). Molecular weight (MWwa) decreased during the incubation period in the F treatment and increased during the UF incubation (Fig. 5). Moreover, there was a decrease in less saturated formulas (H/C < 1.5) in both the F treatment (Fig. 6) and the UF treatment (Fig. S5) and slight increase in the intensities of more saturated formulas for the F (Fig. 6) and UF treatment (Fig. S5) at the end of incubation (t1). Additionally, the highest SPE–DOC recovery, ranging from 74 % to 85 %, occurred in December t0 (Fig. S9).

Additionally, changes in DOM metrics for October incubation show transitional patterns from spring to winter months (Fig. 5) with the decrease in formulas with lower H/C ratios (Fig. 6b).

Molecular weight patterns are shown in mass spectra for winter incubations (Fig. 7c), where a decrease in the relative intensities of higher-molecular-weight compounds is observed at the end of incubation (t1). The molecular ratios of formulas higher than m/z 570 are shown in Fig. S6, composed of formulas in the middle O/C and H/C region of the van Krevelen diagram.

Previous studies have shown that large (up to 5 µm) polymer gels and particulate material can be reformed quickly after they have been removed from filtered seawater and river water despite consecutive filtrations and different filter sizes (0.2, 0.4, or 0.7 µm; Chin et al., 1998; He et al., 2016; Kerner et al., 2003; Passow, 2000, 2002b; Sheldon et al., 1967). Even when bacteria are killed, microgels of sizes between 200 nm to 1 µm can form within 30 min and up to 5 µm after 50 h (Chin et al., 1998; Kerner et al., 2003; Sheldon et al., 1967). The same process cannot be observed in filtered artificial seawater, but the addition of dissolved carbon enhances aggregation (Gruber et al., 2006; Sheldon et al., 1967). Several studies and reviews have pointed out the ability of DOM to assemble spontaneously into polymer gels, which can aggregate to particles (Chin et al., 1998; Engel and Passow, 2001; Passow, 2000). We suggest that the same process occurs during the present study, when we observe an increase in POC after 36 h in filtered (F) water taken during the biologically productive period between April and September. Although the change in POC concentration was small (between -3.10 to +2.28 µM) and varied across individual tanks (Fig. S4; most likely due to the short incubation time), we show, for the first time, the contrasting seasonality of this aggregation process.

DOM concentrations are elevated through phytoplankton production, which is likely why aggregation was observed during the biologically productive period. In situ EPS concentrations, representative for large sugars (> 0.4 µm), follow the same seasonal pattern as protist abundance (Fig. S2t, x). Diatoms dominated the phytoplankton community during the peak Chl a periods in April and June (Fig. S2x). In April, the mucous colony forming prymnesiophyte P. pouchetii also comprised a major fraction of the phytoplankton community of Ramfjorden, and, during this period, EPSs were elevated (0.84 relative absorption at 485 nm; Fig. S2t). It was also in April that we saw the largest increase in EPSs during the 36 h incubation, indicating that the residues from P. pouchetii are especially prone to reform into EPSs. In contrast, in June, when only diatoms dominated (Pseudo-nitzschia sp. and Chaetoceros filiformis), the experimental production of EPSs was approximately 30 % lower despite peak in situ concentrations of EPSs in June (1.91 relative absorption at 485 nm, Fig. S2t).

The increase in POC concentrations in F water accounted for between 1.7 % (April) and 7.9 % (September) of the POC concentrations measured in the field (Table 3), and although our data do not show an antagonistic relationship between POC and DOC, its source is likely to be aggregation from the dissolved pool. Phase shifts from DOM to POM are mainly driven by physical processes such as Brownian motion, chemical changes (e.g., ambient pH, ionic concentrations, temperature, light, bridging with divalent cations) and/or physical stress such as turbulent shear (e.g., filtration), differential settling, surface coagulation (e.g., bubbling), or bacterial motility (Engel and Passow, 2001; He et al., 2016; Kepkay, 1994; Passow, 2000; Timko et al., 2015; Verdugo et al., 2004). During our experiment, shear was probably only introduced during the filtration and at the beginning and the end of the rotation. During incubation, the rotation merely ensured the equal distribution of the material within the tank. Still, within a short amount of time (36 h), POC concentrations increased in F water during the biologically productive period of the year. Other studies showed that up to 35 % of particulate matter present in situ can be formed in filtered water through aggregation (Riley, 1963; Sheldon et al., 1967; Valdes Villaverde et al., 2020); however, this occurred over longer timescales (several days vs. 36 h in our experiment) and through the addition of shear. Keskitalo et al. (2022) demonstrate net aggregation of POC during spring freshet in an Arctic river and POC degradation of between 3–12 d during summer. These examples indicate that, next to environmental conditions, different incubation timescales probably affect DOM–POM processes differently, and more studies are needed to disentangle these effects. For the current study, we focus on the “immediate” behavior of the DOM–POM interactions within a short timescale compared to other experiments.

DOM–POM processes are usually driven by bacterial degradation (dissolution of POM) and physical processes (DOM aggregation to POM, adsorption of DOM to particles, or defragmentation). Our study shows that, although bacterial activity increased during incubation in both seasonal periods, microbial processes during the productive period seem to play a lesser role in the DOM–POM transition compared to physical aggregation. Other studies have similarly shown that, while POC concentrations remain constant after consecutive re-filtrations, bacterial abundances decrease, bacterial activity remains low, and substrates that are preferred over bacterial degradation can accumulate (Engel et al., 2004; Valdes Villaverde et al., 2020). The addition of sodium azide to inhibit microbial activity does not change the coagulation behavior of polymers, but the addition of ethylenediaminetetraacetic acid (EDTA), which disperses microgels and polymers, inhibits coagulation (Chin et al., 1998). These examples support that aggregation mainly stems from physical rather than biological transformation. Similarly, Engel and Passow (2001) and Passow (2000) show that gels > 0.4 µm in size form efficiently under shear and hydraulic stress. Overall, during the biologically productive period, a substantial fraction of particulate material can originate from the DOM pool, and changes in the DOM–POM continuum, in the direction of aggregation, have shown to be dominated by aggregation processes uncoupled from bacterial activity.

Changes in DOM composition during the incubation period for F and UF treatments were similar compared to the start of incubation (t0) and indicated little effect of larger organic size fractions (0.7–90 µm) on the composition of DOM during the 36 h incubations (Figs. 6 and S5). However, DOM compositional changes at t1 relative to t0 were contrasting for the productive versus winter periods regardless of treatment, thus indicating that DOM compositional changes were driven by abiotic process and/or microbial communities in the dissolved fraction (<0.7 µm). Due to similarities between F and UF treatments (Figs. 6 and S5), we primarily focus on discussing F treatment in sections concerning DOM composition unless otherwise stated.

Molecular composition analysis of DOM in September F treatment incubations indicates the removal of more saturated formulas and a decrease in the intensities of these formulas (H/C > 1.4; Fig. 6a), which indicates a decrease in the overall lability of DOM compounds (D'Andrilli et al., 2015, 2023). These changes could be explained by microbial degradation and/or abiotic aggregation of the saturated compounds. Simultaneously, an increase in the average aromaticity (AI_mod wa) of DOM compounds was observed during incubation and is likely due to the removal of saturated molecules, which tend to be low in aromaticity (Koch and Dittmar, 2006a). Additionally, the removal of low-molecular-weight compounds (< 250 Da) was also observed in mass spectra at the start of incubation (t0) versus at the end of incubation (t1; Fig. 7). These low-molecular-weight compounds are mainly composed of higher H/C saturation (H/C > 1.3; Fig. S6); thus, removal of these compounds could explain the lower average H/Cwa ratios observed. Microbial degradation of these compounds (<250 Da) is contrary to the size reactivity continuum, which proposes a higher reactivity of DOM as molecular weight increases (Benner and Amon, 2015).

Previous work has shown the formation of POM via adsorption of hydrophilic and low aromatic DOM (Einarsdóttir et al., 2020). The decrease in the average H/C saturation and the increase in the aromaticity of DOM compounds observed during September incubations could be due to the adsorption of these compounds to POM. However, aggregation of highly saturated DOM is typically observed for larger size fractions, such as polysaccharides (Passow, 2000; Passow et al., 1994), which are outside of our mass spectrometry analysis window. These low-molecular-weight compounds could point to DOM precursor molecules of the larger hydrophilic POM compounds (Orellana and Leck, 2015; Verdugo et al., 2004). The low SPE–DOC percentage recovery observed during September incubations (Fig. S9) supports the increase in these hydrophilic fractions. Although these dissolved fractions were not extracted in our DOM method, their presence could contribute to increases in POC concentrations, as observed during the biologically productive period.

Average oxygenation of molecules (O/Cwa) also increased during incubations in September. This could be due to the removal of formulas of low O/C ratios (<0.5) and high H/C ratios (>1.5), as seen in Fig. S8, instead of the production of highly oxygenated compounds. The preferential biological degradation of compounds with low oxygen numbers compared to oxygen-rich compounds was also observed by Riedel et al. (2016). Additionally, Maie et al. (2008) have shown aggregation of highly oxygenated tannin compounds; however, our experimental results did not show an increase in highly oxygenated compounds, likely due to a limited tannin source in this region, as shown by the low seasonality in the tannin region (Fig. S8).

POC aggregation during the experiment was higher in June compared to in April (increase of 1.2 µM POC in April vs. 2.6 µM in June for F water and an increase of 0.7 µM in April vs. 9.5 µM in June for UF water). Higher aggregation under post-bloom conditions in June compared to in April were not surprising as EPSs accumulate under increasing nutrient limitation and/or with increasing concentrations of senescent cells, as is the case during post-bloom conditions in summer (Engel, 2000; Hellebust, 1965; Mague et al., 1980; Myklestad, 1995; Passow, 2002b; Riebesell et al., 1995; Thornton, 2002). EPSs are released already in the growth phase of phytoplankton during the initiation of a bloom; however, most EPSs are probably present in the form of low-molecular-weight (LMW) compounds in the beginning of a spring bloom (Paulsen et al., 2018). This is supported by higher field DOC concentrations in April than in June, while field concentrations of EPSs, which are representative for large polysaccharides (>0.4 µm), were lower in April compared to in June (Fig. S2t). This might also explain why, in April, both the F and UF treatments had similar increases in POC levels relative to the start of incubation (increases of 1.24 and 0.73 µM, respectively), although initial POC concentrations at the start of incubation (t0) were 11 times higher in UF water than in F water. Despite particle concentrations being high, little precursor material was probably present in UF water to promote aggregation in the beginning of a bloom compared to a post-bloom scenario.

Later, in June, we observed a 9-times-higher aggregation potential of UF water compared to in April, while we only observed a 2-times-higher aggregation potential of F water (Fig. 4a). In summer, towards the end of a bloom, the field concentration of EPSs increase and there is also an increase in the concentration of suspended particles, as shown by an increase in POC and TPM in the field (Fig. S2y, a1). Because coagulation becomes more likely when a critical concentration of particles is reached, aggregation is usually enhanced under post-bloom conditions (Burd and Jackson, 2009; Dam and Drapeau, 1995; Mague et al., 1980; Passow, 2002b; Thornton, 2014). Colloids can be scavenged by larger particles (Druffel et al., 1992; Kepkay, 1994) and can even lead to enhanced carbon sedimentation (Forest et al., 2013; Riebesell et al., 1995). Higher POC and EPS concentrations in the field during June likely increased the chance of organic matter to coagulate in UF water. Interestingly, however, experimental POC concentrations in UF water were higher in April compared to in June. This might indicate that the concentration of precursor material plays a more critical role for aggregation than particle concentration. Our experiment suggests a higher contribution of DOM to the POM pool during summer compared to during spring and a higher aggregation potential in general (as seen in UF water) during a post-bloom scenario compared to during a peak-bloom scenario. It should be noted, however, that experimental EPS concentrations at the start of incubation (t0) in UF water were consistently 1 order of magnitude lower than the respective field EPS concentrations, which indicates that a large fraction of EPSs were retained on the 90 µm sieve and were therefore not included in the experiment, probably because of their sticky nature. Therefore, we likely did not adequately replicate aggregation in truly unfiltered water, and our measurements can be considered to be conservative estimates for aggregation in unfiltered water.

DOM molecular composition analysis during winter incubations (December and February) indicates a decrease in unsaturated DOM (Fig. 6) and an increase in the relative intensities of more saturated DOM (H/C; Fig. S8). Moreover, average aromaticity (AI_mod wa) decreased during this period, which could be due to the removal of low-H/C DOM, which is typically of higher aromaticity (Koch and Dittmar, 2006b). These observations indicate an increase in the average lability of DOM and could be partially due to the dissolution of organic particles observed during this period as this can lead to the production of more labile DOM. This is supported by the reduction in SPE–DOC percentage recovery observed during December incubations (Fig. S9), which indicates hydrophilic material at the end of incubation. Additionally, there was a significant decrease in a group of low-H/C compounds referred to as terrestrial peaks (t-Peaks) in February (t test, p = 0.04, t value = 2.96, df = 4) and October (t test, p = 0.04, t value = 3.02, df = 4). These t-Peaks are a group of compounds that are commonly present in vastly different rivers, as reported by Medeiros et al. (2016; Fig. S7). Removal of these compounds could contribute to the increase in average H/Cwa ratios observed in February and October incubations. This suggests a potential degradation of t-Peak compounds during February and October, which is in contrast to September (t test, p = 0.08, t value = -2.31, df = 4) and December (t test, p = 0.09, t value = 2.21, df = 4), when t-Peaks did not significantly change during incubation (Fig. S7). Arctic winter microbial communities differ substantially from spring, summer, and autumn communities (Marquardt et al., 2016; Vonnahme et al., 2022; Wietz et al., 2021); therefore, the winter community is likely to be better adapted to degrade various carbon sources, while summer communities are specialized in degrading phytoplankton-derived DOM, as shown in Wilson et al. (2017). These findings support that terrestrial DOM is an important carbon source for bacteria during less productive periods in Ramfjorden (Vonnahme et al., 2022), which has implications for the fate of riverine DOM in sub-Arctic fjords and a freshening Arctic Ocean. Additionally, the winter presents a period where heterotrophs have less competition from photoautotrophs for essential nutrients, which could promote degradation of seemingly recalcitrant DOM (Dittmar, 2015).

There was also a removal of relatively higher-molecular-weight compounds (570–700 Da), as shown in mass spectra (Fig. 7c). These higher-molecular-weight compounds were in the middle O/C and H/C ratio regions of the van Krevelen diagram (Fig. S6), which are typically rich in carboxyl groups (Broek et al., 2020; Hertkorn et al., 2006). We suggest that the decrease in these compounds could indicate colloid formation through the adsorption of higher-molecular-weight compounds (>700 Da), as has been previously observed for carboxyl-rich organic matter (Chin et al., 1998). This corresponds to the increase in EPS molecules (<0.45 µm) observed in February incubations. Notably, experimental POC concentrations during the incubations indicate a dissolution and not an aggregation of particles during the winter. This could be due to the low relative intensity of these higher-molecular-weight compounds (570–700 Da); thus, there is no detection of increased POC concentrations during this period.

Throughout the experiment, we expected that DOC and POC would behave antagonistically and, thus, that an increase in POC would see a corresponding decrease in DOC, resulting in a stable amount of total organic carbon (TOC; sum of DOC and POC). However, this was not always the case during our experiments; the most pronounced change in the carbon budget was an increase in both DOC (+180 µM in UF treatment) and POC (+1 µM) in September. This suggests either that our experiment was not a closed system or that there was an interchange with the inorganic C pool. Other studies also report synergistic relationships between POC and DOC during summer incubations and antagonistic relationships during freshet incubations (Keskitalo et al., 2022). This indicates a seasonal component to POC–DOC dynamics. Additionally, other incubation studies report large variability in changes in POC and DOC concentrations (Shakil et al., 2022). These findings highlight the challenges in organic carbon measurements as reported elsewhere (Gardner et al., 2003; Chow et al., 2022) and are likely due to the complex dynamics between POC and DOC, such as the sticky nature of EPS material which can be adsorbed into experimental containers (Chen et al., 2021). Notably, experimental DOC values of ultra-pure water blanks show average DOC concentrations of 31 µM ± 13 (mean ± SD), and DOM molecular analyses of ultrapure water experimental blanks reveal a relatively low number of contamination peaks (average number of peaks: 123 ± 33, mean ± SD), thus not affecting DOM characterization results and ruling out contamination as a source of significant carbon addition. Moreover, the DOC concentrations measured in the experiment are well within the range of the in situ measurements (ranging from 87 to 233 µM; Fig. S2k).

During the winter months, the carbon budget had a net loss of TOC (POC + DOC), as observed in February, which was most likely due to bacterial respiration of OM to CO₂. In both October and December, a decrease in POC (-0.4 and -2 µM, respectively) did reflect an increase in DOC (+74 and +27 µM, respectively); however, this indicates a substantial additional source of DOC. The high increase in TOC observed in September, October, and December could be partly explained by carbon fixation in the dark by phytoplankton, with Vonnahme et al. (2022) and Hoppe et al. (2024) describing low but measurable primary production during winter. This may play a role in the increase in carbon observed in the unfiltered treatments where phytoplankton was abundant. In the filtered treatments, we did, however, observe a rapid increase in the activity of bacteria, presumably due to the fact that, in this fraction, they were released from the grazing pressure from bactivorous protists. Further, viruses are present in the filtered treatment and are at their peak abundance (up to 1.3×107) in autumn. Viruses have a significant role in controlling microbial population (Suttle, 2007), which, during the lytic stage, can lead to a substantial production of DOM through viral lysis of microorganisms (Chen et al., 2022). We suggest that the production of DOC observed in September and October in the filtered treatment may be due to viral lysis of the otherwise rapidly growing bacterial community (relieved from grazing pressure and limited by neither carbon nor nutrients). In the unfiltered treatment, the presence of phytoplankton (new production) may explain the increase in DOC.